|Home | About | Journals | Submit | Contact Us | Français|
EphrinA1 is a glycosylphosphatidylinositol (GPI)-linked ligand for the EphA2 receptor, which is overexpressed in glioblastoma (GBM), among other cancers. Activation of the receptor by ephrinA1 leads to a suppression of oncogenic properties of GBM cells. We documented that a monomeric functional form of ephrinA1 is released from cancer cells and thus explored the mechanism of ephrinA1 release and the primary protein sequence. We demonstrate here that multiple metalloproteases (MMPs) are able to cleave ephrinA1, most notably MMP-1, -2, -9, and -13. The proteolytic cleavage that releases ephrinA1 occurs at three positions near the C terminus, producing three forms ending in valine-175, histidine-177, or serine-178. Moreover, deletion of amino acids 174 to 181 or 175 to 181 yields ephrinA1 that is still GPI linked but not released by proteolysis, underlining the necessity of amino acids 175 to 181 for release from the membrane. Furthermore, recombinant ephrinA1 ending at residue 175 retains activity toward the EphA2 receptor. These findings suggest a mechanism of release and provide evidence for the existence of several forms of monomeric ephrinA1. Moreover, ephrinA1 should be truncated at a minimum at amino acid 175 in fusions or conjugates with other molecules in order to prevent likely proteolysis within physiological and pathobiological environments.
Ephrins are ligands for the Eph receptors, the largest family of receptor tyrosine kinases, and are divided into two classes, A and B, based on the way in which they are attached to the cell membrane (18). EphrinA1 to -A5 are linked to the membrane via a glycosylphosphatidylinositol (GPI) anchor, while ephrinB1 to -B3 are anchored by a transmembrane protein domain followed by an intracytosolic domain (18). Due to their membrane localization, ephrins are able to engage in both forward signaling through the Eph-expressing cell and reverse signaling through the cell on which the ligand is attached (8, 38, 50). While more is known about reverse signaling through ephrinB ligands due to the presence of a cytoplasmic domain (49), recent studies began to shed some light on ephrinA ligands in reverse signaling (13, 28, 31, 41, 45, 46).
Like their ligands, the Eph receptors are divided into two groups. However, their division is based on the sequence homology of their extracellular domains, which determines the ephrin ligand with which the receptor will interact. In the majority of cases, EphA receptors bind to ephrinA ligands, and EphB receptors to ephrinB ligands. There are, however, some exceptions, as in the cases of ephrinA5 binding to and activating EphB2 (23) and EphA4 binding to both ephrinA and ephrinB family members (19, 25, 36, 57).
EphrinA1 is a ligand for the EphA2 receptor. The crystal structure of the ephrinA1/EphA2 complex revealed the insertion of the G-H loop of ephrinA1 into a channel on the surface of EphA2 in a 1:1 ligand/receptor complex (24, 26, 27). Ligand binding causes the autophosphorylation of EphA2 and leads to internalization and degradation of the receptor (62, 68, 70).
Eph-ephrin interactions play an important role in multiple normal physiological processes, such as axon guidance, boundary formation, and topographic mapping, as well as being involved in organization of the vasculature (50). However, some ephrin ligands and Eph receptors play a critical role in multiple human malignancies (66). Previous studies have focused on the ligand as a membrane-bound, GPI-anchored protein, capable of mediating juxtacrine signaling. In fact, ephrinA1 was originally identified as a membrane-bound (7) protein that also existed in a secreted form (5, 29, 56), but subsequent studies suggested the need for membrane attachment or clustering for ephrins to activate their cognate receptors (12). This requirement was thought to be due to the necessity of Eph receptors themselves to undergo clustering in order to be activated (16, 24, 61). Until recently, while there has been evidence put forth demonstrating the existence of soluble ephrins (3, 22, 65), there has been sparse evidence for functionality of a soluble, monomeric form of any member of the ephrinA family. In fact, soluble, unclustered ephrinA5 stimulated autophosphorylation of EphA5 only weakly, and it was proposed that soluble ephrinA5 is actually an antagonist of axon bundling (65). Although functional properties were not tested, ephrinA1 was released from cells, presumably by cleavage, after interaction with exogenous EphA2-Fc (17). Furthermore, ephrinA1 and ephrinA5 exist in soluble forms and are substrates for clustering by tissue transglutaminase (3). This clustering was considered necessary for formation of functional soluble forms of ephrinA family members. For example, ephrinA5 was shown to be oligomerized by tissue transglutaminase in vivo during the differentiation of cultured myoblasts (3).
