In vivo efficacy of HDL-like nanolipid particles containing multivalent peptide mimetics of apolipoprotein A-I.

We have observed that molecular constructs based on multiple apoA-I mimetic peptides attached to a branched scaffold display promising anti-atherosclerosis functions in vitro. Building on these promising results, we now describe chronic in vivo studies to assess anti-atherosclerotic efficacy of HDL-like nanoparticles assembled from a trimeric construct, administered over 10 weeks either ip or orally to LDL receptor-null mice. When dosed ip, the trimer-based nanolipids markedly reduced plasma LDL-cholesterol levels by 40%, unlike many other apoA-I mimetic peptides, and were substantially atheroprotective. Surprisingly, these nanoparticles were also effective when administered orally at a dose of 75 mg/kg, despite the peptide construct being composed of l-amino acids and being undetectable in the plasma. The orally administered nanoparticles reduced whole aorta lesion areas by 55% and aortic sinus lesion volumes by 71%. Reductions in plasma cholesterol were due to the loss of non-HDL lipoproteins, while plasma HDL-cholesterol levels were increased. At a 10-fold lower oral dose, the nanoparticles were marginally effective in reducing atherosclerotic lesions. Intriguingly, analogous results were obtained with nanolipids of the corresponding monomeric peptide. These nanolipid formulations provide an avenue for developing orally efficacious therapeutic agents to manage atherosclerosis.

apoA-I mimetics previously reported (and administered parenterally) ( 21 ). Oral effi cacy was observed in spite of the peptide segments being comprised of L -amino acids, which would make them vulnerable to rapid proteolytic degradation. This nanolipid material, containing a multivalent peptide, provides a new approach for the development of orally effi cacious agents to manage atherosclerosis.

Synthesis of multivalent constructs
All peptide native-ligation reactions involved a 1.5-fold excess of purifi ed peptide relative to the number of thioesters in the scaffold. Ligations were performed in 200 mM MOPS buffer containing 7 M guanidine hydrochloride (Gdn·HCl), 100 mM tris(2carboxyethyl)phosphine (TCEP), pH 7.5, at room temperature for 6-12 h. At completion of the reaction, iodoacetamide ( ‫ف‬ 50fold excess relative to Cys) was added to cap the free thiol moieties on Cys residues. After 5 min, CF 3 CO 2 H was added to quench the reaction, and the product was purifi ed by reverse-phase HPLC. For detailed synthetic procedures for the peptide constructs, see our prior report ( 31 ).

Preparation of vesicles and peptide-lipid nanoparticles
( R )-(+)-1,2-dimyristoyl-sn -glycero-3-phosphocholine (DMPC) was dissolved in 0.5 ml of CHCl 3 in a test tube and dried into a thin fi lm by blowing N 2 gas into the tube while vortexing. The lipid fi lm was further dried under reduced pressure overnight in a desiccator. Multilamellar vesicles (MLVs) were obtained by suspending the dried lipids into PBS [10 mM phosphate (pH 7.4), 136 mM NaCl] with vortexing and/or sonication. MLVs were typically prepared at concentrations of 10 mM. To prepare peptidelipid nanoparticles, a stock solution of peptide in PBS was added to 10 mM MLVs at a 1:10 (helix:lipid) molar ratio, and the solutions were vigorously stirred for 24 h at 22°C. The DMPC unilamellar vesicles (ULVs) (18 mM) were made by extrusion DMPC MLVs through 0.2 m Nucleopore track-etched membranes (Whatman) in the Avanti mini extruder, and sterile fi ltered before ip administration to mice.

