Targeted suppression of microRNA-33 in lesional macrophages using pH low-insertion peptides (pHLIP) improves atherosclerotic plaque regression


 Hypoxia and tissue acidification occur in the macrophage-rich regions of advanced atherosclerotic lesions due to the higher oxygen demand of activated immune cells and insufficient oxygen supply. Our group and others originally identified microRNA-33 (miR-33) as critical regulator of cellular lipid homeostasis and lipoprotein metabolism controlling the development of atherosclerosis. Our prior work has demonstrated that pH Low-Insertion Peptides (pHLIP) can be used to direct miR-33 inhibitors to acidic microenvironments and protect against kidney fibrosis. Here we utilize anti-miR-33 conjugated pHLIP constructs to target macrophages located in atherosclerosis plaques. The inhibition of miR-33 using pHLIP-directed targeting increased collagen content and decreased lesional lipid accumulation within atherosclerotic plaques in a murine model of atherosclerosis regression. Single cell RNA sequencing analysis revealed higher expression of fibrotic genes (Col2a1, Col3a1, Fn1, Dcn, etc) and tissue inhibitor of metalloproteinase 3 (Timp3), and downregulation of matrix metallopeptidase 12 (Mmp12) in macrophages from atherosclerotic lesions targeted by pHLIP- anti-miR-33. These results suggest a potential application of pHLIP for treating advanced atherosclerosis via pharmacological inhibition of miR-33.


Abstract
Hypoxia and tissue acidi cation occur in the macrophage-rich regions of advanced atherosclerotic lesions due to the higher oxygen demand of activated immune cells and insu cient oxygen supply. Our group and others originally identi ed microRNA-33 (miR-33) as critical regulator of cellular lipid homeostasis and lipoprotein metabolism controlling the development of atherosclerosis. Our prior work has demonstrated that pH Low-Insertion Peptides (pHLIP) can be used to direct miR-33 inhibitors to acidic microenvironments and protect against kidney brosis. Here we utilize anti-miR-33 conjugated pHLIP constructs to target macrophages located in atherosclerosis plaques. The inhibition of miR-33 using pHLIP-directed targeting increased collagen content and decreased lesional lipid accumulation within atherosclerotic plaques in a murine model of atherosclerosis regression. Single cell RNA sequencing analysis revealed higher expression of brotic genes (Col2a1, Col3a1, Fn1, Dcn, etc) and tissue inhibitor of metalloproteinase 3 (Timp3), and downregulation of matrix metallopeptidase 12 (Mmp12) in macrophages from atherosclerotic lesions targeted by pHLIP-anti-miR-33. These results suggest a potential application of pHLIP for treating advanced atherosclerosis via pharmacological inhibition of miR-33.

Main
Atherosclerosis is well-known as a chronic in ammatory disease and disorder of lipid metabolism involving the retention of atheroprone lipoproteins and accumulation of monocyte-derived macrophages triggering maladaptive immune response and necrotic core formation. 1 The oxygen supply of vascular cells in atherosclerosis relies on the luminal blood or the adventitial vasa vasorum, and the distance to the deep layer of the intima exceeds the oxygen diffusion threshold resulting in local hypoxia 2,3,4,5 . The uptake of modi ed lipoproteins by macrophages differentiated from recruited monocytes results in the accumulation of macrophage foam cells in the atherosclerotic lesions. Speci cally, hypoxia strongly correlates with the macrophage foam cell clusters surrounding the plaque core 6 . The higher oxygen demand of activated immune cells and insu cient oxygen supply lead to severe hypoxia and tissue acidi cation in the macrophage-rich regions of atherosclerotic lesions 7,8 . Notably, the acidic extracellular environment in macrophage foam cells impairs the expression of ATP-binding cassette transporter ABCA1, resulting in decreased cholesterol e ux and accelerated lipid accumulation 9 .
Our group and others initially identi ed microRNA-33 (miR-33) as crucial regulator of cellular lipid homeostasis and lipoprotein metabolism, controlling downstream target genes including ABCA1 and ABCG1 10,11,12,13,14,15,16 . The bene ts of miR-33 de ciency on atherosclerosis development are attributed to its protective effects in macrophages 12 Therapeutic inhibition of miR-33 in the mice and non-human primates raises plasma high density lipoprotein (HDL) levels and inhibits the progression of atherosclerosis by increasing HDL levels/functionality or enhancing cholesterol e ux through induction of ABCA1 and ABCG1 in macrophages 14,15,16,17 . Notably, speci c disruption of Abca1 targeting by miR-33 is su cient to mimic the effects of miR-33 de ciency on cholesterol e ux and atherogenesis 18 .
