Adiponectin forms a complex with atherogenic LDL and inhibits its downstream effects

Adiponectin, an adipocyte-derived protein, has antiatherogenic and antidiabetic effects, but how it confers the atherogenic effects is not well known. To study the antiatherogenic mechanisms of adiponectin, we examined whether it interacts with atherogenic low density lipoprotein (LDL) to attenuate LDL's atherogenicity. L5, the most electronegative subfraction of LDL, induces atherogenic responses similarly to copper-oxidized LDL (oxLDL). Unlike the native LDL endocytosed via the LDL receptor, L5 and oxLDL are internalized by cells via the lectin-like oxidized LDL receptor-1 (LOX-1). Using enzyme-linked immunosorbent assays (ELISAs), we showed that adiponectin preferentially bound oxLDL but not native LDL. In Chinese hamster ovary (CHO) cells transfected with the LOX-1 or LDL receptor, adiponectin selectively inhibited the uptake of oxLDL but not of native LDL, respectively. Furthermore, adiponectin suppressed the internalization of oxLDL in human coronary artery endothelial cells (HCAECs) and THP-1-derived macrophages. Western blot analysis of human plasma showed that adiponectin was abundant in L5 but not in L1, the least electronegative subfraction of LDL. Sandwich ELISAs with anti-adiponectin and anti-apolipoprotein B antibodies confirmed the binding of adiponectin to L5 and oxLDL. In LOX-1-expressing CHO cells, adiponectin inhibited cellular responses to oxLDL and L5, including nuclear factor-κB activation and extracellular signal-regulated kinas phosphorylation. In HCAECs, adiponectin inhibited oxLDL-induced endothelin-1 secretion and extracellular signal-regulated kinase phosphorylation. Conversely, oxLDL suppressed the adiponectin-induced activation of adenosine monophosphate-activated protein kinase in COS-7 cells expressing adiponectin receptor AdipoR1. Our findings suggest that adiponectin binds and inactivates atherogenic LDL, providing novel insight into the antiatherogenic mechanisms of adiponectin.

