Nascent HDL formation by hepatocytes is reduced by the concerted action of serum amyloid A and endothelial lipase.

Inflammation is associated with significant decreases in plasma HDL-cholesterol (HDL-C) and apoA-I levels. Endothelial lipase (EL) is known to be an important determinant of HDL-C in mice and in humans and is upregulated during inflammation. In this study, we investigated whether serum amyloid A (SAA), an HDL apolipoprotein highly induced during inflammation, alters the ability of EL to metabolize HDL. We determined that EL hydrolyzes SAA-enriched HDL in vitro without liberating lipid-free apoA-I. Coexpression of SAA and EL in mice by adenoviral vector produced a significantly greater reduction in HDL-C and apoA-I than a corresponding level of expression of either SAA or EL alone. The loss of HDL occurred without any evidence of HDL remodeling to smaller particles that would be expected to have more rapid turnover. Studies with primary hepatocytes demonstrated that coexpression of SAA and EL markedly impeded ABCA1-mediated lipidation of apoA-I to form nascent HDL. Our findings suggest that a reduction in nascent HDL formation may be partly responsible for reduced HDL-C during inflammation when both EL and SAA are known to be upregulated.

against 150 mmol/L NaCl, 0.01% EDTA, fi lter sterilized, and stored at 4°C under argon gas. Protein concentrations were determined by the method of Lowry et al. ( 22 ). AdEL was used to express recombinant EL (rEL) in COS-7 cells according to published methods ( 9 ). For in vitro hydrolysis, HDL (0.8 mM fi nal phospholipid concentration, ‫ف‬ 40 g protein) was incubated at 37°C for 24 h in TBS supplemented with the indicated amount of culture supernatant containing human rEL, 10 mg/ml fatty-acid free BSA, and 4 mM Ca 2+ (total volume, 50 l). The extent of HDL hydrolysis was assessed by measuring the amount of FFA released into the reaction mixture by colorimetric assay (Wako).

Plasma HDL-C quantifi cations
Plasma HDL-C was measured using a commercially available kit (Wako) according to the manufacturer's protocol.

Gel electrophoresis and immunoblot analysis
Plasma samples from mice were separated by 4-20% SDS-PAGE and stained with Coomassie Blue or transferred to PVDF membranes (100 min at 100 V, 4°C) and immunoblotted with antimouse apoA-I (Biodesign International) or rabbit anti-human EL (Cayman Chemical). Protein bands were visualized using enhanced chemiluminescence reagents (GE Healthcare) and quantifi ed by densitometry. For other studies, plasma or cell culture supernatants were subjected to nondenaturing gradient gel electrophoresis (GGE). Electrophoresis was carried out in 4-20% polyacrylamide gels for 3.5 h at 200 V, 4°C and the samples were then transferred to PVDF membranes and immunoblotted using anti-mouse apoA-I or anti-human apoA-I (Calbiochem) as indicated in legends to fi gures. For assessment of ABCA1 expression in primary hepatocytes, total cell lysates (10 µg) were separated on a 4-20% polyacrylamide gradient gel, transferred to polyvinylidene fl uoride (PVDF) membranes and immunoblotted with anti-ABCA1 (ABCAM ab7360).

Statistical analysis
Data are presented as mean ± SEM. Statistical analyses to compare differences between plasma HDL levels at selected intervals after viral vector administration were carried out using two-way ANOVA with a Bonferroni posttest. measures of the metabolic syndrome, which is considered a chronic low-grade infl ammatory condition ( 7 ). Furthermore, low-dose endotoxin injection in healthy volunteers produces signifi cantly increased levels of EL in the plasma 12 h after administration ( 15 ). Thus, an interesting question to address is how EL infl uences HDL metabolism during infl ammation, when HDL undergoes substantial changes in lipid and apoliprotein composition ( 16 ). Most notably, HDL becomes enriched in serum amyloid A (SAA), a major acute phase reactant whose secretion from the liver may be increased more than 1,000-fold during infl ammation. The vast majority of SAA in the plasma is found associated with HDL, where it can comprise the major apolipoprotein ( 17 ). Studies by Caiazza et al. ( 18 ) indicate that apolipoproteins on HDL may regulate ELmediated remodeling, such that reconstituted HDLs (rHDL) containing both apoA-I and apoA-II are hydrolyzed more readily by EL compared with particles containing apoA-I only. Whether the presence of SAA on HDL can modulate HDL remodeling by EL has not been investigated. Previous studies from our laboratory showed that in the presence of SAA, perturbation of HDL core and surface lipid by scavenger receptor class B type I (SR-BI) leads to generation of small, lipid-depleted apoA-I that is susceptible to catabolism ( 19 ). In this study, we carried out in vitro and in vivo studies to investigate whether the presence of SAA infl uences EL-mediated metabolism of HDL.

