Cholesterol efflux to apoA-I in ABCA1-expressing cells is regulated by Ca2+-dependent calcineurin signaling.

ATP-binding cassette transporter A1 (ABCA1) is required for the lipidation of apolipoprotein A-I (apoA-I), although molecular mechanisms supporting this process remain poorly defined. In this study, we focused on the role of cytosolic Ca2+ and its signaling and found that cytosolic Ca2+ was required for cholesterol efflux to apoA-I. Removing extracellular Ca2+ or chelating cytosolic Ca2+ were equally inhibitory for apoA-I lipidation. We provide evidence that apoA-I induced Ca2+ influx from the medium. We further demonstrate that calcineurin activity, the downstream target of Ca2+ influx, was essential; inhibition of calcineurin activity by cyclosporine A or FK506 completely abolished apoA-I lipidation. Furthermore, calcineurin inhibition abolished apoA-I binding and diminished JAK2 phosphorylation, an established signaling event for cholesterol efflux to apoA-I. Finally, we demonstrate that neither Ca2+ manipulation nor calcineurin inhibition influenced ABCA1's capacity to release microparticles or to remodel the plasma membrane. We conclude that this Ca2+-dependent calcineurin/JAK2 pathway is specifically responsible for apoA-I lipidation without directly modifying ABCA1 activity.


Cell cultures
Both baby hamster kidney (BHK) cells and RAW 267.4 macrophage cells were maintained in DMEM supplemented with 10% FCS at 37°C in a 5% CO 2 incubator. ABCA1 expression was induced during 16-18 h incubation in DMEM with 1 mg/ml BSA, which included either 5 nM mifepristone or 250 M 8-Br-cAMP, for BHK and RAW cells, respectively. Mock-transfected cells were used as negative controls in experiments with BHK cells, whereas 8-Br-cAMP was withheld for negative controls in experiments with RAW cells.

Cy2-apoA-I cell surface association
Purifi ed apoA-I was conjugated to the Cy2 fl uorophore according to the manufacturer's instructions. Briefl y, apoA-I was dialyzed with 0.1 M Na 2 CO 3 , pH 9.5, and then combined with the Cy2 bis-reactive reagent. Following conjugation, labeled Cy2-apoA-I was separated from unlabeled dye on a P10 BioGel column (Bio-Rad ). The concentration of Cy2-apoA-I was determined using the Lowry protein assay. Cy2-apoA-I (5 g/ml) was incubated with cells for 2 h at 37°C to determine cell association. During each experiment, cells were preincubated with different exogenous compounds for 15 min before the addition of Cy2-apoA-I. After the incubation period (2 h), adherent cells were detached from the plate surface using 4 mM EGTA and 4 mM EDTA in PBS (the use of trypsin was avoided to prevent digestion of surface proteins that may be required for apoA-I cell association). In RAW cells, the cells were suspended before the 2 h treatment, because induction with 8-Br-cAMP reduces the adherence of RAW cells. Finally, degree of Cy2-apoA-I cell association was determined by fl ow cytometry. The amount of Cy2-apoA-I association was expressed relative to negative and positive (ABCA1expressing) controls. Similarly treated cells were also incubated with Cy2-apoA-I on ice for 2 h and Cy2-apoA-I binding was assessed by fl ow cytometry.

Cholesterol effl ux
BHK and RAW 264.7 cells were grown with 1 µCi/ml [1,[2][3] H(N)]cholesterol with DMEM/10% FCS for 2 days to label cells to equilibrium. After 2 days, the growth medium was replaced with DMEM + 1 mg/ml BSA. ABCA1 expression was induced with either 5 nM mifepristone or 250 M 8-Br-cAMP, as indicated above. After expression of ABCA1 (16-18 h), the growth medium was replaced with fresh DMEM + 1 mg/ml BSA to act as the effl ux assay medium. A variety of modulators were included in the medium for 2 or 4 h to measure their effects on cholesterol effl ux. During apoA-I-dependent effl ux, apoA-I (5 g/ml) was included in the effl ux medium. After the duration of the effl ux (2 or 4 h), medium was collected and centrifuged at 500 × g to remove cell debris. The cell-free supernatant was then combined with scintillation liquid and total counts per minute (cpm) were measured. The remaining adherent cells were lysed in 0.5 N NaOH, and the total cpm of the lysate was measured. Cholesterol effl ux was expressed as the ratio between the cpm from the effl ux medium and the total cpm found in the cell lysate and effl ux medium. In some experiments, cholesterol effl ux was Ca 2+ concentration is maintained at low levels (100 nM) relative to the extracellular medium (1-2 mM). This enables cells to rapidly increase cytosolic Ca 2+ levels through Ca 2+ infl ux, often in conjunction with Ca 2+ release from intracellular stores. The rise in cytosolic Ca 2+ then triggers Ca 2+ binding to regulatory proteins, such as calmodulin (CaM). Upon binding of Ca 2+ , CaM undergoes a conformational change that drastically increases its binding affi nity for a wide array of downstream target proteins ( 13 ). Many target proteins of CaM are kinases or phosphatases; these include myosin light chain kinase, CaM-dependent protein kinase (CaMK) I, II, and IV, and calcineurin.