In vivo effi cacy
All procedures involving live animals were approved by the Scripps Research Institute Institutional Animal Care and Use Committee . LDLr Ϫ / Ϫ mice were fed a chow diet until they were 10 weeks old, when they were switched to a high-fat diet (HFD) (Harlan Teklad 94059). At the time that the HFD was started, peptide/DMPC nanoparticles were administered by daily ip injection or by the oral route ad libitum in the drinking water for 10 weeks in the continued presence of the HFD. Mice receiving ip injections of PBS or DMPC ULVs served as controls for the ip groups. Mice receiving drinking water containing 1% sucrose/PBS or DMPC MLVs served as controls for the oral groups. The mice were bled after an overnight fast ( ‫ف‬ 15 h) after 2 weeks of treatment and at the time of harvest (10 weeks); the plasma was used to determine lipoprotein profi les, biomarkers including total cholesterol levels, triglycerides, plasma serum amyloid A (SAA), plasma 15( S )-HETE levels, and plasma alanine aminotransferase (ALT) and aspartate aminotransferase (AST) concentrations.
Atherosclerosis in the aorta from the proximal ascending aorta to the bifurcation of the iliac artery was assessed as described previously ( 40 ). Briefl y, the dissected aorta was pinned fl at on black wax, stained with Sudan IV, and digitally photographed at a are not conducive to traditional medicinal chemistry manipulations to optimize HDL-like function. Thus, there has been considerable interest in employing a design approach that involves shorter peptides that mimic apoA-I ( 21 ). From numerous studies on apoA-I mimetic peptides based on one or two amphiphilic ␣ -helical segments (22)(23)(24)(25)(26), a wide range of peptide sequences have been found that are more-or-less effective in various aspects of apoA-I mimicry. Many of these peptides have no sequence homology to apoA-I, as they capitalize on an amphiphilic ␣helical structural motif, and some are composed of all D -amino acids ( 21 ). Although in vivo effi cacy has been demonstrated for certain peptides in animal models of atherosclerosis, their mechanisms of action remain the subject of active research (27)(28)(29)(30)(31)(32)(33). A leading hypothesis is that the peptides stimulate reverse cholesterol transport (RCT), in a manner similar to apoA-I, by promoting cholesterol effl ux from macrophage cells. Available evidence suggests that the peptides bind to lipoproteins (especially HDLs) in vivo (29)(30)(31), and work in concert with endogenous apoA-I to improve the function of HDL ( 32,33 ). Plasma total cholesterol lowering does not appear to be a general or primary mechanism, because some mimetic peptides that were shown to be atheroprotective in animals did not signifi cantly affect plasma lipid levels ( 34,35 ); further, two mimetic peptides with similar cholesterolreducing properties had different atheroprotective effects ( 36 ). Other proposed mechanisms for these peptides include anti-infl ammatory and anti-oxidant effects, as well as the binding of oxidized or pro-atherogenic lipids, such as lysophosphatidic acid (LPA), in the plasma or intestine (37)(38)(39).
We recently reported on the development of branched multivalent apoA-I mimetic constructs, in which multiple copies of a modifi ed ␣ -helical segment derived from apoA-I were appended to a scaffold ( 31 ). In spite of their abiotic structures and distinctly different molecular topology from the linear arrangement of helical segments in the native protein, these branched, multivalent constructs bound lipids to generate HDL-like nanoparticles, exchanged into native HDLs in human plasma, promoted cellular cholesterol effl ux, and induced remodeling of large mature HDLs to smaller lipid-poor (pre-␤ ) particles. The lipid nanoparticles derived from the multivalent constructs were superior to those from either the corresponding monomeric parent peptide or a monomeric 4F reference peptide in all functional aspects tested ( 31 ). The multivalent peptides showed impressive stability toward proteolytic digestion and had long plasma half-lives in mice. As such, we hypothesized that our multivalent apoA-I mimetics would exhibit improved atheroprotection compared with the parent monomeric peptide. In this report, we describe results of chronic in vivo effi cacy studies in LDL receptor-null (LDLr Ϫ / Ϫ ) mice, a standard mouse model of atherosclerosis, for nanoparticle formulations of the trimeric peptide construct and the parent monomeric peptide. Administration of a nanolipid formulation of the trimeric construct, either ip or po, reduced plasma cholesterol and atherosclerotic lesions in the mice to a degree comparable with the most effective (SEC) using disposable Sephadex G-25 columns (illustra NAP-25 columns, GE Healthcare). Solutions of peptides and purifi ed peptide-DMPC nanoparticles were further concentrated using Amicon centrifugal fi lters (Millipore) with 3 K and 10 K molecular weight cut-offs, respectively. Neomycin and cholestyramine were used as positive control compounds, as these compounds are known to inhibit intestinal cholesterol absorption by precipitating micellar lipids and sequestrating bile acids, respectively (41)(42)(43)(44).
Artifi cial micelles were prepared according to a previously published method with minor modifi cations ( 45,46 ). Briefl y, lipids [0.5 mM cholesterol, 1 mM oleic acid, and 2.4 mM phosphatidylcholine (from egg yolk)] were dissolved in methanol in a test tube and dried into a thin fi lm by blowing N 2 gas into the tube while vortexing. The lipid fi lm was further dried under reduced pressure overnight in a desiccator before adding PBS [10 mM phosphate (pH 7.4), 136 mM NaCl] containing 6.6 mM taurocholate salt. The suspension was sonicated for 1 h and shaken overnight at 37°C. The micelle solution was fi ltered through a 0.2 m syringe fi lter. Each compound, or PBS as negative control, was added to the micelle solution at a volume ratio of 1:4, and incubated for 1 h at 37°C. The solution was then centrifuged at 14,000 rpm for 20 min. The supernatant was collected for the determination of cholesterol and bile acid concentrations by using a cholesterol Amplex assay kit (Invitrogen) and a total bile acids assay kit (Diazyme), respectively.

