Atheroprotective immunization with malondialdehyde-modified LDL is hapten specific and dependent on advanced MDA adducts: implications for development of an atheroprotective vaccine.

Immunization with homologous malondialdehyde (MDA)-modified LDL (MDA-LDL) leads to atheroprotection in experimental models supporting the concept that a vaccine to oxidation-specific epitopes (OSEs) of oxidized LDL could limit atherogenesis. However, modification of human LDL with OSE to use as an immunogen would be impractical for generalized use. Furthermore, when MDA is used to modify LDL, a wide variety of related MDA adducts are formed, both simple and more complex. To define the relevant epitopes that would reproduce the atheroprotective effects of immunization with MDA-LDL, we sought to determine the responsible immunodominant and atheroprotective adducts. We now demonstrate that fluorescent adducts of MDA involving the condensation of two or more MDA molecules with lysine to form malondialdehyde-acetaldehyde (MAA)-type adducts generate immunodominant epitopes that lead to atheroprotective responses. We further demonstrate that a T helper (Th) 2-biased hapten-specific humoral and cellular response is sufficient, and thus, MAA-modified homologous albumin is an equally effective immunogen. We further show that such Th2-biased humoral responses per se are not atheroprotective if they do not target relevant antigens. These data demonstrate the feasibility of development of a small-molecule immunogen that could stimulate MAA-specific immune responses, which could be used to develop a vaccine approach to retard or prevent atherogenesis.


Protocol for induction of atherosclerosis and immunization interventions
Six-month-old male Ldlr Ϫ / Ϫ mice (age 28.4 ± 2.6 weeks, n = 119) were bled for baseline evaluations and matched for age, litter, body weight, and plasma cholesterol and triglycerides and divided into seven immunization groups (n = 15-20/group) as described subsequently. Two weeks later, they were given a primary immunization, at which time a moderate high-cholesterol (HC) diet (regular chow supplemented with 0.5% cholesterol-TD97234; Harlan-Teklad) was initiated and continued for 28 weeks. Mice were given a primary immunization, which consisted of 75 g immunogen (protein) in 100 l sterile PBS emulsifi ed with 50 l of complete Freund's adjuvant (FA; Sigma-Aldrich), and 50 l was injected subcutaneously into both inguinal regions and intradermally at the tail base. Intraperitoneal booster immunizations were given at 2,4,8,14, and 21 weeks following the primary and consisted of 50 g immunogen in 100 l sterile PBS emulsifi ed with 50 l of incomplete FA (Sigma-Aldrich). Mice were bled at baseline and at 5,9,15, and 22 weeks following initiation of diet and at 28 weeks, at the time of euthanization.
The seven intervention groups were injected with the following immunogens or vehicle:

Generation of fl uorescent and nonfl uorescent MDA-modifi ed proteins
MDA-modifi cations of keyhole limpet hemocyanin (KLH; Princeton BioMolecules, NJ), endotoxin-free BSA (BSA; Sigma-Aldrich), and fatty acid-free MSA (Alpha Diagnostics Intl. Inc., San Antonio, TX) were carried out to produce preparations that had either a minimal or maximal amount of fl uorescent MDAderived adducts (see Fig. 1 ).
To produce preparations with minimal fl uorescence, KLH, BSA, or MSA (2 mg/ml PBS) were reacted with freshly generated MDA (50 mol/ml) for 30 min at pH 7.4 and 37°C. Protein content was isolated from free aldehydes using Amicon Ultra concentrators [molecular weight cutoff (MWCO) 10,000 Da; Millipore, Bedford, MA] and sterile fi ltered (0.22 m; Millex GP, Millipore). These preparations are referred to as MDA-modifi ed BSA (MDA-BSA) or MDA-MSA in this article and always have very low fl uorescence as defi ned subsequently.
To produce high-fl uorescence preparations, BSA or MSA (2 mg/ ml PBS) were reacted with acetaldehyde (AA; 200 mol/ml) and freshly generated MDA (100 mol/ml) for ‫ف‬ 4 h at pH 4.80 and 37°C. This incubation of MDA and AA results in the formation of fl uorescent adducts termed MAA adducts, of which the 4-methyl-1,4dihydropyridine-3,5-dicarbonyl (MDHDC)-type adduct is a major product (see Fig. 1 , MAA-MSA). The chemistry of such MAA adducts with proteins has been extensively described [see (15)(16)(17)]. The formation of fl uorescent adducts was followed at max, emission 462 nm ( ex 394 nm), and modifi ed proteins isolated as described above. These fl uorescent preparations are referred to as MAAmodified BSA (MAA-BSA), MAA-MSA, or MAA-modified LDL (MAA-LDL) in this article and always have high fl uorescence as shown in Fig. 1 . Note that because small amounts of AA may spontaneously form as a decomposition product of MDA from its stable bis -diacetal form, there are correspondingly small numbers of MAA epitopes generated in MDA-MSA as well (see Fig. 1 and Results).
As a control for the various MDA-modifi ed immunogens, we used PA-MSA. Like MDA and AA, PA also preferentially modifi es is capable of altering the atherogenic process (1)(2)(3)(4). While there are many potential antigens in the atherosclerotic lesion, much work has focused on oxidized LDL (OxLDL). During the oxidative modifi cation of LDL, a wide variety of immunogenic, oxidation-specifi c neoepitopes are formed, leading to immune activation including auto-Ab formation ( 5 ). Indeed, in mouse models, titers to oxidation-specifi c epitopes (OSEs) of OxLDL parallel the development of lesion formation, as well as regression ( 6 ).
Our laboratory fi rst demonstrated that immunization with malondialdehyde (MDA)-modifi ed homologous LDL (MDA-LDL), an OxLDL model in which native LDL is heavily modifi ed with MDA, led to atheroprotection in rabbits and mice (7)(8)(9). Similar data have been reported by others with MDA-LDL and with other oxidation-specifi c models leading to elevated Ab titers to epitopes of OxLDL ( 10,11 ). These studies led us and others ( 5,(12)(13)(14) to propose the concept that a vaccine to OSEs of OxLDL could be developed to limit atherogenesis. Assuming the safety of such an approach, it would have the advantage of wide applicability and a relatively cost-effi cient approach. However, modifi cation of native human LDL to prepare an atheroprotective immunogen would be cumbersome and impractical for generalized use. Furthermore, when MDA is used to modify LDL, a wide variety of related MDA adducts are formed, both simple and more complex. In order to develop a small-molecule antigen that would reproduce the atheroprotective effects of immunization with MDA-LDL, we sought to determine the responsible immunodominant and atheroprotective adducts.
In this report, we demonstrate that MDA-LDL is not required to generate immunoprotective immunogens, and that complex adducts involving the condensation of two or more MDA molecules with lysine on homologous albumin are equally effective in generating an IgG1, T helper (Th) 2-biased humoral response that was atheroprotective. However, an IgG1, Th2-biased humoral response to an irrelevant antigen was insuffi cient in itself to limit atherogenesis. These data suggest that a hapten-specifi c humoral response to relevant disease-specifi c OSEs on OxLDL is suffi cient to provide atheroprotection and that small-molecule immunogens bearing such OSEs could be used to provide atheroprotective immunity.

