Identification of the F1-ATPase at the Cell Surface of Colonic Epithelial Cells

Background: The F1 domain of F1Fo-ATPase was initially believed to be strictly expressed in the mitochondrial membrane. Results: F1-ATPase is present at the cell surface of colonic epithelial cells and serves as a receptor for a gastrointestinal peptide mediating cell growth. Conclusion: We identified a new localization and a new function for the F1-ATPase in colonic cells. Significance: Cell surface F1-ATPase represents a new target in anti-proliferative strategies. F1 domain of F1Fo-ATPase was initially believed to be strictly expressed in the mitochondrial membrane. Interestingly, recent reports have shown that the F1 complex can serve as a cell surface receptor for apparently unrelated ligands. Here we show for the first time the presence of the F1-ATPase at the cell surface of normal or cancerous colonic epithelial cells. Using surface plasmon resonance technology and mass spectrometry, we identified a peptide hormone product of the gastrin gene (glycine-extended gastrin (G-gly)) as a new ligand for the F1-ATPase. By molecular modeling, we identified the motif in the peptide sequence (E(E/D)XY), that directly interacts with the F1-ATPase and the amino acids in the F1-ATPase that bind this motif. Replacement of the Glu-9 residue by an alanine in the E(E/D)XY motif resulted in a strong decrease of G-gly binding to the F1-ATPase and the loss of its biological activity. In addition we demonstrated that F1-ATPase mediates the growth effects of the peptide. Indeed, blocking F1-ATPase activity decreases G-gly-induced cell growth. The mechanism likely involves ADP production by the membrane F1-ATPase, which is induced by G-gly. These results suggest an important contribution of cell surface F1-ATPase in the pro-proliferative action of this gastrointestinal peptide.

F1 domain of F 1 F o -ATPase was initially believed to be strictly expressed in the mitochondrial membrane. Interestingly, recent reports have shown that the F1 complex can serve as a cell surface receptor for apparently unrelated ligands. Here we show for the first time the presence of the F 1 -ATPase at the cell surface of normal or cancerous colonic epithelial cells. Using surface plasmon resonance technology and mass spectrometry, we identified a peptide hormone product of the gastrin gene (glycineextended gastrin (G-gly)) as a new ligand for the F 1 -ATPase. By molecular modeling, we identified the motif in the peptide sequence (E(E/D)XY), that directly interacts with the F 1 -ATPase and the amino acids in the F 1 -ATPase that bind this motif. Replacement of the Glu-9 residue by an alanine in the E(E/D)XY motif resulted in a strong decrease of G-gly binding to the F 1 -ATPase and the loss of its biological activity. In addition we demonstrated that F 1 -ATPase mediates the growth effects of the peptide. Indeed, blocking F 1 -ATPase activity decreases G-glyinduced cell growth. The mechanism likely involves ADP production by the membrane F 1 -ATPase, which is induced by G-gly. These results suggest an important contribution of cell surface F 1 -ATPase in the pro-proliferative action of this gastrointestinal peptide. F 1 F o ATPase is a multisubunit transmembrane enzyme that was initially localized in the inner side of the mitochondrial membrane. Interestingly, it is now recognized that the F 1 complex, composed by ␣,␤,␥,␦,⑀ subunits, that contains the catalytic site for ATP synthesis or hydrolysis is expressed at the plasma membrane of different cell types and represents a cell surface receptor for apparently unrelated ligands (1). In particular, F 1 -ATPase at the cell surface of human umbilical vascular endothelial cells (HUVEC) 2 or hepatocarcinoma and lung cancer cells has been described as a receptor for angiostatin, a proteolytic fragment of plasminogen. In these cells, angiostatin binds to the cell surface F 1 -ATPase and inhibits ATP production, leading to cell death, anti-tumorigenic, or anti-angiogenic effects (2)(3)(4)(5). Endothelial cell surface F 1 -ATPase also serves as a receptor for apolipoprotein AI (apoAI). This lipoprotein stimulates cell surface F 1 -ATPase activity, leading to ADP production and an increase in proliferation of HUVEC (6). In addition, apoA-I interacts with the F 1 -ATPase at the cell surface of hepatocytes playing a role in HDL endocytosis (7). In the present study using immunofluorescence approaches and confocal microscopy, we demonstrate for the first time the presence of the F 1 -ATPase at the cell surface of colonic epithelial cells. More interestingly, using surface plasmon resonance (SPR) technology coupled to mass spectrometry we have identified a new ligand of the F 1 -ATPase, a peptide hormone product of the gastrin gene, the glycine-extended form of gastrin (G-gly). G-gly is the first gastrin precursor for which growth factor properties were demonstrated (8). Subsequently, numerous publications have reported the proliferative effects of G-gly on normal and cancerous gastrointestinal cells in vitro and in vivo (9 -18). In addition, G-gly induces tubule formation by human vascular endothelial cells in vitro in a similar manner to VEGF, suggesting a potential proangiogenic role for this peptide (19). So far, the membrane receptor for G-gly has not been identified. This gastrin precursor does not bind the receptor specific for the mature form of the hormone, and several publications show that G-gly proliferative effects are independent from the gastrin receptor (8,11,20). However, we and others have previously demonstrated the presence of high affinity G-gly binding sites on gastrointestinal cells, suggesting the existence of G-gly receptors at the cell surface (8,11). Here by molecular modeling and validation of this model with mutagenesis and biological analysis, we have characterized the molecular interaction between G-gly and the F 1 -ATPase. We have also identified the function of the F 1 -ATPase activated by G-gly at the cell surface of colonic epithelial cells.