Abstract Adiponectin, an adipocyte-derived protein, has antiatherogenic and antidiabetic effects, but how it confers the atherogenic effects is not well known. To study the antiatherogenic mechanisms of adiponectin, we examined whether it interacts with atherogenic low density lipoprotein (LDL) to attenuate LDL's atherogenicity. L5, the most electronegative subfraction of LDL, induces atherogenic responses similarly to copper-oxidized LDL (oxLDL). Unlike the native LDL endocytosed via the LDL receptor, L5 and oxLDL are internalized by cells via the lectin-like oxidized LDL receptor-1 (LOX-1). Using enzyme-linked immunosorbent assays (ELISAs), we showed that adiponectin preferentially bound oxLDL but not native LDL. In Chinese hamster ovary (CHO) cells transfected with the LOX-1 or LDL receptor, adiponectin selectively inhibited the uptake of oxLDL but not of native LDL, respectively. Furthermore, adiponectin suppressed the internalization of oxLDL in human coronary artery endothelial cells (HCAECs) and THP-1-derived macrophages. Western blot analysis of human plasma showed that adiponectin was abundant in L5 but not in L1, the least electronegative subfraction of LDL. Sandwich ELI-SAs with anti-adiponectin and anti-apolipoprotein B antibodies confirmed the binding of adiponectin to L5 and oxLDL. In LOX-1-expressing CHO cells, adiponectin inhibited cellular responses to oxLDL and L5, including nuclear factor-κB activation and extracellular signal-regulated kinas phosphorylation. In HCAECs, adiponectin inhibited oxLDL-induced endothelin-1 secretion and extracellular signal-regulated kinase phosphorylation. Conversely, oxLDL suppressed the adiponectin-induced activation of adenosine monophosphate-activated protein kinase in COS-7 cells expressing adiponectin receptor AdipoR1.
Our findings suggest that adiponectin binds and inactivates atherogenic LDL, providing novel insight into the antiatherogenic mechanisms of adiponectin.
Supplementary key words cell biology • oxidized LDL • atherosclerosis • signal transduction • lipoprotein/receptors • adipocyte • ELISA • low density lipoprotein Adiponectin is a well-known adipokine with antidiabetic and antiatherogenic properties. The circulating concentration of adiponectin is characteristically high-much higher than that of other cytokines and hormones (1). Hypoadiponectinemia, characterized by an adiponectin plasma concentration less than 4 μg/ml, has been shown to be associated with abnormal glucose and lipid metabolism and an increased risk of coronary artery disease in men without a history of cardiovascular disease (2). Furthermore, individuals with inherited hypoadiponectinemia, caused by R112C and I164T amino acid substitutions in adiponectin, are known to have an increased risk for metabolic syndrome and coronary artery disease (3,4). Despite the well-established antiatherogenic and antidiabetic properties of adiponectin, the molecular basis for these observations and the significance of adiponectin's high plasma concentration have remained largely unknown.
The antidiabetic and antiatherogenic actions of adiponectin have been extensively studied in mice and cells. In mice, overexpressing adiponectin by introducing a transgene or adenovirus vector carrying the adiponectin gene suppresses atherosclerosis in apolipoprotein E (ApoE)-deficient mice (5). Furthermore, *For correspondence: Tatsuya Sawamura, t-sawamura@umin.ac.jp. deletion of the adiponectin gene in mice leads to insulin resistance and neointimal formation in a cuff injury model (6). Double knockout mice that are deficient for both AdipoR1 and AdipoR2, which are adiponectin receptors through which adenosine monophosphate (AMP)-activated protein kinase (AMPK) and peroxisome proliferator-activated receptor-α (PPARα) are activated (7), show symptoms of insulin resistance and abnormal glucose tolerance. Moreover, the deficiency of AdipoR2 has been reported to suppress atherogenesis in ApoE knockout mice (8). Adiponectin has also been shown to reduce the expression of proatherogenic molecules (9).
With a molecular structure that resembles that of complement C1q, adiponectin forms various multimeric structures, namely a globular form, low-molecular-weight (LMW) trimers, middle-molecular-weight (MMW) hexamers, and high-molecular-weight (HMW) multimerized forms. Another receptor for adiponectin is T-cadherin, a glycosyl-phosphatidylinositol-anchored protein that selectively binds to the hexameric and HMW forms of adiponectin (10). Recently, an association between T-cadherin and adiponectin has been suggested to work against neointimal proliferation and atherosclerosis (11).
We have previously shown that the endogenous protein developmental endothelial locus-1 (Del-1) directly blocks the interaction between atherogenic oxidized low density lipoprotein (oxLDL) and oxLDL receptors and that the overexpression of Del-1 significantly suppresses atherogenesis in mice (12). Therefore, this finding prompted us to examine whether adiponectin works similarly to Del-1 as a direct inhibitor of the interaction between atherogenic LDL and its receptors. A number of "scavenger receptors," including scavenger receptor-A (SR-A), CD36, and lectin-like oxidized LDL receptor-1 (LOX-1), have been cloned and shown to have the ability to mediate the proatherogenic effects of oxLDL (13,14). In addition, increasing attention has been focused on the most negatively charged (but not necessarily oxidatively modified) subfraction of LDL isolated from circulating LDL, called L5, which exerts its atherogenic effects via LOX-1 similar to oxLDL (15). In this study, we present novel evidence showing that adiponectin interacts with atherogenic oxLDL and L5 LDL and attenuates their atherogenicity, providing new insight into the antiatherogenic mechanism of adiponectin.

L5 purification
The collection of human plasma for LDL preparation was approved by the Institutional Review Board of Kaohsiung Medical University Hospital in Taiwan. Plasma LDL was isolated from pooled plasma from 9-10 patients with metabolic syndrome as previously described by using sequential potassium bromide density-gradient ultracentrifugation (density=1.019-1.063 g/ml) and was supplemented with protease inhibitor cocktail (1% penicillin/streptomycin/ neomycin mixture) and 0.5 mM EDTA (16). Isolated LDL was separated into five subfractions (L1-L5) by using an anionexchange fast-protein liquid chromatography system (GE Healthcare, Princeton, NJ) (16).

