Membrane Targeting and Coupling of NHE1-IntegrinαIIbβ3-NCX1 by Lipid Rafts following Integrin-Ligand Interactions Trigger Ca2+ Oscillations*

The cyclic calcium release and uptake during calcium oscillation are thought to result from calcium-induced calcium release (CICR); however, it is unclear, especially in nonexcitable cells, how the initial calcium mobilization that triggers CICR occurs. We report here a novel mechanism, other than conventional calcium channels or the phopholipase C-inositol trisphosphate system, for initiating calcium oscillation downstream of integrin signaling. Upon integrin αIIbβ3 binding to fibrinogen ligand or the disintegrin rhodostomin, sodium-proton exchanger NHE1 and sodium-calcium exchanger NCX1 are actively transported to the plasma membrane, and they become physically coupled to integrin αIIbβ3. Lipid raft-dependent mechanisms modulate the membrane targeting and formation of the NHE1-integrin αIIbβ3-NCX1 protein complex. NHE1 and NCX1 within such protein complex are functionally coupled, such that a local increase of sodium concentration caused by NHE1 can drive NCX1 to generate sodium efflux in exchange for calcium influx. The resulting calcium increase inside the cell can then trigger CICR as a prelude to calcium oscillation downstream of integrin αIIbβ3 signaling. Fluorescence resonance energy transfer based on fluorescence lifetime measurements is employed here to monitor the intermolecular interactions among NHE1-integrin αIIbβ3-NCX1, which could not be properly detected using conventional biochemical assays.

The cyclic calcium release and uptake during calcium oscillation are thought to result from calcium-induced calcium release (CICR); however, it is unclear, especially in nonexcitable cells, how the initial calcium mobilization that triggers CICR occurs. We report here a novel mechanism, other than conventional calcium channels or the phopholipase C-inositol trisphosphate system, for initiating calcium oscillation downstream of integrin signaling. Upon integrin ␣ IIb ␤ 3 binding to fibrinogen ligand or the disintegrin rhodostomin, sodium-proton exchanger NHE1 and sodiumcalcium exchanger NCX1 are actively transported to the plasma membrane, and they become physically coupled to integrin ␣ IIb ␤ 3 . Lipid raft-dependent mechanisms modulate the membrane targeting and formation of the NHE1-integrin ␣ IIb ␤ 3 -NCX1 protein complex. NHE1 and NCX1 within such protein complex are functionally coupled, such that a local increase of sodium concentration caused by NHE1 can drive NCX1 to generate sodium efflux in exchange for calcium influx. The resulting calcium increase inside the cell can then trigger CICR as a prelude to calcium oscillation downstream of integrin ␣ IIb ␤ 3 signaling. Fluorescence resonance energy transfer based on fluorescence lifetime measurements is employed here to monitor the intermolecular interactions among NHE1-integrin ␣ IIb ␤ 3 -NCX1, which could not be properly detected using conventional biochemical assays.
In many excitable or nonexcitable cells, the concentration of free intracellular calcium oscillates with a period ranging from a few seconds to a few minutes. Such calcium oscillations are involved in a wide variety of cellular functions (1,2). It is generally believed that, except for minor variations, the cyclic increase and decrease of calcium results from an autocatalytic release of calcium in a process called calcium-induced calcium release (CICR), 2 followed by a slow negative feedback that terminates calcium release. The cytoplasmic free calcium is then taken up into the organelles to reset the cycle. Despite a general agreement on how calcium oscillation proceeds once the system has been turned on, various different mechanisms have been proposed to explain how the initial calcium mobilization is generated that triggers CICR.
As a general rule, calcium entry through voltage-gated channels in electrically excitable cells (3) or through agonist-receptor interactions in nonexcitable cells, such as epithelial cells, hepatocytes, or oocytes (4), is thought to initiate the CICR process (1,2). Typically, in nonexcitable cells, the binding of an agonist, such as a hormone, a growth factor, or an extracellular matrix, to the corresponding cell surface receptor initiates a series of reactions that end in the activation of phopholipase C (PLC) and the production of the secondary messenger inositol trisphosphate (IP 3 ) (1,2,4). IP 3 is thought to induce calcium release from the internal endoplasmic reticulum or mitochondria store, and governs the CICR mechanisms that modulate calcium oscillations (1,2,4). However, not all calcium oscillations found in nonexcitable cells are initiated by the PLC-IP 3 pathway (5). We report here a novel pathway whereby ion exchangers NHE1 and NCX1, by interacting with integrin ␣ IIb ␤ 3 , can play an active role in mobilizing intracellular calcium that triggers calcium oscillation.