Characterization of gut fl ora by quantitative real-time PCR
Quantitative real-time PCR (qPCR) was performed with universal and group-specifi c bacterial primers (see supplementary  Table I for primer sequences) targeting the 16S rRNA genes of two major intestinal bacterial phyla, Bacteroidetes and Firmicutes (47)(48)(49). Bacterial genomic DNA was extracted from snap-frozen cecum and colon contents by using FastDNA Spin kit feces (MP Bio) according to the manufacturer's instructions. The abundance of total and specifi c intestinal bacterial groups (Bacteroidetes, Firmicutes) was measured by qPCR by using universal and group-specifi c 16S rRNA gene primers (supplementary Table I) and the iTaq Universal SYBR Green Supermix (Bio-Rad). We performed qPCR on a CFX Connect real-time PCR detection system (Bio-Rad). Bacterial DNA was quantifi ed using standard curves constructed with reference bacteria (supplementary Table I) specifi c for each bacterial group analyzed.

Statistical analysis
Data are expressed as the mean ± SD. The difference between groups was assessed by using the one-way ANOVA with a post hoc Tukey-Kramer test, with Prism software (version 5.0d). We considered P < 0.05 as statistically signifi cant.

Peptide agents and HDL-like nanoparticles
The amino acid sequence used in our study ( 31 ) was a conservatively modifi ed derivative of helix 10 of human apoA-I (residues 221 Ϫ 241), which is required in the native protein for lipid binding and cholesterol effl ux. A Cys-Gly dipeptide was appended to the N terminus to allow for synthesis of the multivalent constructs via native chemical ligation. We made two amino acid substitutions by replacing Val-227 with Ala to increase amphiphilicity and Tyr-236 fi xed magnifi cation. Total aortic areas and atherosclerotic lesion areas were calculated by using Adobe Photoshop CS4, Chromatica V, and National Institutes of Health Scion Image software. Results are reported as lesion area as a percentage of total en face aortic area. As a second assessment of atherosclerosis, lesions of the aortic root (heart sinus) were analyzed. Briefl y, hearts were fi xed, frozen, and sectioned on a Leica cryostat. For each aortic sinus cusp, sections were collected from the beginning of the sinus for a distance of 500 m into the sinus. Sections (10 m thick) were stained with oil red O and counterstained with Gill hematoxylin 1 (Fisher). Stained sections were photographed and digitized. Lesion volume in the fi rst 500 m of each cusp was estimated from four sections spaced at 140 m. Lesion volume was calculated from an integration of the measured cross-sectional areas.

Female LDLr
Ϫ / Ϫ mice were administered peptide/DMPC for 10 weeks as described above. On weeks 2 and 6, the monomer and trimer concentration in the plasma of randomly selected mice was measured 2 h and 4 h after ip administration, respectively, by LC-MS in the selected ion monitoring (SIM) mode. These times correspond to the respective t max values for the peptide agents, as determined previously ( 31 ). Details of LC-MS SIM quantifi cation methods are described in our previously published study ( 31 ).

Plasma biomarker (cholesterol, triglyceride, SAA, and 15-HETE) analysis
Plasma total cholesterol and triglyceride were determined using a cholesterol Amplex assay kit (Invitrogen) and a triglyceride quantifi cation kit (BioVision), respectively. Plasma SAA and 15-HETE levels were measured by using a SAA Mouse ELISA kit (Life Technologies) and a 15( S )-HETE EIA kit (Cayman), respectively, following the manufacturer's instructions.

Plasma liver enzyme activities
At the time of harvest (10 weeks), blood ( ‫ف‬ 0.5 ml) was collected by cardiac puncture into EDTA anti-coagulant-coated tubes, and centrifuged at 4°C for 10 min at 5,000 rpm. Plasma samples were stored at Ϫ 80°C until analysis was conducted. Plasma ALT and AST concentrations were measured using Infi nity ALT (GPT) and AST (GOT) liquid stable reagents, respectively, which is a colorimetric kinetic assay (Thermo Scientifi c). Assays were performed in accordance with the manufacturer's recommendations, adjusting the reagent volumes (20 l plasma + 200 l reagent, 0.69 cm light pathlength of the solution in the well) for analysis of samples in 96-well fl at bottom microplate format.

