ATP binding cassette G1-dependent cholesterol efflux during inflammation.

ATP binding cassette transporter G1 (ABCG1) mediates the transport of cellular cholesterol to HDL, and it plays a key role in maintaining macrophage cholesterol homeostasis. During inflammation, HDL undergoes substantial remodeling, acquiring lipid changes and serum amyloid A (SAA) as a major apolipoprotein. In the current study, we investigated whether remodeling of HDL that occurs during acute inflammation impacts ABCG1-dependent efflux. Our data indicate that lipid free SAA acts similarly to apolipoprotein A-I (apoA-I) in mediating sequential efflux from ABCA1 and ABCG1. Compared with normal mouse HDL, acute phase (AP) mouse HDL containing SAA exhibited a modest but significant 17% increase in ABCG1-dependent efflux. Interestingly, AP HDL isolated from mice lacking SAA (SAAKO mice) was even more effective in promoting ABCG1 efflux. Hydrolysis with Group IIA secretory phospholipase A2 (sPLA2-IIA) significantly reduced the ability of AP HDL from SAAKO mice to serve as a substrate for ABCG1-mediated cholesterol transfer, indicating that phospholipid (PL) enrichment, and not the presence of SAA, is responsible for alterations in efflux. AP human HDL, which is not PL-enriched, was somewhat less effective in mediating ABCG1-dependent efflux compared with normal human HDL. Our data indicate that inflammatory remodeling of HDL impacts ABCG1-dependent efflux independent of SAA.

istration Medical Center Institutional Animal Care and Use Committee (Assurance number A3506-01).

Nascent HDL preparations
THP-conditioned acceptor particles (nascent HDL) were generated according to a published protocol ( 6 ). Briefl y, lipid-loaded THP-1 macrophages were incubated for 24 h in RPMI containing 0.2% fatty acid-free-BSA and either 10 µg/ml human apoA-I (Biodesign) or 20 µg/ml SAA (corresponding to human SAA1 ␣ except for the presence of an N-terminal methionine and substitution of asparagine for aspartic acid at position 60 and arginine for histidine at position 71; Biovision). The media was harvested and centrifuged (1500 g for three minutes) to remove detached cells, and the concentration of nascent HDL particles was determined prior to use in effl ux experiments by quantitative immunoblotting using purifi ed apoA-I and SAA as standards.
An important issue to be addressed is how the cooperative interaction between ABCA1 and ABCG1 functions during acute infl ammation. In this condition, serum amyloid A (SAA) is a major acute phase (AP) protein highly induced in the liver ( 9 ). SAA is also induced by infl ammatory stimuli in peripheral cells expressing ABCA1 and ABCG1, such as macrophages and adipocytes ( 9 ). Plasma SAA concentrations can increase up to 1000-fold during an AP response, with peak concentrations exceeding 1 mg/ml. Approximately 95% of AP SAA in the plasma is found associated with HDL, where it composes the major apolipoprotein ( 10 ). In addition, infl ammatory HDL undergoes signifi cant changes in lipid composition, with triglycerides tending to increase ( 11 ). Further, during infl ammation there is concomitant induction ( ‫ف‬ 100-fold) of Group IIA secretory phospholipase A 2 (sPLA 2 -IIA) in the liver, which leads to selective hydrolysis of HDL PL that alters the particle's structure and promotes its catabolism ( 12,13 ).
Lipid-poor SAA has been shown to promote ABCA1dependent cholesterol effl ux similar to apoA-I (14)(15)(16)(17). In this study, we investigated the extent to which SAA and AP HDL promote ABCG1-dependent effl ux. Our data show that SAA acts analogously to apoA-I in effecting sequential effl ux from ABCA1 and ABCG1. With respect to compositional changes in HDL that occur during infl ammation, alterations in PL content and not the presence of SAA impact the ability of AP HDL to promote ABCG1-dependent effl ux.

Human subjects
Blood was collected from healthy volunteers for isolation of normal (N) HDL and from patients undergoing cardiac surgery using a membrane oxygenator (coronary artery bypass, valve replacement), 24 h post-operatively for isolation of AP HDL. Blood was collected only from patients who underwent successful uncomplicated surgery and who gave informed consent. The study was approved by the University of Kentucky Medical Institutional Review Board.