Animals and diets
All mice were 10th-generation male Ldlr Ϫ / Ϫ mice on the C57BL/6J background. Mice were weaned at 21 days of age and fed regular chow consisting of 0.01% cholesterol and 4.4% fat (TD8604; Harlan-Teklad, Madison, WI) until initiation of the diet interventions. Mice were housed in cages equipped with rodent enrichments (igloo and gnawing bone; Bio-Serv, Frenchtown, NJ) in a facility with a 12 h light cycle and had ad libitum access to water and food. All animal experiments were performed according to National Institutes of Health guidelines and were approved by the University of California, San Diego Animal Subjects Committee.
Americas Inc., Torrance, CA), were allowed to react with freshly generated MDA for 60 min at pH 7.4 and 37°C. The molar peptide-to-MDA ratio of the reactants was 1:0.1, 1:1, and 1:10 (mol/ mol). All peptide preparations were dialyzed extensively in Spectra/Por dialysis membranes (MWCO 500 Da; Fisher Scientifi c) against PBS and sterile fi ltered (0.22 m; Millex GP, Millipore). We included a negative control for the peptide-to-MDA (1:10) modifi cation, where the peptide was omitted as the only difference. MDA-modifi ed poly-L -lysine (4-15 kDa; Sigma-Aldrich, St. Louis, MO) was prepared as described previously for MDA-LDL, except that Spectra/Por dialysis membranes with MWCO of 2,000 Da were used.

MAA-modifi ed biotinylated peptide antigen
Other peptides were N-terminally biotinylated and had the following basal primary sequence: biotin-Gly-Asp-Gly-Asp-Gly-Lys(OH), hereafter referred to as Bt-GDGDGK (purchased from Biopeptide Co. Inc., San Diego, CA). It was modifi ed with MAA by allowing the peptide (2 mol primary amino groups/ml PBS) to react with AA (200 mol/ml) and freshly generated MDA (100 mol/ml) for ‫ف‬ 5 h at pH 4.80 and 37°C. The formation of fl uorescent adducts was followed at max, emission 462 nm ( ex 394 nm). The entire reaction mixture was added to a conditioned Strata C18-E SPE tube (8B-S001-HBJ; Phenomenex, Torrance, CA), and the MAA-modifi ed peptide was separated from unreacted, free aldehydes by graduate eluting with increasing concentrations of methanol in distilled water. Elution fractions with the epsilon Ϫ amine of lysines. BSA or MSA (2 mg/ml PBS) were reacted with PA (200 mol/ml) for 24 h at pH 7.4 and 37°C, and modifi ed proteins isolated as described above.
Human LDL (LDL) or mLDL were modifi ed with MDA to generate MDA-LDL or MDA-mLDL as previously described ( 9 ). MAA-modifi ed human LDL (MAA-LDL) was prepared by reacting LDL or biotinylated LDL (2 mg/ml PBS) with AA (200 mol/ml) and MDA (100 mol/ml) under conditions described previously, and formation of fl uorescent adducts similarly monitored. PA-modifi ed LDL (PA-LDL) was prepared as described previously for PA-MSA. Preparation of MDA-LDL with minimal fl uorescence was made as described previously for MDA-MSA. All lipoprotein preparations were dialyzed extensively in Spectra/ Por dialysis membranes (MWCO 12,000-14,000 Da; Fisher Scientifi c) against PBS and sterile fi ltered (0.45 m; Millex HV, Millipore).
In MAA and MDA preparations, >90% of the lysines were modifi ed as judged using the trinitrobenzenesulfonic acid assay ( 18 ), whereas in the PA-modifi ed preparations the degree of modifi cation was ‫ف‬ 50%. All preparations used for immunization and in vitro T-cell assays were tested for endotoxin levels by the chromogenic Limulus amoebocyte assay (QCL-1000; BioWhittaker Inc., Walkersville, MD) and contained <1.5 ng lipopolysaccharides/ mg protein.

MDA-modifi ed peptide antigens
The peptides, Ac-SDKP(OH), Ac-DRLDS(OH), Ac-LKFSKKF (OH), Ac-MDKVLNRE(OH), and Ac-DRVYIHPFHL(OH) (Bachem Emission spectrum (and insert for lower fl uorescent unit (FLU) signal range) upon excitation at 394 nm at 10 g/ml of protein shows the characteristic maxima at 462 nm for the 1,4-dihydropyridine-3,5-dicarbaldehyde-type adducts. Formation of such adducts can be enriched through an MAA-modifi cation, and one of the major products is the MD-HDC-lysine adduct (schematically illustrated on top right). The fl uorescence intensity at 462 nm was ‫ف‬ 20-fold and 100-fold higher in MAA-MSA than in MDA-mLDL and MDA-MSA preparations, respectively. PA-MSA (theoretical major adduct is schematically illustrated on bottom right) and unmodifi ed MSA and mLDL did not yield MAA-specifi c fl uorescence. Note that although we aimed at generating totally nonfl uorescent MDA-MSA, we were not able to modify >80% of the lysines in MSA without generating trace amounts of MAA-type adducts.
instructions with modifi cations to decrease the detection limits for mouse IFN-␥ , IL-4, IL-5, and IL-17A to 1.25 pg/ml.

ELISpot assay for single-cell IL-5, IL-17A, and IFN-␥ secretion
ELISpot assays were conducted essentially as previously reported ( 19 ). In brief, splenocytes from individual C57BL/6J mice immunized as described previously were seeded at 2 × 10 6 cells per well in 96-well ELISpot plates in a fi nal culture volume of 200 l. Triplicate cultures of splenocytes from individual mice were incubated for 24 h at 37°C/5% CO 2 in culture media containing 10 g/ml of anti-CD28 (Clone 37.51 NA/LE; BD Biosciences) with and without titrated amounts of antigens. In parallel wells, splenocytes were stimulated with suboptimal concentrations of Concanavalin A (EMD Chemicals, Gibbstown, NJ) to assess cytokine release capacity under identical and nonspecifi c conditions. ELISpot kits for mouse IFN-␥ , IL-5 (both BD Biosciences), and IL-17A (R and D Systems) were assayed according to manufacturers' instructions, using the BD ELISpot AEC Substrate Set for detection.

Measurement of Ab levels and immune complexes by ELISA
Specifi c Ab levels to given antigens in plasma were determined by chemiluminescent ELISA as previously described ( 9,20 ). In brief, Microfl uor® 2 White "U" Bottom Microtiter® plates (Thermo Labsystems, Franklin, MA) were coated with various antigens at 5 g/ml PBS overnight at 4°C. The plates were blocked with 1% BSA in TBS, serially diluted plasma was added, and the plates incubated for 1.5 h at room temperature. Bound plasma Ig isotype levels were detected with various anti-mouse Ig isotypespecifi c AP conjugates using LumiPhos 530 solution and a Dynex Luminometer (Dynex Technologies, Chantilly, VA). The following goat AP-conjugated secondary Ig isotype-specifi c Abs were used: anti-mouse IgM ( -chain specifi c) (Sigma-Aldrich) and anti-mouse IgG1, IgG2b, IgG2c, and IgG3 (all Jackson Immuno-Research Laboratories Inc., West Grove, PA). Data are expressed as relative light units counted per 100 milliseconds (RLUs/100 ms).
To investigate the fi ne specifi city of the humoral immune responses, we developed a peptide-based ELISA using immobilized NeutrAvidin to capture biotinylated peptides, which were unmodifi ed or MAA modifi ed. In short, plates were coated with NeutrAvidin (10 g/ml PBS) overnight at 4°C. The plates were blocked with 1% BSA in TBS and loaded with the unmodifi ed or semipurifi ed MAA-modifi ed biotinylated peptides at 25 nmol/ml and incubated for 1 h at room temperature. After washing with PBS, serially diluted plasma was added and incubated for 1.5 h at room temperature. The remaining ELISA was performed as described previously.
Circulating apoB immune complexes (ICs) were measured by modifi cations of a sandwich ELISA previously described ( 9,21 ), where apoB-100-containing lipoproteins were captured by the anti-mouse apoB-100-specifi c LF5 monoclonal Ab (mAb) ( 22 ). In short, plates were coated with LF5 mAb (a mouse IgG2a) at 5 g/ml PBS overnight at 4°C. The plates were blocked with 1% BSA in TBS, and plasma at 1:100 dilution was added and incubated for 1.5 h at room temperature. Secondary Ig isotype-specifi c Abs and detection were as described previously. In parallel wells, captured apoB-100 was measured using biotinylated LF3 mAb [1 g/ml ( 22 )], which recognizes another epitope on mouse apoB-100. This was followed by AP-conjugated NeutrAvidin and LumiPhos as described previously. Data are expressed as the Ig/apoB ratio (both measured in RLU/100 ms). The LF3 high fl uorescence were dried down and reconstituted in PBS. Mass spectral analysis confi rmed that the GDGDGK peptide contained the MDHDC adduct (data not shown).