More recently, we documented the existence of soluble, unclustered, monomeric, functional ephrinA1 (70). This monomeric form is released into the extracellular space most likely by proteolysis, because this release was prevented by a broad-range inhibitor of metalloproteases (71). Monomeric ephrinA1 acts in a similar fashion to the artificially clustered ephrinA1-Fc homodimer. Monomeric ephrinA1 induced phosphorylation and internalization of the EphA2 receptor and caused characteristic morphological changes in tumor cells and a decrease in the oncogenic potential of GBM cells (70). Subsequent studies identified soluble ephrinA1 in cultures of HeLa and SK-BR3 cells (2) and in cultures of normal endothelial cells when ephrinA1 was overexpressed (15). The former study demonstrated that this soluble form is important for cellular transformation and promotes growth in these cell lines (2). EphrinA1 is also found in the sera of patients with hepatocellular carcinoma, which is suggestive of release of a soluble form from patients' tumors (11).
Paradoxically, existing data support the role of ephrinA1 in both the initiation and progression, as well as the inhibition, of various cancers (6). In glioblastoma multiforme (GBM), the most common malignant brain tumor in adults, ephrinA1 and EphA2 are differentially expressed (68). Whereas ephrinA1 is downregulated in cell lines and patient tissue specimens, EphA2 is significantly overexpressed (68). Increased receptor expression is associated with increasing astrocytoma grade (69) and decreased patient survival (63). Among breast cancer cells, on the other hand, nearly half produce and release ephrinA1 (44). Therefore, the ephrinA1/EphA2 system is an attractive therapeutic target (14, 59, 67). Our current work defines, for the first time, a possible mechanism of release and the specific region within which cleavage of ephrinA1 takes place. These findings contribute to our knowledge on Eph/ephrins and will impact directly the design of ephrinA1-based anticancer therapeutics and imaging reagents.
U-251 MG and U-87 human GBM-derived cells and SK-BR-3 human mammary adenocarcinoma cells were obtained from the American Type Culture Collection (Manassas, VA) and grown as recommended by the supplier. U-251 MG medium for ephrinA1 transfectants was supplemented with 200 μg/ml Geneticin.
Subconfluent U-251 MG cells were treated with 1 μg/ml recombinant ephrinA1-Fc (R & D Systems, Minneapolis, MN) or 4 μg/ml monomeric ephrinA1 (a generous gift from Dimitar Nikolov, Memorial Sloan-Kettering Cancer Center, New York, NY) or with 1 μg/ml in-house-made recombinant. Phase-contrast microscopy was used to visualize cell rounding at 15 min, 2 h, and 4 h after treatment.
DNA [5 μg; ephrinA1 wild type in pcDNA, ephrinA1Δ(174–181) (see below) in pcDNA, ephrinA1Δ(175–181) in pcDNA, or empty vector control] was transfected into U-251 MG cells growing in Opti-MEM (Invitrogen, Carlsbad, CA) using Lipofectamine 2000 (Invitrogen). Cells were incubated for 24 h and Opti-MEM was replaced with U-251 MG growth medium containing 20% fetal bovine serum (FBS). Once cells reached confluence, they were split into 100-mm dishes and medium was supplemented with 800 μg/ml Geneticin to select clones. Individual clones were isolated and maintained in U-251 MG growth medium containing 200 μg/ml Geneticin.
Subconfluent cell cultures were washed with phosphate-buffered saline (PBS) and cells were lysed with radioimmunoprecipitation assay buffer (RIPA; PBS, 0.5% sodium deoxycholate, 0.1% SDS, and 0.5% Igepal) containing mammalian protease inhibitor cocktail and phosphatase inhibitor cocktail (Sigma, St. Louis, MO). Lysates were passed through an 18-gauge needle and incubated on ice for 30 min. Lysates were centrifuged at 14,000 × g for 10 min to pellet nonsoluble debris. Supernatant was removed and stored at −80°C until use. Total protein concentrations were determined using Bradford dye reagent (Pierce, Hercules, CA) and lysates separated by SDS-PAGE using 12% acrylamide gels. Proteins were transferred to a polyvinylidene difluoride (PVDF) membrane (Pierce, Rockford, IL) and blocked for 1 h with 5% milk in PBS–0.05% Tween 20. PVDF membranes were incubated with primary antibody diluted in 5% milk overnight at 4°C while shaking. Mouse monoclonal EphA2 clone D7 (1:1,000) antibody was purchased from Millipore (Billerica, MA). β-Actin (1:50,000) antibody was purchased from Sigma, and rabbit polyclonal ephrinA1 V-18 antibody (1:500) was purchased from Santa Cruz (Santa Cruz, CA). Alpha tubulin was from NeoMarkers (Fremont, CA). EphrinA5 polyclonal antibody (1:500) was purchased from Abnova (Walnut, CA). Membranes were washed 3 times for 5 min each in PBS–0.05% Tween 20 and incubated with secondary antibody (goat anti-mouse IgG or goat anti-rabbit IgG) conjugated with horseradish peroxidase (Sigma, St. Louis, MO) at a dilution of 1:5,000 in 5% milk for 1 to 2 h. Membranes were then washed as described above, and detection was performed using the Enhanced Chemiluminescence Plus Western blotting detection system (GE Life Sciences, Piscataway, NJ). Membranes were exposed to autoradiographic film for various times.