EXPERIMENTAL PROCEDURES
Cell Culture-HCT116, HT29, and CACO-2 cells were obtained from the American Type Culture Collection (LGC Standards), HUVEC were from Millipore (Tebu Bio), and YAMC cells were kindly provided by Robert H. Whitehead (Vanderbilt University Medical Center, Nashville, TN) and grown as previously described (4,13,45).
Purified Membranes-Cells were scraped and lysed in a phosphate buffer, pH 7.4, through freeze/thaw cycles. After centrifugation for 15 min at 1000 rpm, the pellet was homogenized in 0.3 M sucrose then centrifuged for 2 h 15 min at 27,000 rpm in 10 ml of 1.63 M sucrose (final sucrose molarity, 1.56). The purified plasma membrane were collected at the interface of the sucrose gradient and diluted in 1 ml of a Tris buffer (20 mM, pH 7.4, supplemented by soybean trypsin inhibitor (0.1 mg/ml). The proteins concentration was determined by BCA assay kit (Pierce). F 1 -ATPase and IF1 Preparation-F 1 -ATPase was purified from bovine heart mitochondria as previously described (21) and kindly provided by J. Walker (MRC Mitochondrial Biology Unit, Cambridge, UK). The plasmid encoding the bovine mutated form IF1 (IF1-H49K, histidine 49 switched to lysine) was kindly provided by J. Walker, and protein expression and purification were carried out as previously described (22).
Surface Plasmon Resonance Assays-Studies based on SPR technology were performed on a BIAcore 3000 optical biosensor instrument (BIAcore AB, Uppsala, Sweden). Immobilization of biotinylated peptides was performed on a streptavidincoated sensorchip in HBS-EP buffer (10 mM Hepes, pH 7.4, 150 mM NaCl, 3 mM EDTA, 0.005% surfactant P20). All immobilization steps were performed at a final peptide concentration of 50 ng/ml (flow rate 10 l/min). Injections were stopped when a level of 350 RU was obtained. A channel was left empty and used as a reference surface for nonspecific binding measurements. The analyte was injected over the immobilized peptides for 4 min (flow rate 30 l/min) in a K-Inject mode. Kinetics constants (K a and K d ) were calculated using BIAevaluation 4.0.1 software. K d was calculated as the ratio of K d /K a . For recovery strategy, biotinylated G-Gly peptide was immobilized on 3 channels of a SA sensorchip at a level of 900 RU per channel with an empty channel as reference surface. Solubilized membrane proteins were injected for 6 min (20 l/min flow rate). Four cycles of recovery with a 50 mM triethylammonium, 0.5 M urea solution were performed to obtain 27,000 RU of recovered proteins corresponding to 27 ng of proteins. Recovered proteins were subjected to mass spectrometry analysis for identification (NanoLC/ESI QStar MS/MS, IPBS Toulouse Genopole).
Molecular Modeling-The sequence of G-gly was aligned to several species of IF1 using the ClustalX program (40) and adapted for the figure using the GeneDoc program. The amino acid sequence files with the code P01096, Q9UII2, Q03344, O35143, P01097, and P09940 were downloaded from the Expasy server. The secondary structure and three-dimensional model of G-gly are predicted using the PEP-FOLD server (26). The model of G-gly was rigidly docked to the crystal structure of F1-ATPase (PDB code 2V7Q) using the Glide module of Schrodinger software (Glide, Version 5.5, Schrödinger, LLC, New York, 2009). The box was centered based on the position of IF1 in the crystal structure of 2V7Q, and the size of the docking box, 30 ϫ 30 ϫ 30, was used. The SP scoring function of Glide was chosen, and 40 poses were generated. The selected docking pose was chosen for the optimization of the peptideenzyme interactions using 500-ps molecular dynamics simulations in implicit solvent using the Schrodinger software. The picture was prepared in the Schrodinger software (Maestro, Version 9.1, Schrödinger, LLC, New York, 2010).