Detection of OxLDL uptake by cells
Transfected COS-7 cells, differentiated THP-1 cells, and HCAECs were incubated with DiI-oxLDL or DiI-LDL (3 μg/ ml) in the presence or absence of the indicated concentration of human recombinant full-length adiponectin or BSA in serum-free DMEM for 2 h at 37 • C. To differentiate THP-1 cells, we added 100 nM phorbol 12-myristate 13-acetate to cells for 72 h as described previously (21). For the control uptake assay, adiponectin and DiI-oxLDL or DiI-LDL were added separately. Cells were initially incubated with the indicated concentration of human recombinant full-length adiponectin or BSA in serum-free medium for 2 h at 37ºC. Then, cells were washed twice with PBS and incubated with 3 μg/ml DiI-oxLDL in serum-free medium for another 2 h.

Luciferase reporter assay
LOX-1-AT1-CHO cells were seeded in a 96-well plate and were cotransfected with NF-κB promoter firefly luciferase reporter vector, pGF1-NF-κB (System Biosciences, Palo Alto, CA) and pRL-CMV Renilla luciferase control reporter vector (Promega Corp., Madison, WI) by using Lipofectamine LTX with PLUS reagent kit (cat. no. 15338100, Thermo Fisher Scientific). Briefly, in values converted to the amount per well, 100 ng of pGF1-NF-κB and 10 ng of pRL-CMV vectors were dissolved in 0.2 μl of 10 mM Tris-HCl (pH 8.0)/1 mM EDTA (TE) and added to 25 μl of Opti-MEM medium. Then, 0.1 μl of Plus reagent and 0.2 μl of Lipofectamine LTX were added to the mixture, followed by incubation for 30 min at room temperature to form DNA-Lipofectamine LTX complexes. The mixture was added to growing CHO cells cultured in 100 μl of Ham's F-12/10% FBS in 96-well plates (cat. no. 353072, Corning). After being cultured for 6 h, the cells were starved by culture in 0.1% FBS/Ham's F-12 supplemented with 300 ng/ml doxycycline to induce LOX-1 expression for 24 h. Then, oxLDL or L5 was added to cells and incubated for 20 h in the presence or absence of the indicated concentration of human recombinant full-length adiponectin or BSA. Cells were washed once with PBS and lysed by incubation with 150 μl of Passive Lysis Buffer from the Dual-Luciferase Reporter Assay Kit (cat. no. E1910, Promega) for 15 min at room temperature with mixing. Lysates (10 μl) were loaded onto a 96well white plate (cat. no. 6002290, PerkinElmer), and firefly and Renilla luciferase activities were determined. Luminescence was measured by using a Powerscan four microplate reader (DS Pharma Biomedical, Suita, Japan), with 3 s of mixing time and 10 s of read time for each luciferase. Firefly luciferase activity was normalized with that of Renilla luciferase, and the results were expressed as the ratio of the normalized luminescence intensity of the sample to that of the control (i.e., incubated without oxLDL).

Analysis of ERK and AMP-kinase phosphorylation
Cells were serum-starved for 24 h and treated with oxLDL and human recombinant full-length adiponectin for 5 min (to detect the ERK phosphorylation) or for 15 min (to detect the AMP-kinase phosphorylation). Then, cells were washed once with PBS and lysed by sonication with a Bioruptor (Cosmo Bio, Tokyo, Japan) in Laemmli's sample buffer (Bio-Rad) with a final concentration of 100 mM dithiothreitol. Lysates were heated at 95 • C for 5 min. Proteins were separated by using SDS-PAGE and were transferred to PVDF membranes. The membranes were blocked with Immunoblock solution (KAC) and incubated with primary antibodies. The antigens were visualized by using an Immobilon Western Chemiluminescent HRP Substrate Kit (Millipore). Densitometric analysis was performed with a chemiluminescence detection system (ImageQuant LAS 4000mini; GE Healthcare).