Conditioned medium was collected from subconfluent cell cultures and centrifuged for 10 min at 1,000 × g to remove debris. Supernatant was collected and equal amounts of medium were loaded per lane to analyze ephrinA1 by Western blot analysis. For SK-BR-3 medium analysis, medium was collected and concentrated 10× before Western blotting using 10,000 MWCO Amicon Ultra Centrifugal filter devices (Millipore).
U-251-eA1 cells were plated in 60-mm dishes, and after 24 h in culture, medium was removed from subconfluent cells and replaced with fresh growth medium containing various concentrations of inhibitor or an equal volume of vehicle control. Dishes were incubated overnight and samples of conditioned medium collected for ephrinA1 analysis by Western blotting. GM-6001 (Enzo Life Sciences, Farmingdale, NY) was dissolved in dimethyl sulfoxide, and 4-(2-aminoethyl) benzenesulfonyl fluoride (AEBSF; Sigma) was reconstituted in water.
For analysis of cell surface-bound ephrinA1, 2.5 × 105 cells were resuspended in 100 μl PBS plus 1% bovine serum albumin (PBS-BSA) and incubated on ice for 1 h to block nonspecific binding. Cells were centrifuged at 1,000 × g for 5 min and resuspended in 100 μl PBS-BSA and 4 μg polyclonal ephrinA1 antibody (R & D Systems) conjugated to Alexa Fluor 488 or 4 μg goat IgG conjugated to Alexa Fluor 488 as an isotype control. Antibodies were conjugated using the Alexa Fluor 488 monoclonal antibody labeling kit from Invitrogen (Carlsbad, CA). Cells were incubated on ice for 2 h, followed by two washes with 1 ml PBS-BSA. Cells were resuspended in 500 μl PBS-BSA and subjected to analysis of at least 20,000 events per sample with a Becton Dickinson FACStarplus flow cytometer.
A total of 3 μg ephrinA1-Fc was incubated with 500 μl PBS, U-251 MG cell-conditioned medium, heat-inactivated conditioned medium, medium containing 10% FBS, or serum-free medium for 3 or 6 h at 37°C. Western blot analysis was performed for ephrinA1. For human serum cleavage, 3 μg ephrinA1-Fc was incubated with human serum from human male AB plasma (Sigma) diluted in PBS for a total reaction volume of 500 μl. The reaction was carried out for 1 h at 37°C.
Equal volumes of serum-free medium, 10% serum-containing medium, or GBM cell-conditioned medium were loaded onto 7.5% acrylamide gels containing 1% gelatin. After separation, the gels were washed with 2.5% Triton X-100–50 mM Tris (pH 7.4). Gels were incubated overnight at 37°C in reaction buffer consisting of 10 mM Tris (pH 7.4), 10 mM CaCl2, 150 mM NaCl, and 1 μM ZnCl2. Gels were stained with Coomassie blue to visualize cleared zones of gelatinase activity.
A total of 200 ng human recombinant ephrinA1-Fc (a generous gift from Akiva Mintz, Wake Forest School of Medicine, Winston-Salem, NC) or ephrinA5-Fc (R & D Systems) was incubated with 200 ng of each multiple metalloprotease (MMP) (Enzo Life Sciences) in 50 mM Tris–10 mM CaCl2 buffer in a total volume of 20 μl per reaction. Incubations were at 37°C for 1 h and samples stored at −80°C until Western blot analysis.
Growth medium was removed from subconfluent cells, and the cells were washed 3 times with PBS. UltraDOMA-PF medium (Lonza, Walkersville, MD) was added to cultures and incubated for at least 48 h. Medium was collected from cells and centrifuged at 1,000 × g for 5 min, and supernatant was collected and stored at −80°C until use. A total of 160 ml medium was filtered with a 0.22-μm filter and loaded onto an ephrinA1-agarose affinity column (Santa Cruz) by fast-protein liquid chromatography (FPLC) at a rate of 1 ml/min. Bound ephrinA1 was eluted with 0.1 M glycine (pH 2.7) and the pH of the fractions neutralized with 1 M Tris (pH 9.0). Medium was passed over the column and eluted a total of 5 times, Western blotting of fractions was performed, and ephrinA1-containing fractions were pooled and concentrated.
The sample of ephrinA1 purified by affinity chromatography was subjected to SDS-PAGE and the gel was stained with Coomassie brilliant blue R250. The stained band was manually excised and digested with trypsin. The gel pieces were dried in a small plastic tube under vacuum in a Speedvac for 20 min without heat. Protein disulfide bonds were reduced and alkylated by the serial addition of 10 μl 10 mM dithiothreitol (DTT) followed by 10 μl 50 mM iodoacetamide. The gel pieces were subsequently washed, minced in 200 μl 100 mM NH4HCO3, and dehydrated by the addition of 100% CH3CN. Gel pieces were rehydrated with 6 μl of 0.020-μg/μl trypsin (Promega) in 50 mM NH4HCO3 and incubated for 15 min at 37°C. Then 50 μl of 50 mM NH4HCO3 was added before incubation overnight at 37°C. Duplicate gel pieces were digested with endoproteinase Asp-N, excision grade (EMD, Calbiochem), using the same conditions. This proteinase cleaves before Asp and Glu at pH 8.8.