In light of an early study that documented enhanced anion fl ux in ABCA1-expressing Xenopus oocytes ( 14 ), we attempted to determine whether Ca 2+ , particularly Ca 2+ infl ux, plays an intracellular role in facilitating apoA-I lipidation through signaling events. We found that, in both BHK cells and RAW macrophages, cytosolic Ca 2+ was required for cholesterol effl ux to apoA-I. We provide evidence that apoA-I induced Ca 2+ infl ux into cells. We further demonstrate that calcineurin signaling, the downstream target of Ca 2+ infl ux and CaM activation, was also essential for ABCA1-mediated cholesterol effl ux to apoA-I. Furthermore, inhibition of calcineurin interfered with JAK2 phosphorylation, an established signaling event for cholesterol effl ux to apoA-I, and abolished apoA-I binding. Finally, we demonstrated that neither Ca 2+ manipulations nor calcineurin inhibition affected ABCA1 expression, cellular distribution, basal cholesterol effl ux, or its ability to remodel the plasma membrane. The Ca 2+ -dependent CaM/calcineurin/JAK2 pathway is therefore specifi cally responsible for apoA-I lipidation without directly modifying ABCA1 activity.

Materials and reagents
Cell culture growth media, antibiotics (penicillin and streptomycin), and fetal calf serum (FCS) were purchased from Invitrogen. BHK cells were the generous gift from Drs. Oram and Vaughan (University of Washington, Seattle). These cells carry a mifepristone inducible vector with or without an ABCA1 gene insert. The RAW 264.7 cell line was purchased from the ATCC. Mifepristone was from Invitrogen, and 8-Br-cAMP from Sigma-Aldrich. Ca 2+ -free DMEM medium (cat no. 21068) was purchased from Invitrogen, which contains all the components of normal DMEM including Mg 2+ except no Ca 2+ . The following antibodies were acquired from several vendors: rabbit polyclonal anti-ABCA1 (Novus Biological Inc.), Alexa Fluo 488 goat anti-rabbit IgG (Molecular Probes), Ecl anti-rabbit IgG horseradish peroxidase linked whole antibody from donkey (GE Healthcare), anti-JAK2 rabbit polyclonal IgG (Santa Cruz Biotechnology Inc.), and rabbit polyclonal anti-phosphorylated-JAK2 [pYpY1007/1008] (Invitrogen). Our protease inhibitor cocktail was purchased from Sigma-Aldrich. The following chelators and modulators of ion fl ux were purchased from a variety of sources: 4,4'-diisothiocyano-2,2'stilbene disulphonic acid hydrate disodium salt (DIDS; Sigma-Aldrich), sodium thiocyanate (Sigma-Aldrich), sodium gluconate (Sigma-Aldrich), EDTA (Fisher Scientifi c), EGTA (Fisher Scientifi c), thapsigargin (Calbiochem), ryanodine (Tocris Bioscience), 2-aminoethyl diphenylborinate (2-APB; Sigma), and BAY-K8644 ested in potential intracellular Ca 2+ signaling in cholesterol effl ux to apoA-I. Extracellular Ca 2+ was removed by adding Ca 2+ chelators to the medium or using Ca 2+ -free medium during cholesterol effl ux. EDTA broadly chelates divalent cations (i.e. Mg 2+ and Ca 2+ ), whereas EGTA is much more specifi c for Ca 2+ . Both molecules caused a drastic reduction in cholesterol effl ux, although EGTA was more effective (80% inhibition) ( Fig. 1A ). Effl ux was similarly inhibited when Ca 2+ -free medium was used ( Fig.  1A ). The Ca 2+ -free medium used here (DMEM 21068, Invitrogen) contains normal concentration of Mg 2+ , indicating that the effect of EGTA is specifi c for Ca 2+ . Also, the inhibition can be readily relieved as soon as a normal amount of Ca 2+ (1.8 mM) is reintroduced to the medium ( Fig. 1B ).
We also characterized the dose-dependence relationship between extracellular Ca 2+ concentrations and cholesterol effl ux to apoA-I. We found that the Ca 2+ concentration required for half maximal effl ux was approximately 200 M ( Fig. 1C ). It is particularly interesting that this value is relatively small compared with the normal extracellular concentration of 1.8 mM, suggesting that only a very low concentration of extracellular calcium is necessary for cholesterol effl ux to apoA-I. A similar trend was observed when EGTA was supplemented into the growth medium at increasing concentrations ( Fig. 1D ); the effi cacy of cholesterol effl ux was only dramatically affected at high concentrations of EGTA (>1.75 mM), as it chelates almost all Ca 2+ ions. Together, these fi ndings confi rm the importance of extracellular calcium during effl ux to apoA-I, as previously observed by Smith et al. ( 11 ).