Immunofluorescence and Confocal Microscopy-CACO-2 cells were plated on Transwell-Clear TM filters (Corning, Costar) at high density to obtain confluence in 5-8 days. 8 -10 days post-confluent CACO-2 cells (to obtain well polarized cells) were fixed and permeabilized using standard methods. For all other lines, cells were grown in Lab-Tek chambered cover-glass, and immuno-stainings were performed on intact cells, non-fixed, and non-permeabilized using standard methods. Antibodies used were: anti-␤-ATPase (Molecular Probes), anti-EGF receptor and anti-E-cadherin (Cell Signaling), and DPP-IV (R&D Systems). When mentioned, cells were incubated with 10 nM MitoTracker Red (Molecular Probes) for 15 min at 37°C. After washes in PBS, the cells were visualized by confocal microscopy. For immunostaining on the formaldehyde-fixed, paraffin-embedded tissues (colonic tissue of mouse), heat-induced epitope retrieval was performed in 10 mM citrate buffer, and primary antibodies were applied overnight (anti-␤-ATPase from Molecular Probes and anti E-cadherin from Cell Signaling). The detection was done using Alexa Fluor 488 goat anti-mouse antibody and Alexa Fluor 647 donkey anti-rabbit (Molecular Probes). Control slides, where the primary antibody was replaced by non-immune IgG, were checked for nonspecific reactivity. Slides were analyzed by confocal microscopy (magnificence ϫ40).
ADP Generation Assay-HUVEC in 6-well plates were washed and incubated for 1 h at 37°C in serum-free medium, then [␣-32 P]ATP (0.5 Ci/well) (PerkinElmer Life Sciences) was added for 10 min at 37°C with or without treatments. Nucleotides were then separated by HPLC and quantified as previously described (12).
Measurements of pH i by Flow Cytometry-The pH i was monitored using the pH-sensitive fluorescent probe carboxy-SNARF-1 (Invitrogen) and flow cytometry analysis. Cells were loaded with SNARF by incubating them in a 5 M solution for 20 min at 37°C in cell suspension buffer (CSB; 124.8 mM NaCl, 4.7 mM KCl, 1.2 mM KH 2 PO 4 , 10 mM HEPES, pH 7.4, 1 mM MgCl 2 , 1 mM CaCl 2 , 10 mM glucose). After trypsinization and resuspension in CSB, cells were stimulated with G-gly. The mean fluorescence intensity of 10,000 cells was determined after 10 min of treatment on MACSQuant analyzer (Miltenyi Biotec, Bergish Gladbach, Germany) (excitation 488 nm; emissions B2 channel, 585 nm, and B3 channel, 655 nm). The emission ratio 585/655 was then converted into pH value by using the calibration curve obtained in situ on control cells exposed to calibration buffers containing 10 M nigericin (135 mM KCl, 0.83 mM MgSO 4 , 1.26 mM CaCl 2 , 1.05 mM MgCl 2 , 10 mM glucose, and 10 mM MES, pH 5.5, 10 mM HEPES, pH 7.4, or 10 mM CAPSO, pH 8.7) (23). The dimethyl amiloride (Sigma), an inhibitor of the Na ϩ /H ϩ exchanger, was used as an acidifying positive control (24).

Identification of the Plasma Membrane F 1 F o -ATPase as a
Potential Binding Protein for G-gly-We previously reported G-gly proliferative effects on the human colon cancer cells HCT116 (9). Here we have purified plasma membranes from the HCTT16 cells. Solubilized membrane proteins were then tested for their interaction with G-gly by SPR (BIAcore 3000) using biotinylated G-gly immobilized on the BIAcore sensor chip. Biotin alone was coated as a control. As shown by the sensogram (Fig. 1), we observed a remarkable increase in the interaction with biotinylated G-gly compare with biotin alone indicating the presence of potential G-gly-binding proteins in the purified plasma membranes of HCT116 cells. After a recovery procedure, membrane proteins bound to G-gly were analyzed by mass spectrometry (nanoLC/ESI QStar MS/MS), and we identified, using the Mascot search engine, different subunits that compose the F 1 complex of the F 1 F o -ATPase (Table 1).