However, long-term silencing of miR-33 increases circulating triglycerides levels and lipid accumulation in the liver through upregulation of genes involved in fatty acid synthesis when mice are fed a high-fat diet, indicating deleterious effects of moderate hepatic steatosis and hypertriglyceridemia 19 . Moreover, genetic models of miR-33 de ciency showed a strong predisposition to obesity and metabolic dysfunction 12,13,19,20 Thus, speci c targeting of miR-33 in the macrophages of atherosclerotic lesions could provide an effective therapeutic strategy for atherosclerosis that avoids the deleterious effects in other metabolic tissues.
pH-Low Insertion Peptides (pHLIP) are a novel class of water-soluble membrane molecules that target areas of high acidity at the surface of cells, which have been employed to deliver miRNA inhibitors to the acidic environment of tumors and the kidney 21,22 . Given the hypoxia in macrophage foam cells and the acidic environment of the lipid core in atherosclerosis, we explore the utility of anti-miR-33 peptide nucleic acid (PNA) delivery vectors (anti-miR-33 pHLIP ) for speci c targeting of the macrophages in vascular lesions (Fig. 1a). Notably, near-infrared uorescence imaging showed highly speci c uptake of pHLIP variant 3 conjugated with uorescent Alexa 750 (A750-Var3) in the aortic arch, an atherosclerotic prone area, of hypercholesterolemic mice. Other tissues characterized by the acidic microenvironment such as the kidney also accumulate the peptide. The targeting of VAR3 to atherosclerotic plaques was speci c since a similar peptide with an altered amino acid sequence that prevents insertion across the membranes in acidic conditions (A546-5K-Var3) was unable to target vascular lesions (Fig. 1b). The targeting of A546-Var3 into the aortic arch occurred early (4 hours) and was sustained for 24 hours (Supp Fig. 1a and b). Imaging of whole organs indicated that uptake in the liver was diminished with the A546-Var3 compared to the A546-5K-Var3 mutant (Supp Fig. 1a and b), while similar uorescent density was observed in histologic sections (Supp Fig. 1c). Both techniques demonstrated higher uptake of A546-Var3 in primary renal tubular cells (Supp Fig. 1c). The a nity of A750-Var3 to atherosclerotic aortic arch was con rmed by analyzing aortas in low-density lipoprotein receptor knockout (Ldlr −/− ) and WT mice injected with A750-Var3, A750-5K-Var3 or PBS (Fig. 1c). We further demonstrated the internalization of A750-Var3 in isolated macrophages from atherosclerotic aortas by ow cytometry. The results showed a signi cant uptake of A750-Var3 in macrophages (Lin − CD11b high F4/80 + Ly-6C low ) and less pronounced internalization in monocytes (Lin − CD11b high F4/80 − Ly-6C high ) from atherosclerotic plaques (Fig. 1d). These ndings correlate with our previous studies showing a speci c uptake of pHLIP in tumor associated macrophages. 23 Similar to the results observed in vivo, acidi cation of the media (pH = 6.2) promoted a marked increase of A546-Var3 uptake in mouse macrophages compared to macrophages cultured at a neutral pH (Fig. 2a). Assessment of uorescently labeled constructs (A633-Var3) con rmed the high uptake by in vivo foam cells isolated from Ldlr −/− mice fed a WD for 3 months using both uorescent microscopy and ow cytometry (Fig. 2b). Next, we tested whether the delivery of anti-miR-33 pHLIP was able to enhance the expression of ABC transporters and reduce foam cell formation in vivo.
We found that suppressing miR-33 in macrophages using anti-miR-33 pHLIP resulted in signi cant reduction of neutral lipid accumulation compared to macrophages treated with a non-targeting antisense oligonucleotide conjugated with pHLIP (Src pHLIP ) (Fig. 2c, quanti ed in right panel). The marked reduction in foamy macrophages correlated with an increase of ABCA1 expression in anti-miR-33 pHLIP treated macrophages (Fig. 2d, quanti ed in right panel). Together, these results demonstrate that pHLIP is an effective and highly speci c vehicle to target the expression of miRNAs in macrophages accumulated in atherosclerotic plaques.