Plasma lipid and cytokine analyses
Mice were anesthetized with isofl urane, and retro-orbital blood was obtained via EDTA-coated microcapillary tubes. Plasma cholesterol and triglycerides levels were determined using automated enzymatic assays (Roche Diagnostics, Indianapolis, IN, and Equal Diagnostics, Exton, PA). Total cholesterol exposure was estimated as the area under the cholesterol curve over time divided by days of cholesterol feeding (Origin Pro 7; OriginLab Corporation).
Plasma cytokine levels were determined by a chemiluminescence-based sandwich ELISA. Plasma was assayed at dilutions

Splenocytes from atherosclerotic Ldlr
Ϫ / Ϫ mice after 28 weeks of intervention (without giving a recall boost) were suspended in culture media (RPMI 1640 media containing 10% heat-inactivated fetal bovine serum, 10 mM HEPES buffer, 2 mM L -glutamine, 0.1 mM nonessential amino acid solution, 50 g/ml Gentamicin reagent solution, and 0.05 mM 2-mercaptoethanol; all Invitrogen, Carlsbad, CA) and seeded at 5 × 10 5 cells per well in 96-well fl at-bottom plates in a fi nal culture volume of 200 l. Triplicate cultures of splenocytes from individual mice were incubated for 72 h at 37°C/5% CO 2 in culture media with and without titrated amounts of antigens. In parallel wells, cells were cultured in the presence of 10 g/ml plate-bound anti-CD3 (Clone 145-2C11 NA/LE; BD Biosciences, San Diego, CA).
The amounts of specifi c cytokines in the supernatants were determined using the same chemiluminescence-based sandwich ELISA for determination of plasma cytokine levels.
To characterize the T-cell responses more thoroughly, C57BL/6J mice were immunized (one primary and two booster injections) with the selected immunogens MAA-MSA, PA-MSA, MSA, and PBS. Splenocytes were cultured as described previously, except that all wells (antigen-specifi c as well as nonspecifi c stimulation) contained 10 g/ml of anti-CD28 (Clone 37.51 NA/LE; BD Biosciences). The amounts of specifi c cytokines in the supernatants were simultaneously determined using the BD TM Cytometric Bead Array platform according to the manufacturer's the parametric one-way ANOVA with the Bonferroni's or Newman-Keuls multiple comparison posttest unless otherwise noted (Prism v5; GraphPad Software). In some cases, data were log10-transformed to obtain equal variances among groups. Nonparametric data were expressed as box plots, where horizontal lines indicate median and 25-75 percentiles and whiskers indicate 10-90 percentiles, and analyzed using Kruskal-Wallis ANOVA with Dunn's multiple comparisons posttest. P < 0.05 was considered signifi cant.

Immunogen preparation and characterization
Immunization with MDA-modifi ed proteins generates MDAlysine-specifi c antisera. MDA-LDL contains MDA-derivatized apoB as well as aminophospholipids. We previously demonstrated that the immune response to MDA-LDL was major histocompatibility complex (MHC) class II restricted ( 9 ), suggesting that MDA-modifi ed protein, but not lipid per se, is the immunodominant epitope. To begin to determine the nature of the relevant epitope, we immunized C57BL/6J mice with MDA-mLDL as well as MDA-modifi ed KLH (MDA-KLH), MDA-BSA, and MDA-MSA. With all immunogens, the IgG1 isotype dominated the responses (data not shown), as we previously reported ( 9 ), suggesting a Th2-biased response. Furthermore, >90% of the MDA-LDL-specifi c IgG1 titers in sera from mice immunized with MDA-mLDL could be competed by MDA-KLH (or other MDA-modifi ed proteins) suggesting a haptenspecifi c response to the MDA adduct (data not shown). Because MDA is believed to preferentially modify lysine residues, we tested the specifi city of antisera to MDA-LDL by competition ELISA with a series of MDA-modifi ed arginine-, histidine-, and lysine-containing peptides (see Fig. 2 ). Only MDA-modified lysine-containing peptides competed. Collectively, these results indicated that MDA-LDL and proteins contained immunodominant MDA-derived lysine adducts, which induced a Th2-biased hapten (i.e., MDA-lysine)-specifi c humoral immune response.
Generation of nonfl uorescent and fl uorescent MDA-type adducts. MDA-LDL is a complex antigen. We therefore immunized seven groups of Ldlr Ϫ / Ϫ mice with a variety of model MDA antigens to provide insight into the nature of its atheroprotective epitopes. Our fi rst goal was to determine whether the molecular structure of the MDA-lysine adducts infl uenced their atheroprotective effects, and in particular, whether MDA-LDL itself was required to make an effective immunogen. We generated two different MDA-derived MSA preparations. MSA is a 67 kDa protein that contains 51 lysine residues and, when modifi ed with MDA, most likely results in a mixture of MDA-derived adducts. Therefore, we focused on generating fl uorescent and nonfl uorescent MDA-derived MSA preparations (see Methods and Fig. 1 ). The nonfl uorescent preparations are hereafter called MDA adducts (e.g., MDA-MSA) and represent simple MDA adducts with proteins prepared by a 30 min incubation with MDA at pH 7.4 (see Fig. 1 ). In contrast, the fl uorescent preparations represent more and LF5 mAbs were kindly provided by Stephen Young of University of California, Los Angeles.

Competition ELISA
We have previously shown that MDA-mLDL immunization induces strong hapten-specifi c anti-MDA-epitope IgG1 humoral responses and confers atheroprotection in Ldlr Ϫ / Ϫ mice ( 9 ). To test which MDA-modifi ed amino acids were immunodominant in antisera from MDA-mLDL-immunized mice, we utilized the MDA-modifi ed peptides described previously as competitors to inhibit IgG1 binding to plated MDA-LDL in competition ELISA. In brief, plates were coated with MDA-LDL at 1 g/ml PBS overnight at 4°C. Plasma from C57BL/6J mice immunized with homologous MDA-mLDL (1:80,000 dilution) and competitors (250 g/ml) were mixed and preincubated for 1 h at room temperature before addition to appropriate wells. After 1.5 h incubation, bound IgG1 was detected as described previously and expressed as a ratio of binding in the presence of competitor (B) divided by binding in the absence of competitor (B 0 ).
In other competition ELISAs, competitors were serially diluted before overnight incubation at 4°C with fi xed dilution of plasma. Plasma-competitor solutions were added to antigen-coated ELISA plates without pelleting of ICs in order to produce inhibition curves as B/B 0 as described ( 9,19 ).

Atherosclerosis analysis and immunocytochemistry
The extent of atherosclerosis was quantifi ed by computer-assisted image morphometry in Sudan IV-stained en face preparations of the entire aorta and in cross-sections through the aortic origin of paraffi n-embedded hearts, as previously described ( 23 ). It should be noted that among 119 total mice analyzed for atherosclerosis, 2 mice were statistical outliers (>3× SD) in the en face analysis, but not in any other analyses. Therefore, for this analysis only, one MDA-mLDL-immunized (36.7% lesion area) and one FA-immunized (43.5% lesion area) mouse were excluded. A modifi ed trichrome stain, consisting of hematoxylin, fuchsin red/picric acid, and aniline blue, was utilized for crosssectional analysis of plaque burden and gross composition, as previously described ( 23 ). Techniques for the immunochemical analysis of immunoglobulin isotypes in lesions are presented in the supplementary fi gures.