Because several publications have demonstrated that binding of two ferric ions to G-gly is essential for the biological activity of the peptide (25,26), we analyzed the interaction between the solubilized membrane proteins and G-gly employing SPR in the presence or absence of desferrioxamine (DFO), a ferric ions chelator, which is known to inhibit the proliferative effect of G-gly in the HCT116 cells (9). As expected, the sensogram in Fig. 2A shows a decrease (42%) of the G-gly binding to the solubilized proteins in the presence of DFO, suggesting a role of ferric ions in the interaction of G-gly to its potential binding proteins. Eluted proteins were resolved by SDS-PAGE and silver-stained (Fig. 2B). Interestingly, staining of the bands around 50 kDa was decreased by the pretreatment with DFO, indicating a decrease in the interaction of G-gly with these proteins. We specifically cut the corresponding bands from the gel and identified the proteins by mass spectrometry. The ␣and ␤-subunits of the F 1 -ATPase were clearly identified with a good score ( Table 2).
Direct Interaction of G-gly with F 1 -ATPase-We used the SPR technology to verify the direct interaction between purified F 1 -ATPase (Fig. 3A) and G-gly. The sensograms in Fig. 3B show a direct interaction between G-gly coated on the BIAcore sensor chip and the purified F 1 -ATPase used as an analyte. Binding of F 1 -ATPase to the immobilized G-gly was dose-dependent (3-50 nM) allowing us to determine the affinity between this multisubunit complex and the gastrin precursor, G-gly (K D ϳ 1.7 nM).
Computational Study of the G-gly Interactions with the F 1 -ATPase-We examined the molecular basis of interactions between G-gly and F 1 -ATPase by molecular modeling. The prediction of the secondary and three-dimensional structures of G-gly using the peptide structure prediction PEP-FOLD server shows that G-gly is a likely ␣-helix. We noted that G-gly has an E(E/D)XY motif also found in IF1, an ␣-helical natural inhibitor of F 1 -ATPase. Fig. 4A shows the sequence alignment of IF1 from different species and human G-gly in which the E(E/D)XY motif can be identified. Because the sequence fragment of IF1 corresponding to G-gly is a helix, it additionally supports the helical nature of G-gly predicted by the web server. The crystal structure of IF1 in the complex with F 1 -ATPase in which IF1 is bound to ␣ and ␤ subunits and makes few contacts with the ␦-subunit is available (PDB code 2V7Q). Given that our experimental studies predicted the direct interactions of G-gly with mainly ␣ and ␤ subunits and possibly the ␦-subunit, we docked the ␣-helical structure of G-gly to F 1 -ATPase, specifying the docking box around the position of IF1 in the crystal structure. Among generated docking poses of G-gly, we selected for detailed examinations, the pose in which the residues of the E(E/D)XY motif form contacts similar to IF1, thus preserving the molecular recognition pattern, which has been recently shown to be an essential component to the binding energy that  holds IF1 and F 1 -ATPase together (27). The position of G-gly and its interactions with F 1 -ATPase are shown in Fig. 4, B-D).
We propose that G-gly is bounded to the ␤-subunit through several interactions: (i) the salt bridges involving Glu-9 and -10 of G-gly and Arg-408 of the ␤-subunit, (ii) a hydrogen bond between Glu-9 and Tyr-381, (iii) hydrophobic and van der Waals interactions of Leu-5, Tyr-12, and Phe-17 with Met-393, Lys-401, Ile-388, Pro-453, and Gln-455 of the ␤-subunit, and (iv) hydrogen bonds between Gly-18 and His-477 and Gly-7 and Ala-401. G-gly also interacts with the ␣-subunit through hydrophobic and van der Waals interactions of Trp-4, Ala-11, and Trp-14 with Phe-403, Val-400, Leu-417, and Gln-396 of the ␣-subunit. There is no direct contact of G-gly with the ␦-subunit in this binding mode, although some docking solutions provided few contacts with the ␦-subunit.
The salt bridge between Glu-9 and Arg-408 corresponds to similar interactions between Glu-30 of IF1 and Arg-408. These interactions have been recently shown by mutagenesis to play an important role in bovine and yeast IF1 binding to F 1 -ATPase (27). In particular, mutation of bovine Glu-30 (yeast E25) to alanine strongly decreases the binding (27).