Integrin ␣ IIb ␤ 3 is the most abundant membrane protein found in platelets (6). Many functions of platelet are mediated * Microscopy for this study was supported by the Biophotonics Interdisciplinary by integrin ␣ IIb ␤ 3 (7,8), and these are tightly associated with the control of intracellular calcium concentrations (9). Binding of integrin ␣ IIb ␤ 3 with plasma fibrinogen (Fg) or von Willebrand factor, for example, elicits a complicated series of calcium events, including calcium oscillations, which modulate both the inside-out and outside-in signaling pathways and regulate platelet thrombus formation (9). Calcium oscillations could also be induced in human platelets by plating the cells on a substrate coated with Fg or rhodostomin (rho), a disintegrin protein isolated from snake venom (10,11). Like platelets, Chinese hamster ovary cells expressing exogenous integrin ␣ IIb ␤ 3 on their plasma membrane (CHO ␣ IIb ␤ 3 ) also exhibit active calcium oscillations when plated on substrates coated with Fg or rho (12). Using CHO ␣ IIb ␤ 3 cells as an experimental model, we report here that calcium oscillations downstream of integrin ␣ IIb ␤ 3 signaling could be readily triggered by the combined function of two ionic exchangers, the sodiumproton exchanger NHE1 and the sodium-calcium exchanger NCX1, which are actively recruited from intracellular vesicles to plasma membranes and form molecular complexes with integrin ␣ IIb ␤ 3 by a lipid microdomain-or lipid raft-dependent mechanism.

Cell Models, Preparation of Substrates, and Pharmacological
Treatments-The CHO ␣ IIb ␤ 3 cell line and human integrin ␤ 3 cDNA was a gift from Dr. M. H. Ginsberg (The Scripps Research Institute, La Jolla, CA). Purifications of recombinant rho and isolations of human platelets from volunteers were as described previously (11). The fluorescently labeled proteins were made by using pDNR-Dual donor vector that contained gene constructs of interest and acceptor vector (pLP-AcGFP1-C or pLP-AcmRFP1-C) that contained fluorescence protein tags (BD Biosciences Clontech). A donor vector encoding an mRFP gene (13) (pLP-AcmRFP1-C) was constructed in the laboratory by replacing the GFP gene in pLP-AcGFP1-C with mRFP. Gene transfection was done using Fugene 6 (Roche Applied Science) following the manufacturer's protocol.
To prepare cell growth substrates, coverglasses or 5-m beads were incubated with Fg, rho, or poly-L-lysine (PLL) at 4 ϫ 10 Ϫ7 M for 90 min at room temperature, washed with PBS three times to remove the excessive protein, and then back-coated with 5% bovine serum albumin. The substrates were prepared freshly before the experiments.
Live Cell Imaging-For calcium/sodium imaging, suspended cells were labeled either with 2 M Fluo-4 or with 2 M Sodium Green (Molecular Probes) at 37°C for 10 min before cell plating and observed under time lapse fluorescence microscopy (16,17). Fluorescence signals were quantified by MetaMorph program (Universal Imaging, Downingtown, PA). In Fig. 1, a and c, -fold changes of intracellular calcium/sodium concentration were calculated by dividing the fluorescence intensity taken over time (F n ) with reference fluorescence intensity taken at 10 min after cell plating (F 10 ). In Fig. 1d, fluorescence intensity taken at 40 min (F 40 ) was divided by F 10 . The cell showing calcium oscillation was defined by having more than one calcium transient within a 10-min observation period.
TIRFM was performed using an Olympus system, equipped with a single mode TIRF fiber illuminator and a numerical aperture 1.45 objective lens. The images were acquired by a highly sensitive charge-coupled device camera (Hamamatsu Photonics KK ORCA ER) controlled by the Aqua Cosmos program (Hamamatsu Photonics KK). For live cell imaging, a climate chamber was used to maintain cell viability and activity.
FLIM-FRET-For FLIM-FRET measurements done in fixed cells, immunofluorescence staining was performed using fluorescein isothiocyanate and TRITC fluorochrome as FRET donor and acceptor, respectively. For live cell FLIM-FRET experiments, proteins of interest were conjugated with GFP or mRFP as a FRET pair was applied. The FLIM-FRET platform was constructed in the laboratory, as described previously (18). Briefly, time-correlated single photon counting was built on a modified two-photon laser-scanning microscope (Olympus FV300). Two-photon fluorescence was excited using an 800-nm mode-locked femtosecond laser (Mira F-900; Coherent Inc., Palo Alto, CA) that operated at 76 MHz. Fluorescence emission was detected using a photon-counting photomultiplier (H7422P-40; Hamamatsu Photonics KK) and SPC-830 PC board (Becker & Hickl GmbH, Berlin, Germany). Images were taken at 256 ϫ 256 pixel resolution. Data analysis via model function fitting, along with the instrument response function deconvolution and colorcoding was conducted by SPCImage version 2.8 software package (Becker & Hickl). In Fig. 3e, the histogram of all image pixels was plotted as a function of fluorescence lifetime value and smoothed by the Sabitzky-Golay algorithm to approximate a normal distribution (upper panel), from which mean and S.D. were quantified (Fig. 3e, lower panel, and Fig. 4c).
