ABCA1 promotes the efflux of bacterial LPS from macrophages and accelerates recovery from LPS-induced tolerance.

Macrophages play important roles in both lipid metabolism and innate immunity. We show here that macrophage ATP-binding cassette transporter A1 (ABCA1), a transporter known for its ability to promote apolipoprotein-dependent cholesterol efflux, also participates in the removal of an immunostimulatory bacterial lipid, lipopolysaccharide (LPS). Whereas monocytes require an exogenous lipoprotein acceptor to remove cell-associated LPS, macrophages released LPS in the absence of an exogenous acceptor by a mechanism that was driven, in part, by endogenous apolipoprotein E (apoE). Agents that increased ABCA1 expression increased LPS efflux from wild-type but not ABCA1-deficient macrophages. Preexposure of peritoneal macrophages to LPS for 24 h increased the expression of ABCA1 and increased LPS efflux with a requirement for exogenous apolipoproteins due to suppression of endogenous apoE production. In contrast, LPS preconditioning of ABCA1-deficient macrophages significantly decreased LPS efflux and led to prolonged retention of cell-surface LPS. Although the initial response to LPS was similar in wild-type and ABCA1-deficient macrophages, LPS-induced tolerance was greater and more prolonged in macrophages that lacked ABCA1. Our results define a new role for macrophage ABCA1 in removing cell-associated LPS and restoring normal macrophage responsiveness.

defi cient mice have revealed that LPS can persist in cells and tissues, and if not inactivated, the resulting chronic LPS stimulation produces prolonged macrophage tolerance to LPS and other microbial stimuli accompanied by harmful immunosuppression ( 28 ) and chronic pathologies involving other cells and tissues ( 29,30 ). The regulatory mechanisms that maintain the tolerant state are exceedingly complex (reviewed in Ref. 24 ) and include several negative feedback loops that inhibit the Tlr4 signaling pathway at multiple levels. Other regulatory mechanisms, such as those involved in chromatin modifi cations, have positive or negative effects that are gene-specifi c. The recovery from tolerance presumably occurs when the macrophage has disposed of enough of its bioactive LPS to turn off Tlr4 signaling ( 28 ).
Our previous studies have revealed another mechanism that removes cellular LPS. We found that HDL and other plasma lipoproteins promote the release of LPS from the monocyte cell surface ( 31 ), a process that requires an LPSbinding acceptor (e.g., HDL) and is accelerated by an LPS transfer protein, principally soluble CD14 (sCD14) ( 31,32 ). LPS effl ux from monocytes is accompanied by decreased cytokine responses to LPS that has already bound to the cell; the abilities of sCD14 ( 32 ) and LBP ( 33 ) to attenuate cell responses can occur by different mechanisms that both move LPS away from MD-2/Tlr4.
In the present study, we show that LPS effl ux from macrophages occurs by a different mechanism that does not require exogenous LPS acceptors or transfer proteins. We hypothesized that endogenous (macrophage-produced) apoE might play a role in LPS effl ux, because it (a) is one of the most highly expressed macrophage genes ( 34,35 ), (b) plays a role in cholesterol and phospholipid effl ux (36)(37)(38), and (c) can bind LPS ( 39 ). We also hypothesized that ABCA1 is involved, because previous studies have shown that ABCA1 is a receptor for apoE and other exchangeable apolipoproteins ( 40,41 ) and that ABCA1 promotes apolipoprotein-mediated cholesterol and phospholipid effl ux ( 1,2 ). Our results reveal the complexity of macrophage LPS effl ux, and they defi ne new roles for ABCA1 and apolipoproteins in removing cell-associated LPS and accelerating macrophage recovery from LPSinduced tolerance.