Measurement of endothelin-1
Endothelin-1 (ET-1) was measured as described previously (12). Briefly, HCAECs were grown in 96-well dishes (cat. no. 353072, Corning) and cultured at 37 • C. Cells were serumstarved in EGM-2 for 24 h and were incubated with oxLDL (10 μg/ml) and the indicated concentration of human recombinant full-length adiponectin in 50 μl of medium for 4 h at 37 • C. Conditioned medium was collected and centrifuged for 10 min at 1,500 rpm (362 g) by using a Centrifuge 4-15C (Qiagen, Germany), and the supernatant was collected. The concentration of ET-1 in HCAEC-conditioned medium was measured by using an ET-1 Assay Kit (cat. no. 17165, Immuno-Biological Laboratories, Gunma, Japan).

Quantitative reverse transcriptase-polymerase chain reaction
Cells were serum-starved in EGM-2 for 24 h and treated with oxLDL (10 μg/ml) and the indicated concentration of human recombinant full-length adiponectin for 16 h. RNA isolation and cDNA synthesis were performed by using a TaqMan Gene Expression Cells-to-CT Kit (cat. no. AM1728, Thermo Fisher Scientific). Briefly, cells were washed once with PBS and incubated with 50 μl of lysis solution supplemented with DNase I, both from the kit, for 5 min at room temperature. Then, 5 μl of stop solution from the kit was added to the lysate, and the solution was incubated for 2 min at room temperature. For the reverse transcriptase (RT) reaction, 10 μl of the solution was mixed with 40 μl of RT master mix (25 μl of RT buffer, 2.5 μl of enzyme mix, and 12.5 μl of nuclease-free water) and incubated for 1 h at 37 • C. The reaction mixture was heated for 5 min at 95 • C to inactivate the RT. The resultant cDNA solution was then subjected to gene expression analyses. Gene expression was analyzed on an Applied Biosystems 7900HT Fast Real-Time PCR System (Applied Biosystems, Foster City, CA) by using TaqMan Gene Expression Master Mix (cat. no. 4369016, Thermo Fisher Scientific) and TaqMan Gene Expression Assays (Thermo Fisher Scientific) for monocyte chemoattractant protein-1 (MCP-1; Hs00234140m1), intercellular adhesion molecule-1 (ICAM-1; Hs99999152m1), and 18S (4319413E).

Statistical analysis
All data are presented as the mean ± standard error of the mean. A Student's t-test was used for the comparison of two data sets. One-way analysis of variance and the Tukey post hoc test were used for the comparison of multiple data sets. Statistical calculations were performed in GraphPad Prism, version 7.0, for Windows (GraphPad Software Inc., San Diego, CA). A probability value (P) < 0.05 was considered statistically significant.

Declaration of Helsinki
This study complied with the Declaration of Helsinki. The locally appointed ethics committee approved the research protocol, and informed consent was obtained from study participants.