Isolation and Characterization of Cholesterol-enriched Cell Fractions-Protocols for isolating lipid rafts were as previously described (19) with some modifications. Briefly, CHO ␣ IIb ␤ 3 cells were plated on different substrates as indicated at 37°C for 20 min, lysed with lysis buffer (20 mM Tris-HCl at pH 7.5, 0.5 mM EDTA, 0.1% protease inhibitor cocktail, 2 mM phenylmethylsulfonyl fluoride, and 1% Triton X-100) at 4°C for 4 h. A 10 -60% continuous sucrose gradient was used to fractionate total cell lysate into 13 fractions, collected from the top. The cholesterol content within each fraction was analyzed by the Amplex Red cholesterol assay kit (Molecular Probes). The protein content within each fraction was first precipitated with 10% tricholoroacetic acid and then analyzed by 6.5% SDS-PAGE and Western blotting, using 1:1000 anti-integrin ␣ IIb , 1:1000 anti-NHE-1 (Chemicon), and 1:500 anti-NCX-1 (Swant, Bellinzona, Switzerland) as primary antibodies, followed by peroxidase-conjugated secondary antibody, and detected by an enhanced chemiluminescence system (Pierce).

Net Calcium Influxes Generated by Combined NHE1-NCX1
Effects Induced Calcium Oscillations in CHO ␣ IIb ␤ 3 Cells FIGURE 1. NHE1 and NCX1 were involved in calcium oscillations of CHO ␣ IIb ␤ 3 cells or human platelets when grown on Fg or rho substrates. a, calcium imaging was performed using Fluo-4 in single live CHO ␣ IIb ␤ 3 cell grown on control PLL, Fg, or rho substrates. Changes of intracellular calcium concentration were quantified over time using fluorescence intensity taken at 10 min after cell plating as the reference. b, the percentage of CHO ␣ IIb ␤ 3 cells or human platelets exhibiting active calcium oscillations was calculated. Cells grown on the substrates as indicated were pretreated with calcium-free medium (5 mM EGTA), 3 M Tg, 20 M BAPTA-AM, 2 M U73343 or U73122, 5 M xestospongin C, 25 M EIPA, 80 M bepridil, 40 M KBR7943, or 1, 5, or 10 M M␤CD before calcium imaging. Data shown are mean Ϯ S.D. from more than 100 cells in three separate experiments. c, sodium imaging was done using Sodium Green. CHO ␣ IIb ␤ 3 cells grown on rho substrate were pretreated with mock solution (black trace), 25 M EIPA alone (blue trace), 80 M bepridil alone (red trace), or both (purple trace) at 37°C for 10 min before sodium imaging. Changes of intracellular sodium concentration over time were quantified in single live cell. d, -fold changes of intracellular sodium concentration were calculated by dividing the fluorescence intensity of Sodium Green measured at 40 min by that taken at 10 min after cell plating (F 40 /F 10 ). CHO ␣ IIb ␤ 3 cells grown on rho substrate were pretreated with drugs as indicated. Data shown are mean Ϯ S.D. from at least 85 cells in three separate experiments. e, differential interference contrast images of CHO ␣ IIb ␤ 3 cells grown on rho substrate, pretreated or not with the drugs as indicated for 20 min. Bar, 5 m. *, p Յ 0.0001.
Grown on Fibrinogen/Rhodostomin Substrates-CHO ␣ IIb ␤ 3 cells adhered strongly to the surface coated with Fg (12) or the disintegrin rho (supplemental Fig. 1a). The interactions between integrin ␣ IIb ␤ 3 and such ligands could further induce cell spreading (supplemental Fig. 1, b and c, and supplemental Movie 1), which coincided with the occurrence of intracellular calcium oscillations that started at ϳ10 min after plating CHO ␣ IIb ␤ 3 cells on a substratum coated with either Fg or rho ( Fig. 1a and supplemental Movie 2); growing CHO ␣ IIb ␤ 3 cells on the control PLL-coated substrate, on the other hand, did not result in any noticeable calcium transient.
More than 60% of the adherent CHO ␣ IIb ␤ 3 and human platelet cells on Fg/rho substrates exhibited calcium oscillations (Fig. 1b, open bars). Treating these cells with calcium-free medium, Tg, or BAPTA-AM effectively inhibited calcium oscillation (Fig. 1b, gray and black bars), suggesting that both extracellular and intracellular calcium pool participated in the cal-cium oscillation triggered by Fg/rho. Interestingly, inhibition of calcium mobilization through the conventional PLC-IP 3 pathway by U73122 or xestospongin C only partially inhibited the calcium oscillations triggered by Fg/rho (Fig. 1b, dark blue and purple bars), whereas control homolog U73343 had no obvious effect (Fig. 1b, light blue  bars). Also, Goncalves et al. (20) had previously reported that platelets isolated from PLC knock-out mice still exhibited calcium oscillation when triggered by Fg, despite having a delayed onset. Taken together, these findings suggested the presence of additional pathway(s), other than PLC, that might be involved in triggering the calcium oscillations downstream of integrin ␣ IIb ␤ 3 .