Measurement of liver cholesterol and triglyceride levels
Lipid extracts of liver tissue were assayed for cholesterol and triglyceride according to the manufacturer's protocols using the cholesterol Amplex assay kit (Invitrogen) and the triglyceride quantifi cation kit (BioVision), respectively. Briefl y, liver tissue was homogenized in 5% NP-40 in water (1:15 w/v). Samples were slowly heated to 80°C for 10 min. Insoluble materials were removed by centrifugation (13,000 rpm, 10 min). Cholesterol and triglyceride concentrations in the supernatant were determined by the enzyme based fl uorometric assays.

Artifi cial micelle precipitation assay
Peptides and peptide-DMPC nanoparticles were tested for interference on cholesterol solubility in artifi cially prepared micelles. Peptide-DMPC nanoparticles were prepared as described above, and purifi ed by size-exclusion chromatography Encouraged by these results and the impressive in vitro proteolytic stability of the trimer nanoparticles ( 31 ), we carried out further cholesterol-reduction studies by using an oral route of administration. The peptide/DMPC nanoparticles were provided ad libitum in the drinking water, by dissolving them in PBS at concentrations such that each mouse would receive on average a 75 mg/kg dose of peptide over the course of 1 day. Fresh water was prepared every 2 days, and we confi rmed by HPLC analysis that the peptide materials did not degrade in PBS over this time period. After 2 weeks of oral administration, the trimer nanoparticles (n = 10) reduced plasma total cholesterol levels by 43%, compared with the PBS control (n = 15) ( Fig. 2A ). As with the ip administration, SEC lipoprotein fractionation indicated that VLDL and LDL, but not HDL, levels were reduced in the animals ( Fig. 2B ). On the contrary, peptide treatment caused an increase in observed HDL-cholesterol levels compared with the PBS control ( Fig. 2B , inset). An additional control group administered DMPC liposomes in the drinking water (n = 8) had unchanged plasma total cholesterol levels compared with the PBS group ( Fig. 2 ). At a 10-fold lower oral dose ( ‫ف‬ 7.5 mg/kg/day) over a 2 week period, the trimer nanoparticles (n = 7) reduced total plasma cholesterol by 19% compared with the PBS control (n = 15) ( Fig. 2 ).
A nanolipid formulation of the corresponding monomeric peptide was also studied according to the above in vivo protocol. Two weeks of daily 40 mg/kg ip injections of monomer nanoparticles (n = 5) reduced plasma total cholesterol levels by a surprising 30%, compared with the PBS control (n = 5 or 7) (supplementary Fig. I), despite having an inferior plasma residency compared with the trimer (area under the curve = 110 ± 10 and 500 ± 40 M·h for monomer and trimer, respectively) ( 31 ). Even more surprising, oral administration of the monomer nanoparticles (n = 10) for 2 weeks reduced plasma total cholesterol levels by 32% compared with the PBS control (n = 15) ( Fig. 2A ). It is remarkable that oral administration of the monomer formulations effectively diminished cholesterol levels in vivo, considering that the simple linear peptide synthesized from all L -amino acids would be expected to undergo rapid digestion in the gut. Plasma pharmacokinetic (PK) levels of both the monomer and trimer peptide were below the limit of detection (0.5 and 0.2 M for monomer and trimer, respectively) during the oral administration studies.

In vivo reduction of atherosclerotic lesions
The results in our cholesterol-reduction studies suggested that the trimer and monomer nanolipids would have potential as atheroprotective agents. Thus, we sought to establish the effect of the peptide/DMPC nanoparticles on the development of atherosclerotic plaques in the LDLr Ϫ / Ϫ mice. Groups of animals were treated for 10 weeks via daily ip injections ( ‫ف‬ 40 mg/kg) or oral administration ad libitum in the drinking water with nanolipid formulations of trimer and monomer ( ‫ف‬ 75 mg/kg/day). The 40 mg/kg value for the ip route is an average daily with Trp for improved spectrophotometric properties. The 23-residue peptide, Cys-Gly-Val-Leu-Glu-Ser-Phe-Lys-Ala-Ser-Phe-Leu-Ser-Ala-Leu-Glu-Glu-Trp-Thr-Lys-Lys-Leu-Gln-CONH 2 ( Fig. 1 ), is an amphiphilic ␣ -helix with an overall neutral charge at physiological pH. The trimeric peptide construct ("trimer") comprises three copies of this 23-mer, each attached by the peptide's N terminus to a small organic scaffold, with the Cys sulfur groups capped by CH 2 C(O)NH 2 . For this peptide agent, as well as its corresponding monomer, to serve as a reference, HDL-like nanoparticles were prepared by incubating the peptide with DMPC MLVs at 22°C for 12-16 h, which produced peptide-lipid nanoparticles in the size range of 10-15 nm. All of the in vivo effi cacy experiments reported herein were conducted by using such a peptide-lipid nanoparticle formulation.