Adenoviral vectors
SAA was expressed alone or in combination with EL or SR-BI in livers of mice using adenoviral vector-mediated gene transfer. The indicated dose of AdSAA, a vector encoding mouse CE/J SAA isotype ( 20 ), AdSR-BI, a vector encoding mouse SR-BI ( 21 ) or AdEL, a vector encoding human endothelial lipase [generously provided by Dr. Daniel Rader (9)] was administered via tail vein injection in 100 l PBS. For estimates of EL expression 72 h after adenoviral vector infusion, mice were injected i.p. with 2.5 U heparin in 100 l PBS 30 min prior to plasma collection. In the case of the AdSAA + AdSR-BI experiments, the dose of AdSR-BI increased hepatic SR-BI expression ‫ف‬ 3-fold, which resulted in a ‫ف‬ 50% decline in HDL levels, whereas hepatic SAA was increased ‫ف‬ 8-fold without impacting HDL levels when expressed by itself ( 19 ).

HDL isolation and hydrolysis by EL in vitro
HDL (d = 1.063 to 1.21 g/ml) was isolated by density gradient ultracentrifugation from plasma of untreated C57BL/6 mice or mice 24 h after injection with LPS ( 17 ). HDL was then dialyzed

Effect of SAA on EL-mediated HDL metabolism in vivo
We next investigated whether SAA alters EL-mediated metabolism of HDL in vivo. As reported by Maugeais et al. ( 4 ), overexpression of EL in mice by adenoviral vector reduces plasma HDL-C in a dose-dependent manner ( Fig.  3A ). For our studies, we selected a dose of AdEL (0.5 × 10 10 particles) that resulted in a reproducible 30-40% reduction in HDL-C 3 days after viral vector injection. We also selected a dose of AdSAA (2.5 × 10 10 particles) that only modestly enriched HDL particles with SAA ( Fig. 2A ). C57BL/6 mice were injected with mixtures of AdEL, Ad-SAA, and AdNull to provide a total viral dose of 3 × 10 10 particles/mouse. Infusion of AdSAA along with AdEL resulted in a modest but not statistically signifi cant reduction in the amount of EL in postheparin plasma compared with AdEL alone ( Fig. 3B ). Plasma HDL-C concentrations were determined at selected intervals up to 72 h after injection, when adenoviral vector gene expression was maximal. Whereas AdEL resulted in the expected ‫ف‬ 35% reduction in HDL-C 72 h after injection , infusion of Ad-Null or AdSAA had no effect on HDL-C levels at any time point during the course of the experiment ( Fig. 3C ). In contrast, HDL-C was reduced ‫ف‬ 90% when AdSAA and AdEL were administered together, despite comparable levels of EL in plasma. This result suggests a synergistic effect of SAA and EL on HDL metabolism.