To delineate the putative alternative pathway, we noticed that NHE1 and NCX1 have been reported to be involved in calcium mobilization of platelet (21,22) and activation of platelets (or inside-out signaling) (23-25). In our system, treating CHO ␣ IIb ␤ 3 cells with bepridil and KB R7943, which inhibited the function of sodium-calcium exchanger NCX1 could effectively inhibit the calcium oscillation induced by Fg or rho substrates (Fig. 1b, red bars). NCX1 could potentially generate a local increase of calcium (to trigger CICR) by operating in a "reverse" mode, using sodium efflux in exchange for calcium influx (26). The question then was how the high local sodium concentration was generated that drove NCX1 in the reverse mode. Indeed, treating CHO ␣ IIb ␤ 3 cells grown on Fg or rho substrates with bepridil did cause a progressive increase of sodium (Fig. 1, c and d, and supplemental Movie 3) and cell swelling as more water flowed inside the cell (Fig. 1e). We noticed that the sodium increase and cell swelling caused by bepridil could be reversed by co-treating the cells with EIPA that perturbed the function of sodium-proton exchanger NHE1 (Fig. 1, c-e). Furthermore, EIPA treatment alone also prevented calcium oscillation (Fig.  1b, pink bars). These results suggested that NHE1 could use proton efflux in exchange for sodium influx; the neighboring NCX1 could then take the local high sodium to drive calcium influx that triggered CICR and calcium oscillation (27)(28)(29)(30)(31). Such combined NHE1-NCX1 activity appeared to also function in human platelets, since either EIPA or bepridil could effectively inhibit calcium oscillation induced by the cells' binding to Fg (Fig. 1b). Targeting of NHE1 and NCX1 from intracellular vesicles to the plasma membranes. a, Western blotting analysis revealed the presence of NHE1 and NCX1 in total cell lysate of cultured CHO ␣ IIb ␤ 3 cells. b and c, CHO ␣ IIb ␤ 3 cells plated on PLL, Fg, or rho substrates for 20 min were subjected to immunofluorescence staining to reveal the subcellular distributions of integrin ␣ IIb and either NHE1 (b) or NCX1 (c). Note that NHE1 and NCX1, originally found mainly in intracellular vesicles when plating CHO ␣ IIb ␤ 3 cells on control PLL substrate, were targeted to the plasma membranes when grown on Fg or rho substrates (arrowheads). d, a CHO ␣ IIb ␤ 3 cell expressing NHE1-mRFP and integrin ␤3 -GFP was held in contact with a bead (asterisk) coated with rho. Time lapse fluorescence imaging was done by a confocal microscope. Time interval for the image sequence shown is 40 s. Note the progressive approaching and close apposition by the NHE1-mRFP-containing vesicular structures toward the plasma membrane in contact with the bead (arrowheads). e, a CHO ␣ IIb ␤ 3 cell expressing NHE1-GFP was plated on Fg substrate for 10 min before being examined under TIRFM. A vesicle containing NHE1-GFP was noticed to approach the evanescent field of TIRFM (red arrowheads) and then fused with the plasma membrane as the fluorescence signals dispersed locally (yellow arrowheads). The time interval for the image sequence shown is 759 ms. Bar, 5 m.
Targeting of NHE1-NCX1 to the Plasma Membrane When Plating CHO ␣ IIb ␤ 3 Cells on Fibrinogen/Rhodostomin Substrates-Endogenous NHE1 and NCX1 proteins in CHO ␣ IIb ␤ 3 cells were revealed by Western blotting analysis (Fig. 2a). Although the amounts of ion exchanger did not change significantly when the cells were plated on different substrates, their subcellular distributions varied. In Fig. 2, b and c, immunofluorescence staining showed that the ion exchangers were present mainly in the cytoplasm as punctate vesicles but were absent from the cell membrane of CHO ␣ IIb ␤ 3 cells when plated on PLL control substrate for 20 min. In contrast, CHO ␣ IIb ␤ 3 cells grown on Fg or rho substrates for 20 min resulted in recruitment of NHE1-NCX1 from the intracellular vesicular compartments to the plasma membranes (arrowheads), where they co-localized with integrin ␣ IIb ␤ 3 .
In Fig. 2d, the dynamic redistribution process of fluorescently labeled NHE1 (NHE1-mRFP) was carefully monitored by confocal microscopy in a live CHO ␣ IIb ␤ 3 cell that was held in contact with a bead coated with rho (asterisk). As shown in the zoomed-in image sequence, vesicular or tubular structures that contained NHE1-mRFP were noticed to appose the plasma membrane to where the bead bound (arrowheads). We have also employed total internal reflection fluorescence microscopy (TIRFM) to monitor fusion by an intracellular vesicle to the plasma membrane (supplemental Movie 4). A typical example of such vesicle fusion event is shown in Fig. 2e. A vesicle containing GFP-labeled NHE1 was seen entering the evanescent field of TIRFM (Fig. 2e, red arrowheads); the vesicle's fluorescence then locally dispersed in a pattern that was characteristic of vesicle fusion (yellow arrowheads). In contrast, we seldom detected vesicle in the evanescent field; nor did we see any discernible occurrence of vesicle fusion if plating CHO ␣ IIb ␤ 3 cells on PLL substrate (supplemental Movie 5).