Loss of Interaction between the F 1 -ATPase and G-gly with a Mutation in the E(E/D)XY Domain-On the basis of our SPR studies and molecular modeling, we propose that binding of G-gly into F 1 -ATPase may occur in a similar fashion to IF1. To confirm this hypothesis, we mutated the Glu-9 residue of G-gly to alanine and analyzed by SPR the interaction between this mutated form of the peptide and the F 1 -ATPase. As shown by the sensograms in Fig. 5, the binding of G-gly coated on the BIAcore sensor chip to the purified F 1 -ATPase used as an analyte is abolished when Glu-9 of G-gly is mutated to alanine. The natural receptor of the mature amidated form of gastrin is the CCK2 receptor, a G protein-coupled receptor mainly localized in the stomach. However, we verified the potential interaction

TABLE 2 Proteins identification by mass spectrometry (nanoLC/ESI QStar MS/MS)
Solubilized plasma membrane proteins bound to G-gly immobilized on a sensor chip in the presence or absence of DFO were eluted, resolved by SDS-PAGE, and silver-stained. The bands around 50 kDa were cut and analyzed by mass spectrometry. MW, molecular weight; pI, isoelectric point.   NOVEMBER 30, 2012 • VOLUME 287 • NUMBER 49 between gastrin-amide (G-NH 2 ) and the F 1 -ATPase. As expected, we did not observe any interaction by SPR between the F 1 -ATPase and G-NH 2 (Fig. 5).

Localization of F 1 -ATPase at the Cell Surface of Normal and Tumor Colonic Cells-
The expression of F 1 -ATPase at the plasma membrane of normal and tumor colonic epithelial cells has never been reported. We performed immuno-staining with an antibody against the ␤-subunit of F 1 -ATPase on non-fixed non-permeabilized living cells. In these conditions, the ␤F 1 antibody interacts only with proteins localized at the cell surface. Analysis by confocal microscopy shows a ␤F 1 staining on the cell surface of two colon cancer cell lines, HCT116 and HT29, in which we previously reported a biological effect of G-gly (Fig. 6, a and d) (9, 28). These results suggest that the orientation of the F1-ATPase in the membrane is such that the F1 component is extracellular and available to the antibody. These patch-like structures did not colocalize with Mito-Tracker Red, a specific marker of mitochondria (Fig. 6, g and h).
Using an antibody directed against a membrane receptor (EGF receptor) as a control, we observed a staining at the cell surface similar to that of plasma membrane ␤F 1 (Fig. 6, b and e). In contrast, when cells were permeabilized (Fig. 6i), the F 1 -ATPase staining was intracellular, localized to mitochondria as expected.
Because the proliferative effects of G-gly have also been observed on normal colonic epithelial cells (11, 29 -31), we analyzed the expression of F 1 -ATPase at the cell surface of a nontransformed colon cell line, YAMC, on which a previous report has demonstrated the presence of specific binding sites for G-gly and a proliferative effect of the peptide (11). Similarly to what we observed for colon cancer cell lines, confocal microscopy analysis on non-fixed, non-permeabilized living cells show ␤F 1 staining on the cell surface of YAMC that does not colocalize with MitoTracker Red (Fig. 6j). In addition, G-gly proliferative effects have been previously reported on normal epithelial cells from the colonic mucosa of mice (22). Here we confirmed on mouse colonic tissue sections the plasma membrane localization of the F 1 -ATPase on normal colonic epithelial cells. Indeed, by confocal microscopy using antibodies that detect specifically, ␤F 1 or E-cadherin, known to be localized on the basolateral membrane, we observed a co-staining of the two proteins at the membrane of colonic epithelial cells (Fig. 7A). To confirm the location of the F 1 -ATPase on the basolateral membranes of polarized colonic epithelial cells, we used specific markers of, respectively, the apical or basolateral membranes on a monolayer of polarized CACO-2 cells. Confocal X-Y (top and bottom of the cells) (Fig. 7B) and Z sections (Fig.  7C) show the absence of apical labeling with the ␤F 1 antibody, whereas the staining for DPP-IV, a marker of apical membranes, is positive. On the contrary, we observed the presence of basolateral labeling for F1-ATPase as well as for E-cadherin.