Complex formation between adiponectin and atherogenic LDL in vitro and in human plasma
To examine whether oxLDL and adiponectin interact, human recombinant full-length adiponectin and the globular form of adiponectin were added to native LDL and oxLDL immobilized on anti-apoB monoclonal antibody HUC20 on the solid phase of plates (18). The extent of adiponectin binding to LDL preparations was detected by using an anti-adiponectin antibody. Both the full-length and globular forms of adiponectin were readily bound to immobilized oxLDL, whereas the binding of adiponectin to native LDL was significantly lower (Fig. 1A). The apparent Kd values were calculated as 1.41 ± 0.17 μg/ml (equivalent to 15.7 ± 1.9 nM of human full-length adiponectin trimer) for human recombinant full-length adiponectin and 2.31 ± 0.28 μg/ml (equivalent to 45.3 ± 5.5 nM of human globular adiponectin trimer) for the human recombinant globular form of adiponectin, with the assumption that the binding is 1:1 reversible binding. The anti-apoB monoclonal antibody HUC20 equally recognized native LDL and oxLDL, as confirmed by using an anti-apoB polyclonal antibody (Fig. 1B), indicating that the difference in binding between adiponectin and native LDL or oxLDL was due to the preferential recognition of oxLDL by adiponectin.
We then explored the possibility of complex formation between endogenous atherogenic LDL and adiponectin in human plasma. We have previously demonstrated that LDL can be separated according to charge into five subfractions (L1-L5) by using anionexchange chromatography (15). We examined the presence of adiponectin in L5, the most atherogenic and most electronegative subfraction of LDL, and in L1, the least electronegative subfraction of LDL (22). L1 is the most abundant subfraction of LDL, whereas L5 composes only 1%-2% of LDL in healthy individuals (22). Western blot analysis showed that the presence of adiponectin was prominent in L5 but hardly detectable in L1 ( Fig. 2A). Because the adiponectin receptor T-cadherin has been reported to be associated with exosomes, we performed a second ultracentrifugation of L1 and L5 subfractions to remove any exosomes. After ultracentrifugation, the supernatant of L5 retained adiponectin, excluding the possibility of exosome contamination or microparticleassociated adiponectin (Fig. 2B). To confirm that the loading conditions for Western blot analysis were the same for L1 and L5, we analyzed different amounts of L1 and L5 both after electrophoresis and after Western blotting for adiponectin and apoB (Fig. 2C). Coomassie Brilliant Blue staining of electrophoresed proteins showed similar amounts of loading between L1 and L5, although L5 proteins were degraded, which is a property of L5 (Fig. 2C, left). Western blotting for apoB showed similar amounts of apoB in L1 and L5, although apoB was degraded in L5 (Fig. 2C, right). Under these conditions, adiponectin was detectable only in the lanes with L5 (Fig. 2C, middle), even when 5-fold more L1 than L5 was analyzed.
Furthermore, the results of sandwich ELISAs showed that the complex between adiponectin and L5 was simultaneously recognized by both anti-adiponectin and anti-apoB antibodies (Fig. 2D, right). This finding indicates that adiponectin and L5 are not merely cosegregated but that they are in the same molecular complex. A control ELISA with anti-apoB polyclonal antibody showed that the anti-apoB monoclonal antibody HUC20 equally recognized L1 and L5 (Fig. 2D,  left), indicating that the difference in binding between L1 and L5 to adiponectin was due to the preferential recognition of L5 by adiponectin.