Although removing extracellular Ca 2+ could infl uence protein-protein interaction at the cell surface or endocytosis ( 11 ), it also eliminates Ca 2+ infl ux, a key event for many intracellular signaling processes. To test whether intracellular Ca 2+ is specifi cally required, ABCA1-expressing BHK cells were preloaded with increasing concentrations of BAPTA-AM (0-200 M) for 15 min. BAPTA-AM is a membrane permeable precursor of a Ca 2+ chelator that does not bind Ca 2+ in its native form. However, once inside cells, the AM ester is hydrolyzed by esterases, which traps BAPTA inside cells and also enables BAPTA to chelate Ca 2+ . BAPTA-AM, therefore, provides a means to specifically buffer intracellular Ca 2+ without compromising extracellular Ca 2+ levels. We found that BAPTA-AM abolished cholesterol effl ux in a concentration-dependent manner ( Fig. 1E ). Signifi cantly, cholesterol effl ux was completely inhibited at the highest concentration of BAPTA-AM (200 M) used. Cells remained viable under these conditions (not shown). Also, by increasing the extracellular Ca 2+ concentration (thus increasing Ca 2+ infl ux), we could partially rescue the cholesterol effl ux from BAPTA-AM treated cells (data not shown). This demonstrates for the fi rst time that intracellular Ca 2+ is critically required for ABCA1dependent cholesterol effl ux to apoA-I. It also implies that removing extracellular Ca 2+ most likely affects intracellular Ca 2+ levels, thus altering cholesterol effl ux.
Importantly, we found that ABCA1-expressing RAW macrophages also require Ca 2+ in cholesterol effl ux to expressed as a percentage relative to the negative controls (without apoA-I) and positive controls (with apoA-I alone).

Immunofl uorescent staining
BHK cells were plated and grown in glass-coverslip-bottom microscopy dishes to 50-70% confl uency. Cells were washed with PBS then fi xed with 4% paraformaldehyde in PBS for 10 min, followed by permeabilization with 0.1 mg/ml saponin in PBS for 30 min. Cells were blocked with 5% calf serum and 50 mM NH 4 Cl in PBS for 20 min. The primary ABCA1-specifi c antibody was then added at a concentration of 1:500 in a solution of 5% calf serum/ PBS for 30 min. After washing with PBS and incubating with 5% calf serum/PBS for 20 min, secondary antibody (Alexa Fluo 488 goat anti-rabbit IgG) was then added at a concentration of 1:200 for 30 min, followed by a 45 min incubation in 5% calf serum/PBS. The cellular localization of immunofl uorescence was observed and recorded using a C1 confocal module on a Nikon TE2000-E inverted fl uorescent microscope with a 60× objective. Images from ABCA1 and mock cells were taken using identical settings.

Ca infl ux into BHK cells
BHK cells were grown to approximately 75% confl uency in 24well plates. The culture plates were placed in a 37°C water bath and incubated in 500 l of prewarmed DMEM + 1 mg/ml BSA for 5 min. In some of wells, 5 g/ml apoA-I was included in the medium. DMEM (50 l) with 55 Ci/ l [ 45 Ca] was then added to triplicate wells for 4 min. Immediately after 4 min incubation, cells were washed with ice cold DMEM + 5 mM EGTA to stop channel activity and remove surface bound [ 45 Ca]. Cells were lysed with 0.5 N NaOH and the total cell associated 45 Ca cpm was measured using a ␤ -counter to determine the total cellular infl ux of [ 45 Ca]. To determine the amount of nonspecifi c [ 45 Ca] association with cells, the same experiments were carried out in ice-cold buffers to prevent channel activity. This nonspecifi c association was subtracted from the total cell associated cpm to produce values for 45 Ca infl ux. Finally, the total protein content of each well was determined using the Lowry assay and all values were presented as a ratio of [ 45 Ca] cpm/ g protein.

Methylthiazol tetrazolium assay
After overnight induction with various concentrations of mifepristone, BHK cells in 12-well dishes were rinsed and returned to DMEM + 10% FCS before treating with various reagents for 2 h. Then 20 l of methylthiazol tetrazolium (MTT) solution (5 mg/ ml) was added directly into each well. Cells were incubated at 37°C for 3 h. Medium was carefully removed with a syringe to avoid disturbing formazan crystals formed during the incubation. DMSO (100 l) was added to dissolve the crystals. The absorbance was then measured at 550 nm.