In vivo reduction of plasma cholesterol
To determine the effi cacy of the peptide-containing nanolipids in an animal model of atherosclerosis, we carried out a series of atheroprotection studies with the LDLr Ϫ / Ϫ mouse model, which has a human-like lipoprotein profi le and is a widely used animal model for this purpose ( 50,51 ). The mice were fed a chow diet until ‫ف‬ 10 weeks of age, at which time they were switched to a HFD (1.25% cholesterol, 15.8% fat, and no cholate), whence treatments with the peptide/DMPC nanoparticles were commenced. After 2 weeks of daily 40 mg/kg ip injections, treatment with the trimer nanoparticles (n = 5 or 8) reduced plasma total cholesterol levels by 30-40% compared with the PBS control (n = 5 or 7), as reported in our prior publication ( 31 ) (supplementary Fig. I). When the mice dosed for 2 weeks with the trimer nanoparticles were kept for an additional 2 weeks without any further treatment ("wash out"), their plasma cholesterol levels rose to match those of the PBS control group (supplementary Fig. IA). Thus, the 23-mer peptide construct had just a short-term reversible effect on plasma cholesterol levels. As determined by SEC fractionation of pooled plasma samples, the reductions in plasma total cholesterol stemmed mainly from reduced levels of VLDLs and LDLs (supplementary Fig. IC). 2 h, respectively, after ip administration ( Fig. 4 ). These times correspond to the respective t max values for the peptide agents, as determined previously ( 31 ). The observed peptide concentrations on weeks 2 and 6 were 22 ± 6 M (n = 5) and 22 ± 3 M (n = 10) for the trimer and 12 ± 4 M (n = 5) and 12 ± 3 M (n = 10) for the monomer, respectively; these values were consistent with our earlier PK studies ( 31 ). At the completion of the study, whole aorta atherosclerotic lesions were reduced in the ip cohort by 49% for the monomer nanoparticles (n = 12) and by 55% for the trimer nanoparticles (n = 15) compared with the PBS control (n = 10) ( No signifi cant differences were observed between the four groups regarding food intake, water intake, spleen weight, liver weight, whole animal weight (supplementary Fig. IV), or plasma liver enzyme levels (supplementary Fig. V), consistent with an absence of toxicity.
In the oral cohort, the trimer/DMPC nanoparticles (n = 10; ‫ف‬ 75 mg/kg/day) reduced the development of whole aorta atherosclerotic lesions by 58% compared with the PBS control (n = 18) and by 55% compared with the DMPC liposome control (n = 8) ( . Surprisingly, the monomer/DMPC nanoparticles (n = 10; ‫ف‬ 75 mg/kg/ day) likewise reduced whole aorta and aortic sinus lesions compared with the PBS and DMPC liposome controls by 49-74% ( Fig. 6 , supplementary Figs. VI, VII). No signifi cant differences were observed between the four groups regarding food intake, water intake, spleen weight, whole animal weight, or plasma liver enzyme levels; however, liver weights were increased in the orally dosed peptide-treated groups (supplementary Figs. VIII, IX). To determine whether the increased liver weights were due to an accumulation of lipids in the liver, we measured cholesterol and triglyceride levels in liver homogenates from the mice. Interestingly, the levels of both lipids were reduced in the peptide-treated animals compared with the controls (supplementary Fig. X). Plasma PK levels of both the monomer and trimer peptide were below the limit of detection (0.5 and 0.2 M for monomer and trimer, respectively) during these studies.
To further defi ne the oral effi cacy of the trimeric construct, we carried out a 10 week drinking water study at a 10fold lower dose (n = 8; ‫ف‬ 7.5 mg/kg/day). In this group, 2 week plasma cholesterol levels were reduced by 24% ( Fig. 2 ), 10 week whole aorta lesion areas were not signifi cantly changed ( Fig. 6 ), and 10 week aortic sinus lesion volumes were reduced by 50% compared with the controls ( Fig. 6 ). In the low-dose trimer group, no signifi cant differences were dose over the 10 week period, accounting for dosing the animals with a constant daily quantity of peptide (not adjusted for mouse body weight). A schematic diagram of the study design is presented in Fig. 3 . On weeks 2 and 6 of the study, the trimer and monomer concentration in the plasma of randomly selected mice was measured 4 h or