Antisera inhibition of modifi ed LDL binding to macrophages
Binding of biotinylated OxLDL ligands to J774 mouse macrophages plated in microtiter wells was assessed by a chemiluminescent binding assay as recently described by Montano et al. ( 24 ). In brief, isolated human LDL was biotinylated according to the manufacturer's protocol (Cat# 21326; Pierce Biotechnology) prior to modifi cations to prepare biotinylated MAA-LDL, CuOxLDL, MDA-LDL, and PA-LDL ligands. The biotinylated MAA-LDL ligand at fi xed concentrations (2.5 g/ml) was incubated with serially diluted competitors (antisera) and controls in 1% BSA-PBS at concentrations/dilutions indicated in fi gures. The ligand-competitor solutions were incubated overnight at 4°C, and the following morning, the extent of binding of the biotinylated ligands was detected with NeutrAvidin-conjugated AP (Pierce Biotechnology), LumiPhos 530 (Lumigen), and a Dynex Luminometer (Dynex Technologies). Data were recorded as RLUs/100 ms and expressed as a ratio of binding in the presence of competitor (B) divided by binding in the absence of competitor (B 0 ).

Statistical analyses
All experimental data are expressed as mean ± SEM unless otherwise noted. Differences in mean values were analyzed with MIP-2 above native LDL control. Thus, MAA epitopes, but not PA epitopes, are proinfl ammatory.

Immunization protocol and impact on atherosclerosis
Experimental protocol. To determine which of the MDAtype adducts were atheroprotective, we immunized seven groups of Ldlr Ϫ / Ϫ mice with various models. Five intervention groups (n = 15-20 mice/group) were immunized with modifi ed proteins plus FA, one group with FA alone, and one group with PBS alone. Immunogens used in the fi ve experimental groups were as follows: MDA-mLDL, MDA-MSA, MAA-MSA, PA-MSA, and MSA, respectively.
We chose to study 28-week-old Ldlr Ϫ / Ϫ mice because such mice have preexisting early to intermediate atherosclerotic lesions before modulation of their immunity and would therefore represent a model for the human target group of such atheroprotective immunizations. After baseline lipid evaluations, mice were given a primary immunization (with complete FA) and begun on a moderate HC diet (regular chow supplemented with 0.5% cholesterol), which was continued for 28 weeks. This diet generates an almost exclusive elevation of LDL-cholesterol, without associated hypertriglyceridemia, weight gain, and alterations of insulin/glucose associated with high-fat "Western"-style diets ( 23 ). Five booster immunizations (with incomplete FA) were given during the 28 weeks. Mice gained weight equally in all groups, and plasma cholesterol and triglycerides were similar throughout the study ( Table 1 ). At the end of 28 weeks, terminal blood samples were obtained, and the extent of lesion formation was determined.
Atherosclerosis analyses. The extent of en face lesion formation in the entire aorta for all seven groups is graphed in Fig. 3 as percent of aortic lesion area, as well as percent of the complex malondialdehyde-acetaldehyde adducts, hereafter referred to as MAA adducts (e.g., MAA-MSA, Fig. 1 ), which exhibit a characteristic fl uorescence profi le ( em,max 462 nm, ex 394 nm) as described in Methods. In both preparations, typically ‫ف‬ 90% of the lysines were modifi ed. Fluorometric analysis indicated a >100-fold higher fl uorescent intensity for MAA-MSA compared with the MDA-MSA or native MSA preparations ( Fig. 1 ). It should be noted that although the MAA-MSA preparation was greatly enriched in MAA as judged by specifi c fl uorescence, it is still likely that other MDA-derived adducts are also present. In turn, although MDA-MSA is termed a nonfl uorescent preparation, Fig. 1 shows that very small amounts of fl uorescent MDA-derived adducts are present.
Generation of PA-lysine adduct control. It has been suggested that a Th2-biased response per se might be atheroprotective independent of the immunogen ( 3 ). To study this, we generated a similar lipid peroxidation-derived adduct with PA, derivatized to MSA to yield PA-MSA ( Fig. 1 ). PA contains only one aldehyde group, resulting in simpler chemistry than with MDA, and forms only adducts with primary amines via Schiff bases. We used PA-MSA as an "adduct-carrier" control in the intervention study, which we anticipated would serve as a nonrelevant Th2-biased immunogen control.
Thiele and colleagues ( 25 ) and we ( 26 ) have shown that MAA-modifi ed proteins are proinfl ammatory. To confi rm that our immunogens demonstrated this property, we incubated them with RAW cells, a mouse macrophage line. As