Animals
C57BL/6 mice were obtained from Jackson Laboratories. Mice lacking SAA1.1 and SAA 2.1 were generated by targeted deletion of both mouse acute phase SAA genes SAA1 and SAA2 (InGenious Targeting Laboratory Inc., Stony Brook, NY) using embryonic stem cells derived from C57BL/6×129 SVEV mice ( 18 ). The targeting vector contained a neo cassette that replaced ‫ف‬ 10.1 kb of SAA1 and SAA2 , including exon 2 of both oppositely orientated genes. SAA null (SAAKO) mice and littermate controls [wild-type (WT)] were maintained in a pathogen-free facility under equal light-dark cycles with free access to water and food. To elicit an AP response, 12-16 week-old mice were injected intraperitoneally with 6 µg lipopolysaccharide (LPS) ( Escherichia coli 0111:B4, Sigma Chemical Co.) per gram of body weight. After 24 h the mice were humanely euthanized, and plasma was collected for preparation of HDL. All procedures were carried out in accordance with PHS policy and approved by the Veterans Admin-mation of nascent HDL. Thus, our results were analogous to a previous report that ‫ف‬ 15% of lipid-free apoA-I converts to nascent HDL when incubated with THP-1 macrophages under similar conditions ( 24 ). In contrast to apoA-I, virtually all of the SAA migrated as larger-sized HDL particles after incubation with THP-1 macrophages ( Fig. 1B ). The size distribution of THP-1-conditioned apoA-I and SAA was distinct.

THP-1-conditioned SAA stimulates ABCG1-mediated cholesterol effl ux
Using the ABCG1-inducible BHK in vitro model system ( 5, 7 ), we next assessed the ability of lipid-poor and THP-1-conditioned SAA to serve as acceptors for ABCG1-independent and ABCG1-dependent cholesterol effl ux. Effl ux to lipid-poor and THP-1-conditioned apoA-I was measured for comparison ( Fig. 2 ). In BHK cells without ABCG1 induction, cholesterol effl ux to lipid-poor apoA-I was negligible. Lipidation of apoA-I by THP-I conditioning resulted in signifi cantly increased ABCG1-independent effl ux. Lipid-poor SAA was more effi cient than apoA-I in mediating ABCG1-independent effl ux, similar to what has been reported for untransfected HeLa cells ( 16 ) and HepG2 cells ( 17 ). However, lipidation of SAA had no effect on ABCG1-independent effl ux. As expected, lipid-poor apoA-I was not an effective acceptor for ABCG1-dependent cholesterol effl ux, which was defi ned as the difference in effl ux by BHK cells induced to express ABCG1 (Total) and control BHK cells (ABCG1-independent) ( 5,6 ). When compared with lipid-poor apoA-I, ABCG1-dependent effl ux to THP-1-conditioned apoA-I was signifi cantly increased (6.5-fold), confi rming an earlier report that THP-1 macrophages convert lipid-poor apoA-I to nascent HDLs that are suffi ciently lipidated to serve as substrates for ABCG1-mediated cholesterol export ( 6 ). Interestingly, THP-1 macrophages had a similar effect on lipid-poor SAA, such that ABCG1-dependent effl ux was 9.5-fold higher for THP-1-modifi ed SAA compared with lipid-poor hexane/isopropyl alcohol (3:2 v/v) for 30 min at room temperature and counted for radioactivity. Effl ux of cellular [ 3 H] cholesterol to media was expressed as the percentage of total radioactivity in media and cells. ABCG1-specifi c values were calculated as the difference between the effl ux values in mifepristone-treated and control cells.

HDL hydrolysis
Mouse HDL (0.6-1.0 mg/ml HDL protein) was hydrolyzed by human recombinant sPLA 2 -IIA in Tris-buffered saline (pH 7.4) containing 10 mg/ml fatty-acid free BSA and 2 mM CaCl 2 . Lipolysis was terminated after 24 h incubations at 37°C by the addition of EDTA (fi nal concentration 20 mM). Mock hydrolyzed HDL was generated under the same conditions but omitting sPLA 2 -IIA. Hydrolysis reactions were carried out as reported in the literature (21)(22)(23) and approximated physiological conditions. The extent of PL hydrolysis was assessed by measuring the amount of free fatty acids released (Wako Chemicals).