Effect of SAA on EL-mediated HDL remodeling in vivo
To investigate whether the increased EL-mediated HDL metabolism in SAA-expressing mice was accompanied by enhanced HDL remodeling, plasma from AdEL and AdEL + AdSAA-treated mice were fractionated by nondenaturing GGE and immunoblotted for apoA-I. The reduction in HDL-C levels that occurred with EL and SAA coexpression was accompanied by a marked loss of apoA-I protein at 48 and 72 h ( Fig. 4A ). Indeed, quantifi cation of apoA-I protein by densitometry revealed that plasma concentrations of apoA-I 72 h after AdEL + AdSAA infusion were 84% lower compared with plasma of mice expressing EL alone (not shown). Interestingly, there was no evidence that surface perturbation of HDL by EL in the presence of SAA resulted in the formation of small, lipid-depleted particles ( Fig. 4A ). These fi ndings contrast to what we have previously shown to occur with HDL core remodeling by SR-BI ( 19 ) and repeat here for illustrative purposes ( Fig. 4B ). In this case, coexpression of SAA and SR-BI by adenoviral vector leads to enhanced reduction in HDL-C compared with expression of SR-BI or SAA alone ( 19 ) (data not shown), and this effect is associated with enhanced generation of small, lipid depleted apoA-I that is susceptible to catabolism ( Fig. 4B ).

Impact of EL and SAA on HDL formation
The marked reduction in apoA-I in AdEL + AdSAAtreated mice prompted us to consider the possibility that HDL biogenesis may be altered when EL and SAA were coexpressed. The initiating step in HDL assembly involves lipidation of apoA-I to form nascent HDL particles. Studies in gene-targeted mice indicate that the major site of

EL remodeling of normal and SAA-containing mouse HDL in vitro
Previous studies demonstrated that EL hydrolyzes rHDL or HDL isolated from human plasma without generating lipid-poor apoA-I ( 10 ). In this study, we assessed whether EL remodels mouse HDL in a similar manner. Incubations with EL resulted in dose-dependent hydrolysis of normal mouse HDL as indicated by an increase in the release of FFAs ( Fig. 1A ). At maximal hydrolysis, ‫ف‬ 53% of the HDL phospholipids was hydrolyzed. Analysis by nondenaturing GGE indicated that lipid-poor apoA-I (which migrates faster than the 7.1 nm standard) is not released from mouse HDL, even when >50% of phospholipids are hydrolyzed ( Fig. 1B ).
Studies by Caiazza et al. ( 18 ) demonstrated that rHDLs containing both apoA-I and apoA-II are better substrates for EL hydrolysis compared with rHDLs containing apoA-I only, suggesting that apolipoproteins modulate EL-mediated HDL remodeling. Thus, it was of interest to determine whether the presence of SAA on HDL alters the ability of EL to remodel the particle to liberate apoA-I. To prepare HDLs for these studies, mice were injected with either LPS to induce an acute phase response, or an adenoviral vector (1.5 × 10 11 particles) that expresses high levels of SAA in the absence of infl ammation (AdSAA). Analysis by SDS-PAGE indicated that HDL isolated from LPS or AdSAA-injected mice were highly enriched in SAA compared with mice infused with a control adenoviral vector ( Fig. 2A ). The HDL preparations were incubated with EL and then separated by nondenaturing GGE. Similar to what occurs with human and normal mouse HDL, EL hydrolysis of SAA-containing HDL does not result in the dissociation of lipid-poor apoA-I ( Fig. 2B ). metabolism by reducing the extent to which apoA-I is lipidated in the liver, an effect that is amplifi ed when hepatic cells secrete SAA. The combined effect of SAA and EL to impede apoA-I lipidation may partly explain the reduction in HDL-C that is known to occur during infl ammation, when both EL and SAA are induced.