Analysis of plasma Ab titers and lipoprotein ICs
To ensure the effectiveness of the immunizations in each mouse, titers were measured after primary and two booster injections (week 5) for IgG1, IgG2c, and IgM titers to respective immunogens. All mice immunized with MDA-mLDL, MDA-MSA, MAA-MSA, and PA-MSA showed high-titered IgG1 reactivity to respective immunogens, whereas titers in MSA-and FA-injected mice were similar to PBS-injected mice (data not shown). Thus, after only 5 weeks of immunization, there were high-titered, hapten-specifi c IgG1 responses against respective adducts in each immunized group, indicating Th2-biased cellular immunity.
At the end of the study, a detailed analysis of fi nal Ab titers and specifi city was conducted. Formal Ab dilution curves for individual mice were conducted for each isotype and each antigen (supplementary Figs. II and III). IgG1 titers to respective OSEs were uniformly high, while IgG2c titers were low, indicative of Th2-biased responses, and pertinent results of the IgG1/IgG2c ratios to each antigen are summarized in Fig. 4 . In general, MAA-MSA immunization generated and sustained a stronger and relatively more IgG1-dominant haptenspecifi c humoral immune response to both MAA-BSA and MDA-LDL than the other immunogens. For example, the IgG1/IgG2c ratio was 40 for the MAA-MSA-immunized mice versus ‫ف‬ 8 for the MDA-mLDL-immunized group. By analogy, PA-MSA immunization induced high IgG1/IgG2c ratios for binding to propanal-modifi ed BSA (PA-BSA), indicating strong Th2 bias for this immunization strategy also.
Mice immunized with MAA-MSA showed high IgG1 titers to the complex MDA/MAA epitopes as exemplifi ed by binding to MAA-BSA and MDA-LDL, but much lower binding to the nonfl uorescent MDA epitopes on MDA-BSA (supplementary Fig. III). Importantly, MDA-MSA-immunized mice showed higher binding to MAA-BSA than to MDA-BSA (note different scales), while the binding to MDA-BSA was not different than that with sera from MAA-MSA-immunized mice. Collectively, these data indicate that complex, fl uorescent MDA/MAA epitopes are immunodominant.
Because of the specifi c modifi cation protocol used, MDA-LDL contains a greater distribution of MDA-derived epitopes, both nonfl uorescent and fl uorescent epitopes, than MDA-MSA and MAA-MSA ( Fig. 1 ). It was therefore expected that mice immunized with MDA-mLDL would show hapten-specifi c IgG titers to the entire spectrum of MDAderived epitopes, that is, binding to MDA-BSA, MDA-LDL, and MAA-BSA (supplementary Fig. III). In contrast, mice immunized with PA-MSA showed high titers to PA-BSA, but only weak titers to MDA or MAA epitopes. At this high dilution, no titers to any OSEs were seen in MSA-, FA-, or PBSimmunized mice, although titers were measured at much lower dilutions (supplementary Fig. III).
We also analyzed pooled antisera for responses of other isotypes. IgG2b titers to MDA/MAA were relatively strongly induced by MDA-mLDL immunization, while in contrast, IgG3 titers were not induced (data not shown). IgM titers appeared to follow the same pattern with highest IgM binding to antigens in MDA-mLDL-immunized mice, less PBS control (actual data presented in Table 1 ). These data indicate that the most effective atheroprotective immunogen was the MAA-MSA, leading to a 37% reduction of lesion formation compared with the PBS control (10.7 ± 0.8% vs. 17.1 ± 1.4% lesion area, P < 0.01). As these mice were already 28 weeks old at the initiation of the protocol, and the intervention phase lasted another 28 weeks, this represents a substantial reduction in the rate of progression. Lesion formation in the MAA-MSA group was also signifi cantly decreased from three other control groups, FA, MSA, and PA-MSA controls ( P < 0.05). Although the MAA-MSA group had less atherosclerosis than the two MDA groups, this difference was not signifi cant. The FA group showed a small (though nonsignifi cant) protective effect, consistent with prior observations by us ( 20 ) and others ( 27 ) that FA and other adjuvants have minor atheroprotective properties. The MDA groups had signifi cantly less atherosclerosis than the PBS group, but not compared with the FA control in this experiment.
Quantifi cation of lesion formation at the aortic root revealed that only the MAA-MSA group showed signifi cantly reduced lesion formation at this site, a 24% reduction compared with PBS (0.299 ± 0.025 vs. 0.393 ± 0.027 mm 2 / section, P < 0.05) ( Table 1 ).
We investigated lesion composition by a modifi ed trichrome stain (as described in Methods) of the MAA-MSA, MDA-LDL, and PBS and MSA control mice but did not fi nd any obvious differences in composition between lesions of comparable size. We also investigated the deposition of endogenous IgG isotypes in the aortas of the variously immunized mice, and signifi cant differences were found, as described subsequently. MAA-lysine, while mice immunized with MDA-mLDL and MDA-MSA showed lesser titers ( Fig. 5B ). None of the other groups showed any MAA-specifi c binding, including the PA-MSA-immunized mice. Similarly, low titers of IgG2b and IgG2c to MAA were seen in the MAA-immunized and, to lesser extent, MDA-immunized mice ( Fig. 5C, D ). In contrast, the highest titers of IgM Abs were seen in the MDA-mLDL-immunized group ( Fig. 5F ). Combined, the data in Fig. 5A-F show that MAA-MSA immunization induces a very strong Th2-biased IgG1-dominant MAA-epitope-specifi c humoral response with a formal titer >>1:50,000 in the MAA-MSA-immunized mice. In further studies, we used the synthetic MDHDC-lysine peptide to perform a competitive ELISA using pooled plasma from MAA-MSA-immunized mice. The MDHDC peptide was able to fully compete the IgG1 binding to the plated MDHDC peptide ( Fig. 5G ) and was nearly as effective in competing binding to plated MAA-BSA ( Fig. 5H ). Thus, MDHDC is an immunodominant epitope of antisera of MAA-MSA-immunized mice. binding in MAA-MSA-immunized mice, and the least IgM binding in the remaining fi ve groups, which had fairly equal IgM binding (supplementary Fig. III). We also measured titers of the atheroprotective IgM natural Ab (NAb) E06, which binds to the phosphocholine (PC) of oxidized phospholipids (OxPLs ). Levels were highest in the MAA-MSA group among the active immunizations, and nearly two to three times as high as seen with other modifi ed LDLs or MSA immunogens (supplementary Fig. IV). Interestingly, the E06 titers were as high as that seen after atheroprotective immunization with Streptococcus pneumoniae ( 20 ).

MAA-specifi c Ab titers
To characterize the fi ne specifi city of the induced Abs, we coated microtiter plates with a small lysine-containing peptide bearing a chemically synthesized MAA (MDHDC) epitope (shown in Fig. 5 ) and measured isotype-specifi c binding to the MAA-lysine epitope ( Fig. 5 ). Mice immunized with MAA-MSA showed very high-titered IgG1 binding to . Ratios were calculated using the same plasma dilution within an immunization group, although different dilutions may have been used as more appropriate for a given antigen/immunization group (most ratios were calculated at 1:10,000 plasma dilution). For the ratio to be meaningful, both IgG1 and IgG2c binding were selected to be in the linear range. Data are given as mean ± SEM of 15-20 mice per group as indicated in Table 1 . other mice were similar. No other important discriminatory fi ndings were observed with the other IC analyses.

MAA-specifi c antisera inhibit binding of MAA-LDL to macrophages and immunostain atherosclerotic plaques
We previously showed that mAbs to OSEs of OxLDL inhibit the binding and uptake of OxLDL or MDA-LDL to

Presence of ICs with apoB lipoproteins in plasma
We measured isotype-specifi c ICs with apoB-containing lipoproteins (supplementary Fig. V). Levels of IgM/apoB ICs were signifi cantly higher in mice immunized with MDA-mLDL, MDA-MSA, MAA-MSA, PA-MSA, and MSA than in mice injected with FA and PBS. Interestingly, mice immunized with MAA-MSA had the highest IgG1/apoB levels, while levels of these ICs in that observed with plasma from PA-MSA-immunized mice ( Fig. 6C ). We then assessed the ability of plasma from each of the seven immunized groups to compete, using antisera pools at 1:100 dilution ( Fig. 6D ). Even at this high dilution, plasma from MAA-MSA mice was a highly effective competitor, followed closely by plasma from MDA-mLDL-immunized mice, competing as effectively as a 40-fold excess of unlabeled MAA-LDL, and much more effectively than any of the other plasma pools. In part, the degree of competition of the other plasmas is due to presence of endogenous anti-MAA Abs, and in part to the presence of nonimmunoglobulin plasma proteins that bind MDA, such as complement factor H (CFH), which we recently reported is a major MDA/MAA binding macrophages, which retards foam cell formation ( 20,21,24,28 ). To determine whether the antiserum from the immunized mice had this capacity, we added varying dilutions of plasma to macrophages in microtiter wells and tested their ability to inhibit the binding of MAA-LDL ( Fig. 6 ). MDA-LDL is well known to bind to macrophage scavenger receptors ( 29 ), and in a preliminary study, we showed that biotinylated MAA-LDL also bound to J774 macrophages in a saturable and specifi c manner ( Fig. 6A, B ), whereas PA-LDL did not. Next, we showed that plasma from MAA-MSA mice inhibited MAA-LDL binding in a dose-dependent manner, consistent with their high MAA-specifi c Ab titer. At equivalent dilutions of plasma, this competition was much greater than reveals that the MAA-MSA-immunized mice had a remarkably blunted IFN-␥ response ( Fig. 8C ).
Such whole spleen cytokine release assays are subject to a number of experimental variables that may limit their sensitivity to detection of antigen-specifi c responses. Therefore, we immunized C57BL/6J mice with MAA-MSA and PA-MSA, MSA, and adjuvant and again examined the ability of splenic T cells from these mice to release IL-5 and IFN-␥ into the supernatant in dose-dependent, antigen-specifi c recall assays under more optimal in vivo and in vitro conditions. Preliminary studies validated that the expected humoral Ab responses occurred as were observed with the immunized Ldlr Ϫ / Ϫ mice (data not shown).
Splenocytes were harvested 7 days after the last booster injection and were costimulated with an anti-CD28 Ab. Splenocytes responded with dose-dependent, antigen-specifi c release of IL-5 ( Fig. 9A ), IFN-␥ ( Fig. 9B ), and IL-17A (supplementary Fig. VIIIA) in response to MAA-MSA, which generally was more stimulatory than PA-MSA. Because there is a differential capacity for release of different cytokines in response to stimulation, the data were normalized and plotted as the ratio of antigen-specifi c responses of IFN-␥ and IL-5 as a percentage of the responses to stimulation with anti-CD3/anti-CD28, indicative of maximal capacity of cytokine release by T cells in the splenic culture ( Fig. 9C ). Clearly, there was a heightened release of IL-5 compared with IFN-␥ in MAA-immunized mice, and in a similar comparative plot of IL-5 versus IL-17A, again there was a strong IL-5 bias (supplementary Fig. VIIIB).
As an alternative approach to assess the Th bias of antigenspecifi c T cells, we performed ELISpot assays of splenic cells from the same groups of immunized mice and counted splenic cells that released IL-5 and IFN-␥ after stimulation with their respective immunogens. To normalize the antigenspecifi c ELISpot data in a similar manner, we compared each value with the number of spot-forming cells (SFCs) after stimulation with the nonspecifi c T-cell stimulant, Concanavalin A, at a limiting concentration. There were considerably more IL-5 SFCs compared with IFN-␥ SFCs ( Fig. 9D ) in the MAA-MSA-immunized mice. In contrast, IL-5 SFCs were only minimally increased after PA-MSA stimulation. A similar analysis for IL-17A SFCs also revealed strong bias toward IL-5 SFCs (supplementary Fig. VIIIC) in MAA-MSA-immunized mice.
These data strongly support a strong antigen-specifi c Th2biased response for T cells of MAA-MSA-immunized mice, which also was associated with a suppression of IFN-␥ release.