DISCUSSION
We have previously conducted research to systematically evaluate the importance of multivalency in the function of synthetic apoA-I-like materials, especially their effects on native HDL particle distribution, cholesterol disposition, and in vivo stability ( 31 ). Our positive results with certain multivalent peptide constructs encouraged us to undertake in vivo effi cacy studies in LDLr Ϫ / Ϫ mice, a standard animal observed from controls regarding food intake, water intake, spleen weight, whole animal weight, or liver weight (supplementary Fig. VIII).

Analysis of plasma biomarkers from chronic effi cacy studies
To explore the mechanism of action of the peptides, we determined the levels of several plasma markers taken after 10 weeks of treatment. Interestingly, plasma cholesterol and triglyceride levels were reduced by 10 weeks of ip peptide nanoparticle administration, but no signifi cant differences were observed for these lipids in the orally treated animals ( Fig. 7 ), despite both routes reducing the development of atherosclerotic lesions. Thus, both routes of administration initially reduced plasma cholesterol levels (after 2 weeks of treatment) ( Fig. 2 , supplementary Fig. I), but by 10 weeks of treatment, the cholesterol levels in only the ip group remained depressed relative to controls. The plasma SAA level, which is an indication of systemic infl ammation, was signifi cantly reduced by oral administration of both the peptide nanoparticles ( Fig. 7 ). In the ip treatment regime, the monomer nanoparticles signifi cantly reduced the plasma SAA level compared with DMPC, but the reduction brought about by the trimer nanoparticles was not statistically signifi cant (supplementary Fig. XI). The level of plasma 15( S )-HETE, an oxidized metabolite of arachadonic acid, was not affected by either ip or oral administration of the peptide nanoparticles ( Fig. 7 , supplementary Fig. XI).