Gradient gel electrophoresis and Western blots
Aliquots containing 25-50 ng lipid free apoA-I or SAA, or apoA-I or SAA in conditioned media from THP-1 macrophages, were separated by nondenaturing gradient gel electrophoresis (GGE). Electrophoresis was carried out in 4-20% polyacrylamide gels for 3.5 h at 200 V at 4°C, and the samples were then transferred to PVDF membranes (100 min at 100 V at 4°C) and immunoblotted using either anti-human apoA-I (Calbiochem) or anti-human SAA (Behring, Germany) antibodies, as indicated. Bound antibodies were detected by enhanced chemiluminescence (GE Healthcare, NJ). To assess the effect of PL hydrolysis on the size distribution of mouse HDL, mock-hydrolyzed and sPLA 2 -hydrolyzed HDL (1 g HDL protein) were separated by GGE as described above and immunoblotted using an anti-mouse apoA-I antibody (Biodesign International).

Statistical analyses
Data are expressed as mean ± SEM. After testing for normalcy and equal variance, data was analyzed for statistical signifi cance as indicated in the fi gure legends. Signifi cance was set at = <0.05; = <0.01; and = <0.001.

THP-1 macrophages convert lipid-free SAA to nascent HDL particles
Previous studies determined that lipid-poor SAA stimulates ABCA1-dependent cholesterol effl ux (14)(15)(16)(17). In the current study, we investigated whether SAA is lipidated through the action of ABCA1 to generate nascent HDL particles in a manner analogous to ABCA1-mediated lipidation of apoA-I ( 6 ). THP-1 macrophages were treated with PMA and cholesterol loaded to upregulate ABCA1 expression, and then incubated with 10 µg/ml lipid-free apoA-I or SAA for 24 h. The media from the cells were subjected to nondenaturing GGE followed by immunoblotting to assess the extent of lipidation of the THP-1conditioned apoA-I and SAA ( Fig. 1A , B ). Lipid-free apoA-I migrated as a predominant band below the smallest size standard, whereas lipid-free SAA migrated as two distinct bands on the nondenaturing gel, possibly due to its propensity to aggregate. After THP-1 conditioning, a portion of apoA-I migrated as larger-sized particles, indicating for- SAA2.1, and assessed their ability to stimulate effl ux through ABCG1. In effl ux assays using fi ve separate HDL preparations, there was no signifi cant difference in ABCG1-dependent effl ux stimulated by N WT HDL (2.06 ± 0.13% of total label) and N SAAKO HDL (1.95 ± 0.20% of total label; data not shown). This result was not surprising, given that HDLs isolated from untreated WT and SAAKO mice do not differ in their lipid or apolipoprotein content ( 18 ). However, AP SAAKO HDL was even more effective in mediating ABCG1-dependent cholesterol effl ux compared with AP WT HDL ( Fig. 3 , gray bars). In effl ux assays using fi ve separate N and AP HDL preparations from WT and SAAKO mice, incubations with SAA-containing AP WT HDL resulted in a ‫ف‬ 17% increase in ABCG1dependent effl ux compared with the corresponding N HDL, whereas ABCG1-dependent effl ux to AP SAAKO HDL was increased more than 2-fold (data not shown). Thus, it appears that the presence of SAA per se does not account for the enhanced capacity of mouse AP HDL to serve as an acceptor for ABCG1-mediated effl ux and that other modifi cations of HDL that occur during infl ammation can signifi cantly infl uence cellular cholesterol effl ux by ABCG1.

Phospholipid depletion of HDL reduces ABCG1-mediated effl ux
We recently reported that AP WT HDL has decreased protein and increased PL content compared with N WT HDL, and these differences are even more pronounced in SAA. This enhanced ABCG1-dependant effl ux suggests that the increased size of THP-1-conditioned SAA ( Fig.  1B ) is likely due to lipidation rather than extensive aggregation. Taken together, our results suggest that ABCA1 and ABCG1 can act sequentially to mediate cellular cholesterol export to SAA.