Inhibition of the cellular uptake of oxLDL with physiologic concentrations of adiponectin
We next explored how the physical interaction between adiponectin and atherogenic LDL affects the biologic activity of atherogenic LDL. When we added DiI-labeled native LDL to COS-7 cells transfected with LDLR, the presence of adiponectin (0.3-30 μg/ml) did not affect the uptake of native LDL by the transfected cells (Fig. 3A). Because LOX-1 is a receptor for oxLDL (14,23), we examined the effect of adiponectin on the uptake of DiI-labeled oxLDL by COS-7 cells transfected with LOX-1. The addition of adiponectin (0.3-30 μg/ml) dose-dependently suppressed the uptake of oxLDL (Fig. 3B), indicating that adiponectin was selectively The dose-dependent binding of AdN (human recombinant fulllength or globular form) to oxLDL or LDL was detected by using an enzyme-linked immunosorbent assay (ELISA) with antiadiponectin polyclonal antibody. Native LDL (black circles) or oxLDL (red circles) (1 μg/ml) was first captured by using an anti-apoB monoclonal antibody (ie, HUC20), and the indicated concentration of human recombinant full-length (left) or globular form (right) AdN was applied to the well. After incubation, the plate was washed, and bound recombinant human AdN was detected by using anti-AdN polyclonal antibody. B: A control ELISA with anti-apoB polyclonal antibody was performed to show that the solidphase monoclonal antibody HUC20 (used in A) equally recognizes LDL (orange circle) and oxLDL (blue circle). As a negative control, IgY was used to instead of HUC20 (LDL, orange triangle; oxLDL, blue inverted triangle). *P < 0.05, determined by using the Tukey test. LDL, low density lipopritein; oxLDL, copper-oxidized low density lipopritein.  bound to oxLDL but not to native LDL and that the binding of adiponectin to oxLDL interfered with the receptor-mediated uptake of oxLDL in cells.
We next examined whether the inhibitory action of adiponectin on oxLDL could also be observed in endothelial cells and macrophages, which are two crucial cell types in the initiation and progression of atherosclerosis. Similar to what we observed in COS-7 cells, adiponectin suppressed the cellular uptake of DiI-labeled oxLDL in cultured HCAECs and in THP-1derived macrophages (Fig. 3C, D).
Because various other oxLDL receptors may be expressed in these cells, we compared the effect of adiponectin on oxLDL uptake mediated by SR-A and CD36, as well as by LOX-1, in COS-7 cells (Fig. 3E). We found that adiponectin did not suppress oxLDL uptake mediated by CD36 but clearly suppressed that by SR-A, the major oxLDL receptor in macrophages that mediates foam cell formation (13). Notably, the concentration of adiponectin (10-30 μg/ml) that effectively inhibited oxLDL uptake corresponded to the physiologic concentration of plasma adiponectin.
To examine the possibility that free adiponectin in the adiponectin-oxLDL mixture suppresses the cellular uptake of DiI-oxLDL, we performed experiments in which we added adiponectin and oxLDL separately (Fig. 4). After the addition of adiponectin, the cells were washed to remove any remaining adiponectin in the culture media before the addition of oxLDL. As shown in Fig. 4A, in COS-7 cells transfected with LOX-1, SR-A, or SD36, preincubation with adiponectin did not inhibit DiI-oxLDL uptake. These results represent a natural scenario because COS-7 cells do not express adiponectin receptors. Similarly, in THP-1-derived macrophages and HCAECs, preincubation with adiponectin did not inhibit DiI-oxLDL uptake (Fig. 4B, C). Careful analysis in HCAECs showed a slight trend of decreasing oxLDL uptake after preincubation with increasing amounts of adiponectin, but this trend did not reach statistical significance. Therefore, the possibility remains that, via its receptors, adiponectin inhibits oxLDL uptake to a minor extent in HCAECs.

OxLDL-induced signal transduction inhibited by adiponectin
To confirm the inhibitory effect of adiponectin on atherogenic LDL, we analyzed the downstream signal transduction induced by oxLDL and L5 in CHO cells and endothelial cells. We found that oxLDL and L5 induced ERK phosphorylation in LOX-1-expressing CHO cells and that adiponectin dose-dependently inhibited these effects (Fig. 5A, B). Using a luciferase reporter assay, we found that oxLDL and L5 also activated NF-κB in LOX-1-expressing CHO cells and that adiponectin dose-dependently inhibited this reaction (Fig. 5C, D). Furthermore, in endothelial cells, adiponectin suppressed oxLDL-induced ERK phosphorylation (Fig. 6A), the induction of MCP-1 and ICAM-1 expression (Fig. 6B), and ET-1 secretion (Fig. 6C) in a dose-dependent manner.

Adiponectin inhibited by oxLDL
Last, we examined whether oxLDL conversely affects the action of adiponectin. It is known that adiponectin increases AMPK phosphorylation in endothelial cells or COS-7 cells transfected with AdipoR1 (7). When we examined the effect of oxLDL on adiponectininduced AMPK phosphorylation, we found that oxLDL significantly inhibited adiponectin-induced AMPK phosphorylation in dose-dependent manner in both COS-7 cells and HCAECs transfected with Adi-poR1, whereas native LDL had no effect (Fig. 7A, B). The more potent effects of oxLDL in HCAECs than in COS-7 cells may have been due to the combined effects of oxLDL via its receptors and its direct interaction with adiponectin in an aqueous phase.