The trypsin and Asp-N digests were analyzed by mass spectrometry (MS) following separation by reverse-phase chromatography. Aliquots were injected onto a Waters nanoAcquity high-performance liquid chromatograph (HPLC) (Milford, MA) attached to a nano-electrospray ionization source on an LTQ-Orbitrap mass spectrometer (ThermoFisher Scientific, San Jose, CA). Mass spectra from 300 to 1,600 m/z were acquired in the Orbitrap. The six most abundant ions were selected for concurrent MS-MS analysis in the linear ion trap.
The MS-MS data generated were searched against human sequences found in Swiss-Prot release 57.15 using Mascot (Matrix Science, London, United Kingdom). An error-tolerant search method was used that matches MS-MS spectra for peptides that result from nonstandard cleavages. Carboxamidomethyl-cysteine and methionine (oxidation) were used as fixed and variable modifications, respectively.
A total of 400,000 cells were plated in 60-mm cell culture dishes and grown overnight. Cells were washed twice with PBS, 0.5 U/ml phosphatidylinositol-specific phospholipase C (PI-PLC; Enzo Life Sciences) diluted in PBS was added to each dish, and dishes were incubated at 4°C for 20 min on a rocker. The PBS was collected from each dish, centrifuged for 5 min at 12,000 × g, and assayed for released ephrinA1 by Western blotting.
Recombinant ephrinA1-(19–175) [eA1-(19–175)] was produced using the baculovirus expression vector system from BD Biosciences (San Diego, CA). Sf9 insect cells were cotransfected with BD BaculoGold linearized baculovirus DNA and the baculovirus transfer vector pAcGP67B containing eA1-(19–175)-His. High-titer viral stocks were prepared by infecting Sf9 cells at a low multiplicity of infection (MOI) and harvesting the supernatant 5 days postinfection. Sf9 cells were then incubated with 2 ml of the high-titer viral stock, and the supernatant was harvested for protein purification. The pH of the supernatant was adjusted to 7.4, and the sample was filtered and loaded onto a HisTrap FF column (GE Healthcare, Piscataway, NJ) by FPLC. The column was equilibrated before being loaded with lysis buffer (50 mM NaH2PO4, 300 mM NaCl, 10 mM imidazole; pH 7.4), and His-tagged ephrinA1-175 was eluted from the column in a stepwise gradient with elution buffer containing 50 mM NaH2PO4, 300 mM NaCl, and 250 mM imidazole (pH 7.4). Fractions of 1 ml were collected and separated by SDS-PAGE and stained with Coomassie blue. EphrinA1-containing fractions were pooled and concentrated.
RNA was isolated from cells using the RNeasy kit (Qiagen, Valencia, CA). A quantity of 2 μg of total RNA was used to prepare cDNA with the High Capacity RNA to cDNA kit (Applied Biosystems, Foster City, CA). The ephA2 transcript levels were studied with real-time quantitative PCR based on the TaqMan methodology using the 7000 real-time PCR system (Applied Biosystems). Gene expression assays (Applied Biosystems) were used to quantify the target and control genes. These assays are mixtures of an unlabeled PCR primer and a TaqMan MGB probe (6-carboxyfluorescein [6-FAM]) at the 5′ end with a nonfluorescent quencher at the 3′ end, designed over an exon-exon boundary to specifically detect cDNA sequences. The assay identification numbers of the selected genes are Hs00171656_m1 for EphA2 and Hs99999903_m1 for β-actin. The expression of β-actin was used to normalize the cDNA template. A quantity of 50 ng of cDNA, universal TaqMan PCR master mix (Applied Biosystems), and gene expression assay primers were combined in a final volume of 25 μl. The amplification reactions, in triplicate, were performed as follows, using a 7000 real-time PCR system according to the manufacturer's instructions: 95°C for 10 min followed by 40 cycles of 95°C for 15 s and then 60°C for 1 min. Data were analyzed by the comparative cycle threshold (CT) method and graphed as fold change over control, with the control being U-251 parental cells.
Wounds were made in a confluent monolayer of cells with a sterile 200-μl tip, and growth medium was added. Phase-contrast microscopy pictures were taken of the same field at 0 h and 10 h. Distance of the wound was measured in five places for each of the three wounds for each cell type at each time point using ImagePro Plus software, and the distance traveled was calculated for graphical representation.