Experimental subjects and cells
Wild-type (WT) C57BL/6J mice (stock #002207), mice with targeted deletions of the ABCA1 gene (ABCA1 KO or ABCA1 Ϫ / Ϫ ) (B6:DBA/1 background; stock #003897) and the apoE gene (apoE KO or apoE Ϫ / Ϫ ) (C57Bl/6J background; stock # 002052) were obtained from the Jackson Laboratory (Bar Harbor, ME). Mice with targeted deletions of the LXR ␣ and LXR ␤ genes (C57BL/6:129Sv background) were generated in the Mangelsdorf lab as previously described ( 42 ). ABCA1 Ϫ / Ϫ mice derived from a different source ( 43 ) were also utilized after being backcrossed onto the C57Bl/6:129/Sv background as described ( 44 ). Peritoneal macrophages were isolated by lavage 4 days after intraperitoneal injection of 1.5 ml of sterile Brewer Thioglycollate evidence that transporter-dependent downregulation of infl ammatory responses is mechanistically linked to sterol effl ux. The control of ABCA1 expression is complex and highly regulated by multiple transcriptional and posttranscriptional processes, including those that infl uence protein activity and stability ( 1 ). Transcriptional activation is induced by oxysterols via the activation of liver X receptors LXR ␣ and LXR ␤ , which form heterodimers with retinoid X receptors (RXR) and bind to response elements in the ABCA1 gene promoter ( 8 ). LXR activation also induces ABCG1 and apoE ( 9 ). ABCA1 mRNA and protein levels are also increased by cAMP ( 10 ), interleukin (IL)-10 ( 11 ), transforming growth factor (TGF) ␤ ( 12 ), and tumor necrosis factor (TNF) ( 13 ). However, reports on the effects of LPS are confl icting. In one study, LPS increased ABCA1 expression by an LXR-independent pathway in the human THP-1 premonocyte cell line and in mouse liver in vivo, whereas ABCG1 expression was decreased ( 14 ). In other studies, LPS decreased macrophage ABCA1 expression (15)(16)(17)(18).
LPS (also called endotoxin) is among the most potent of the immunostimulatory bacterial lipids. Host recognition of LPS by Toll-like receptor 4 (Tlr4) induces a large array of genes whose products have infl ammatory, antiinfl ammatory, and antimicrobial activities. Sensitive host responses to LPS are orchestrated by several proteins. LPSbinding protein (LBP) ( 19 ), a secreted protein found in the plasma and extravascular fl uids, transfers LPS to CD14 ( 20,21 ), an LPS binding receptor that exists in soluble or membrane-bound forms. CD14 greatly increases host cell sensitivity to LPS by transferring LPS to the MD-2/Tlr4 receptor complex ( 22,23 ). Sentinel responses to LPS often confer resistance to infection, whereas the response to LPS during uncontrolled infection contributes to the pathophysiology of severe sepsis and septic shock. Therefore, animal hosts have evolved mechanisms for controlling LPS responses. Endotoxin tolerance or gene reprogramming (reviewed in Refs. 24 and 25 ) is thought to be an important mechanism for minimizing infl ammationinduced damage during recovery from bacterial infection. During the tolerant or gene reprogrammed state, the vast majority of infl ammation-inducing genes are suppressed, whereas other subsets of LPS-responsive genes (such as those involved in antimicrobial effectors mechanisms) remain LPS-inducible due to chromatin modifi cations ( 26 ). The tolerant state has been associated with a high risk of secondary infection and mortality, and recovery from tolerance has been associated with survival. Endotoxin tolerance has also been identifi ed in noninfectious settings such as smoking, trauma, surgery, pancreatitis, alcoholic cirrhosis, and cancer, some of which may be associated with gut translocation of endotoxin.
One important mechanism for recovery from endotoxin tolerance involves the ability of macrophages and other phagocytes to rapidly internalize LPS and partially degrade it over a period of days. This critical catabolic step is performed by acyloxyacylhydrolase (AOAH), a phagocyte enzyme that inactivates LPS by removing secondary fatty acyl chains from its lipid A moiety ( 27 ). Studies with AOAH-In experiments that involved LPS conditioning, the cells were preincubated with unlabeled O14 LPS for 24 h and washed before adding the [ 3 H]LPS. To distinguish between cell-surface bound and internalized [ 3 H]LPS, the cells were incubated on ice with 0.02% proteinase K in PBS for 30 min to release LPS that was bound to surface proteins as previously described ( 47 ). This treatment did not cause detachment of the macrophages from the culture dish and did not cause membrane permeability to trypan blue. Radioactivity was measured in the proteinase K supernatants and in the cells after detaching the cells in PBS with a cell scraper. In experiments in which the effl ux time course exceeded 1 h, the effl uxed and cell-associated [ 3 H]LPS or [ 3 H/ 14 C] LPS was precipitated with ethanol as previously described ( 30 ) to measure LPS, which is ethanol-insoluble, and to exclude ethanolsoluble degradation products such as free fatty acids.
LPS effl ux from human monocytes or macrophages in suspension was performed as previously described ( 31,32

ApoE inhibition
To inhibit apoE expression by small inhibitory RNA (siRNA) knockdown, human macrophages were taken from Tefl on beakers on day 2 of the above protocol, and 3 × 10 6 cells/tube were electroporated with 2 µM apoE siRNA (sense 5 ′ -GGAGUU-GAAGGCCUACAAAtt-3 ′ ; antisense 5 ′ -UUUGUAGGCCUUCAA-CUCCtt-3 ′ ; ID #41694, Ambion, Inc., Austin, TX) or control siRNA (Negative control #1 siRNA, Ambion #4611) using the Amaxa Human Monocyte Nucleofector Kit and electroporator according to the manufacturer's instructions (Lonza, Basel, Switzerland). The mRNA analysis and measurement of apoE secretion were performed after an additional three days of culture in complete medium. The amount of apoE produced by the control macrophages in 48 h varied among the three donors (150, 410, and 607 ng/ml/10 6 electroporated cells), whereas apoC-I production was only 2-7% that of apoE and was not inhibited by the siRNA (data not shown).
In other experiments, antibodies to apoE were used to inhibit endogenous apoE activity. [ 3 H]LPS was bound to cultured human macrophages as described above. The cells were washed with cold SFM and incubated for 30 min at 4°C in SFM containing control or anti-apoE antibodies [20 g/ml each of mouse monoclonal antibody (H11004M) and rabbit polyclonal (K741180B) antibody (Meridian Life Science/BioDesign, Saco, ME); controls contained nonimmune control mouse IgG1 and rabbit IgG (Sigma)]. The mixtures were then warmed to 37°C for 15 min to allow the release of [ 3 H]LPS.