The effect of EL on HDL surface remodeling has been extensively investigated by Rye and colleagues ( 10,18 ). Kinetic studies in vitro indicated that the V max of phospholipid hydrolysis is signifi cantly greater for rHDLs containing both apoA-I and apoA-II compared with rHDL containing only apoA-I ( 18 ), whereas rHDL containing only apoA-II undergoes minimal EL hydrolysis ( 10,18 ). Given that rHDL preparations used in these studies were of similar size and lipid composition, these fi ndings strongly suggested that apolipoproteins on HDL are a major determinant of EL hydrolysis. On the other hand, Broedl et al. ( 29 ) reported that adenovirus expression of EL was less effective in altering HDL metabolism in apoA-I/apoA-II double transgenic mice compared with apoA-I single transgenic mice, indicating that apoA-II reduces the ability of EL to alter HDL metabolism in vivo. In the current study, we demonstrated that enrichment with SAA has no discernible effect on the ability of EL to remodel mouse HDL in vitro. We also provide evidence that SAA amplifi es the effect of EL on HDL metabolism in vivo, but through a mechanism that appears to be independent of HDL remodeling in the circulation.
Based on in vitro studies using rHDLs, EL-mediated phospholipid hydrolysis leads to the formation of smaller HDL particles without dissociating either lipid-poor apoA-I or apoA-II ( 10 ). Thus, the impact of EL on HDL remodeling appears to be analogous to what occurs with phospholipase A 2 , where surface hydrolysis leads to the formation of small HDL particles without releasing apoA-I ( 30 ). In contrast, the triglyceride lipase activity of hepatic lipase remodels HDL to form smaller particles and generate lipid-poor apoA-I ( 11 ). In the current study, we found no evidence that enrichment with SAA promotes the dissociation of lipid-poor apolipoproteins from EL-hydrolyzed HDL either in vitro or in vivo. Thus, it does not appear that SAA impacts EL-mediated intravascular HDL remodeling. Nevertheless, SAA signifi cantly enhanced the ability of EL to reduce HDL-C and apoA-I levels in vivo.
lipidation is the liver, where ABCA1 mediates the effl ux of cholesterol and phospholipid from hepatocytes to lipidfree apoA-I ( 24 ). To examine the effect of concomitant EL and SAA expression on HDL biogenesis, primary hepatocytes were isolated from C57BL/6 mice 24 h after adenoviral vector infusions, and the impact of EL and/or SAA expression on the ability of hepatocytes to convert lipidpoor apoA-I to nascent HDL was assessed ( Fig. 5 ). As shown previously, the interaction of lipid-free apoA-I with hepatocytes leads to the formation of multiple discretely sized particles ranging from ‫ف‬ 7.4 to 20 nm in size ( 25 ). These nascent HDLs are known to differ in apoA-I, phospholipid, and cholesterol content ( 26,27 ). Interestingly, compared with hepatocytes from AdNull-treated mice, nascent HDL formation by hepatocytes expressing EL alone was slightly reduced, indicating that EL may alter HDL biogenesis as well as the metabolism of mature HDL. Expression of SAA had no apparent effect on apoA-I lipidation. However, concomitant expression of EL and SAA resulted in a marked reduction in the conversion of lipidpoor apoA-I to nascent HDL. This difference in apoA-I lipidation was not due to differences in ABCA1 expression ( Fig. 5B ). Taken together, our data indicate that EL and SAA act synergistically to interfere with HDL formation.