Downregulation of the EphA2 receptor and cell rounding are characteristic responses to ephrinA1. Therefore, monomeric ephrinA1 used in crystallographic studies of the structure of the EphA2/ephrinA1 complex (27) was tested on U-251 MG GBM cells, which naturally overexpress EphA2 and display very low levels of ephrinA1 (68). Treating U-251 MG cells with either monomeric ephrinA1 (Fig. 1A) or the artificially clustered, homodimeric ephrinA1-Fc (Fig. 1B) caused a marked decrease in immunoreactive EphA2 protein levels in a time-dependent manner. We also investigated the ability of monomeric ephrinA1 to induce cell rounding of GBM cells, indicative of EphA2 activation by the ligand (47, 70). Treatment of subconfluent cells with both monomeric and recombinant dimeric ephrinA1-Fc led to a drastic change in cell morphology, reflected by cell rounding as early as 15 min and diminishing by 4 h after monomeric ephrinA1 treatment or 2 h after dimeric ephrinA1 treatment (Fig. 1C). Therefore, monomeric ephrinA1, shown previously by crystallography to be in complex with EphA2 in a 1:1 ratio (27), is functionally comparable to its artificially clustered form, although the monomer evoked longer retention of a rounded cell shape.
We next sought to determine the way in which ephrinA1 is released from cancer cells. A previous report from our laboratory documented the involvement of proteolysis in the release of ephrinA1 (70). Therefore, we treated cells with various protease inhibitors and analyzed the media for released ephrinA1 (Fig. 2). Treatment of ephrinA1-transfected U-251 MG cells (U-251-eA1) with a broad-spectrum matrix metalloprotease (MMP) inhibitor, GM-6001, caused a dose-dependent decrease in ephrinA1 release into the medium (Fig. 2A). Additionally, treatment of SK-BR-3 cells, a breast cancer cell line which endogenously expresses ephrinA1, led to a dramatic dose-dependent decrease in ephrinA1 release (Fig. 2B), suggesting proteolysis as one mechanism by which ephrinA1 is released from cancer cells. In addition, a serine protease inhibitor, 4-(2-aminoethyl) benzenesulfonyl fluoride (AEBSF), also had a potent, dose-dependent inhibitory effect on ephrinA1 release into the medium (Fig. 2C). Given that serine proteases activate various MMPs (1, 51, 55), it is plausible that there is interplay between the serine and metalloproteases leading to ephrinA1 release. Importantly, the inhibition of ephrinA1 release into the cell medium is not due to a decrease in the overall production of ephrinA1, as total-cell lysates showed a corresponding, relative to tubulin, increase in protein after treatment with both GM-6001 (Fig. 2D) and AEBSF (Fig. 2E).
To examine whether a protease(s) present in the extracellular environment may be responsible for ephrinA1 cleavage, we incubated ephrinA1-Fc with U-251 MG cell-conditioned medium and analyzed the samples for the presence of monomeric ephrinA1 cleaved from ephrinA1-Fc (Fig. 3A). Conditioned medium, heat-inactivated conditioned medium, or unconditioned medium produced immunoreactive bands at a size of cleaved ephrinA1, approximately 23 kDa, at 3 h and at 6 h (Fig. 3A). Human serum, a more physiological milieu, also cleaved ephrinA1 from ephrinA1-Fc when diluted in PBS to a concentration as low as 0.1% (Fig. 3B). Of interest, MMPs such as MMP-2 and MMP-9 are present in serum-containing medium, as demonstrated by gelatin zymography (Fig. 3C).
Because GM-6001 inhibited ephrinA1 cleavage, and MMPs are likely candidates for components of conditioned media able to cleave ephrinA1, we next performed an acellular cleavage assay to determine which MMPs may participate in this process. EphrinA1-Fc was incubated with various MMPs for 1 h. Western blot analysis demonstrated that, among the panel of MMPs tested, the collagenases MMP-1 and MMP-13, and the gelatinases MMP-2 and MMP-9, cleaved recombinant ephrinA1-Fc in the most robust manner in the assay repeated more than five times (Fig. 3D), suggesting the involvement of these specific four MMPs in cleavage of ephrinA1 from the cell membrane.
To investigate whether other ephrinA proteins are susceptible to cleavage leading to their release into the extracellular environment, lysate and medium from the naturally ephrinA5-expressing U-87 GBM cell line were analyzed by Western blotting for the presence of ephrinA5. An immunoreactive band corresponding to ephrinA5 was detected in both the lysate and the medium (Fig. 4A), documenting that ephrinA5 is also released from cancer cells. In addition, ephrinA5-Fc was incubated with a panel of recombinant MMPs at 37°C for 1 h, and Western blot analysis demonstrated cleavage of ephrinA5 from ephrinA5-Fc (Fig. 4B). Although ephrinA1-Fc and ephrinA5-Fc are both susceptible to MMP cleavage, the patterns of MMPs that cleave each protein most robustly differ between the two ligands.