Fig. 3. Schedule of in vivo studies. Female LDLr
Ϫ / Ϫ mice were used as the animal model. We had observed enhanced proteolytic and in vivo stability for the trimer construct, which would be expected to provide a functional advantage over the monomer ( 31 ). Nevertheless, we decided to conduct parallel studies with both constructs to have a basis for comparison. This experimental approach required scaling up the synthesis of the trimer and monomer to supply gram quantities for the chronic in vivo effi cacy studies. We were surprised to fi nd that both peptides, administered ip as nanolipid formulations, were similarly effi cacious in lowering plasma cholesterol and reducing the development of atherosclerosis. Remarkably, we also found that the nanoparticles of both peptides were effective when administered orally ad libitum in drinking water, despite having undetectable plasma concentrations. Thus, plasma level and duration are apparently not important determinants of anti-atherogenic activity for these peptides, at least under the experimental conditions employed. These fi ndings point to the possibility of two different in vivo mechanisms of action, one based in the plasma compartment and one based in the gastrointestinal (GI) tract. In this regard, recent studies ( 37,38,52,53 ) involving the 4F and 6F apoA-I mimetic peptides suggest that the intestine may be a key site of action.
An important observation of our studies is that the multivalent construct and monovalent peptide, which contain a peptide sequence that is completely unrelated to the apoA-I mimetic 4F peptide, behave so similarly to 4F in terms of in vivo effi cacy in the absence of measurable plasma concentrations. Our fi ndings provide further evidence  about a lower reduction in total cholesterol levels. Likewise, with the 4F peptide, the two highest oral doses improved HDL infl ammatory index in humans in spite of low plasma levels, whereas lower doses did not improve HDL function ( 38,54 ). On the other hand, there appear to be some important differences in the function of our peptide agents compared with other apoA-I mimetic peptides. First, the present peptide agents cause more substantial reductions in plasma cholesterol levels than are typically seen for apoA-I mimetics. Compared with DMPC, the 10 week daily 40 mg/kg ip monomer and trimer nanoparticles reduced plasma cholesterol levels by 41 and 28%, respectively ( Fig. 7 ), while cholesterol levels were unchanged by daily 8 week administration of D4F at 45 mg/kg po ( 38 ), or daily 6 week administration of ATI-5261 at 30 mg/kg ip ( 55 ), and just modestly reduced (9-14%) by daily 3 week or 6 week administration of mR18L at 50 mg/kg po ( 56 ). Plasma cholesterol levels were reduced 20% by tomatoes transgenic for 6F with daily dosing at 45 mg/kg po over 13 weeks ( 52 ). Furthermore, infusions of rHDLs or apoA-I failed to lower total cholesterol levels ( 6,57,58 ). Lipoprotein analyses ( Fig. 2 , supplementary Fig. I) indicated that the cholesterol reduction promoted by our peptides was from non-HDL particles, while HDL cholesterol levels were either unchanged or increased by peptide treatment. Our previous in vitro studies in human plasma indicated that these peptides bound to HDLs, and while they did not appear to bind larger lipoprotein particles, we did not rigorously rule out their binding to larger LDL or VLDL particles ( 31 ). Therefore, the mechanism of LDL-and VLDL-cholesterol lowering may not require direct binding of the peptide material to those lipoproteins. Some or all of the observed reductions in atherosclerotic lesions may result from plasma lipid lowering. However, it is diffi cult to reconcile this hypothesis with the intriguing observation that the oral route of administration was similarly atheroprotective to the ip route, despite cholesterol and triglyceride levels only being lowered at the early (2 week) time point for the orally dosed groups. We are working to understand the different time courses of lipid lowering promoted by the peptide agents when administered by different routes. A related question is whether other apoA-I mimetics, when given orally, lower LDL-cholesterol levels in the initial few weeks of administration but not over longer time periods.
A second difference between the peptides described here and the 4F/6F peptides is their disparate modulatory effect on oxidized fatty acids in vivo. Recently emerging data for 4F/6F have pointed to a mechanism of atheroprotection in which the peptides modulate the level of oxidized phospholipids and fatty acids to reduce infl ammation ( 37,53,59 ). Consistent with this hypothesis, these peptides were found to reduce markers of systemic infl ammation, such as SAA, and the levels of oxidized lipids and fatty acids, such as 15-HETE, in the intestines and plasma ( 37,53,59 ). Although our monomer and trimer peptide nanoparticles did reduce the plasma level of SAA (both peptides reduced SAA levels by oral administration, while for the viewpoint that the effectiveness of apoA-I mimetic peptides is not dependent on their plasma concentration ( 38 ). Additionally, our results refl ect dose-dependent in vivo effi cacy despite the low or unobserved plasma concentrations, with higher peptide doses being more effective than lower doses. The 75 mg/kg/day oral dose of the trimer nanoparticles was effective in reducing both aortic root and en face lesion development, but the 10-fold lower dose was only effective in the aortic root and brought In all panels, individual data points are shown along with the mean ± SD; ns, not signifi cantly different; * P р 0.05; ** P р 0.01; *** P р 0.001 by one-way ANOVA, post hoc Tukey-Kramer test.