DISCUSSION
HDL comprises a polydisperse population of lipoproteins that includes larger spherical particles as well as smaller lipid-poor discoidal HDL. The synthesis of new HDL particles occurs primarily in the liver, where apoA-I is secreted in a lipid-poor/lipid-free form. The addition of phospholipids and cholesterol to apoA-I occurs extracellularly through the action of the lipid transporter ABCA1. The resulting disc-shaped particles, conventionally designated pre ␤ -HDL, are then converted to mature spherical HDL. In plasma, these particles are in a dynamic equilibrium that involves active and continuous remodeling mediated by a number of cellular receptors, lipid transport proteins, and modifying enzymes. It is well documented that EL plays a major role in HDL metabolism through its capacity to remodel circulating HDL particles and hence promote their catabolism ( 3,4,28 ). In this study, we show for the fi rst time that EL may also modulate HDL Fig. 2. EL remodeling of acute phase mouse HDL in vitro does not liberate lipid-poor apoA-I. HDL was isolated from plasma of mice 24 h after injection of LPS (AP) or 72 h after infusion of a control adenovirus (Adnull; 1.5 × 10 11 particles) or a low dose (2.5 × 10 10 particles) or high dose (1.5 × 10 11 particles) of AdSAA. A: Aliquots corresponding to 3 g HDL protein were separated by SDS-PAGE and stained with Coomassie Blue. B: AP-HDL and HDL from mice infused with Adnull or high dose AdSAA were incubated for 24 h with or without 40 l EL as indicated and then separated by nondenaturing GGE and immunoblotted using anti-mouse apoA-I. of lipid-poor apoA-I to spherical HDL. Our fi ndings suggest that the entry of nascent HDL into the mature HDL pool may be further impeded during infl ammation, when both EL and SAA are present.
The mechanism by which SAA and EL interact to disrupt nascent HDL formation is unclear. ApoA-I lipidation by ABCA1 is known to occur through a multi-step process that is initiated by the high affi nity binding of apoA-I to ABCA1 ( 31 ). The consequent activation and stabilization of ABCA1 leads to the formation of high capacity binding sites on the plasma membrane where lipid-poor apoA-I can effectively solubilize membrane phospholipids and cholesterol to form nascent HDLs ( 32,33 ). Although both apoA-I binding sites are dependent on ABCA1, the high capacity binding site does not involve a direct interaction between apoA-I and ABCA1. Thus, factors that interfere with ABCA1 activation or the formation of the high capacity binding site would be expected to disrupt apoA-I-mediated cholesterol and phospholipid effl ux and HDL biogenesis. There is evidence to suggest that both SAA and EL modulate ABCA1-dependent Our data indicate this synergism may be attributed to the combined effect of SAA and EL to interfere with the conversion of lipid-poor apoA-I to nascent HDL at the surface of the liver, thereby reducing the production rate of mature HDL. This conclusion is consistent with the fi ndings of Maugeais et al., ( 4 ) who measured the plasma clearance rate of HDL in mice with adenoviral vector-mediated EL overexpression. Although AdEL signifi cantly increased the fractional catabolic rate of mature HDL, the increase in the rate of clearance clearly could not account for the substantial decrease in plasma apoA-I levels that was observed in AdEL-infused mice. This points to an alteration in the rate of HDL production as an important factor contributing to the effect of EL in lowering HDL levels. The authors speculated that EL hydrolyzes nascent HDL phospholipids and consequently reduces the maturation  effl ux, although both factors are in general thought to promote the removal of cellular cholesterol by ABCA1 (34)(35)(36). However, earlier studies did not investigate whether enhanced cholesterol effl ux was accompanied by increased apoA-I lipidation, a key event in HDL biogenesis.
In summary, we report the novel fi nding that coexpression of SAA and EL by adenoviral vector-mediated gene transfer reduces HDL-C in mice to an extent that is significantly more pronounced than when EL is expressed alone. Notably, the cooperative effect of SAA and EL was not accompanied by the appearance of small, lipid-depleted HDLs in the plasma, as would be expected to occur if HDL was undergoing enhanced intravascular remodeling. Our data indicate that apoA-I lipidation by mouse hepatocytes is markedly reduced when EL and SAA are coexpressed. Taken together, our fi ndings suggest that a reduction in nascent HDL formation may be partly responsible for reduced HDL-C during infl ammation, when both EL and SAA are known to be upregulated. Current strategies for raising HDL-C to lower cardiovascular disease risk include increasing HDL production ( 37,38 ). Our fi nding that infl ammatory factors may serve to negatively impact nascent HDL formation may be important to consider in order to maximize the success of such strategies.
The authors thank Vicky Noffsinger for excellent technical assistance.  (1 ×  10 10 particles), AdSAA (5 × 10 10 particles), or AdEL + AdSAA as indicated. Primary hepatocyte cultures were prepared 24 h after infusion, treated with the LXR agonist T0901317 for 8 h to upregulate ABCA1, and then incubated for 18 h with 15 g/ml lipid-free apoA-I. A: Culture supernatants were separated by nondenaturing GGE and immunoblotted using anti-human apoA-I. Lipid-free apoA-I was included on the gel for comparison. B: Lysates of cells prepared before or after T0901317 treatment (10 µg cell protein) were separated on a 4-20% polyacrylamide gradient gel and immunoblotted using anti-ABCA1.