We next sought to elucidate the molecular form of ephrinA1 purified from the culture medium of GBM cells by mass spectrometry. Conditioned medium was collected from U-251-eA1 cells and five rounds of affinity chromatography were used to capture ephrinA1 from the medium. EphrinA1-containing fractions, as determined by Western blot analysis (Fig. 5A), were pooled and concentrated. Concentrated samples were subjected to SDS-PAGE and stained with Coomassie blue (Fig. 5B). The band corresponding to the position of ephrinA1 was excised, digested with proteases, and analyzed by tandem mass spectrometry. The Mascot (Matrix Science, London, United kingdom) search of MS-MS data obtained from peptide digests from the samples treated with Asp-N (Fig. 5C) and trypsin (Fig. 5D) matched peptides from the UniProtKB entry EFNA1_HUMAN (ephrin-A1) with statistically significant molecular weight search (MOWSE) scores. Four peptide cleavage products of ephrinA1 not arising from tryptic or Asp-N digestion were found in digests of the purified ephrinA1 sample. Three digest peptides were consistent with three different C-terminal fragments of ephrinA1: a peptide ending in valine-175 (LAADDPEVRV), with a monoisotopic mass of neutral peptide [Mr(calc)] of 1,165.57 (Fig. 6A); a peptide ending in histidine-177 (LAADDPEVRVLH), with an Mr(calc) of 1,333.70 (Fig. 6B); and a peptide ending in serine-178 (DDPEVRVLHS), with an Mr(calc) of 1,165.57 (Fig. 6C). Moreover, a peptide was identified ending in arginine-174 (LAADDPEVR), with an Mr(calc) of 984.49 (Fig. 6D). Although this is likely the result of trypsin cleavage during digestion, we did not rule out the possibility that this also represents a preexisting C terminus from ephrinA1 cleavage from the cell. Note that the peptides described in Fig. 5A to toCC do not correspond to the known cleavage specificity of the digestive proteases. Therefore, these peptides must have arisen from the existing C termini of soluble ephrinA1. Thus, ephrinA1 is cleaved from the cell membrane at at least three sites of cleavage, generating forms of soluble ephrinA1 within the extracellular environment ending in valine, histidine, serine, and possibly arginine (Fig. 6E).
Following observation by mass spectrometry analysis of three forms of ephrinA1, ending in amino acids 175, 177, and 178, we engineered ephrinA1 mutants lacking amino acids 174 to 181 and 175 to 181. An ephrinA1 was also generated with arginine 174 replaced by an alanine (R174A), and expression vectors were transfected into U-251 MG cells in order to determine whether this arginine was required for recognition by cellular proteases or was the result of trypsin digestion. To ensure that the mutants remain GPI linked like the wild-type form and therefore accessible to proteases within the extracellular environment, U-251 MG cells were transfected with the ephrinA1 mutants with deletions of amino acids 174 to 181 or 175 to 181 [eA1-Δ(174–181) or eA1-Δ(175–181), respectively] and retaining serine-182, the site of GPI anchorage. Amino acids 183 to 205 (hydrophobic tail) were also preserved because of their critical role in retaining the polypeptide at the endoplasmic reticulum (ER) membrane until GPI modification occurs and the protein is trafficked to the cell membrane (30). Flow cytometry performed on parental U-251 MG cells, wild-type-transfected cells, and each ephrinA1 mutant demonstrated the membrane localization of ephrinA1 in all cases (Fig. 7A). In addition, we documented that the membrane-bound status of ephrinA1 was indeed through a GPI linkage. Cells transfected with proteolytic-site-deficient mutants were treated with phosphatidylinositol-specific phospholipase C (PI-PLC), which releases GPI-linked proteins. Western blot analysis demonstrated that all the mutants were GPI linked to the membrane, just like wild-type ephrinA1 (Fig. 7B).
We next compared release of ephrinA1 from the cell surface of the mutants and that from wild-type transfectants. For reasons that are as of yet unclear, ephrinA1 was expressed at a higher level in eA1-Δ(175–181) transfectants (Fig. 8A). The levels of expression of ephrinA1 in wild-type, eA1-Δ(174–181), and R174A transfectants, however, were comparable (Fig. 8A), but only the wild type and R174A were released into medium (Fig. 8B). Given that mutation of arginine-174 had no effect on ephrinA1 release, we conclude that this peptide was generated during tryptic digestion before mass spectrometry analysis.
We next investigated whether differences in cell migration exist between wild-type ephrinA1 and eA1-Δ(174–181) transfectants. The eA1-Δ(174–181)-transfected cells were chosen because they display ephrinA1 protein levels similar to those of wild-type-transfected cells (Fig. 8A). There was no difference in migration as detected in a wound healing assay between wild-type and deletion mutants of ephrinA1 (Fig. 9A). The lysates from each transfectant were also collected, and Western blot analysis demonstrated downregulation of the EphA2 receptor in both the wild-type ephrinA1 and eA1-Δ(174–181) transfectants that was similar to that of vector-transfected cells (Fig. 9B). However, we found ephA2 expression to be significantly elevated in proteolytic-site-deficient mutants compared to the wild-type counterpart (Fig. 9C). Therefore, while gene expression is increased, there could be a feedback mechanism leading to the different regulation of the EphA2 receptor, including its activity.