SAA present on mouse AP HDL is not responsible for an enhanced capacity of AP HDL to stimulate ABCG1-mediated effl ux
During an AP response, the majority ( ‫ف‬ 95%) of SAA in plasma is associated with HDL ( 10 ). Thus, it was of interest to determine whether SAA-containing AP HDL differed in its ability to serve as an acceptor for ABCG1-dependent effl ux compared with N HDL. Accordingly, HDL was isolated from plasma of N mice (N WT HDL) and mice 24 h after injection with LPS (AP WT HDL). Whereas SAA was not detectable in N WT HDL, it comprised approximately 40% of the apolipoprotein associated with AP WT HDL (data not shown). SAA-bearing AP HDL exhibited a modest but signifi cant reduced capacity to elicit ABCG1-independent cholesterol effl ux compared with N WT HDL. However, this same HDL elicited signifi cantly more ABCG1-dependent effl ux compared with N WT HDL ( Fig. 3 , white and black bars).
To dissect out the role of SAA versus other modifi cations of AP WT HDL in mediating enhanced ABCG1dependent effl ux, we isolated N and AP HDL from mice lacking the two major AP SAA isoforms, SAA1.1 and and then induced to express ABCG1 as described in "Materials and Methods." Cellular cholesterol effl ux stimulated by 5 h incubations with either lipid-poor apoA-I or SAA or conditioned media from THP-1 cells incubated with apoA-I or SAA (10 µg/ml apolipoprotein) was determined. ABCG1-dependent effl ux represents the difference between BHK cells treated with mifepristone (Total) and untreated cells (ABCG1-independent). Values are the mean ± SEM of triplicate determinations. Total and ABCG1-independent effl ux to lipid-poor apoA-I and SAA, or THP-1-conditioned apoA-I and SAA data were analyzed using one way ANOVA with Tukey-adjusted pairwise comparisons. Different lower case letters identify different means among apoA-I groups; different capital letters identify different means among SAA groups ( P р 0.004). In addition, the ABCG1-dependent component was compared across groups in posthoc tests.
indicates P < 0.001. Similar results were obtained in two additional experiments. ABCG1, ATP binding cassette transporter G1; apoA-I, apolipoprotein A-I; BHK, baby hamster kidney; SAA, serum amyloid A.

DISCUSSION
It has long been recognized that plasma HDL concentrations are inversely related to the risk of atherosclerotic AP SAAKO HDL that lacks SAA ( 18 ). We therefore speculated that alterations in the surface composition of N and AP HDL may account for differences in ABCG1-dependent effl ux. To investigate whether increased PL content of AP HDL enhances its ability to accept cellular cholesterol through ABCG1, N WT HDL (0.92 nmol PL/µg HDL protein), AP WT HDL (0.97 nmol PL/µg HDL protein), and AP SAAKO HDL (1.18 nmol PL/µg HDL protein) were incubated in the presence or absence of sPLA 2 -IIA and then tested for their ability to stimulate ABCG1dependent effl ux ( Fig. 4A-C ). PL hydrolysis was assessed by measuring the release of FFA (data not shown) and was calculated to result in approximately 30-50% loss of PL from HDL particles. PL hydrolysis was also confi rmed by size reduction in sPLA 2 -treated HDLs ( Fig. 4A-C , inserts). For each of the three HDL preparations, hydrolysis by sPLA 2 signifi cantly reduced ABCG1-mediated cellular cholesterol effl ux, indicating that decreasing the PL content of HDL irrespective of the presence of SAA has a negative effect on ABCG1-mediated effl ux.
To substantiate this fi nding, AP WT HDL was incubated with increasing concentrations of sPLA 2 -IIA and then assessed in ABCG1-dependent effl ux assays. As expected, incubations with increasing concentrations of sPLA 2 -IIA resulted in more extensive PL hydrolysis, as assessed by the release of FFA ( Fig.  5A ) and the generation of progressively smaller HDLs (data not shown). In BHK cells lacking ABCG1, PL hydrolysis did not alter the ability of AP WT HDL to serve as an acceptor for cholesterol effl ux ( Fig. 5B ). However, for cells induced to express ABCG1, hydrolysis by sPLA 2 decreased effl ux in a dosedependent manner, indicating that ABCG1-mediated effl ux is reduced as PL content is decreased.