Cell Surface F 1 -ATPase Activity Is Stimulated by G-gly-Several works have reported the presence F 1 -ATPase at the cell surface of HUVEC and demonstrated its role in angiogenesis. In particular, Radojkovic et al. (6) reported that F 1 -ATPase on the surface of endothelial cells serves as a receptor for apoAI. This . Probing G-gly-F 1 -ATPase interactions using molecular modeling. A, shown is sequence alignment of IF1 from Bos taurus, Homo sapiens, Mus musculus, Rattus norvegicus, Saccharomyces cerevisiae, Pichia jadinii, and the human G-gly. B, shown is the hypothetical location of G-gly in the x-ray structure of F 1 -ATPase bound to IF1. The x-ray structure is shown in ribbonlike representation, where ␣-subunits binding ATP or ADP or an ␣-nucleotide low affinity subunit are in red, green, and gray, respectively; ␤-subunits binding ATP or ADP or a ␤-nucleotide low affinity subunit are in brown, blue, and yellow, respectively; ␦-subunit is in cyan, and G-gly is in purple. The G-gly carbon atoms of backbone and side chains are shown in gray stick. C and D, shown is a close view of G-gly interactions with the ␣ and ␤ subunits. The hydrogen bond interactions are highlighted with the dotted purple line in C, and all favorable contacts are in the dotted green line in D. lipoprotein stimulates cell surface F 1 -ATPase activity, leading to ADP production and an increase in proliferation of HUVEC. Recent work showed that G-gly is capable to increase tubule and node formation by HUVEC (19). We, therefore, hypothesize that G-gly might stimulate the cell surface F 1 -ATPase activity in a manner similar to what is observed with apoAI. Extracellular ADP production by HUVEC or colon cancer cells were analyzed by measuring the generation of [␣-32 P]ADP from [␣-32 P]ATP in the cell media 10 min after the addition of the peptide, as previously described (6). In the presence of G-gly we observed an increased hydrolytic activity on HUVEC (ϩ46% Ϯ 6) (Fig. 8A) similar to that previously described for apoAI on the same cells (6). The ability of IF1 protein to specifically inhibit the hydrolytic activity of F 1 -ATPase is well documented. The efficiency of this natural inhibitor is pH-dependent. We, therefore, used a mutated form of IF1 (IF1-H49K) known to FIGURE 6. Localization of the ␤-subunit of F 1 -ATPase on colonic epithelial cells by confocal microscopy. Cells were stained for ␤-F 1 - ATPase (green, a, d, g,  h, i, and j), mitochondria (MitoTracker: red, g, h, and j), or a plasma membrane protein (EGF receptor ((EGFR)) (red, b and e) as described under "Experimental Procedures." Experiments (a-h and j) were performed on intact, non-fixed, non-permeabilized colonic epithelial cells normal (YAMC) or cancerous (HCT116, HT29) cells. Experiment i was performed on fixed and permeabilized HCT116 cells. Magnification, ϫ40). Results are representative of three-six independent experiments. NOVEMBER 30, 2012 • VOLUME 287 • NUMBER 49 be active at all pH values. As previously reported in the HUVEC cells, IF1-H49K decreases the basal extracellular ADP production (6) but more interestingly reverses ATP hydrolysis and ADP production in response to G-gly (Fig. 8A). Similar results were obtained on colonic epithelial cancer cells (ϩ21.5 Ϯ 2.3) (Fig. 8B), indicating that the membrane F 1 -ATPase is also functional on this cellular model and leads to ATP hydrolysis.

Colonic Cell Growth via Cell Surface F 1 -ATPase
Cell surface ATPase on endothelial cells has been previously shown to regulate the intracellular pH (32). Here we have analyzed whether G-gly induces intracellular acidification on colonic epithelial cell. pH i was measured using the pH-sensitive fluorescent probe carboxy-SNARF-1 and flow cytometry analysis as described under "Experimental Procedures." Experiments were performed in neutral extracellular pH conditions for which we observed ADP production and cell proliferation in response to G-gly. As shown in Fig. 8C, pH i is not significantly modified after 10 min of G-gly treatment as compared with control (respectively, 7.23 Ϯ 0.03 and 7.25 Ϯ 0.05). Similar results were obtained for longer times of stimulation (1-48 h; data not shown). The dimethyl amiloride, an inhibitor of the Na ϩ /H ϩ exchanger, was used as an acidifying positive control. As expected we observed at 10 min an acidification of the pH i with dimethyl amiloride compared with the control condition (6.84 versus 7.29, p value ϭ 0.0001).