DISCUSSION
In this study, our findings provide novel insight into the mechanism of adiponectin's antiatherogenic action, which may also explain the high concentration of adiponectin observed in human plasma. Although how adiponectin exerts its antiatherogenic effects has previously been poorly understood, signal transductionmediated effects of adiponectin have been believed to play a role, such as AMP-kinase activation and the subsequent reduction of adhesion molecule expression (24). Here, we show that atherogenic LDL and adiponectin directly interact at physiologic concentration of adiponectin, which leads to the inhibition of atherogenic effects caused by oxLDL and L5, such as decreased ERK and NF-κB activation, MCP-1 and ICAM-1 expression, and ET-1 release (Fig. 6) (11).
In the general context of atherosclerosis research, showing that an endogenous molecule inhibits the action of atherogenic LDL by direct molecular interaction is an important finding. Previously, we showed that Del-1 directly interacts with oxLDL to inhibit its atherogenic action in vitro, and the overexpression of Del-1 in vivo significantly decreased the formation of atheroma AdN is detectable only in the L5 subfraction. The electrophoresis of samples was performed in reduced and heated conditions, so the electrophoretic pattern here is different from that in (A), (B). D: Detection of adiponectin-LDL complexes in L1 (black and gray triangles), L5 (blue, light blue, and green circles), or whole LDL (orange triangles) by using a sandwich enzyme-linked immunosorbent assay (ELISA) with anti-apoB monoclonal antibody (HUC20) and anti-AdN polyclonal antibody. E: A control ELISA with apoB polyclonal antibody was performed to confirm that the solid-phase anti-apoB monoclonal antibody (HUC20) used in (D) equally recognizes L1 (orange circles) and L5 (blue circles). As a negative control, IgY was used instead of anti-apoB monoclonal antibody (L1, orange triangles; L5, blue inverted triangles). HMW, high-molecular-weight adiponectin; LDL, low density lipopritein; LMW, low-molecular-weight adiponectin; MMW, middle-molecular-weight adiponectin. induced by a high-fat diet (Paigen diet) (12). Importantly, although the overexpression of Del-1 did not alter the plasma concentration of oxLDL, it significantly decreased the receptor binding activity of oxLDL. To our knowledge, this was the first report of a protein exerting antiatherogenic effects in such a manner. However, to produce enough Del-1 to show an effect, these experiments were performed in genetically engineered mice; therefore, we could not conclude whether this mechanism operates under physiologic conditions. Adiponectin is recognized as an adipokine but is present in human plasma at a much higher concentration (5-10 μg/ml, 50-100 nM as a homotrimer) (25) than other hormones and cytokines, such as angiotensin II (1 nM), insulin (<1 nM), and IL-6 (<1 pM). A functional difference (P < 0.05) compared with BSA at the same concentration. The DiI-oxLDL or DiI-LDL uptake values were expressed as the percentage of DiI-oxLDL or DiI-LDL fluorescence intensity of cells when incubated without AdN normalized with the number of nuclei. The fluorescence intensity of cells incubated without DiI-oxLDL or DiI-LDL was set to 0%. explanation for this observation has been previously lacking. The phagocytosis of dead cells was shown to occur via adiponectin at concentrations similar to those found in plasma (1,26). In addition, genetic analyses of a healthy middle-aged cohort of individuals showed that hypoadiponectinemia was associated with the risk of atherosclerosis (2)(3)(4). Notably, the binding of adiponectin to its receptors occurs at much lower concentrations than its plasma range. This suggests that the presence of adiponectin at its high physiologic concentration has an antiatherogenic role independent from its receptor-mediated properties. In the present . Cell extracts were subjected to Western blot analysis of phosphorylated ERK (pERK) and total ERK (42 kDa) (upper panels). The bands were quantified to calculate the ratio of pERK to total ERK (n = 3). C, D: Adiponectin suppressed atherogenic LDL-induced NF-κB activation in LOX-1-AT 1 -CHO cells. C, D: Adiponectin inhibited NF-κB activation induced by oxLDL (C) or L5 (D) in LOX-1-AT1-CHO cells. NF-κB activation was analyzed by using a luciferase reporter assay. Data are expressed as the mean ± SEM (n = 3). *P < 0.05, determined by using the Tukey test. LDL, low density lipopritein; oxLDL, copper-oxidized low density lipopritein. study, adiponectin at physiologic concentrations (3-30 μg/ml) interacted with oxLDL and exerted antiatherogenic actions, providing evidence for the first time to our knowledge that adiponectin is a physiologic endogenous antagonist for oxLDL. Notably, we detected adiponectin-L5 complex formation even after the purification of L5. Figure 2 shows a complex between adiponectin and oxLDL already formed in vivo. Because L5 was purified by means of ultracentrifugation, dialysis, and anion-exchange chromatography,   . Cell extracts were subjected to Western blot analyses for phosphorylated AMPK (pAMPK) and total AMPK (62 kDa). Bands were quantified to calculate the ratio of pAMPK to total AMPK. Data are expressed as the mean ± SEM (n = 3). *P < 0.05, determined by using the Tukey test. AMPK, adenosine monophosphate-activated protein kinase; oxLDL, copper-oxidized low density lipopritein.