FLIM-FRET Measurements Revealed Direct Molecular Interactions between Integrin ␣ IIb ␤ 3 and Ion Exchanger NHE1 or NCX1-Despite the strong functional link between integrin ␣ IIb ␤ 3 and NHE1 or NCX1 (Fig. 1) and their co-localizations on the plasma membrane (Fig. 2), extensive tests using conventional co-immunoprecipitation assays failed to demonstrate their protein-protein interactions in a convincing way. Does integrin ␣ IIb ␤ 3 bind to the ion exchangers? To further explore this question, we employed the FRET technique to FIGURE 3. FLIM-FRET experiments revealed intermolecular binding between integrin ␣ IIb ␤ 3 and NHE1 or NCX1. a, CHO ␣ IIb ␤ 3 cells were plated on different substrates as indicated at 37°C for 20 min and then subjected to immunofluorescence staining. NHE1 labeled with fluorescein isothiocyanate was used as FRET donor, whereas integrin ␣ IIb labeled with TRITC was used as a FRET acceptor. In donor-only control, the FRET acceptor was omitted. Both the FRET donor's fluorescence intensity (Int) and lifetime (Lt) were recorded and displayed in gray scale or pseudocolor (ranging from 1200 to 2500 ps), respectively. In zoom-in insets, strong FRET focal areas (red spots as indicated by arrowheads) were noted along the plasma membrane when CHO ␣ IIb ␤ 3 cells were plated on Fg or rho substrates but not on the control PLL substrate. b and c, NCX1-integrin ␣ IIb and NCX1-NHE1 pair was used as the FRET pair, respectively. d, the same setting as in a, but the FRET donor's fluorescence intensity and lifetime were recorded before and after photobleaching the FRET acceptor. e, FRET donor's fluorescence lifetime values of all image pixels from the FLIM images as shown in a-c were plotted as histograms (thin traces), to which the Sabitzky-Golay algorithm was applied to approximate normal distributions (thick traces). From individual histograms' normal distributions, Means Ϯ S.D. were calculated from at least seven sets of cells in three separate experiments (lower panel). *, p Ͻ 0.0001; **, p Յ 0.005; bar, 5 m. reveal any putative intermolecular interactions that could not be properly studied by conventional biochemical methodology. Notably, FRET in this study was determined by quantifying the donor fluorochrome's FLIM. The reduction of FRET donor's FLIM when present very close to (i.e. Ͻ10 nm) the acceptor fluorochrome was indicative of FRET, or intermolecular binding between the paired molecules. The use of FLIM-FRET should be comparable with and, in some cases, complementary to fluorescence intensity-based FRET measurements that apply the ratio imaging method (see "Discussion") (18,32).
Typical FLIM-FRET results between integrin ␣ IIb and NCX1 or NHE1 or between NCX1 and NHE1 are shown in Fig. 3. In Fig. 3a, CHO ␣ IIb ␤ 3 cells were subjected to immunofluorescence staining; NHE1 stained with fluorescein isothiocyanate fluorochrome was used as FRET donor, whereas TRITC-labeled integrin ␣ IIb was used as FRET acceptor. The fluorescence intensity images of FRET donor are shown in gray scale, whereas the corresponding FLIM images (ranging from 1200 to 2500 ps) are displayed in pseudocolor. Note that FLIM images of the CHO ␣ IIb ␤ 3 cells grown on Fg or rho substrates significantly shifted to yellow/red across the entire cell, as compared with the cells grown on the control PLL substrate or the cells containing FRET donor only. In the latter two cases, long lifetime values were obtained, and the FLIM images were pseudocolor-coded in blue/green. Note also that in CHO ␣ IIb ␤ 3 cells grown on Fg or rho substrates, certain focal spots along the plasma membranes exhibited particularly strong FRET (red spots indicated by arrowheads). These were the regions where strong intermolecular interactions between NHE1 and integrin ␣ IIb took place. Similar findings were obtained using NCX1 as FRET donor and integrin ␣ IIb as FRET acceptor (Fig.  3b). Interestingly, FRET between NCX1 and NHE1 (Fig. 3c) was less evident than FRET between individual ion exchangers and integrin ␣ IIb .
To confirm that the reduction of NHE1 donor's FLIM was actually due to the energy transfer (or FRET) to the integrin ␣ IIb acceptor, we photobleached the acceptor fluorochrome and repeated the donor's lifetime measurement (Fig. 3d). Indeed, photobleaching of the FRET acceptor effectively prevented the reduction of donor fluorochrome's lifetime, suggesting that detecting decrease of the donor's lifetime was an accurate way to reveal the occurrence of FRET. To quantify the degrees of FRET, lifetime values of individual image pixels were calculated and displayed in histograms or as mean Ϯ S.D. (Fig.  3e). Using NHE1-integrin ␣ IIb and NCX1-integrin ␣ IIb as FRET pairs, we noticed that CHO ␣ IIb ␤ 3 cells plated on the Fg (red symbols) or rho (blue symbols) substrates exhibited profound donor lifetime reductions, compared with the cells grown on the PLL substrate (black symbols), or the cells containing donor fluorochrome only (gray symbols). Note also from these results that FRET between the NCX1 and NHE1 pair was less evident than FRET pairs between integrin ␣ IIb and individual ion exchangers.