AP human HDL is not a better acceptor for ABCG1-dependent effl ux than normal human HDL
In humans, the AP response is characterized by a marked increase in plasma sPLA 2 -IIA activity that does not occur in C57BL/6 mice due to a frame shift mutation in the mouse sPLA 2 -IIA gene ( 25 ). Thus, it was of interest to determine if AP human HDL is enriched in PL similar to AP WT HDL and if AP human HDL is altered in its ability to serve as an acceptor for ABCG1-mediated cholesterol effl ux. For these studies, HDL was isolated from the plasma of normal volunteers (N hu HDL) and from patients 24 h after cardiac surgery (AP hu HDL). Despite marked enrichment with SAA ( Fig. 6A ), the protein and lipid content of AP hu HDL was not signifi cantly altered compared with N hu HDL ( Table 1 ). As previously reported, AP hu HDL migrated as slightly larger particles compared with N hu HDL when subjected to nondenaturing GGE ( Fig. 6B ). This fi nding is likely due to the enrichment of the HDL surface with SAA ( 10 ). Compared with N hu HDL, AP hu HDL evoked a modest but signifi cant decrease in ABCG1-dependent cholesterol effl ux ( Fig. 6C ). Taken together, our data indicate that the impact of acute infl ammation on ABCG1-dependent effl ux to mouse and human HDL is different and may be related to the effect of the AP response on the PL content of the respective HDLs. 3 H] cholesterol, induced to express ABCG1, and then incubated for 5 h with the indicated HDLs (25 µg/ml). ABCG1-mediated effl ux was calculated as the difference between ABCG1-expressing cells and control cells that were not induced to express ABCG1. Values represent the mean ± SEM of triplicate determinations. Inserts: Aliquots of mock-hydrolyzed ( Ϫ ) and sPLA 2 -hydrolyzed (+) HDL were separated by nondenaturing GGE and immunoblotted with anti-apoA-I to show the size reduction of hydrolyzed HDLs. Signifi cance between mock-and sPLA 2 -hydrolyzed samples was determined by Student t -test. ABCG1, ATP binding cassette transporter G1; AP, acute phase; BHK, baby hamster kidney; N, normal; SAAKO, mice with targeted deletion of SAA1.1 and SAA2.1; sPLA 2 -IIA, Group IIA secretory phospholipase A 2 ; WT, wild-type. ABCG1-dependent effl ux. This is in agreement with an earlier report that ABCG1-dependent effl ux correlates with the PL content of acceptor particles ( 6 ).
We showed that lipid-free SAA is highly effective in stimulating ABCA1-dependent effl ux, consistent with previous reports ( 14-17, 27, 28 ). This fi nding is not unexpected, given that nascent HDL generated through ABCA1 appears to require the presence of amphophilic ␣ -helices as a key conformation in the acceptor polypeptide, rather than a specifi c amino acid sequence ( 27,28 ). The predicted secondary structure of human SAA1 indicates two amphophilic ␣ -helical segments ( 29 ). Like apoA-I, lipidpoor SAA has been shown to stabilize ABCA1 protein, providing additional evidence that SAA interacts with ABCA1 in a manner similar to apoA-I ( 15 ). Based on their migration on nondenaturing gels, the interaction of ABCA1 with lipid-poor SAA appears to generate nascent HDLs with a size distribution that is distinctly larger compared with nascent HDL generated by apoA-I ( Fig. 1 ). Consistent with our analysis, Abe-Dohmae et al. ( 14 ) determined by size-exclusion chromatography that apoA-Icontaining nascent HDL elutes in two distinct peaks, both of which are smaller than nascent HDL generated by SAA.  3 H] cholesterol and then induced to express ABCG1 as described in "Materials and Methods." Cellular cholesterol effl ux stimulated by 5 h incubations with 25 µg/ml N and AP human HDL was measured. ABCG1-mediated effl ux was calculated as the difference between ABCG1-expressing cells and control cells that were not induced to express ABCG1 (C). Values (mean ± SEM) were obtained from eight experiments performed with three preparations of human N and AP HDL and are expressed relative to the corresponding N hu HDL. Signifi cance was determined by Student t -test. ABCG1, ATP binding cassette transporter G1; AP, acute phase; BHK, baby hamster kidney; GGE, gradient gel electrophoresis; N, normal. cardiovascular disease. The protective effect of HDL is largely attributed to its key role in RCT, whereby excess cholesterol from peripheral cells is transported back to the liver for excretion in the bile. During acute infl ammation, HDL undergoes substantial changes in lipid and apolipoprotein composition that could alter its ability to carry out individual steps in the RCT pathway. In this study, we focused on the impact of infl ammation on ABCG1-dependent cellular cholesterol effl ux. We utilized a well-established in vitro system that provides inducible expression