PBS controls. There are some previous indications that DMPC can reduce the development of atherosclerosis in vivo ( 62,63 ). In view of these data, it is possible that the lipid component of the nanoparticles is contributing somewhat to the observed atheroprotective effects. However, these contributions would appear to be minimal, based on the absence of effects in the aortic root upon ip administration and a lack of any effects with oral administration. Liver weights increased in the peptide-treated animals (supplementary Figs. IV, VIII), without an associated increase in plasma liver enzymes (supplementary Figs. V, IX). In light of the reduced plasma total cholesterol levels in these animals, we determined lipid levels in liver tissue samples to test the hypothesis that excess cholesterol had been removed from plasma and deposited in the liver. Instead, we found that cholesterol and triglyceride levels were lower in the livers of treated animals (supplementary Fig. X). Enlarged livers can be a sign of infl ammation; however, the treated animals exhibited generally reduced levels of plasma SAA, which indicates lower systemic infl ammation. It remains unclear why liver size was increased by treatment with the peptide nanoparticles.
It is important to understand the possible mechanisms behind the cholesterol lowering and atheroprotection. Previously, we showed that our peptides promote remodeling of HDLs to lipid-poor particles in vitro and in vivo (parenteral dosing), and induce cellular cholesterol effl ux in vitro ( 31 ). Therefore, the mechanism of anti-atherogenicity could involve the promotion of RCT, at least in part. However, in the case of oral administration, it is diffi cult to reconcile a RCT mechanism operating in the vascular system with the virtually undetectable plasma concentrations of the peptides. Further, the monomeric peptide was 3-fold less effective than the trimeric construct in promoting cholesterol effl ux in vitro ( 31 ), but both agents were similarly effi cacious in vivo in preventing atherosclerosis. An alternative consideration is that the peptides act in the intestine, as suggested previously for the 4F and 6F peptides ( 37,38,52,53 ), and one can propose several possible mechanisms. A fi rst possibility is that the peptide nanoparticles are remodeling the maladaptive composition of the gut microbiota associated with a HFD (64)(65)(66)(67)(68)(69)(70)(71)(72)(73) to yield a rebalanced bacterial community that would ultimately reduce the progression of atherosclerosis. To test this hypothesis, we measured by qPCR the number of 16S rRNA gene copies ( 47-49 ) for a number of representative gut bacterial strains in both the vehicle-treated and peptide nanoparticle-treated animals (supplementary Table I). No signifi cant differences were observed between groups (supplementary Fig. XII), indicating that modulation of the microbiota is unlikely to be a meaningful cause. A second possibility is that the peptide nanoparticles prevent the absorption of dietary cholesterol, such as by precipitating cholesterol or bile acids in the gut ( 74 ). To explore this hypothesis, albeit indirectly, we prepared synthetic micelles containing cholesterol and bile acids designed to mimic the micelles found in the intestine ( 45,46 ), and mixed these synthetic micelles with the peptide nanoparticles. However, we did not observe any only the monomer caused a reduction by ip injection), plasma 15( S )-HETE was not affected ( Fig. 7 ). Future mechanistic studies will be needed to establish the affi nity and potential modulatory effects of our peptides on oxidized phospholipids or fatty acids, to clarify the similarities and differences in their function with that of other apoA-I mimetic peptides.
It is a striking result that our trimer and monomer peptides, synthesized from L -amino acids, exhibited marked in vivo effi cacy when administered orally. Especially surprising was the frank anti-atherosclerotic effi cacy of the monomeric peptide, because it is not resistant to degradation, unlike the trimer ( 31 ). Basically, the monomer, as a simple linear peptide comprised of native L -amino acids, would not be expected to survive in the GI environment. However, there are previous suggestions from the literature that apoA-I mimetic peptides comprised of L -amino acids can exert biological effects on oral administration. The 4F peptide, when administered orally together with niclosamide as a means of protecting the peptide from degradation, improved the HDL-infl ammatory index in apoE-null mice ( 60 ). Even in the absence of niclosamide, a relatively high dose of 4F (100 mg/kg) administered to mice in chow signifi cantly decreased plasma LPA levels ( 61 ). Most recently, the 6F peptide, produced in transgenic tomatoes and fed to LDLr Ϫ / Ϫ mice fed a HFD, resulted in improved biomarkers, lower levels of plasma cholesterol, and a reduction in atherosclerotic lesions ( 52,53 ). The lipid nanoparticle formulation used in the present work may have played a role in protecting our peptides from destruction in the GI tract.
One factor that could be responsible for the observed similarity in effi cacy of our trimer-and monomer-based nanoparticles, despite the trimer exhibiting better PK properties ( 31 ), is the dosing regimen. Perhaps, the doses used in our studies were suffi ciently high to mask differences in effi cacy, that is, the animals may have been overdosed. To investigate this possibility, we evaluated the trimer and monomer nanoparticles at a 10-fold lower dose (7.5 mg/kg ip) in 2 week daily-dosing experiments. As expected, the magnitude of cholesterol reduction was less at the lower dose ( Fig. 2 ); however, both agents again reduced plasma cholesterol levels to a similar degree, suggesting that overdosing was not a factor at the 75 mg/kg dose level.
We observed some surprising effects of DMPC treatment compared with the PBS controls when the agents were administered by ip injection. At the two-week timepoint, the DMPC unilamellar vesicles modestly increased the level of plasma total cholesterol compared with PBS controls (supplementary Fig. I), although there was no difference between these groups in plasma cholesterol or triglyceride levels after 10 weeks ( Fig. 7 ). Further, ip DMPC apparently reduced the development of en face aortic lesions in the LDLr Ϫ / Ϫ mice, while the development of aortic root lesions was not affected by ip DMPC treatment ( Fig. 5 ). In the analogous studies involving oral administration, DMPC MLVs did not affect the development of lesions or plasma lipid levels to any degree compared with