Statistics
Statistical comparisons between groups were performed with PRISM software (GraphPad). Data are mostly presented as mean ± SD. The statistical signifi cance of differences between groups was analyzed by Student's t -test. Differences were considered signifi cant at a P -value < 0.05.

RESULTS
An early study revealed that ABCA1 expression on the plasma membrane of Xenopus oocytes increases anion fl ux ( 14 ). Also, Smith et al. ( 11 ) reported that extracellular Ca 2+ is required for apoA-I lipidation. Thus, we were inter- H radioactivity were counted and presented as percentage of cholesterol in the medium relative to the total cholesterol (medium and cellassociated). In B, cholesterol effl ux was presented as a percentage relative to untreated cells (control). Data is presented as mean ± SD of triplicate wells, representative of at least three experiments performed.
We found that cholesterol effl ux to apoA-I was not inhibited by thapsigargin at concentrations (100 nM to 10 M) effectively used by others (15)(16)(17)(18) ( Fig. 3A ). Ryanodine and 2-APB, which inhibit RyR and InsP 3 R , respectively, also did not inhibit cholesterol effl ux to apoA-I at wellestablished inhibitory concentrations ( 19,20 ) ( Fig. 3A ). Combinations of thapsigargin, ryanodine, and/or 2-APB were also tested with no measurable effects (data not shown). Therefore, we conclude that ER Ca 2+ stores did not contribute to the intracellular BAPTA-sensitive pool of Ca 2+ that participates in ABCA1-dependent cholesterol effl ux to apoA-I. The BAPTA-sensitive pool of Ca 2+ most likely comes from the extracellular medium.
We next tested whether apoA-I stimulates Ca 2+ infl ux from the extracellular medium. We initially attempted to measure changes in intracellular free Ca 2+ using the cell permeable fl uorescent probe, Fura 2-AM, but did not detect any signifi cant changes in intracellular free Ca 2+ (supplementary Fig. I ). We took this as a sign that the magnitude of Ca 2+ infl ux triggered by apoA-I could be relatively low such that fl uorescent probes may not be sensitive enough to detect it. Consequently, we resorted to a highly sensitive 45 Ca 2+ method to analyze the net infl ux of Ca 2+ into the cells from the extracellular medium. In these experiments, cells were briefl y exposed to 45 Ca 2+ containing medium either in the presence or in the absence of apoA-I. The short duration of 45 Ca 2+ pulse was to ensure that Ca 2+ effl ux does not signifi cantly contribute to the analysis ( 21 ). Also, extra care was taken to strip off Ca 2+ bound on the cell surface to ensure true measurement of Ca 2+ infl ux to the cytoplasm. Specifi cally, the net 45 Ca 2+ infl ux was measured by cellassociated 45 Ca 2+ radioactivity after 4 min incubation in the medium containing 45 Ca 2+ . We found that 45 Ca 2+ infl ux into ABCA1-expressing BHK cells was nearly doubled in the presence of apoA-I, whereas apoA-I did not alter 45 Ca 2+ infl ux in mock-BHK cells ( Fig. 3B ). This demonstrates that apoA-I induced Ca 2+ infl ux in ABCA1-expressing cells. Such Ca 2+ infl ux from the extracellular medium could be critical for cholesterol effl ux to apoA-I. To test this, we used BAY-K8644, an agonist of plasma membrane L-type Ca 2+ channels ( 22 ), to further increase Ca 2+ infl ux. BAY-K8644 (1 M) was indeed able to increase cholesterol effl ux to apoA-I by 35% ( Fig. 3C ). Together, these fi ndings demonstrate that apoA-I stimulates Ca 2+ infl ux in ABCA1expressing cells. This infl ux likely initiates intracellular signaling events required for cholesterol effl ux to apoA-I.
We next looked into the potential target of intracellular Ca 2+ in the process of cholesterol effl ux. Intracellular Ca 2+ is commonly employed as a second messenger to regulate a variety of signaling pathways ( 23 ). For example, CaM is the primary target of Ca 2+ signaling in eukaryotic cells ( 23 ). We therefore fi rst examined whether CaM is involved in cholesterol effl ux process. Cells were treated with W-7 during cholesterol effl ux to apoA-I. W-7 is a CaM antagonist and prevents the binding of Ca 2+ -bound CaM with its downstream substrates ( 24 ). W-7 indeed caused a significant inhibition of cholesterol effl ux to apoA-I at relative low concentrations (0.5 and 1 M) ( Fig. 4A ). Both BHK cells and macrophages were highly sensitive to W-7, which apoA-I. Cholesterol effl ux from these macrophages shared the same sensitivity to EGTA and Ca 2+ free condition ( Fig. 2A ) or to BAPTA-AM ( Fig. 2B ). The minimal requirement for extracellular Ca 2+ was also in the low-micromolar range, similar to what was observed in BHK cells (data not shown). These results suggest that Ca 2+ , particularly intracellular Ca 2+ , is a common factor required for cholesterol effl ux to apoA-I, regardless of cell types.