Molecular Interactions between Integrin ␣ IIb ␤ 3 and Ion Exchangers Also Occurred in Intracellular Vesicular
Compartments-As shown in Fig. 3a, FRET between integrin ␣ IIb ␤ 3 and ion exchangers was found not only on the cell membrane but also in the intracellular vesicular compartments. To further explore this phenomenon, time lapse FLIM imaging (Fig. 4) using NHE1-GFP and integrin ␤ 3 -mRFP as the donor and acceptor of FRET pair, respectively, was performed on a live CHO ␣ IIb ␤ 3 cell, which was held in contact with a 5-m rho-coated bead (asterisks). The resulting donor's intensity and FLIM images were shown in gray scale and pseudocolor (ranging from 1600 to 2600 ps), respectively. We noticed that FRETpositive areas (yellow/red) did appear in the cytoplasm as punctate intracellular vesicles, and these progressively increased over time (Fig. 4a), whereas the donor-only control remained free of FRET. Interestingly, such FRET between integrin ␤ 3 and NHE1 in the intracellular vesicular compartments could be effectively reduced by treating the cell with Tg that depleted intracellular calcium store or partially with M␤CD that disrupted lipid raft integrity (Fig. 4b). On the other hand, genistein, a tyrosine kinase inhibitor; LY294002, a phosphatidylinositol 3-kinase inhibitor; calphostin C, a protein kinase C inhibitor; or nocodazole, a microtubule toxin, had no obvious effect on FRET between integrin ␤ 3 -NHE1 in the intracellular vesicular compartments. These results indicated that following the CHO ␣ IIb ␤ 3 cell's binding to a Fg/rho substrates intermolecular interactions between integrin ␣ IIb ␤ 3 and ion exchangers took place not only on the plasma membrane, where the ligands bound to the integrin receptors, but also in vesicles many m away from the ligand-receptor binding sites.
Lipid Microdomain-mediated Targeting of NHE1 and NCX1 to the Plasma Membrane and Their Interactions with Integrin ␣ IIb ␤ 3 -Previous reports indicated that lipid rafts were involved in integrin downstream signaling (33,34) and platelet activation (35). We have shown in Fig. 4b that disruption of lipid rafts by M␤CD affected intermolecular interactions between integrin ␣ IIb ␤ 3 and ion exchanger NHE1 or NCX1. The same drug treatment also reduced calcium oscillations in CHO ␣ IIb ␤ 3 cells grown on Fg/rho substrates in a dose-dependent manner (Fig. 1b, green bars). These results suggested a role played by lipid rafts, or lipid microdomains, in modulating the observed calcium oscillations in CHO ␣ IIb ␤ 3 cells grown on Fg/rho substrates. The presence of lipid rafts was revealed by staining CHO ␣ IIb ␤ 3 cells' GM1 with fluorescently labeled cholera toxin B. We noticed that in CHO ␣ IIb ␤ 3 cells grown on control PLL substrate, GM1, like NCX1 or NHE1, was present as punctate intracellular vesicles but absent from the plasma membranes (Fig. 5a). In CHO ␣ IIb ␤ 3 cells plated on Fg surfaces, GM1 appeared to transport to the plasma membrane (arrowheads) together with NHE1 (Fig. 5a) or NCX1 (supplemental Fig. 2a). Furthermore, disruption of lipid rafts by M␤CD profoundly inhibited targeting of GM1 to the plasma membrane, and as a result, NHE1 (Fig. 5b) and NCX1 (supplemental Fig. 2b) were retained in the cytoplasmic vesicles (green arrowheads). Plating CHO ␣ IIb ␤ 3 cells on rho substrate also resulted in membrane targeting of GM1, and such recruitment to the plasma membrane was lipid raft-dependent (data not shown).