Cell Surface F 1 -ATPase Mediates G-gly Proliferative Effects on Endothelial Cells and Colon Cancer Cells-To demonstrate the role of F 1 -ATPase in mediating the growth factor effect of G-gly, we performed proliferative assays on HUVEC and colon cancer cells in the presence of the F 1 -ATPase inhibitor, IF1-H49K. Stimulation of HUVEC (Fig. 9A) or HCT116 cells (Fig.  9B) by G-gly induced a significant increase of cell proliferation. These effects were completely abolished when the cells were pretreated with the F 1 -ATPase inhibitor, indicating that the effects of G-gly on cell survival are lost when the F 1 -ATPase is inhibited. To confirm the involvement of cell surface F 1 -ATPase in the growth factor effect of G-gly, proliferative assays were also performed in the presence of an anti-␤-F 1 -ATPase blocking antibody that was shown to bind to the cell surface   Fig. 9, the blocking antibody could abolish G-gly-induced cell proliferation on endothelial cells (Fig. 9A) and colon cancer cells (Fig. 9B) similarly to what we observed with IF1 on these cellular models. These results confirm that the cell surface F 1 -ATPase plays an important role in mediating growth factor effects of G-gly in both endothelial cells and colon cancer cells. We also used angiostatin, another ligand of membrane F 1 -ATPase known to inhibit endothelial cell proliferation. By treating colon cancer cells with a concentration of angiostatin previously used in other studies (6), we confirmed its ability to inhibit cell proliferation. In addition, G-gly proliferative effects were abolished when the colon cancer cells were pretreated with angiostatin (Fig. 9B).
On the basis of SPR studies and molecular modeling, we proposed that binding of G-gly into F 1 -ATPase involves the E(E/ D)XY motif conserved in the G-gly and IF1 sequences. Because mutation of the Glu-9 residue of G-gly into alanine strongly decreases the binding of the peptide to the F 1 -ATPase, we also tested the ability of the mutated G-gly to stimulate cell proliferation. As shown in Fig. 9C, the proliferative effect observed when colon cancer cells are stimulated with 1 nM G-gly is totally absent when a similar concentration of the mutated peptide is used (G-gly-E9A). Higher concentrations of G-gly-E9A were also tested in proliferation assays (10 -1000 nM), but none was efficient to stimulate cell proliferation (data not shown). As  . Cell surface F 1 -ATPase mediates G-gly proliferative effects on HUVEC and colon cancer cells. Proliferation of HUVEC (A) or HCT116 cells (B and C) was measured as described under "Experimental Procedures" by using an MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) test. Cells were treated with the native form of G-gly at 10 nM for HUVEC (A) or 1 nM for HCT116 (B and C). Cells were also treated with G-gly mutated on Glu-9 at 1 nM (C) or the amidated form of gastrin (G-NH 2 ) at 1 nM. When indicated, cells were pretreated with IF1-H49K, a natural inhibitor of the hydrolytic activity of F 1 -ATPase, an antibody against the ␤-subunit of F1-ATPase (50 g/ml) or angiostatin (1 M). Results are expressed as the percentage of change in cell proliferation as compared with untreated cells (set at 0). Data are the mean Ϯ S.E.; n ϭ 4 -6. NOVEMBER 30, 2012 • VOLUME 287 • NUMBER 49 expected, the proliferation of HCT116 cells that do not express the CCK2 receptor was not stimulated by gastrin-amide (Fig.  9C).

DISCUSSION
Although F 1 F o ATPase was initially believed to be strictly expressed in the inner membrane of mitochondria, recent studies have shown that the F1 complex, containing the catalytic site for ATP synthesis or hydrolysis, is expressed at the cell surface of endothelial cells and hepatocytes as well as hepatoma and lung cancer cells (1). This complex has been characterized as a cell surface receptor for different ligands. In the present study we have identified by mass spectrometry in the purified plasma membranes from colorectal cancer cells different subunits of the F1 complex (␣, ␤, ␦). In addition, using immunofluorescent approaches and confocal microscopy, we confirmed the presence of the F 1 -ATPase at the cell surface of different colon cancer cell lines and in the plasma membrane of normal colonic epithelial cells. F 1 -ATPase, at the cell surface of endothelial cells as well as hepatoma and lung cancer cells, has been described as a receptor for angiostatin that mediates cell death through the inhibition of ATP production (2)(3)(4)(5). In contrast, apoAI, another ligand of endothelial cell surface F 1 -ATPase, induces cell proliferation and survival by stimulating ADP production (6). Here, using SPR technology coupled to mass spectrometry, we have identified a gastrointestinal peptide, the glycine-extended form of gastrin, as a new ligand for the F 1 -ATPase that is localized at the cell surface of colonic epithelial cells. Because the gastrin precursor, G-gly, does not stimulate gastric acid secretion as does the mature form of the hormone, it was initially considered without biological activity. However, we were the first to demonstrate a proliferative effect of this hormonal precursor on tumor cells (8). G-gly is now clearly established as a growth factor for normal and cancerous gastrointestinal cells particularly from colon origin. However, the receptor that mediates the biological activity of this precursor peptide and might represent a new potential target in gastrointestinal cancers was unknown. In the present study we clearly demonstrate using purified F 1 -ATPase and synthetic G-gly in a SPR approach the direct interaction between G-gly and the F 1 -ATPase. Interestingly the affinity calculated in these experiments was similar to that previously reported in pharmacological studies for the G-gly high affinity binding sites (8,11). In addition, by molecular modeling we identified the motif in the G-gly sequence (E(E/D)XY) that directly interacts with F 1 -AT-Pase as well as the amino acid residues in the ␤-and ␣-subunits of the F 1 -ATPase that bind this motif. We confirmed this molecular model by mutation of the E(E/D)XY motif, in particular, replacement of the Glu-9 residue by an alanine resulted in a strong decrease of G-gly binding to the F 1 -ATPase.