Together, the above results suggest that cholesterol effl ux to apoA-I requires Ca 2+ infl ux across the plasma membrane and consequently induces a rise in cytosolic Ca 2+ . Alternatively, such a rise could be initiated by Ca 2+ release from intracellular Ca 2+ stores. The sarco/endoplasmic reticulum Ca 2+ -ATPase (SERCA) is a Ca 2+ pump that loads the ER in an ATP-dependent manner, whereas the ryanodine receptor (RyR) and the inositol 1,4,5triphosphate-gated Ca 2+ release channel (InsP 3 R) control the release of Ca 2+ from the ER. Thus, we used thapsigargin as a specifi c inhibitor of the SERCA pump. Thapsigargin binds SERCA irreversibly and prevents refi lling of the ER with Ca 2+ . Under this condition, the ER cannot be replenished with Ca 2+ after it is released through the RyR and InsP 3 R channels. Therefore, thapsigargin inhibits the contribution of ER stores to the cytoplasmic pool of Ca 2+ . apoAI. Interestingly, earlier studies reported that CsA abrogates ABCA1-dependent cholesterol effl ux ( 27,28 ). Best known as an immune-suppressor in vivo, CsA mechanistically inhibits calcineurin by forming a complex with cyclophilins, the effector protein of calcineurin ( 29 ). When CsA was included in the medium with apoA-I, cholesterol effl ux to apoA-I was potently inhibited with complete inhibition at 10 M ( Fig. 4B ). To further verify the specifi city of CsA on calcineurin, we used another unrelated specifi c calcineurin inhibitor, FK506. FK506 binds to FKBP12, another effector protein that interacts with calcineurin at a slightly different site from that of cyclophilins ( 30 ). Fig. 4C shows that FK506 also functioned as a potent inhibitor. Cholesterol effl ux to apoA-I was dose dependently inhibited and was completely abolished at 25 M of FK506. Moreover, both CsA and FK506 were equally effi cient at inhibiting effl ux to apoA-I in RAW cells ( Fig. 4D ). Thus, by using two highly specifi c cal-prevented us from testing the effect of higher W-7 concentrations.
Activation of CaM by Ca 2+ is known to modulate the function of various downstream targets, including CaM-KII, myosin light chain kinase, and calcineurin ( 23 ). Although CaM is a relatively abundant protein, the pool of free CaM is limited due to a wide range of CaM targets in cells. There is likely intense competition among CaM targets for the Ca 2+ -CaM complex, and the action of these targets may be at least partially determined by their respective affi nity to Ca 2+ -CaM ( 25 ). Among these targets, calcineurin has the highest affi nities to Ca 2+ -CaM [dissociation constant ( K d ) = ‫ف‬ 0.1 nM] ( 26 ) and therefore should be most sensitive to Ca 2+ fl uctuations or CaM activation. Given our observation described above that Ca 2+ infl ux triggered by apoA-I is of low amplitude (thus could not be detected by Fura-2), we considered calcineurin as a likely candidate in cholesterol effl ux to cell association was affected. This was achieved by fl ow cytometry detection of fl uorescent-labeled apoA-I (Cy2-apoA-I ). Cy2-apoA-I retained its potency in inducing cholesterol effl ux as native apoA-I (data not shown). As expected, ABCA1 expression induced high levels of apoA-I cell association, whereas Cy2-apoA-I did not signifi cantly bind mock-BHK cells ( Fig. 6A ). Excess unlabeled apoA-I completely abrogated Cy2-apoAI binding to ABCA1-expressing cells ( Fig. 6B ), further confi rming the specifi city of Cy2-apoA-I. We found that EGTA (2 mM) signifi cantly reduced Cy2-apoA-I cell association to 42% relative to cells bathed in normal Ca 2+ medium ( Fig. 6C ). The addition of 50 M BAPTA-AM to ABCA1expressing cells similarly reduced Cy2-apoA-I cell association. The representative apoA-I binding is shown in Fig. 6D . Because BAPTA-AM does not alter extracellular Ca 2+ concentrations, diminished apoA-I cell association could not be simply due to disruptions in ligand-receptor binding on the cell surface. Buffering intracellular Ca 2+ by BAPTA-AM must have perturbed intracellular processes required for apoA-I binding. Consistent with this notion, we found that both calcineurin inhibitors abolished apoA-I cell association ( Fig. 6E, F ). Importantly, we found that apoA-I cell association under the experi-cineurin inhibitors that are structurally and mechanistically distinct, we conclude that the Ca 2+ -dependent CaM/ calcineurin signaling pathway is essential for cholesterol effl ux to apoA-I.