Impact of immunizations on B-cell and T-cell populations
We also examined the impact of MAA-MSA and PA-MSA immunizations on T-and B-cell populations compared with adjuvant-immunized controls (supplementary Table I Table I), nor did we fi nd differences in Treg cells that were CD25 + or CD25 Ϫ (data not shown).
protein ( 26 ). These data strongly suggest that the induced anti-MAA Abs have the ability to block modifi ed LDL binding.
To demonstrate that the induced immune sera recognized endogenous epitopes in lesions, we immunostained aortic sections from mice of the various groups for deposition of endogenous isotype-specifi c Abs (supplementary Fig. VIA). Interestingly, the aortas of mice immunized with MAA-MSA showed the greatest content of IgG1 Abs, consistent with the very high IgG1 titers to MAA. A semiquantitation of these data and other Ab isotypes is presented in supplementary Fig. VIB, which shows that IgG1 deposition in MDA-MSA-and MAA-MSA-immunized mice was the most prominent, consistent with the high IgG1 titers. Interestingly, the MAA-MSA showed the least IgG2c titers, possibly refl ecting an actual decrease in this Th1-biased IgG isotypes. It should be noted that the PA-MSA-immunized mice also showed considerable IgG1 deposition, presumably to a PA-like epitope.

T-cell responses to immunizations
The humoral Ab responses to the MAA-lysine and other modifi cations were dominated by IgG1 isotypes, suggesting a Th2-biased response. To directly examine the T-cell response, we measured IFN-␥ and IL-5 cytokine responses in splenic T-cell cultures prepared from the immunized and atherosclerotic Ldlr Ϫ / Ϫ mice after 28 weeks of intervention. We did not give the mice a recall injection 3 days prior to euthanization, as is common practice, because we wanted to evaluate the actual immune status in the mice at time of euthanization. Initially, we examined the T-cell memory response to generalized activation with anti-CD3 ( Fig. 7 ). Remarkably, the most prominent response was a generalized suppression of IFN-␥ release in mice immunized with MDA-mLDL, MDA-MSA, MAA-MSA, and PA-MSA ( Fig. 7A ). There did not appear to be differences in IL-5 release ( Fig. 7B ), and a comparison of the two expressed as the IFN-␥ /IL-5 ratio ( Fig. 7C ) shows the marked suppression of IFN-␥ by MAA immunization.
We next examined antigen-specifi c cytokine release to each of the different antigens ( Fig. 7D-G ). Only splenocytes of mice immunized with MDA-mLDL ( Fig. 7D ) or MAA-MSA ( Fig. 7F ) showed antigen-specifi c release of IL-5 in this assay. IFN-␥ responses were mixed and unremarkable, though again, there appeared to be blunted responses to MAA-MSA ( Fig. 7G ).
We also measured cytokines in plasma after 28 weeks of intervention ( Fig. 8 ). In general, there were large variations in each of the cytokines measured even within a given immunization group. Increased IFN-␥ levels were observed in mice immunized with MDA-MSA and MDA-mLDL and, to a lesser extent, PA-MSA, but remarkably, levels of IFN-␥ were statistically suppressed in the MAA-MSA group ( Fig. 8A ). With respect to Th2 cytokines, IL-5 levels were highest in MDA-MSA-and MDA-mLDL-immunized mice ( Fig. 8B ), and IL-4 and IL-10 were also increased signifi cantly only in the two groups of MDA-immunized mice (supplementary Fig. VII). A comparison of the IFN-␥ /IL-5 ratio in individual mice again lesions ( Fig. 10A , B ), confi rming similar observations recently reported by others ( 17,30 ). We have postulated that because such OSEs are proinfl ammatory [e.g., see Methods and supplementary Fig. I and ( 25, 26 )], multiple innate pattern recognition receptors, such as NAbs, have been conserved to mediate homeostasis ( 5 ). We have recently shown that in mice, up to 20% of innate IgM NAbs bind to

MAA epitopes and NAbs to MAA are present in humans
The work presented here supports the concept that MAA is an immunodominant epitope among the many MDA-derived adducts and its use as an immunogen can lead to atheroprotective immunity. Using a recently cloned human monoclonal Fab Ab specifi c for MAA, we demonstrate the presence of MAA epitopes in human atherosclerotic molecules with lysine to form MAA-type adducts generate immunodominant epitopes that lead to atheroprotective responses. We further demonstrate that a hapten-specifi c humoral response is suffi cient, and thus, MDA-LDL is not a requisite immunogen, as MDA-and MAA-modifi ed homologous albumin were equally effective immunogens. This report provides a detailed investigation into the nature of the Abs that appear to provide atheroprotection and the cellular immune responses associated with the immunizations. From these studies, a number of relevant observations on the desirable atheroprotective properties of a vaccine approach can be deduced.

First: a hapten-specifi c humoral response against a disease-relevant OSE is suffi cient to provide atheroprotective immunity
Thus, modifi cations of autologous LDL are not needed to provide the relevant antigens. In vivo, there are likely to be a variety of OSE that represent such hapten-specifi c epitopes. For example, mAbs that bind OxPLs, MDA, MAA, and 4-hydroxy-2-trans-nonenal bind to OxLDL and also to experimental and human atherosclerotic tissues ( 31,32 ). The PC of OxPL, found on both OxLDL and apoptotic cells, is an example of such an OSE, and raising the titer of the PC-specifi c NAb E06 by immunization was atheroprotective ( 20 ). Similarly, immunization of apoE-defi cient mice with PC-KLH also reduced lesion formation ( 33 ). In analogy, we recently showed that infusion of human Fab IK17, which binds to MDA and MAA, is also atheroprotective ( 21 ). These data suggest that hapten-specific MAA-type epitopes ( 19 ). To demonstrate that humans also possess MAA-specifi c NAbs, we examined IgM of newborn umbilical cord plasma. Because this IgM is solely derived from the newborn, it is thought to represent naïve NAbs in humans. As shown in Fig. 10C , newborn IgM bound specifically to MAA epitopes in a hapten-specifi c manner with relatively high avidity ( K d of 2.6 × 10 Ϫ 9 M) and were also competed for in part by MDA-LDL.