We have also demonstrated the role of the F 1 -ATPase in mediating the growth factor effect of G-gly on both colon cancer cells and HUVEC. Indeed, blocking the F 1 -ATPase with a specific antibody or IF1, a natural inhibitor of the enzyme, abolished G-gly proliferative effects in these cells. The mechanism likely involves ADP production by the membrane F 1 -ATPase. Indeed, in the presence of G-gly the membrane F 1 -ATPase functions as an ATP hydrolase and produces extracellular ADP. These effects are reversed in the presence of IF1, known to inhibit the hydrolytic activity of F 1 -ATPase. In addition, G-gly activity is also abolished by mutation of the E(E/D)XY motif in the G-gly sequence, which directly interacts with the F 1 -ATPase. In contrast, G-gly does not regulate pH i in the experimental conditions (neutral pH i ) for which we observed ADP production and cell proliferation. Some membrane F 1 -ATPase ligands such as angiostatin have been previously reported to induce cell death by decreasing pH i but only in a low extracellular pH environment (33).
In vivo, transgenic mice overexpressing circulating G-gly exhibited increased colonic proliferation compared with wildtype controls (31). Similarly, continuous infusion of G-gly into rats or gastrin-deficient mice also resulted in an increase of colonic proliferation (29). These results suggest the existence on colonic epithelium of a putative receptor or binding protein for G-gly that mediates the proliferative effects of the peptide.
Here we have shown that the F 1 -ATPase, which directly binds G-gly, is present at the plasma membrane of colonic epithelial cells in a mouse model. It is also interesting to notice that in both in vivo models (G-gly transgenic mice or G-gly infusion), treatment of the animals with DFO, a ferric ions chelating agent, reverses the proliferative effects induced by G-gly, suggesting that the biological activity of the peptide is dependent on the presence of ferric ions (34). In vitro studies confirm that G-gly binds two ferric ions necessary to its biological activity (9,25,26). In accordance with these previous data, our results show that the binding of G-gly to its potential receptor is decreased in the presence of ferric ions chelator.
Several studies have suggested a link between F 1 -ATPase and colorectal cancer. Indeed, F 1 -ATPase has been shown to be up-regulated in primary colorectal tumors compared with the adjacent normal mucosa (35). In addition, an association between F 1 -ATPase overexpression with poor survival of the patients has been reported in stage III colorectal cancer (36). F 1 -ATPase expression is also up-regulated in liver metastasis of colorectal cancer as compared with primary tumor as well as in tumor spheroids of colon carcinoma cells compared with cells grown in monolayer (36,37). It is, therefore, interesting to notice that the new ligand of F 1 -ATPase that we have identified in the present study is also overexpressed in these tumor cells. Indeed, the gene coding for gastrin is a target of oncogenic pathways frequently activated in colon cancer such as APC/␤catenin or Ras pathways (38 -41). Therefore, high concentrations of gastrin precursors, including G-gly have been observed in colon tumors and in blood of patients with colorectal cancer (42)(43)(44)(45). These data suggest that the cell surface F 1 -ATPase may contribute to a pro-tumoral autocrine loop in colorectal cancer cells. Interestingly, an anti-tumor strategy using monoclonal antibodies against the ␤-subunit of F 1 -ATPase has been previously proposed for hepatoma because these antibodies have been shown to reduce growth of these tumors (46). In view of our results, targeting cell surface F 1 -ATPase might be also a new therapeutic approach for colon cancer. In colon cancer, both F 1 -ATPase and its ligand G-gly seem to be overexpressed; an alternative or complementary strategy to F 1 -ATPase blocking could be the use of antibodies against the gastrin precursors.