Since ephrinA1 mutants lacking amino acids 174 to 181 and 175 to 181 were not released from the cell surface, we next tested the functionality of a truncated form of ephrinA1, eA1-(19–175). We used the baculovirus system in Sf9 cells followed by His tag purification to produce recombinant eA1-(19–175). U-251 MG cells were treated with 1 μg/ml eA1-(19–175) or 1 μg/ml homodimeric ephrinA1-Fc, and both forms induced morphological changes (Fig. 10A), consistent with ligand activation of the EphA2 receptor. Cells were also treated with eA1-(19–175) in an identical manner and total-cell lysates analyzed by Western blot analysis for EphA2. eA1-(19–175) downregulated the receptor in a time-dependent manner compared to time-matched controls (Fig. 10B). Therefore, eA1-(19–175) is fully functional as far as forward signaling is concerned and would be an attractive therapeutic or imaging ligand for targeting the EphA2 receptor.
We have demonstrated that ephrinA1, present on the cell membrane as a GPI-anchored protein, is cleaved at three positions within the C terminus, releasing three forms of the protein, eA1-(19–175), -(19–177), and -(19–178) (Fig. 11). Our findings also reveal that ephrinA1 is a substrate for cleavage by multiple MMPs which are likely responsible for proteolysis of membrane-bound ephrinA1 from cancer cells. These findings strongly suggest, for the first time, the mechanism of release and the forms of soluble ephrinA1.
In support of previous studies revealing the functionality of a soluble form of ephrinA1 cleaved from cancer cells (2, 70), the current study demonstrates downregulation of EphA2 and alteration of cell morphology in response to exogenous monomeric ephrinA1 treatment. Notably, the ephrinA1 protein used in those experiments is the same form as was used for crystallography studies in which it was shown that ephrinA1 and EphA2 bind in a 1:1 ratio (27). Functionality of soluble ephrinA2 has also been documented, and interaction between soluble ligand and receptor influences osteoclast differentiation (31). Interestingly, while we and others have documented functional soluble ephrinA1 activity (2, 70), clustered ephrinA1 was shown to suppress Slit2-mediated angiogenesis in endothelial cells, while the soluble form of ephrinA1 abrogated this effect (15). This seemingly contradictory role for ephrinA1 may highlight the importance of cell type and context in ephrin and Eph function.
We investigated the protease(s) possibly responsible for ephrinA1 cleavage using broad-spectrum inhibitors and determined that members of the MMP family are likely the responsible proteases. Inhibition of serine proteases also led to a decrease in ephrinA1 release likely due to serine proteases activating the MMP(s) (1, 51, 55). We demonstrated that monomeric ephrinA1 is released from cancer cells via cleavage by at least four different MMPs: MMP-1, -2, -9, and -13 (Fig. 11). In addition, mass spectrometry analysis of cleaved ephrinA1 purified from the medium of ephrinA1-transfected cells revealed three forms of ephrinA1, which is not a surprising finding considering the broad specificity of MMP cleavage and the functional overlap within the MMP family (35, 48, 54).
MMPs are a family of zinc-dependent endopeptidases known to play fundamental roles in tumor progression (20). Multiple MMPs are expressed highly in GBM and other human malignancies. Extensive research on the gelatinase family of MMPs (MMP-2 and MMP-9) has identified them as critical players in various stages of cancer growth, metastasis, and angiogenesis (34). MMP-2 and MMP-9 are overexpressed in GBM (10, 37, 53). In fact, the expression pattern of MMP-2 and MMP-9 correlates with increased invasion in vitro, and in vivo studies verified that increased MMP-2 and MMP-9 expression correlates with increasing tumor grade (40). Additionally, expression of collagenases, such as MMP-1 and MMP-13, is deregulated in various cancers (9). MMP-1 expression has been documented to occur in GBM cells (4, 21), and increased expression correlates with increased tumor grade (58, 71) and tumorigenicity (52), and decreased survival in GBM patients (71). Moreover, increased MMP-13 production in glioma cells in response to various stimuli leads to an increase in migration and invasion (43). Also of note, MMP-13 is a critical player in the activation of other MMPs (39).
Not only do MMPs cleave components of the extracellular matrix to facilitate cancer cell migration and invasion, but also they are responsible for the shedding of multiple membrane-bound proteins (20, 48). In fact, interaction of ephrinB1 with EphB receptors leads to reverse signaling within the ephrin-expressing cell, causing an increase in MMP-8 secretion and cleavage of the ligand (60). Additionally, the EphB2 receptor is cleaved by MMP-2 and MMP-9, and this release is ephrinB2 induced (42). EphrinA2 is also cleaved by a member of the MMP family (22, 31), and stimulation of EphA2 by the soluble form of ephrinA2 leads to a functional ligand-receptor interaction and osteoclast differentiation (31). On the other hand, ADAM13 cleavage of ephrinB1 and ephrinB2, as demonstrated by Wei et al., is not dependent on ligand-receptor interaction. In fact, they postulate that cleavage of the ligand prevents ligand-induced receptor activation and forward signaling into the Eph-expressing cell (64).