The CaM/calcineurin signaling pathway is known to activate nuclear factor of activated T-cells (NFAT) to initiate gene regulations ( 23 ). Because all the experiments reported here were performed within 2 h, gene regulation is not likely to be a signifi cant factor. We nevertheless wanted to rule out the possibility of ABCA1 downregulation. We found that ABCA1 protein expression levels and cellular distribution remained unchanged after Ca 2+ manipulation or calcineurin inhibition ( Fig. 5A , B ). This is largely consistent with an earlier conclusion made by Smith et al. ( 11,27 ) that inhibition of cholesterol effl ux to apoA-I by either Ca removal or CsA is not due to ABCA1 downregulation. However, we did not detect increased ABCA1 expression as observed by earlier studies ( 11,27 ), which could be due to the short duration of our experiments (2 h vs. 4 h) or cell type differences.
The successful lipidation of apoA-I is known to require prolonged interaction between plasma membrane and apoA-I, which generates a measurable apoA-I association with cells. We therefore examined whether apoA-I mental conditions described here is positively correlated with the effi ciency of cholesterol effl ux to apoA-I ( Fig. 6G ), consistent with the interdependence between apoA-I association and cholesterol effl ux observed among ABCA1 mutants ( 31 ). This relationship suggests that the Ca 2+ -dependent CaM/calcineurin signaling pathway is directly responsible for maintaining apoA-I interaction with ABCA1-expressing cells. Finally, specifi c inhibition of calcineurin signaling with CsA and FK506, respectively, did not alter ABCA1 expression or localization (data not shown). Furthermore, none of the treatments, i.e., EGTA, Ca 2+ , CsA, or FK506, affected cell viability, as evidenced by the MTT test (supplementary Fig. II ). Also, consistent with previous observations by Takahashi et al. ( 11 ), removal of Ca 2+ from the medium did not affect apoA-I binding at 4°C. In fact, 4°C apoA-I binding was intact under all treatment conditions (supplementary Fig. III ).
Similarly, we tested Cy2-apoA-I cell association in RAW macrophages. Control conditions indicated that Cy2-apoA-I interacts preferentially with ABCA1-expressing RAW cells ( Fig. 7A ), and effective competition occurred in the presence of unlabeled apoA-I ( Fig. 7B ). Removal of either intracellular or extracellular Ca 2+ caused signifi cant reductions in apoA-I association ( Fig. 7C, D ). Again, specifi c inhibition of calcineurin signaling with  at 10 M signifi cantly inhibited JAK2 phosphorylation ( Fig. 8A ). Also, if the effect of CsA is to prevent calcineurin activation via Ca 2+ infl ux, removal of extracellular Ca 2+ should also jeopardize JAK2 activation. We indeed found similarly decreased JAK2 phosphorylation when extracellular Ca 2+ was removed ( Fig. 8B ). Our results thus suggest that CaM/calcineurin likely operates upstream of JAK2 signaling and, together, they act to enhance apoA-I interactions with ABCA1-expressing cells and thus facilitate cholesterol effl ux.
CsA and FK506 abolished apoA-I association in macrophages ( Fig. 7E, F ). We therefore conclude that the CaM/calcineurin signaling pathway is essential for both apoA-I cell-association and cholesterol effl ux to apoA-I in ABCA1-expressing cells.
Interestingly, Oram et al. reported that JAK2 signaling is also necessary for apoA-I cell-association and cholesterol effl ux to apoA-I (7). JAK2 becomes phosphorylated rapidly after ABCA1-expressing cells interact with apoA-I (7,32), and JAK2 phosphorylation positively correlates with apoA-I cell association or cholesterol effl ux to apoA-I (31). Also, cells lacking JAK2 were defective in cholesterol effl ux to apoA-I (7). Therefore, we analyzed JAK2 phosphorylation in cells treated with calcineurin inhibitor CsA. CsA  The expression levels of phosphorylated Jak2 (P-Jak2) and total Jak2 was determined by immunoblotting. A: Duplicates of ABCA1-BHK cells were induced with mifepristone overnight and then incubated with or without 10 M CsA for 2 h before cell lysis and immunoblotting. B: ABCA1-expressing BHK cells were also incubated with Ca 2+ -free medium for 2 h before immunoblotting. The optical density of the immunoblots was quantifi ed and expressed as mean ± SD. Untreated cells served as control (100%). Results are representative of three experiments with identical conditions. PKA inhibitor PKI inhibited both processes. It is also noteworthy that CsA seemed to increase apoA-I-independent effl ux ( Fig. 9A ), which may refl ect the trapping effect of CsA. CsA was reported to inhibit ABCA1 turnover on the plasma membrane, resulting in increased ABCA1 on the cell surface ( 27 ). The largely unperturbed basal activity in ABCA1-expressing cells also suggests that the manipulations used here did not cause general cytotoxicity. Cells indeed remained perfectly viable as judged by the MTT test (supplementary Fig. II ).