LPS response and recovery from tolerance
Peritoneal macrophages were cultured for 16 h in 24-well plates (4 × 10 5 cells/well) as described above. The adherent cells were incubated with 10 ng/ml of unlabeled E. coli O14 LPS or control buffer for 24 h in complete medium. The cells were then washed twice, incubated with fresh complete medium for 0-5 days, and challenged with 10 ng/ml LPS for 2 h in experiments to measure mRNA or for 6 h in experiments to measure cytokine protein production. The plates were placed on ice, the supernatants were removed, and the cells were lysed with RLT buffer (RNeasy kit, Qiagen, Valencia, CA) to isolate total RNA. In experiments that measured cytokine protein production, the cells were lysed in PBS containing 1% Triton X-100, and the cell Medium (Difco). The peritoneal cells were counted using a hemocytometer and were cultured in 24-well plates (4 × 10 5 cells/ well) in 0.5 ml of RPMI 1640 (Cellgro; Fisher Scientifi c) containing 10% heat-inactivated FBS (Hyclone, defi ned; Fisher Scientifi c) with penicillin and streptomycin (complete medium) ( 45 ) for 3-4 h to obtain adherent macrophages. The animal protocol was approved by the Institutional Animal Care and Use Committee (University of Texas Southwestern Medical Center). Human peripheral blood was obtained by informed consent from healthy volunteers according to protocols approved by the Institutional Review Board. HDL (1.063 < d < 1.21 g/cc) was prepared from freshly drawn blood of three healthy volunteers by ultracentrifugal fl otation ( 31 ). Normal human monocytes were prepared from peripheral blood mononuclear cells (PBMC) isolated on Histopaque 1077 (Sigma) by adherence to plastic for 1-2 h. Human macrophages were cultured either in suspension or attached to culture plates in complete medium containing 50 ng/ml recombinant human macrophage colony-stimulating factor (M-CSF)(Sigma) for 5-7 days. For cultures in suspension, the adherent monocytes were lifted from 10 cm culture plates by incubating the cells briefl y with PBS containing 1 mM EDTA and were then cultured in Tefl on beakers as described above. THP-1 cells (a human premonocyte cell line) were cultured in 0.05 M 1,25 dihydroxyvitamin D 3 (VD 3 ) (BioMol, Plymouth Meeting, PA) for 3 days to induce mature monocyte characteristics ( 33 ). THP-1 cells were differentiated into adherent macrophages by adding 100 nM phorbol myristate acetate (PMA) to the above culture for the fi rst 24 h followed by 48 h in VD 3 alone. After culture in PMA, the cells spread and adhered tightly to the plate, and they stopped proliferating.  In experiments using T0901317 and/ or cAMP, the cells were preincubated with these compounds in complete medium (or medium containing 5% lipoprotein-deficient FBS) for 16-24 h, and the drugs were also added to the appropriate SFM-containing cultures during the LPS effl ux assay. peritoneal macrophages and human THP-1 macrophages compared with that of THP-1 monocytes and normal human monocytes.

LPS effl ux assays
To investigate the role of apoE in LPS effl ux from macrophages, we fi rst used real-time PCR to quantitate apoE mRNA expression. As shown in Table 1 , we found that cultured human macrophages expressed the highest levels of apoE among the human cells tested, and as expected, monocytes expressed low levels of apoE. To test the hypothesis that endogenous apoE might play a role in LPS effl ux, we attempted to inhibit its activity with anti-apoE antibodies and its expression using siRNA. As shown in Fig. 1B , preincubation of cultured human macrophages with antibodies to apoE signifi cantly decreased [ 3 H]LPS effl ux from the cells in serum-free medium that did not contain any exogenous apoE. ApoE siRNA knockdown experiments yielded similar results. As shown in Fig. 1C , incubation of macrophages with apoE siRNA decreased apoE protein production by 95%; however, this decreased LPS effl ux by only 32%. The data indicate that although endogenous apoE plays a signifi cant role in LPS effl ux, the majority of LPS effl ux from cultured human macrophages is apoE-independent.
To determine if exogenous apolipoproteins could serve as LPS acceptors during macrophage LPS effl ux, we tested the activities of purifi ed apoE and certain other exchangeable apolipoproteins that are normally found in HDL. As shown in Fig. 1D , low concentrations of exogenous apoE and apoC-I increased LPS effl ux from THP-1 macrophages, whereas apoC-II was relatively ineffective at low concentrations. ApoA-I, the only apolipoprotein tested that is not expressed by macrophages, also promoted LPS effl ux at low apolipoprotein concentrations. Although apoC1 mRNA was highly upregulated in cultured human macrophages (data not shown), apoC-I protein production by these cells was only 2-7% that of apoE (data not shown).
We next tested the hypothesis that ABCA1 plays a role in LPS effl ux. The effects of agents that augment the expression of ABCA1 were compared in peritoneal macrophages from ABCA1 +/+ and ABCA1 Ϫ / Ϫ mice. cAMP induces ABCA1 expression by an LXR-independent mechanism (Ref. 10 and data not shown), and T0901317 ( 48 ) is a nonsteroidal activator of the nuclear receptor LXR, which in turn activates ABCA1 transcription ( 8 ). The effects of these agents on ABCA1 expression are shown in the legend below supplementary Table II. To minimize the basal expression of ABCA1 and other LXR target genes, we fi rst preincubated the macrophages in lipoprotein-defi cient serum (i.e., serum that does not contain cholesterol or oxysterol ligands of LXR). T0901317 increased LPS effl ux in WT macrophages in SFM in the absence of exogenous acceptors ( Fig. 2A ), whereas it had little or no effect in macrophages that lacked ABCA1 ( Fig. 2B ). T0901317 also increased cholesterol effl ux to an extent similar to that of LPS effl ux in WT macrophages (Refs. 49 , 50 , and data not shown). T0901317 decreased the level of protease-sensitive (cell-surface) LPS in WT ( Fig. 2C ) but not in ABCA1 Ϫ / Ϫ macrophages ( Fig.   2D ), and it decreased protease-resistant (predominantly protein was measured by the BCA (bicinchoninic acid) method (Pierce Chemical Co.) using BSA as a standard.

mRNA analysis
Total RNA (RNeasy kit) was reverse transcribed by Superscript II (Invitrogen, Carlsbad, CA) using random hexamer primers. In some experiments, total RNA was reverse transcribed using the iScript cDNA Synthesis Kit containing oligo(dT) and random hexamer primers (BioRad, Hercules, CA). Relative levels of mRNA were quantitated using Sybr Green Master Mix in a Model 7300 Real Time PCR System (Applied Biosystems, Foster City, CA) as previously described ( 45 ). Analysis of decreasing amounts of total RNA confi rmed that real-time PCR measurements were within the dynamic range for each gene. The data were analyzed by the ddCt method after validation of each primer pair by the standard curve method; the results were normalized by levels of 18S rRNA. PCR primers used are shown in supplementary Table  I. Melting point analyses performed at the end of each PCR run revealed a single amplifi cation product for each primer pair. The fold change in mRNA level was measured relative to the unstimulated WT control for each gene. The baseline mRNA levels were undetectable for some genes; in these cases, we arbitrarily assigned a threshold cycle number of 40.