EphrinA5 and ephrinA2 are cleaved by the metalloprotease ADAM10 (a disintegrin and metalloprotease 10) (22, 32). Analysis of ADAM10 substrates revealed a conserved motif that is also present in ephrinA1 (17, 22). However, this motif lies within the G-H loop of ephrinA1, the region known to interact with and bind the EphA2 receptor. We have never detected an immunoreactive band of ephrinA1 present in conditioned medium that would correspond to the size of ephrinA1 cleaved toward the middle of the protein. Additionally, we have demonstrated the functionality of released ephrinA1, which would not be the case if cleavage of the ligand occurred within the receptor binding domain. Furthermore, cleavage studies implicating ADAMs in ephrin proteolysis indicate the requirement of receptor-ligand interaction in order for cleavage to occur (17, 22, 32, 33). Conversely, we demonstrated that ephrinA1 release is not Eph receptor dependent. In fact, ephrinA1 is present in conditioned medium of cancer cells even when plated at a density at which cell-cell contact is less likely to occur (70). Incubation of ephrinA1-Fc with GBM cell-conditioned medium in an acellular assay caused cleavage, as did incubation with nonconditioned, serum-containing medium, implicating a secreted protease in ephrinA1 release. In a physiologically relevant scenario, ephrinA1-Fc is also cleaved by human serum. Long-lasting effects of exogenous ephrinA1-Fc treatment in cell culture in previous studies may be due, in part, to monomeric ephrinA1 formed from cleavage of the homodimeric protein by secreted MMPs. While MMPs cleaved ephrinA1-Fc and also ephrinA5 in an acellular cleavage assay, incubation of ephrinA1-Fc with ADAM10 did not produce smaller ephrinA1 immunoreactive fragments (data not shown).
Similar to ephrinA1, ephrinA5 is also released from cells and has been demonstrated to be present in medium from cells endogenously expressing ephrinA5 as well as in cells in which ephrinA5 has been overexpressed (3). While previous studies have reported the cleavage of ephrinA5 from cells by ADAM10 (32, 33), our acellular cleavage assay suggests that it may be susceptible to cleavage by various other MMPs as well. Of note, while MMP cleavage patterns between ephrinA1 and ephrinA5 were different, both were cleaved readily by members of the collagenase class of MMPs (MMP-1, -8, and -13). Unlike ephrinA1, ephrinA5 was also readily cleaved by MMP-7.
When residues 174 to 181 or 175 to 181 of ephrinA1 are deleted, ephrinA1 is no longer released from membranes, thus confirming that proteolysis of ephrinA1 occurs within the region encompassing these amino acids. Comparison of proteolytic-site-deficient and wild-type ephrinA1-transfected cells did not show significant differences in the migratory capacity of the cells or in EphA2 protein levels. This is to be expected since extensive cell-cell contact occurs within these assays and, although ephrinA1 is not released from proteolytic-site-deficient mutants, it remains on the membrane and could still activate EphA2 upon contact or be a continuous partner for proteins involved in reverse signaling. Although the protein levels of EphA2 are comparable, the wild-type and proteolytic-site-deficient mutants display various levels of ephA2. Therefore, in this system, EphA2 may not be subjected solely to genetic regulation. Thus, this represents a complex system that needs to be explored in a systematic manner.
In addition, the longest proteolysis-resistant form of ephrinA1, ending at amino acid 175, retains the ability to activate the EphA2 receptor. This form of ephrinA1 or similarly truncated or mutated forms can be exploited for pharmaceutical targeting.
EphA2, the primary receptor for ephrinA1, is overexpressed in multiple human malignancies, making it a promising target for new cancer therapeutics (66). In GBM, a disease of dismal prognosis, EphA2 is highly overexpressed while the ligand ephrinA1 is nearly absent. On the other hand, EphA2 is not expressed in the normal adult brain (68). Therefore, ephrinA1 conjugated to an agent such as a toxin could be delivered specifically to GBM cells while sparing normal brain tissue (67). In order for the ephrinA1-based therapy to successfully reach its target, however, it is critical that proteolysis of the protein not occur within the tumor microenvironment, which would release the cytotoxic or imaging agent from its targeting ligand before encountering cancer cells. Our study has demonstrated that a form of ephrinA1 ending in amino acid 175 would represent a prototype cleavage-resistant functional binding unit serving such a purpose.
This work was supported by grants R01 CA74145 and CA139099 from the National Cancer Institute and by the Brain Tumor Center of Excellence, Wake Forest University.
We thank Dimitar Nikolov of Memorial Sloan-Kettering Cancer Center for the kind gift of monomeric ephrinA1 and for advice and protocols on baculovirus protein production. We also thank Sara Ferluga for her work on the optimization of the baculovirus system for ephrinA1 production in our laboratory.
Published ahead of print 11 June 2012