C B A
most of the adiponectin associated with L5 in plasma may have become dissociated during these processes, especially with the presumably low-affinity binding between adiponectin and L5 that occurs under a high physiologic concentration of adiponectin. However, even under such conditions, we still detected adiponectin in L5. Although we clearly showed that the inhibitory effect of adiponectin on oxLDL stems from oxLDL-adiponectin complex formation, it is not our intention to negate the various effects of adiponectin mediated by its receptors. Of note, Ouchi and Walsh (1) have reported that the 2-day incubation of macrophages with 30 μg/ml adiponectin reduced SR-A expression. Rather, the present study sheds light on a previously overlooked role of adiponectin that potentially occurs in an aqueous phase, such as in blood.
Despite the well-known antiatherogenic properties of adiponectin, complex in vivo phenomena related to adiponectin have been previously reported, referred to as the "adiponectin paradox." For example, too much adiponectin has been shown to have negative effects (27). In addition, in obese individuals, adiponectin concentrations are reduced in the plasma, despite the overabundance of adipose tissue (28). Furthermore, in elderly people or cohorts with prevalent cardiovascular disease, adiponectin levels have been positively associated with all-cause/ cardiovascular disease mortality (29). The reasons for this remain unknown, despite various possible explanations, including the compensatory secretion of adiponectin. "Adiponectin resistance" has also been speculated as a cause, although the basis of this is unknown (27,30).
To help understand these phenomena, the recent advances made toward understanding the role of HDL in atherosclerosis should be considered for the sake of comparison. According to a recent epidemiologic study, an HDL level that is too high is a risk factor for cardiovascular disease, despite that HDL is well recognized as an antiatherogenic lipoprotein. HDL has also been shown to have proatherogenic properties in patients with coronary artery disease or with a high risk of coronary artery disease (31). These findings have suggested that, under such conditions, HDL becomes dysfunctional by losing its antiatherogenic properties (i.e., paraoxonase-1 activity and cholesterol efflux capacity) and gains proatherogenic properties (i.e., binding to the atherogenic receptor LOX-1) (32).
In the present study, the binding of adiponectin to atherogenic LDL also attenuated the function of adiponectin. Thus, adiponectin bound by atherogenic LDL becomes dysfunctional. It is unknown to what extent this binding contributes to the dysfunction of adiponectin under human physiologic conditions, but the cosegregation of adiponectin and atherogenic LDL in human plasma suggests that this mechanism functions in humans. This may constitute the mechanism of adiponectin resistance.
In conclusion, we have demonstrated a novel antiatherogenic mechanism of adiponectin by which it directly inhibits atherogenic LDL (Fig. 8). Thus, this study provides a biologically relevant explanation for the well-known properties of adiponectin. Human studies are warranted for further investigating this new aspect of adiponectin biology.

Data availability
The authors declare that all data supporting the findings of this study are contained within the manuscript.

Note added in proof
Very recently, while our manuscript was under review at the Journal of Lipid Research, Ye et al. reported in PNAS (29:17381-17388; 2020) that adiponectin binds to anionic phospholipids and sphingolipids.
Addition of any more material should be requested at PNAS.