Western blotting and ELISA
Total membrane fractions of peritoneal macrophages were prepared as previously described ( 45 ), and 10 g of membrane protein in SDS-PAGE sample buffer was warmed to 95°C for 5 min for apoE samples and 25°C for ABCA1 samples. The samples were fractionated on SDS-PAGE gels and transferred to Immobilon-P membranes as described ( 45 ). The membranes were probed with affi nity purifi ed rabbit anti-ABCA1 antibodies (#NB400-105, Novus Biologicals, Littleton, CO) or anti-apoE antibodies ( 45 ) and detected by enhanced chemiluminescence ( 45 ). Secreted human apoE and apoC-I were measured by ELISA as described ( 45 ). TNF, IL-6, and RANTES [chemokine (C-C motif) ligand 5 (Ccl5)] were measured in culture supernatants by ELISA. The RANTES ELISA set was from R and D Systems (Minneapolis, MN), and the other sets were from BD Biosciences (San Jose, CA).

Statistics
The data were analyzed using Prism 5.0 software from Graph-Pad (San Diego, CA) with error bars denoting standard deviation (SD), standard error of the mean (SEM), or 95% confi dence interval (CI). Signifi cant differences from controls were determined as two-tailed P values of the unpaired t -test.

ABCA1 and endogenous apoE promote LPS effl ux from macrophages
We previously showed that the rapid release of cell surface-bound LPS from human monocytes requires an LPS transfer protein and a lipoprotein acceptor (e.g., soluble CD14 and HDL) ( 31 ). Here we show that the differentiation of monocytes into macrophages alters the mechanism of LPS effl ux. Fig. 1A shows that after normal human monocytes had differentiated into macrophages by culturing them for 5-7 days, they released cell-associated LPS much more rapidly in the absence of plasma components and without an exogenous acceptor. LPS effl ux under these conditions was also signifi cantly elevated in mouse (Ctrl) ABCA1 +/+ macrophages was similar to that of ABCA1 Ϫ / Ϫ macrophages, suggesting the existence of an ABCA1-independent effl ux mechanism. Nevertheless, the ABCA1 Ϫ / Ϫ Ctrl group retained signifi cantly more cell-surface internalized) LPS after 60 min in WT macrophages ( Fig.  2E ) but not in ABCA1 Ϫ / Ϫ macrophages ( Fig. 2F ). Under these lipoprotein-defi cient conditions that minimize ABCA1 expression, the rate of LPS effl ux in unstimulated  Human monocytes were differentiated into macrophages in complete medium, and mRNA levels for apoE, ABCA1, and ABCG1 were measured by real time PCR as described in "Materials and Methods." The data are shown as fold-change from levels in normal human monocytes (mean ± SD, n = 6). Normal human monocytes and macrophages were derived from three healthy volunteers. ABCG1, ATP-binding cassette transporter G1; apo, apolipoprotein.
a P < 0.05 from the monocyte group of the same cell type. b P < 0.01 from the monocyte group of the same cell type. c P < 0.001 from the monocyte group of the same cell type.
basal ABCA1 expression in the LXR ␣ ␤ Ϫ / Ϫ cells ( 8 ) accompanied by only a slight decrease in apoE expression ( 9 ). We conclude from these experiments that ABCA1 and apoE promote LPS effl ux from macrophages in the absence of an exogenously added LPS acceptor. Under preincubation conditions that minimize ABCA1 expression, the data also reveal LPS effl ux mechanisms that are independent of ABCA1 or apoE in unstimulated ABCA1 Ϫ / Ϫ or apoE Ϫ / Ϫ macrophages, respectively.

LPS induces ABCA1 expression and LPS effl ux in the presence of exogenous apolipoproteins
To determine the effects of prior exposure of macrophages to LPS, we cultured ABCA1 +/+ and ABCA1 Ϫ  Table II, cAMP also increased LPS effl ux in WT macrophages, but it had the reverse effect of slightly decreasing LPS effl ux in ABCA1 Ϫ / Ϫ macrophages. T0901317 had similar effects in the three WT mouse strains (DBA, C57Bl/6, and B6/129) used in these experiments. In apoE Ϫ / Ϫ mice, T0901317 induced less LPS effl ux (17% increase) than in apoE +/+ macrophages (41% increase), indicating the induction of an endogenous apoE-dependent effl ux mechanism consistent with our results in human macrophages ( Fig. 1B, C ), suggesting a role for endogenous apoE. As expected, T0901317 had no effect on LPS effl ux in LXR ␣ ␤ Ϫ / Ϫ macrophages, indicating that the effects of this drug were mediated by LXR activation. The observed 60% increase in the basal level of LPS effl ux in unstimulated LXR ␣ ␤ Ϫ / Ϫ versus LXR ␣ ␤ +/+ macrophages may be due to increased Ϫ / Ϫ was no signifi cant difference in LPS uptake (i.e., the total initial cell-associated LPS) between ABCA1 +/+ and ABCA1 Ϫ / Ϫ macrophages (data not shown). As shown in Fig. 3A , naïve ABCA1 +/+ macrophages (preincubated without LPS) promoted signifi cantly more [ 3 H]LPS effl ux in 1 h than naïve ABCA1 Ϫ / Ϫ macrophages, and exogenously added apoE increased LPS effl ux slightly in cells from both strains. Preincubation of the macrophages with LPS decreased [ 3