We next tested ABCA1's capacity to remodel the plasma membrane. Functional ABCA1 causes caveolin to redistribute from largely clustered structures, presumably caveolae, to the general area of the plasma membrane ( Fig. 9C , panels A and B), due mainly to generating more nonraft membrane microdomains ( 36 ). Disrupting Ca 2+dependent CaM/calcineurin signaling had little effect on this basal ABCA1 function; caveolin still diffusely decorated the plasma membrane in CsA-treated cells, Finally, because disrupting the Ca 2+ -dependent CaM/ calcineurin signaling pathway appeared to specifi cally perturb apoA-I binding to ABCA1-expressing cells but not ABCA1 protein expression or distribution, we wondered whether other ABCA1 functions also require Ca 2+dependent CaM/calcineurin signaling. We and others reported recently that ABCA1 expression leads to basal cholesterol effl ux, independent of apoA-I, to produce microparticles ( 33,34 ). Also, functional ABCA1 remodels the plasma membrane, also independent of apoA-I (36,37). For the purpose of clarity, we collectively term these effects as "basal ABCA1 activity." We treated cells with the various reagents described above and analyzed microparticle production and membrane remodeling. We found that apoA-I-independent effl ux is completely insensitive to the perturbations of Ca 2+ -dependent CaM/ calcineurin pathway ( Fig. 9A ), whereas the same perturbations abolished cholesterol effl ux to apoA-I ( Fig. 9B ). Interestingly, the only shared pathway is PKA, because calcineurin has the highest affi nity for Ca 2+ -CaM ( K d = ‫ف‬ 0.1 nM), whereas many other targets, such as CaMKII ( K d = 50 nM), require much higher concentrations of Ca 2+ -CaM ( 26 ). It is tempting to suggest that low aptitude Ca 2+ infl ux may serve to specifi cally activate the CaM/calcineurin pathway without signifi cantly affecting other Ca 2+ -CaM activated processes, which may offer an explanation as to why a general CaM inhibitor, W-7, can only partially block cholesterol effl ux to apoA-I ( Fig. 4A ).
The antagonists of the two ER Ca 2+ channels, RyR and IP 3 R, and the SERCA pump had no effect on effl ux ( Fig.  3A ). This was in contrast to our observations that BAY-K8644, which opens L-type Ca 2+ channels, increased cholesterol effl ux to apoA-I by 35% ( Fig. 3C ). Interestingly, the Ca 2+ ionophore, A23187, did not further enhance cholesterol effl ux (data not shown). Unlike infl ux through regulated channels, A23187 binds Ca 2+ and shields it against the hydrophobicity of the lipid bilayer, causing unregulated entry of Ca 2+ into cells. This lack of effect from A23187 may indicate that low aptitude Ca 2+ infl ux induced by apoA-I alone is suffi cient to activate calcineurin or, alternatively, only regulated infl ux of Ca 2+ through specifi c plasma membrane channels can facilitate cholesterol effl ux to apoA-I. Extracellular Ca 2+ is known to be necessary for cholesterol effl ux to apoA-I (11). We now attribute this requirement to Ca 2+ infl ux that activates the calcineurin pathway. In support of this view, intracellular Ca 2+ buffering by BAPTA-AM was equally potent as removal of extracellular Ca 2+ . Interestingly, the effect of CsA in cholesterol effl ux to apoA-I has also been documented previously ( 27 ), although the link to calcineurin was not defi ned. Because both CsA and FK-506, two unrelated and highly specifi c calcineurin inhibitors, completely blocked ABCA1-mediated cholesterol to apoA-I ( Fig. 4 ), calcineurin activity is likely one of the essential components that enables apoA-I lipidation. This conclusion was further supported by the fi nding that blocking calcineurin activation by inhibitors or Ca 2+ removal inhibited JAK2 phosphorylation ( Fig. 8 ). JAK2 phosphorylation is a well-documented essential component for cholesterol effl ux to apoA-I. Thus, our fi ndings here place Ca 2+ infl ux and calcineurin activation upstream of JAK2 phosphorylation.
It is intriguing to see that both Ca 2+ manipulation and calcineurin inhibition abolish apoA-I cell association ( Fig.  6 ) yet preserve 4°C apoA-I binding (supplementary Fig.  III ). This further supports our conclusion that Ca 2+ and calcineurin function along the same pathway. It also suggests that initial binding of apo-A-I to ABCA1 is not sufficient to lipidate apoA-I and subsequent steps have to take place to produce effect cholesterol effl ux .