of ABCG1 and thus allows for direct measurements of ABCG1-dependent cellular cholesterol effl ux ( 5,7,26 ). We determined that, similar to apoA-I, lipid-poor SAA supports the cooperative interaction of ABCA1 and ABCG1 to produce nascent HDL. We also showed that infl ammatory remodeling of mouse HDL modestly enhances ABCG1dependent effl ux and that the presence of SAA on infl ammatory HDL is not required for this effect. Our data indicate that PL enrichment of HDL, such as what occurs during an AP response in mice, has a positive effect on diabetic patients, reduced expression of ABCG1 is associated with increased cholesterol accumulation in macrophages ( 33 ). The precise relevance of the interaction between ABCA1 and ABCG1 in humans merits further investigation given the known species differences in the regulation of key genes involved in cholesterol homeostasis (34)(35)(36).
The impact of acute infl ammation on macrophage RCT in vivo has been investigated in mice ( 37,38 ). For these studies, 3 H-cholesterol-labeled J774 macrophages or primary mouse macrophages were administered into the peritoneal cavity of normal and LPS-injected mice, and the movement of 3 H-cholesterol from these cells into plasma, liver, and feces was monitored. The results of both studies indicated that acute infl ammation impairs macrophageto-feces RCT. A likely contributing factor to this reduction in RCT was decreased hepatic expression of ABCG5 and ABCG8, major transport proteins mediating biliary cholesterol secretion ( 37,38 ). Whether SAA per se contributes to decreased RCT during the AP response was addressed by Annema et al. ( 37 ), who determined that adenovirus overexpression of mouse SAA (but not human SAA) results in a signifi cant reduction in fecal excretion of the macrophage-derived 3 H-cholesterol tracer. However, there was no evidence that the rate of movement of 3 H-cholesterol from macrophages to the plasma or the liver was impaired as a result of SAA overexpression. Taken together, in vivo studies suggest that the integrated effect of infl ammation is to retard macrophage-to-feces RCT, although the impact of individual components of the AP response on specifi c steps in the RCT pathway has not been completely delineated. Our data indicate that reduced macrophage-to-feces RCT during acute infl ammation in mice is not likely due to a detrimental effect on ABCG1-dependent effl ux.
Studies focusing on the ability of AP-HDL to promote macrophage cholesterol effl ux in vitro have been carried out. In one study, N and AP human HDL promoted cholesterol effl ux from THP-1 cells in a similar manner, but enrichment of the N HDL with SAA ex vivo reduced cellular cholesterol effl ux by 30% ( 39 ). McGillicuddy et al. ( 38 ) reported that infl ammatory remodeling of mouse HDL impairs its capacity to serve as an acceptor for macrophage cholesterol effl ux. They also reported that HDL isolated from humans subjected to experimental endotoxemia was less effective in stimulating cholesterol effl ux from cholesterol-loaded J774 macrophages compared with normal human HDL. Similarly, Annema et al. ( 37 ) concluded that effl ux from cholesterol-loaded THP-1 macrophages to plasma or HDL isolated from sepsis patients was markedly reduced compared with healthy controls. On the other hand, on the basis of their studies in J774 macrophages, Kisilevsky et al. concluded that SAA on AP HDL promotes macrophage effl ux by mobilizing intracellular cholesterol stores, thereby facilitating its transport out of cells ( 40 ).Thus, available data regarding the effect of infl ammatory remodeling and/or SAA enrichment of HDL on macrophage cholesterol effl ux is somewhat confl icting and likely infl uenced by the cell system utilized and the source of the HDL ligand. To date, the impact of They also identifi ed distinct differences in the migration of the two HDLs when separated by agarose gel electrophoresis, refl ecting the respective isoelectric point values of apoA-I and SAA. Thus, nascent HDLs generated by apoA-I and SAA clearly possess different physiochemical properties.