H]LPS effl ux in ABCA1
Ϫ / Ϫ cells but not in ABCA1 +/+ cells. After LPS pretreatment, exogenous apoE induced a 77% increase in LPS effl ux in control macrophages, whereas apoE had no effect in cells that lacked ABCA1. ApoA-I was similarly effective at promoting LPS effl ux in ABCA1 +/+ cells that were preincubated with LPS (supplementary Fig. I).
The most striking effect of LPS preconditioning was its ability to prolong the retention of LPS on the cell surfaces of ABCA1 Ϫ / Ϫ macrophages, whereas it decreased LPS retention on the surfaces of cells that expressed ABCA1 ( Fig.  3B ). The addition of apoE caused an additional decrease in cell-surface LPS in LPS-pretreated ABCA1 +/+ macrophages but had no effect on ABCA1 Ϫ / Ϫ macrophages.
The ability of LPS preconditioning to augment the LPS effl ux-promoting activity of apoE may be due to the ability of LPS to increase ABCA1 expression. Therefore, we tested the effects of LPS on macrophage expression of both ABCA1 and apoE. As shown in Table 2 , exposure of macrophages to LPS for 16-24 h increased ABCA1 mRNA levels over 5-fold, whereas apoE mRNA was decreased by 43% and was suppressed somewhat more after 48 h; T0901317 increased ABCA1 mRNA to high levels, as expected. As shown in Fig. 4 , Western blots of the total membrane fraction of macrophages showed that LPS induced a striking increase in ABCA1 protein by 24 h, which decreased only slightly after 48 h. As expected, LPS pretreatment decreased apoE secretion by over 60% ( Table 2 ), and it also decreased cellular apoE in macrophage membrane fractions ( Fig. 4A, B ).
We conclude that after exposure of macrophages to LPS, ABCA1-defi cient macrophages retain LPS on the cell surface for a longer period of time than WT macrophages. Moreover, LPS preconditioning appears to decrease the activity of ABCA1-independent LPS effl ux mechanisms, and that preconditioning makes apoE-mediated LPS effl ux completely dependent upon ABCA1. The data also suggest that the ability of LPS to induce ABCA1 and to suppress apoE expression both contribute to the enhanced ability of exogenous apolipoproteins to promote LPS effl ux in LPS-pretreated WT macrophages.
Macrophages rapidly internalize LPS and retain it for an extended period of time ( 28,47 ). To determine whether  Fig. 2 and expressed as percentage of the total initial cell-associated LPS. Error bars denote mean ± SD for 3-4 experiments performed in duplicate. Asterisks denote signifi cant differences as indicated (* P < 0.05, ** P < 0.01, and *** P < 0.001). apo, apolipoprotein; LPS, lipopolysaccharide; SFM, serum-free medium.
conditioning, some of these genes became hyporesponsive to subsequent LPS challenge ( Fig. 5A , Fig. 6A , and supplementary Table III), and the other genes either responded normally to subsequent LPS challenge or were "primed" and hyper-responsive to subsequent LPS challenge ( Fig.  5B , Fig. 6B , and supplementary Table III). The responsiveness of the nontolerizable genes is due to chromatin modifi cations induced during LPS preincubation, which render these genes permissive for response to LPS ( 26 ). The analysis of IL-1 ␤ , a representative tolerizable gene ( Fig. 5A ), shows that LPS induced similar IL-1 ␤ mRNA levels in ABCA1 +/+ and ABCA1 Ϫ / Ϫ macrophages that had not been preincubated with LPS. In contrast, cells that received a macrophages. After LPS loading at the beginning of the effl ux time course (time 0), only 14% of the total cellular [ 3 H/ 14 C]LPS could be removed from the cell surface by protease treatment, suggesting that the remaining ‫ف‬ 86% of the cellular LPS had been internalized. After three days, the ABCA1 +/+ cells had released approximately 40% of the cellassociated LPS into the culture medium, whereas ABCA1 Ϫ / Ϫ cells release approximately 30%. Throughout days 1-6, very little ( ‫ف‬ 5%) of the total LPS could be identifi ed on the cell's surfaces. We conclude that although ABCA1 appears to have a more robust role in removing LPS from the cell surface ( Fig. 3 ), the data also suggest that ABCA1 plays a significant role in the effl ux of intracellular LPS (supplementary The intracellular enzyme AOAH partially deacylates LPS over a period of days by removing two of its six secondary fatty acyl chains, a process that greatly decreases LPS bioactivity ( 27,28 ). To determine the extent of LPS deacylation during the six-day effl ux time course, we measured the 3 H to 14  C]LPS loading period, only approximately 20% of the secondary fatty acyl chains had been removed. Deacylation increased after one additional day. However, the loss of secondary fatty acyl chains from the LPS did not exceed approximately 50% during the six-day period, suggesting that a signifi cant amount of fully acylated bioactive LPS was inaccessible to AOAH in both ABCA1 +/+ and ABCA1 Ϫ / Ϫ macrophages. These results support the hypothesis that ABCA1 plays a role in removing both bioactive (fully acylated) and attenuated or inactive (partially deacylated) LPS from macrophages.

Delayed recovery from LPS-induced tolerance in ABCA1-defi cient macrophages
To determine how prolonged retention of LPS by ABCA1 Ϫ / Ϫ macrophages affects cell function, we measured the ability of macrophages to recover from LPSinduced tolerance after preincubation with LPS for 24 h. Using real time PCR, we measured steady-state mRNA levels of a panel of ten LPS-responsive genes. After LPS pre-