Perhaps most interestingly, this Ca 2+ -activated calcineurin pathway is highly specifi c for ABCA1-mediated cholesterol effl ux to apoA-I. Interrupting this pathway has no infl uence on ABCA1 basal function, i.e., basal cholesterol effl ux and plasma membrane remodeling (Fig. 9) . Calcineurin, a phosphatase by nature, most likely exerts its function by regulating the phosphorylation status of other cellular proteins; this potentially includes ABCA1. However, because basal function of ABCA1 is totally insensi-un distinguishable from untreated ABCA1-expressing cells ( Fig. 9C , panel C). These observations demonstrate that apoA-I specifi cally requires the CaM/calcineurin signaling pathway and, in the situation where calcineurin activity is inhibited, only apoA-I-related activities are diminished. ABCA1 still retained its capability to remodel the plasma membrane or generate microparticles. Together, we conclude that the CaM/calcineurin signaling pathway is a part of the cellular machinery that ABCA1 employs to specifi cally execute cholesterol effl ux to apoA-I.

DISCUSSION
In this study, we report that Ca 2+ plays a critical role in ABCA1-dependent cholesterol effl ux to apoA-I. We provide evidence that either removal of extracellular Ca 2+ or buffering intracellular Ca 2+ severely impaired ABCA1-mediated cholesterol effl ux to apoA-I. apoA-I initiated Ca 2+ infl ux from the extracellular medium in ABCA1-expressing cells, which potentially triggers intracellular signaling events. Indeed, we found that the Ca 2+ -activated CaM/calcineurin signaling pathway was required in cholesterol effl ux to apoA-I. Both CsA and FK506, two structurally distinct and highly specifi c inhibitors of calcineurin, completely abolished cholesterol effl ux to apoA-I and blocked apoA-I association with ABCA1-expressing cells. We also provide evidence that calcineurin signals through JAK2, an established signaling event required for cholesterol effl ux to apoA-I. Interestingly, the CaM/calcineurin pathway infl uences only apoA-I lipidation and not the basal activity of ABCA1. Together, our results establish a novel signaling pathway that ABCA1 employs specifi cally to transfer cholesterol on to apoA-I.
Although the link between ion fl ux and ABCA1 activity has not been fi rmly established, many studies have used the chloride channel inhibitor, DIDS, as an ABCA1 antagonist. We found that either chloride channel inhibition with DIDS or chloride anion replacement with gluconate Ϫ and SCN Ϫ moderately inhibited cholesterol effl ux (data not shown). This suggested that the chloride channel might only be indirectly involved and the modest inhibition may refl ect disruptions to other plasma membrane ion channels. Indeed, we found Ca 2+ channels played a much more pronounced role than chloride channels. This is supported by the following observations: a ) apoA-I triggered Ca 2+ infl ux, and b ) manipulations that prevent this infl ux (removal of extracellular Ca 2+ ) or dampen intracellular Ca 2+ rise (BAPTA-AM) impaired cholesterol effl ux to apoA-I. The dose response of extracellular Ca 2+ ( Fig. 1C, D ) also indicates that a relatively low concentration of extracellular Ca 2+ is suffi cient to produce effi cient cholesterol effl ux. Consistent with this, a relatively high concentration of BAPTA-AM was needed to completely inhibit effl ux ( Fig.  1E ). These observations collectively suggest that the Ca 2+ fl ux that initiates the CaM/calcineurin pathway is of low aptitude, unlike what is often required with excitable cells or cells stimulated by hormones ( 23 ). It explains why we were only able to detect apoA-I-stimulated Ca 2+ infl ux using 45 Ca 2+ ( Fig. 3B ). This is also consistent with the fact that tive to calcineurin activity, it is not likely that ABCA1 itself is a direct target of calcineurin. Consistent with this view, JAK2 does not directly phosphorylate ABCA1 or contribute to membrane remodeling ( 7 ). This is in sharp contrast with PKA. ABCA1 can be directly phosphorylated by PKA, and preventing PKA-dependent phosphorylation of ABCA1 by point mutations partially impairs phospholipid effl ux to apoA-I (38). PKA is also essential for ABCA1 basal function (Fig. 9) . We speculate that calcineurin regulates the phosphorylation of downstream target proteins, other than ABCA1, to support apoA-I lipidation. This also implies that apoA-I lipidation on the plasma membrane is not merely a biophysical event where apoA-I interacts with membrane bilayers. In addition to ABCA1, a cohort of proteins, made ready by phosphorylation/dephosphorylation by JAK2, calcineurin, and others, is required to participate in lipidating apoA-I. Disruption on this apoA-I activated cellular machinery could lead to premature release of apoA-I, thus diminished apoA-I cell-association, and failure of apoA-I lipidation.
In conclusion, we show for the fi rst time that apoA-I induces Ca 2+ infl ux from the extracellular medium. This rise in intracellular Ca 2+ activates the CaM/calcineurin/JAK2 pathway to maintain apoA-I cell association, resulting in effi cient cholesterol effl ux to apoA-I.