DISCUSSION
Our laboratory fi rst demonstrated that immunization with homologous MDA-LDL led to atheroprotection in rabbits and mice, data confi rmed and extended by others [reviewed in ( 3,5 )]. Assuming the safety of such an approach, these studies led us to propose the concept that a vaccine to OSEs of OxLDL could be developed to limit atherogenesis. However, modifi cation of native human LDL to use as an atheroprotective immunogen would be cumbersome and impractical for generalized use. Furthermore, when MDA is used to modify LDL, a wide variety of related MDA adducts are formed, both simple and more complex. In order to develop a small-molecule antigen that would reproduce the atheroprotective effects of immunization with MDA-LDL, we sought to determine the responsible immunodominant and atheroprotective adducts.
In this report, we demonstrate that fl uorescent adducts of MDA involving the condensation of two or more MDA  Table 1 , except that only 11 of 19 PBS plasmas were analyzed). P values are given in plots as * P < 0.05; ** P < 0.01; and *** P < 0.001. Fig. III). The Th2 bias is all the more remarkable in that immunization with MDA-modifi ed heterologous carrier proteins such as MDA-BSA and MDA-KLH lead to haptenspecifi c anti-MDA IgG1 titers, but mixed IgG1 and IgG2c anti-carrier titers (data not shown). Moreover the use of complete FA, particularly in C57BL/6 mice, is known to strongly stimulate Th1-biased responses ( 34 ). Because preparations of MDA-LDL yield both simple (nonfl uorescent) and complex (fl uorescent) adducts, we sought in this study to defi ne which of these were immunodominant among the MDA-LDL-specific Abs. We first showed that MDA-type modifi cations of lysine were the relevant immunogenic modifications ( Fig. 2 ) and that, among these, the highly fl uorescent MAA-type lysine humoral responses to a variety of such OSEs may be of potential benefi t.

Second: MAA epitopes are immunodominant structures of the MDA-LDL immunogen and lead to a Th2-biased IgG1 response and atheroprotection
We previously showed that immunization of mice with homologous MDA-LDL was MHC-II restricted and was strongly Th2 biased ( 9 ). In the current studies, we found that immunization with MDA-mLDL, MDA-MSA, MAA-MSA, and PA-MSA all lead to similar high-titered, almost exclusive IgG1 titers to their respective hapten immunogens, compared with IgG2c titers, again suggesting strong Th2-biased immune responses ( Fig. 4  human IgM NAbs. As shown in Fig. 10C , these human IgM bound to MAA in a hapten-specifi c manner and with relatively high avidity. We have also recently shown that the innate protein CFH bound strongly to MAA-LDL and MAA-BSA, neutralizing their proinfl ammatory effects ( 26 ). Further, we show that MAA-LDL is specifi cally bound by macrophage scavenger receptors ( Fig. 6A, B ) consistent with prior reports showing that other MAA-modifi ed proteins are also bound by scavenger receptors ( 35,36 ).

and supplementary
These data demonstrate that MAA is a danger-associated molecular pattern (DAMP), the target of multiple innate pattern recognition receptors (PRRs): NAbs, scavenger receptors, and CFH. This implies that MAA-specifi c innate responses have been conserved via natural selection to maintain homeostasis, as previously suggested ( 5 ). Indeed, we [supplementary Fig. I and ( 26 )] and others ( 25 ) have shown that MAA-modified proteins induce strong proinfl ammatory responses, and thus, NAbs and CFH likely serve to neutralize some of these proinfl ammatory responses, and scavenger receptors likely effect removal of these modifi ed structures (albeit in the context of also initiating infl ammatory responses). We also show the prominent presence of MAA epitopes in human atherosclerotic lesions ( Fig. 10A ), confi rming recent reports ( 17,30 ). Importantly, MAA-epitopes are also prominent OSEs on apoptotic cells, as demonstrated by the binding epitopes were immunodominant. For instance, antisera from MAA-MSA immunization bound to both MAA-BSA and MDA-LDL (supplementary Fig. III), and conversely, MDA-MSA immunization also induced both MDA-and MAA-specifi c titers, even though the MDA-MSA preparation had only low amounts of fl uorescent MAA-lysine adducts ( ‫ف‬ 2 FLU; Fig. 1 ). Furthermore, using the pure, chemically synthesized MAA (MDHDC) epitope (shown in Fig. 5 ), we show that in the MAA-MSA antisera, the anti-MDHDC IgG1 titers were >100-fold higher compared with MDA-MSA antisera ( Fig. 5B ), and in turn, in the MDA-MSA antisera they were higher than any of the four control groups. The induced IgG1 in the MAA-MSA-immunized mice were remarkably specifi c as shown by competition immunoassays ( Fig. 5G ). These data support the conclusion that MAA-lysine adducts are immunodominant, leading to a Th2-biased IgG1 response.
Third: MAA is a DAMP and a major target of multiple innate immune pattern recognition receptors; therefore, adaptive responses to this OSE are likely to be benefi cial We pursued an approach to target adaptive responses to MAA as we previously found that up to 20% or more of all IgM in both mouse plasma and newborn human cord blood specifi cally bound to MAA ( 19 ). IgM in newborn cord blood is fully of fetal origin, and therefore, represent binding, followed closely by that of MDA-mLDL-immunized mice. Indeed, the induced antisera were as effective as a 40-fold excess of unlabeled MAA-LDL ( Fig. 6 ).
On the other hand, we were surprised that the antisera from the MDA-MSA immunizations were not as effective in blocking MAA-LDL binding, and yet this immunization reduced atherogenesis, obviously implying that other properties of the Abs, and/or changes in cellular immunity, were also important to the atheroprotective effects of immunizations. We noted above that MAA-modifi ed proteins were proinfl ammatory, and that CFH had the capacity to block some of these effects. In a similar manner, it is also possible that an important if not major effect of the MAA-specifi c Abs is related to their ability to bind and block such proinfl ammatory effects.
To characterize the cellular immune responses, we measured cytokine release (IFN-␥ , IL-4, IL-5, IL-10, and IL-17) following dose-dependent 72 h stimulation of primary splenic cultures with each of the modifi ed antigens and MSA. In response to nonspecifi c stimulation with anti-CD3 ( Fig. 7 ), we found impaired splenic capacity to release the Th1-cytokine IFN-␥ , a well-known proatherogenic cytokine, in T cells of mice obtained from all four adduct-carrying immunogens, but did not see any impairment of Th2-cytokine release. This suggests that immunizations with the adduct-carrying immunogens in general induced a Th2 response that blunted a Th1 response (e.g., via IL-4). On the other hand, in response to antigen-specifi c stimulation, we saw a pronounced Th2 response with antigen-specifi c cytokine release for IL-5, and this only occurred following MDA-mLDL and MAA-MSA stimulation ( Figs. 7D, F , and 9 ). The released IL-5 corresponds to 60%-80% of the maximal amount released following anti-CD3 stimulation. These data with MAA-MSA recapitulate what we previously reported for MDA-mLDL immunization ( 9 ). Indeed, in that report we showed that the pronounced IL-5 release, a known B-1 cell stimulant, subsequently led to stimulation of the E06 NAb. In the current studies, MAA-MSA immunization led to the strongest stimulation of IL-5 and, similarly, led to a strong induction of E06 titers (supplementary Fig. IV); indeed, we speculate that these titers contributed signifi cantly to the atheroprotection seen with the MAA immunizations. Finally, we note that we did not see an obvious impact of the immunizations on changes in T-cell populations (supplementary Table I), and in particular, we did not see changes in various Treg populations, as has been reported by others, with other immunization strategies ( 40,41 ).
Thus, we believe that the protective effects of immunization with MAA were due to the induction of IgG1 Abs that both bound to and blocked the uptake of OxLDL by macrophages and also blocked proinfl ammatory effects of MAA-modifi ed proteins. One might speculate that the predominant IgG1 ICs formed with the induced Abs (e.g., supplementary Fig. VIB), which are known to bind with higher affi nity to inhibitory Fc ␥ receptor (Fc ␥ R)IIB receptors versus those of Ig2c isotypes, for example, might contribute also to the anti-infl ammatory activity ( 42 ). In addition, the prominent Th2-biased cellular response appeared to of LRO4, an MAA-specifi c mouse NAb ( 37 ), and such apoptotic cells are proinfl ammatory if not promptly cleared ( 38 ). Indeed, as shown in Fig. 10C , MAA is a prominent target of human IgM NAbs in newborn cord blood. We have cloned a panel of MAA-specifi c NAbs from a phage display library prepared from lymphocytes isolated from the cord blood, and remarkably, each of these MAA-specifi c NAbs bind apoptotic cells (data not shown), observations similar to those recently reported by Wang and colleagues ( 30 ). All these data suggest that MAA-epitopes are prominent OSEs that mediate proinfl ammatory events and play a major role in atherogenesis, as well as other infl ammatory diseases ( 26,39 ). Such natural selection of PPRs infers that immune responses to MAA, a DAMP, confers overall benefi t and thus is desirable. Our current data are thus consistent with the concept that MAAs are indeed immunodominant epitopes of MDA-type modifi cations, and that targeting them by induction of adaptive responses, as occurs with this immunization strategy, should not only provide a degree of atheroprotection, but will likely also provide overall benefi cial effects.