ND ND ND
Peritoneal macrophages from wild-type mice were incubated as described in Fig. 3 in the presence or absence of LPS or 5 µM T0901317 (T) for the indicated times. mRNA levels were measured by real time PCR as described in "Materials and Methods." To measure secreted apoE, the medium was replaced with fresh medium after the indicated incubation times, and apoE was measured after 8 h in the culture supernatants by Western blotting. The data are shown as fold-change from levels in unstimulated control macrophages (Ctrl) and are expressed as mean ± SD for mRNA (n = 5) and secreted apoE (n = 3). ABCG1, ATP-binding cassette transporter G1; apo, apolipoprotein; Ctrl, unstimulated control macrophages; LPS, lipopolysaccharide; T, T0901317; ND, not done. a P < 0.01 from Ctrl. b P < 0.001 from Ctrl. of mRNA levels in naïve cells (764% in ABCA1 +/+ versus 340% in ABCA1 Ϫ / Ϫ cells). Similar results were obtained with most of the other genes analyzed, as shown in Fig. 6 (day 0) and supplementary Table III (day 0). To measure the recovery of macrophages from tolerance, we incubated naïve or LPS-conditioned cells in fresh medium for 1-5 days and subsequently challenged them with LPS for 2 h. Fig. 6 shows mRNA levels after LPS challenge of LPS-conditioned macrophages expressed as the percentage of mRNA levels in LPS-challenged naïve macrophages. Thus, a level of 100% would indicate a return to normal responsiveness for each experimental group. The results show that, for tolerizable genes, ABCA1 Ϫ / Ϫ macrophages generally had lower responses after LPS-conditioning and delayed recovery from tolerance compared with ABCA1 +/+ macrophages ( Fig. 6A ). Nontolerizable genes also showed decreased LPS responsiveness in ABCA1 Ϫ / Ϫ macrophages during the same period after LPS conditioning ( Fig. 6B ).
Differences were most pronounced after one day of recovery, at which time seven genes showed signifi cantly lower responsiveness to LPS challenge in ABCA1 Ϫ / Ϫ macrophages.
After two days of recovery, four genes remained signifi cantly less responsive in the ABCA1 Ϫ / Ϫ group. After fi ve days of recovery, three of the tolerizable genes (TNF, IL-1 ␤ , and IL-12b) had not fully recovered to normal responsiveness in both groups, and the most profoundly tolerized gene, IL-12b, remained more suppressed in the ABCA1 Ϫ / Ϫ group.
Similar results were obtained for each time point when the data were expressed as fold changes from unstimulated naïve macrophages (supplementary Table III).
To analyze the recovery from tolerance by measuring cytokine protein production, we challenged naïve or LPSconditioned macrophages with LPS in the same manner as the above mRNA experiments, except that the challenge was continued for 6 h to allow time for protein production of TNF, IL-6, and RANTES, which were measured in the culture supernatants by ELISA. As shown in Fig. 7A , TNF was profoundly tolerized in both ABCA1 +/+ and ABCA1 Ϫ / Ϫ macrophages. However, the ABCA1 Ϫ / Ϫ cells remained less responsive to LPS than the ABCA1 +/+ cells throughout the fi ve-day recovery period. The ABCA1 +/+ cells fully regained their ability to produce IL-6 after one day of recovery, whereas the ABCA1 Ϫ / Ϫ cells remained tolerant; although both groups of macrophages were equally responsive to LPS challenge after two days of recovery, the ABCA1 +/+ group became signifi cantly primed after fi ve days, whereas the responsiveness of the ABCA1 Ϫ / Ϫ did not change ( Fig.   7B ). In contrast, both groups of macrophages produced slightly more RANTES when challenged immediately after LPS conditioning (day 0). After one day of recovery, the ABCA1 Ϫ / Ϫ cells became signifi cantly hyporesponsive for RANTES compared with the ABCA1 +/+ group and remained so during the fi ve days of recovery. When the data were expressed as total cytokine production (ng/100 µg cell protein) in LPS-conditioned macrophages (supplementary Fig. IIID-F), TNF production was almost completely ablated after LPS challenge in ABCA1 Ϫ / Ϫ cells and remained so after two days of recovery; after fi ve days of recovery, the response of the ABCA1 Ϫ / Ϫ cells was still tolerizing dose of LPS for 24 h had a diminished response to subsequent LPS challenge. The response to LPS challenge was signifi cantly more diminished in cells that lacked ABCA1 both in terms of fold change from unstimulated control macrophages (0.31 ratio of ABCA1 Ϫ / Ϫ to ABCA1 +/+ ) and percentage of mRNA levels in LPS-challenged naïve macrophages that had not been pretreated with LPS (58% in ABCA1 +/+ versus 18% in ABCA1 Ϫ / Ϫ cells). The analysis of IL-12a, a representative nontolerizable gene ( Fig. 5B ), shows that LPS challenge induced similar mRNA levels in naïve ABCA1 +/+ and ABCA1 Ϫ / Ϫ macrophages, whereas LPS preconditioning resulted in an increased response to LPS challenge in both groups. However, the response to LPS was signifi cantly diminished in ABCA1 Ϫ / Ϫ macrophages as expressed by both the fold change of mRNA level from the unstimulated control (0.58 ratio of ABCA1 Ϫ / Ϫ to ABCA1 +/+ ) and the percentage Ϫ / Ϫ and ABCA1 +/+ groups are denoted (* P < 0.05, ** P < 0.01, *** P < 0.001). IL, interleukin; LPS, lipopolysaccharide. In summary, the data show that LPS-conditioned ABCA1 Ϫ / Ϫ macrophages were hyporesponsive to LPS in their ability to produce TNF and RANTES protein throughout the fi ve-day recovery period, whereas the cells regained their ability to produce normal levels of IL-6 after two days of recovery. Although differences in responses to LPS challenge between LPS-preconditioned ABCA1 +/+ and ABCA1 Ϫ / Ϫ macrophages showed a great deal of gene-specifi c variation in mRNA and protein levels, the ABCA1 Ϫ / Ϫ cells showed a general pattern of delayed recovery from tolerance and/or decreased responsiveness to LPS challenge compared with ABCA1 +/+ cells.