We also conducted biochemical assays to assess the recruitments of integrin ␣ IIb ␤ 3 and ion exchangers to lipid rafts when plating CHO ␣IIb␤3 cells on Fg or rho substrates, as compared with the control PLL substrate. As shown in Fig. 5c, CHO ␣ IIb ␤ 3 FIGURE 5. Lipid rafts played important roles in mediating membrane targeting of NHE1/NCX1 and their interactions with integrin ␣ IIb ␤ 3 . a, CHO ␣ IIb ␤ 3 cells plated on PLL or Fg substrates at 37°C for 20 min were stained for integrin ␣ IIb (red), NHE1 (green), and ganglioside GM 1 (blue). On the control PLL substrate, only integrin ␣ IIb was present along the plasma membrane. Plating CHO ␣ IIb ␤ 3 cells on Fg substrate caused evident targeting of NHE1 and GM 1 from the intracellular vesicles to the cell membrane (arrowheads). b, the same setting as in a, but the cells were pretreated with 10 M M␤CD for 10 min. The lipid raft disruptor effectively inhibited membrane targeting by GM 1 and NHE1. c, CHO ␣ IIb ␤ 3 cells plated on different substrates as indicated were fractionated into 13 fractions by sucrose gradient fractionation. High cholesterol in the three fractions from the top indicated that they were DRM, or lipid raft-enriched. The contents of integrin ␣ IIb , NHE1 and NCX1 within each fraction were determined by Western blotting. Note that the plating of CHO ␣ IIb ␤ 3 cells on Fg substrate resulted in recruitments or accumulation of integrin ␣ IIb , NHE1, and NCX1 in DRM fractions, as compared with the cells grown on PLL substrate. d and e, CHO ␣ IIb ␤ 3 cells on Fg substrate were subjected to FLIM-FRET experiments using FRET pairs as indicated. The results of donor fluorochrome's FLIM images and means Ϯ S.D. of fluorescence lifetime values (from at least 10 sets of cells in three separate experiments) showed close intermolecular appositions between NHE1 or NCX1 with caveolin-1 (Cav 1) but not with CD45. *, p Ͻ 0.0001; bar, 5 m. cells plated on PLL or Fg substrates for 20 min were subjected to detergent treatments; the detergent-resistant membrane (DRM) that was lipid raft-enriched could be isolated using sucrose gradient fractionation protocols (19,33). The content of cholesterol, as well as integrin ␣ IIb , NHE1, and NCX1, within each fraction was determined. We found that high cholesterol contents, indicative of DRM, were enriched in the first three fractions from the top. From the Western blotting analyses, we noticed that when CHO ␣ IIb ␤ 3 cells were grown on Fg substrate, there were significantly higher contents of integrin ␣ IIb , NHE1, and NCX1 in the DRM fractions (or these proteins' distributions in fractions shifted to the DRM fractions) than when growing CHO ␣ IIb ␤ 3 cells on the control PLL substrate.
Caveolin-1 is another protein marker for lipid rafts. Using NHE1-caveolin 1 or NCX1-caveolin 1 as FRET pairs, we concluded from the FLIM-FRET experiments shown in Fig. 5d and from mean FLIM values shown in Fig. 5e that close intermolecular apposition between NHE1/NCX1 and caveolin-1 did occur (Fig. 5e, red bars), not only along the cell membrane (Fig. 5d, arrowheads) but also over the entire cell. As controls, there was no significant FRET taking place when the FRET acceptor was omitted in the experiments (Fig. 5e, black bar) or when CD45, a membrane protein excluded from the lipid raft, was used as the FRET acceptor (blue bars).

DISCUSSION
We report here a novel mechanism downstream of integrin ␣ IIb ␤ 3 signaling whereby a net calcium influx generated by the combined effect of NHE1 and NCX1 ion exchangers is capable of inducing calcium oscillation in a CHO ␣ IIb ␤ 3 cell model, as well as in native human platelets. The current working model is depicted in Fig. 6. In the control condition, integrin ␣ IIb ␤ 3 is present mainly on the plasma membrane (Fig. 6, 1) and in some intracellular vesicles. NHE1 and NCX1, on the other hand, are found mostly in intracellular vesicular compartments (Fig. 6, 1Ј) (see Fig. 2a). Upon binding to ligands, such as Fg or rho disintegrin (Fig. 6, 2), integrin ␣ IIb ␤ 3 on the cell membrane elicits signaling events that trigger intermolecular binding between the integrin ␣ IIb ␤ 3 and the ion exchangers that can occur on the intracellular vesicles (Fig. 6, 3) or on the plasma membranes (Fig. 6, 4). The transports and/or interactions of the integrins and ion exchangers are mediated by a lipid raft-dependent mechanism. Close molecular apposition between NHE1 and NCX1 render both ion exchangers functionally coupled. NHE1 can generate local high intracellular sodium concentration, which facilitates NCX1 to operate in reverse mode using sodium efflux to drive calcium influx (Fig. 6, 5) (see Fig. 1). The resulting increase of intracellular calcium represents an alternative pathway to trigger CICR as a prelude to calcium oscillation downstream of integrin ␣ IIb ␤ 3 signaling. Our experimental results strongly suggest that plasma membrane targeting and intermolecular coupling among integrin ␣ IIb ␤ 3 and ion exchangers NHE1 or NCX1 are dependent on the integrity of lipid rafts (see Figs. 4 and 5).
Note that although Fg and the disintegrin rho could both trigger calcium oscillations downstream of integrin ␣ IIb ␤ 3 , there are differences between these two ligands. As shown in Fig. 1b, calcium oscillation induced by rho could be totally abolished by inhibiting NCX1 function using bepridil, whereas calcium oscillation triggered by Fg was only partially inhibited by the drug. The difference may be due to the fact that the much larger Fg could bind to membrane receptors other than integrin ␣ IIb ␤ 3 heterodimer (41), whereas the much smaller rho protein can only interact with integrin ␣ IIb ␤ 3 (11).