Fourth: immunization with MDA/MAA confers atheroprotection via both humoral and cellular mechanisms
Immunizations with MDA-LDL and MDA-MSA, as well as MAA-MSA, reduced atherosclerosis equally, compared with PBS ( Fig. 3 and Table 1 ). However, among these, MAA-MSA appeared to provide better atheroprotection, as this was also statistically different from the FA-as well as PA-MSA-immunized mice, and among all the immunogens, it was the only one that provided atheroprotection when examined at the aortic root ( Table 1 ). It should be realized that the intervention in these mice only began when the mice were already 28 weeks old and already had atherosclerotic lesions, and thus, the bar to show effective reduction of progression was fairly high. We chose this strategy as this might be the situation in adult humans who likely already have existing lesions. Even under these circumstances, there was substantial reduction in progression over a prolonged 28-week course of HC feeding.
Although these studies do not provide a defi nitive answer as to the mechanisms by which this immunization strategy provided atheroprotection, we think it is likely a combination of induced changes in both humoral and cellular immunity.
We have previously shown that many Abs to OSEs have the capacity to block the binding and uptake of OxLDL to macrophages. This is true for the atheroprotective IgM NAb E06 ( 20 ), as well as the human Fab Ab IK17, which binds to OxLDL as well as both MDA-LDL and MAA-LDL ( 21 ). Plasma from these mice strongly inhibited the binding of Ox-LDL to macrophages in parallel with the atheroprotection. In the latter study, IK17 was injected intravenously and thus directly demonstrates the potential of such an Ab to inhibit atherogenesis. Further, because the IK17 Fab lacks the Fc effector portion of Abs, this strongly suggests that IK17's binding and blocking ability alone is suffi cient to provide atheroprotection. In the current work, antisera from MAA-immunized mice were the most effective in inhibiting macrophage ( 3,14,40,41,47,48 ). We emphasize that our approach is quite different in that the principle of the immunization strategy here is to generate an active humoral and cellular response to OSEs (e.g., to haptens of proteins generated by adduct formation between products of lipid peroxidation and various proteins, including apoB of OxLDL). The principle of this strategy is to generate humoral and cellular responses that effectively provide protective immunity against OSEs, in much the same way that immunization to pathogenic epitopes helps prevent and/or ameliorate infectious diseases. However, it is possible that the two basic strategies are related in that the generation of OSEs on apoB, for example, might well lead to the formation of conformational epitopes on apoB at a distance from the oxidized lipid-apoB adduct, to which an adaptive response is initiated that has proinfl ammatory and atherogenic consequences. Thus, theoretically, enhanced adaptive responses to the OSEs might indirectly lead to a reduction in the formation of such conformational epitopes. Obviously, a great deal is left to learn about these complex immune responses and their relationship to both enhancement and amelioration of infl ammation and atherogenesis.

Summary
Our report demonstrates the feasibility of development of a small-molecule immunogen that could stimulate the atheroprotective properties of MAA-specifi c Abs and their associated cellular immune responses, which could eventually lead to the generation of a vaccine that could be used to retard or even prevent the development of atherosclerosis.
dampen IFN-␥ release and led to increases in IL-5, which led to expansion of atheroprotective NAbs such as E06. (As discussed subsequently, the Th2 responses by themselves were not atheroprotective.) At the same time, we recognize that many questions remain to fully understand the mechanisms by which such immunizations are atheroprotective. For example, what is the fate of the OxLDL bearing OSE to which the induced IgG1 binds? Are they cleared by the reticuloendothelial system outside the artery? Our previous studies with immune-mediated clearance of nonenzymatically glycated LDL (glyLDL) may be of relevance. We previously showed that LDL in euglycemic humans, and to a greater extent in diabetics, is nonenzymatically glycated. We further showed that there is a prominent hapten-specifi c IgG response to the glucitollysine adduct formed, which leads to enhanced plasma clearance of glyLDL ( 43,44 ). To determine the fate of the glyLDL bound by anti-glucitollysine Abs, we studied the tissue sites of degradation of apoB of glyLDL in rabbits immunized with homologous glyLDL. The high plasma IgG titers to glyLDL redirected degradation of the modifi ed LDL primarily to the reticuloendothelial cells of the liver and other tissues, and consequently, the net deposition in the artery was reduced ( 45 ). In a similar manner, it is possible that LDLs bearing OSEs are similarly cleared and contribute to the atheroprotective effects noted, but future studies will be needed in this setting to test this speculation.

Fifth: nonspecifi c IgG1-Th2-biased immune responses per se are not atheroprotective
It has been suggested that a Th2-biased response per se might be atheroprotective independent of the immunogen ( 46 ). To study this, we generated a similar, Th2-biased response to a lipid peroxidation-derived adduct, but one that we did not expect to be abundant in vivo, nor relevant to atherogenesis. We derivatized MSA with PA because it is a lipid peroxidation product of n-3 polyunsaturated fatty acids and a small 3-carbon aldehyde like MDA but contains only one aldehyde group, resulting in simpler chemistry. PA modifi es lysines via the Schiff base reaction and thus forms PA-lysine adducts ( Fig. 1 ). In this study, immunization with PA-MSA conferred no atheroprotective benefi t, even though it generated a pronounced IgG1, Th2-biased humoral response ( Fig. 4 ), though the cellular Th2 response ( Fig. 9D ) was quite weak. The PA-MSA antisera did immunostain atherosclerotic tissue to some extent, but it was unable to inhibit binding of MAA-LDL to macrophages ( Fig. 6 ). Notably, PA-LDL did not appear to be proinfl ammatory as indicated by the failure to stimulate MIP-2 release when incubated with macrophages (supplementary Fig. I). These data suggest that an irrelevant IgG1, Th2-biased humoral immune response per se is not atheroprotective.

Sixth: implications regarding autoimmunity to apoB
Recently, a number of investigators have reported a variety of immunization strategies to essentially abrogate presumed immune responses to epitopes of native apoB