DISCUSSION
Our data show for the fi rst time that ABCA1 plays signifi cant roles in both the effl ux of bacterial LPS from macrophages and in promoting macrophage recovery from LPS-induced tolerance. The data also show that endogenous apoE can serve as an LPS acceptor during LPS effl ux in resting macrophages, whereas exogenous apolipoproteins (e.g., apoA-I and apoE) serve this function during macrophage activation (i.e., LPS preconditioning), a state in which endogenous apoE production is suppressed and ABCA1 expression is induced.
Although reports of the effects of LPS on macrophage ABCA1 expression are confl icting (14)(15)(16)(17)(18), our experiments consistently showed that LPS increased ABCA1 expression at the mRNA and protein levels in peritoneal macrophages. In keeping with a previous report ( 14 ), our fi ndings that LPS also inhibited the expression of the LXR target genes ABCG1 and apoE under the same conditions suggest that the positive effects of LPS on ABCA1 expression occurred through an LXR-independent pathway. We also found that LPS administration to WT mice increased ABCA1 mRNA levels in the liver (R. L. Kitchens, unpublished observations) as previously reported ( 14 ). Another study showed that TNF also increased ABCA1 expression in peritoneal macrophages ( 13 ), whereas it had the opposite effect in the J774 macrophage cell line ( 16 ). Others found that LPS decreased the expression of ABCA1 in RAW ( 15 ) and J774 ( 16 ) macrophage cell lines. In other macrophage studies, the inhibition of ABCA1 expression by LPS required the activation of interleukin-1 receptor associated kinase (IRAK)-1 ( 18 ), and Tlr4 and Tlr3 agonists inhibited the expression of ABCA1 and other LXR target genes by inhibiting LXR activation by an interferon regulatory factor (IRF)3-dependent mechanism, whereas TNF and Tlr2 signaling had no effects ( 17 ). The reasons for these divergent fi ndings in the literature are unclear. One potential factor may be related to the effects of lipids in the macrophage culture medium. The studies that show increased ABCA1 expression in response to LPS ( 14 ) and TNF ( 13 ) were performed in the presence of serum (0.5-10% FBS), whereas those that show LPS-induced inhibition of ABCA1 expression were performed under serum-free conditions ( 15,16 ), in medium containing lipoprotein-defi cient serum ( 17 ), or under low (1%) serum conditions ( 18 ). Our signifi cantly decreased below that of ABCA1 +/+ cells. Likewise, RANTES production was signifi cantly reduced in LPS-conditioned ABCA1 Ϫ / Ϫ macrophages throughout the fi ve-day recovery period, whereas IL-6 production was reduced only on days 0 and 1. In naïve macrophages (supplementary Fig. IIIA-C), RANTES production was the same in both ABCA1 groups through day 5, and IL-6 production was not signifi cantly different through day 2. However, naïve ABCA1 Ϫ / Ϫ cells produced more IL-6 than ABCA1 +/+ cells on day 5, which explains the lack of priming for IL-6 production in the ABCA1 Ϫ / Ϫ cells ( Fig. 7B ). Unexpectedly, TNF production was signifi cantly decreased in naïve ABCA1 Ϫ / Ϫ cells through day 2 (supplementary Fig. IIIA) in contrast to IL-6 and RANTES, which were not significantly different. Ϫ / Ϫ macrophages were preincubated for 24 h with or without LPS as described in Fig. 5 , washed twice, incubated in fresh complete medium for 0-5 days, and challenged with 10 ng/ml LPS for 6 h. TNF (A), IL-6 (B), and RANTES (C) were measured in the culture supernatants by ELISA as ng/100 µg total cellular protein. The responses of the LPS-preconditioned cells are expressed as percentages of the naïve cell responses. The ABCA1 +/+ data were derived from 15 determinations from a total of 11 mice, and the ABCA1 Ϫ / Ϫ data were from 8 determinations from 8 mice (cells from more than one mouse were mixed in some experiments). Error bars denote mean ± SEM. Asterisks denote signifi cant differences between the groups at each time point (* P < 0.05, ** P < 0.01, *** P < 0.001). IL, interleukin; LPS, lipopolysaccharide; TNF, tumor necrosis factor. tary Fig. II) suggest that ABCA1 also promotes the effl ux of internalized LPS over a period of days. During the sixday time course of these experiments, very little ( ‫ف‬ 5%) of the total LPS could be identifi ed on the cell surface, although we cannot rule out the possibility that some of the protease-resistant LPS may be inserted into the cell membrane or bound to a proteinase K-resistant surface protein.
It is unclear whether ABCA1 transports LPS from intracellular compartments or if ABCA1 removes LPS from the cell surface as it gradually recycles from intracellular compartments; the former mechanism is suggested by the lack of accumulation of cell-surface LPS in ABCA1 Ϫ / Ϫ macrophages under these conditions. As the LPS effl ux assays in this study measure the bulk fl ow of cell-associated LPS, they cannot account for the amounts of LPS that move from specialized microenvironments that may exist in the cell surface or intracellular compartments which may or may not contribute to Tlr4 signaling ( 52 ). Therefore, it is diffi cult to hypothesize how much of the total LPS must be removed from macrophages to reverse the LPS-induced attenuation of Tlr4 signaling. These experiments also showed that only half maximal deacylation of secondary fatty acyl chains had occurred even by day 6 in both cellassociated and effl uxed LPS, suggesting that ABCA1 promotes the effl ux of both fully and partially deacylated LPS.
Our data show that LPS-induced tolerance occurred to a greater degree in macrophages that lacked ABCA1 and that these cells were delayed in their recovery from the tolerant state. We hypothesize that this is due to decreased LPS effl ux in LPS-preconditioned ABCA1 Ϫ / Ϫ macrophages, although we cannot formally rule out other explanations. Differences in tolerance between ABCA1 Ϫ / Ϫ and ABCA1 +/+ macrophages were not due to differences in the initial responses of the naïve macrophages to LPS. We consistently found that initial responses of ABCA1 Ϫ / Ϫ cells to LPS was the same as those of ABCA1 +/+ cells for almost every gene that we tested (supplementary Table III and Fig. IIIA-C). These fi ndings are similar to a previous report ( 5 ) in which LPS response differences were rarely significant between ABCA1 Ϫ / Ϫ and ABCA1 +/+ macrophages.