Despite strong functional links and close intermolecular apposition between integrin ␣ IIb ␤ 3 and NHE1 or NCX1 as detected by FRET experiments, stable physical binding among these molecules could not be unequivocally demonstrated using conventional biochemical assays, such as co-immunoprecipitation experiments. These results suggest that the intermolecular interactions among components of the proposed FIGURE 6. A model summarizes the findings of this study. Most of NHE1 and NCX1 (blue ovals) were present in intracellular vesicles in control CHO ␣ IIb ␤ 3 cells (1Ј). Upon binding to ligands, such as Fg or rho disintegrin (purple circles) (2), integrin ␣ IIb ␤ 3 (green symbols) on the cell membrane elicit signaling events that trigger intermolecular binding between the integrin ␣ IIb ␤ 3 and the ion exchangers that can occur on the intracellular vesicles (3) or on the plasma membranes (4). The transports and/or interactions of the integrins and ion exchangers are mediated by a lipid raft-dependent mechanism (red-shaded membrane). Functional coupling between NHE1 and NCX1 is able to generate net calcium influx (5) that triggers CICR as a prelude to calcium oscillation downstream of integrin ␣ IIb ␤ 3 signaling. NHE1-integrin ␣ IIb ␤ 3 -NCX1 molecular complex are either too transient or too weak to sustain the preparation procedures associated with the biochemical assays or too sensitive to the detergent treatment typically applied during co-immunoprecipitation experiments. We demonstrate here that biophotonic approaches, such as FLIM-FRET, could overcome this difficulty and disclose intermolecular interactions that cannot be otherwise revealed using conventional biochemical methodologies. FRET can also be applied to monitor dynamic intermolecular interactions in live cells or even in situ of a subcellular focal area where cellular events of interest may be only short lived. Note that in the current study, FRET is determined by recording the reduction of the FRET donor's lifetime as the FRET acceptor approaches. Such a FLIM-FRET imaging platform should be complementary to the conventional ratio-imaging FRET that is based on fluorescence intensity measurements. FLIM-FRET is less sensitive to the noise caused by fluorescence photobleaching or local fluctuations of fluorochrome concentrations, but its current temporal resolution, at less than a frame/s, is less advantageous than the ratio imaging that can easily reach at least 10 to even hundreds of frames/s. FRET occurs if the intermolecular proximity is less than 10 nm, which is generally accepted as the distance to define intermolecular binding (18,32). It is interesting to know that FRET does occur between integrin ␣ IIb ␤ 3 and NHE1 or NCX1 but not much between NHE1 and NCX1, despite the fact that both ion exchangers are co-localized and functionally coupled to each other. We believe that NHE1 and NCX1 may be brought into proximity by individually binding to integrin ␣ IIb ␤ 3 heterodimer or by restricting their distributions within a membrane microdomain confined by the lipid rafts.
We have reported previously that plating CHO ␣ IIb ␤ 3 cells on Fg or rho substrates could elicit transport of integrin ␣ IIb ␤ 3 proteins from intracellular vesicular pool to the cell membrane (14); it appears that NHE1 and NCX1 can also undergo such targeting from intracellular vesicular compartments to the plasma membrane. A major presence of ion exchanger in the intracellular vesicles found in this study is rather new. Previously, NCX had been located to mitochondria (42) or the nuclear envelope (43); however, it is unlikely that NCX located at such organelles could be recruited to the plasma membrane. Also, NHE had been located to the late endosomal compartment in yeast (44). The identities of the vesicular compartments where NHE1 or NCX1 resides, as shown in Fig. 2, have not been determined, but the ion exchangers' vesicular localizations and their transport to the cell membrane are associated with the lipid raft microdomain (Fig. 5, a and b) (43,45,46). It is generally accepted, despite a lack of mechanistic description, that integrity of the lipid raft is essential for many functions downstream of integrin signaling (33), including calcium mobilization (47) and thrombosis (35). Our results support this notion by raising a mechanistic example showing that lipid rafts, by "corralling" NHE1 and NCX1 to the vicinity, are able to facilitate functional coupling between the two "interacting" molecules and trigger important signaling downstream of integrin.
Another unexpected finding made in this study is that binding of ligands to integrin ␣ IIb ␤ 3 on the plasma membrane could trigger molecular coupling among the integrin and ion exchangers in vesicles many m away from the plasma membrane (Fig. 4a). Mechanisms underlying such "long range signal transduction" are still unknown. Moreover, the families of NHE and NCX ion exchangers have been shown to play important roles in various pathophysiological processes, including ischemia-induced reperfusion injuries in myocardium (48) or myocardial stretching (49), salt-dependent hypertension (50), and primary hypertension (51). Our results link NHE and NCX to integrin signaling and may help delineate some of the currently unidentified mechanisms underlying these ion exchanger-related disease processes.