The majority of SAA in plasma is found associated with HDL, predominantly in the denser HDL 3 subfraction ( 10 ). The process by which SAA-containing HDL is formed during acute infl ammation is not entirely clear. One proposed mechanism is that lipid-free SAA secreted by hepatocytes associates with existing spherical HDL particles through a remodeling process that may involve the displacement of apoA-I ( 10,30,31 ). However, induction of an AP response in apoA-I-defi cient mice leads to the formation of large, spherical HDL particles in which >90% of the protein is SAA, suggesting that SAA is capable of sequestering lipid to form HDL in the absence of apoA-I and other apolipoproteins ( 30,31 ). On the other hand, adenoviral vectormediated expression of SAA in apoA-I-defi cient mice in the absence of infl ammation results in circulating SAA that is mostly in a lipid-poor form, suggesting that components of the AP response are required for the biogenesis of SAA-rich HDL in the absence of apoA-I ( 32 ). Studies in ABCA1-defi cient mice demonstrate that the formation of SAA-containing HDL is dependent on ABCA1 ( 15 ), consistent with several reports, including the current study, that incubating cells with exogenous SAA results in robust cholesterol release in an ABCA1-dependent manner (14)(15)(16). In this study, we show for the fi rst time that nascent HDL generated by SAA serves as an effi cient substrate for ABCG1-mediated cellular cholesterol effl ux. The extent to which ABCG1 contributes to the biogenesis of SAA-containing HDLs in vivo requires further study. What also is not entirely clear is whether ABCG1 plays a major role in cholesterol effl ux in humans. Although extensive evidence indicates that ABCA1 and ABCG1 act in an additive manner to promote RCT in mice (1)(2)(3), the importance of this interaction in humans is likely more complex and less well studied. LXR-stimulated cholesterol effl ux in human cholesterol-loaded macrophages (THP-1 cells or peripheral blood monocytes) was reported to be independent of ABCG1 ( 4 ). Others, however, have shown that in type II infl ammation on ABCG1-dependent effl ux has not been specifi cally addressed.
In the current study, we determined that infl ammatory remodeling of mouse HDL, particularly PL enrichment, had an enhancing effect on ABCG1 effl ux that was more pronounced in mice lacking SAA ( Fig. 3 ). In humans, infl ammatory remodeling of HDL appeared to negatively impact effl ux through ABCG1, as HDL isolated from patients undergoing an AP response due to cardiac surgery was modestly defi cient in its ability to stimulate ABCG1dependent effl ux compared with HDL from healthy controls ( Fig. 6 ). We recently reported that acute infl ammation in WT mice results in a modest increase in HDL PL content, and AP HDL from SAA DKO mice is even more PL-enriched ( 18 ). In contrast, compositional analyses indicated that the AP response in humans is not associated with PL enrichment of HDL (Table 1), possibly due to the induction of sPLA 2 -IIA that occurs during the AP response in humans but not C57BL/6 mice. Thus, our data suggests that increasing the PL content of HDL has a positive effect on ABCG1-dependent effl ux. This conclusion is substantiated by our fi nding that sPLA 2 -IIA hydrolysis of normal and AP mouse HDL signifi cantly decreased effl ux through ABCG1 ( Figs. 4 and 5 ). However, the difference in ABCG1dependent effl ux between normal and AP HDLs may not be entirely attributable to differences in PL content. For example, sPLA 2 hydrolysis of AP WT HDL resulted in a 31% decrease in PL content, and this was associated with a 30% decrease in effl ux ( Fig. 4B ). A similar 33% difference in ABCG1-dependent effl ux was observed between N WT HDL and AP WT HDL ( Fig. 3 ), despite only a 5% difference in PL content between these two HDLs. Studies to compare macrophage RCT in WT and human sPLA 2 -IIA transgenic mice have demonstrated a decreased rate of appearance of 3 H-cholesterol in plasma of transgenic mice, indicating that sPLA 2 may decrease macrophage cholesterol effl ux during infl ammation ( 37 ). These results appear to contradict the work of Sankaranarayanan et al. ( 26 ), who reported that HDL 3 incubated with multilamellar vesicles to enrich the particles with PL were not altered in their ability to serve as substrates for ABCG1-dependent effl ux. The reason for the discrepant results is not clear. It is possible that enriching HDL with PL ex vivo generates a particle whose structure is not equivalent to HDLs that are remodeled during acute infl ammation in vivo.
In summary, our data indicate that the extensive remodeling of HDL that occurs during acute infl ammation has modest effects on ABCG1-dependent effl ux. Interestingly, infl ammatory remodeling of mouse HDL increases ABCG1dependent effl ux, whereas effl ux stimulated by human HDL appears to decrease during infl ammation. We provide evidence that this differential effect is likely due to differences in the PL content of AP HDL from the respective species. The presence of SAA on AP HDL does not appear to alter the capacity of the particle to serve as a substrate for ABCG1. The capacity of lipid-poor SAA to promote sequential effl ux from ABCA1 and ABCG1 may be important for cholesterol fl ux out of atherosclerotic lesions, where each of these three factors is present.