Others have reported modest but signifi cant increases in ABCA1
Ϫ / Ϫ macrophage responses to LPS ( 4, 6 ), but the basis for these confl icting fi ndings in the literature remains unclear.
Our hypothesis that ABCA1-mediated LPS effl ux is responsible for the rapid recovery from tolerance is also supported by the observation that signifi cant differences between the response to LPS challenge in ABCA1 Ϫ / Ϫ and ABCA1 +/+ macrophages always reveal a lower response in ABCA1 Ϫ / Ϫ cells during the tolerant state in all genes tested. Even nontolerizable genes that were initially primed by preexposure to LPS (e.g., RANTES, IL-12a, and Fpr1) also showed decreased responses to LPS challenge in ABCA1 Ϫ / Ϫ cells. These results are consistent with fi ndings in AOAH Ϫ / Ϫ macrophages in which LPS remains bioactive for weeks; in these experiments, both nontolerizable genes (e.g., RANTES and IL-1RA) and tolerizable genes were hyporesponsive to LPS challenge in LPS-preconditioned experiments were performed in the presence of 10% FBS. However, in two of our preliminary experiments performed in medium containing lipoprotein-defi cient serum, we did not see an increase in macrophage ABCA1 protein levels in response to LPS (data not shown). Considerably more work will be required to establish whether lipid signaling can promote the ability of LPS to induce ABCA1 expression.
Other potential factors may include differences in macrophage subpopulations, which may express ABCA1 either constitutively or inducibly ( 51 ). Prior exposure of macrophages to LPS reduced the activity of one or more ABCA1-independent LPS effl ux mechanisms ( Fig. 3 ), but their identities remain unclear.
Although we cannot rule out the possibility that ABCG1 plays a role in LPS effl ux, our measurements of ABCG1 mRNA levels under certain conditions do not point to a role for this protein ( Tables 1 and 2 ). We also considered the possibility that rapid LPS release from human macrophages could be due to CD14 shedding from the membrane. However, fl ow cytometry analysis of the macrophages showed no decrease in mCD14 density after LPS effl ux, and the effl uxed LPS could not be immunoprecipitated by anti-CD14 antibodies that were immobilized on agarose beads (data not shown), suggesting that CD14 shedding did not play a signifi cant role. We also found that rapid LPS effl ux occurred regardless of the physical state of the LPS that was bound to mCD14 (i.e., LPS monomers versus aggregates ( 47 )), suggesting that LPS effl ux was not due to LPS release from LPS aggregates on the cell surface (data not shown).
The degree of ABCA1-dependency of LPS effl ux appears to vary with the activation state of the macrophage. In unstimulated macrophages, the rates of LPS effl ux in cells that lacked ABCA1 or apoE were similar to those of the WT controls, although the ABCA1 Ϫ / Ϫ cells retained more LPS on their surfaces than ABCA1 +/+ cells ( Fig. 2D, C ; Fig. 3B ; and supplementary Table II). However, the rates of LPS effl ux increased in the WT controls upon stimulation with agents that increased ABCA1. The unstimulated baseline effl ux rates may be similar due to low ABCA1 expression in the WT macrophages and possibly to the induction of a compensatory phenotype in ABCA1 Ϫ / Ϫ or apoE Ϫ / Ϫ cells in which redundant effl ux mechanisms may be increased. However, after LPS preconditioning, LPS effl ux became more dependent upon both ABCA1 and exogenous LPS acceptors (e.g., apoA-I and apoE) ( Fig. 3 ). This was due, in part, to the ability of LPS to suppress endogenous apoE expression and to induce ABCA1, and it may also be due to the suppression of one or more redundant LPS effl ux mechanisms. Under these conditions, exogenous apoE had no ability to increase LPS effl ux in macrophages that lacked ABCA1, and these cells retained LPS on their surfaces even in the presence of apoE ( Fig. 3B ). These data suggest that ABCA1 can effi ciently remove cell-surface LPS from LPS-preconditioned cells over relatively short periods of time (e.g., 1 h). However, the results of LPS preconditioning experiments in which the cells were loaded with radiolabeled LPS for 24 h (supplemen-

AOAH
Ϫ / Ϫ macrophages ( 28 ). Although LPS-induced chromatin modifi cations in nontolerizable genes make them permissive for responsiveness to LPS, prolonged exposure to LPS may decrease their responsiveness to LPS challenge by negative feedback mechanisms that attenuate but do not ablate Tlr4 signaling ( 24 ). Taken together, our data are consistent with the hypothesis that prolonged interaction of LPS with Tlr4 in ABCA1 Ϫ / Ϫ macrophages attenuates Tlr4 signal responses during the state of LPSinduced tolerance and gene reprogramming. Our fi ndings raise the possibility that macrophage expression of ABCA1 may play a physiologically signifi cant role in live animals by removing LPS from macrophages and helping to restore normal innate immune responsiveness. By accelerating the recovery of macrophages from the tolerant state, ABCA1 may help to limit the period of time that animals are immunosuppressed and more vulnerable to further attack by pathogens after an initial encounter with Gram-negative (LPS-bearing) bacteria. Although increased infection rates have not been described in patients with Tangier disease, the rarity of this genetic disorder would make this very diffi cult to measure. ABCA1 may also speed the recovery of macrophages from tolerance during noninfectious conditions that involve either chronic exposure to endotoxin or gut translocation of endotoxin; these conditions may include smoking ( 53 ), trauma and surgery ( 25 ), hepatic cirrhosis ( 54 ), HIV infection ( 55 ), and asthma ( 56 ). As LPS is environmentally ubiquitous and gut translocation of LPS occurs normally at low levels, ABCA1 may also be instrumental in maintaining normal macrophage responsiveness in healthy individuals by keeping macrophages clear of bioactive LPS. Confi rmation of the in vivo signifi cance of this novel role of ABCA1 awaits carefully controlled studies in experimental animals.