We next assessed the e cacy of inhibiting miR-33 expression in lesional macrophages during the regression of atherosclerosis. To this end, Ldlr −/− mice were fed a Western diet (WD) for 3 months to establish atherosclerotic plaques, then switched to a chow diet (CD) and injected with Src pHLIP and anti-miR-33 pHLIP (1 mg/Kg) weekly for one month (Fig. 3a). Consistent with previous studies, switching Ldlr −/− mice to CD resulted in a signi cant decrease of circulating total cholesterol (TC) and triglycerides (TG) level and increase in plasma HDL-C (Supp Fig. 2a). However, there was no difference between mice treated with anti-miR-33 pHLIP and Src pHLIP , suggesting that the hepatic delivery of anti-miR-33 using pHLIP was not su cient to impact regulation of HDL biogenesis by miR-33 in the liver (Supp Fig. 2a). were found in mice treated with anti-miR-33 pHLIP compared with Src pHLIP -treated mice (Supp Fig. 3b).
We next sought to determine whether delivery of anti-miR-33 by pHLIP peptides could contribute to the regression of established atherosclerotic plaques. Consistent with our previous studies, 24  To assess the potential mechanisms by which anti-miR-33 pHLIP regulates macrophage function and regression of atherosclerosis, we isolated live whole cells from enzyme-digested aortas of anti-miR- and broblasts (Cluster 6 and 9) (Fig. 4a, Supp Fig. 6). Given the cellular delivery of anti-miR-33 by pHLIP into macrophages and monocytes in atherosclerotic lesions (Fig. 1d), we further analyzed the phenotype of these Mono/Mac populations during the regression of atherosclerosis (Supp Fig. 6a and b). Speci c gene expression pro les differentiated 5 aortic monocyte and macrophage populations from the Mono/Mac clusters (Fig. 4b, Supp Fig. 7a). These included Trem2 high Mac (Cluster 0: Trem2, Cd9, Spp1, Lgals3), F10 + Mono (Cluster 2: F10, Ccr2, low H2-Eb1), In ammatory Mac (Cluster 3: Il1b, Cxcl2, Nfkbiz, S100a9, S100a8) and Stem-like Mac (Cluster 4: Top2a, Ube2c, Cenpf, Stmn1) (Supp Fig. 7b). Interestingly, besides these well-established cell populations found in atherosclerotic lesions (Trem2 high Mac, F10 + Mono, In ammatory Mac and Stem-like Mac), 25,26 we observed one speci c macrophage cluster (Cluster 1, herein referred to as "ECM high Mac") that expressed monocyte and macrophage genes (Cd14, Cd68, Adgre1 and Csf1r) (Supp Fig. 7a) and was highly enriched for extracellular matrix (ECM)-associated genes, including Col1a2, Col3a2, Col1a1, Fn1, Eln, Lum, Bgn and Dcn (Fig. 4c, Supp Fig. 7c). The percentage of ECM high Mac were increased in the mice treated with anti-miR-33 pHLIP as compared to Src pHLIP (Fig. 4d). This intriguing nding indicate that during the regression of atherosclerosis suppression of miR-33 in macrophages promotes a pro-brotic phenotype that favour plaque stabilization. Notably, the population of in ammatory macrophages (Cluster 3) characterized by the high expression of pro-in ammatory cytokines and chemokines was decreased, and the "stem-like macrophages" in cluster 4, enriched for cell cycle genes and highly proliferative, were increased in the mice treated with anti-miR-33 pHLIP as compared to Src pHLIP (Fig. 4d). We next performed pathway enrichment analysis associated with changes in gene expression in Mono/Mac cells. Among all the pathways that were signi cantly altered in response to anti-miR-33 pHLIP treatment, we found upregulated brosis, M2 polarization (IL-4 Signaling) and antigen presentation pathways (Fig. 4e). In addition to an increased signature with extracellular matrix (Fig. 4f), we also observed a decrease of in ammatory genes (Cxcl2 and Tnfsf9) and an increase of antigen presentation genes (H2-Eb1, H2-Aa and H2-Ab1) in the mice treated with anti-miR-33 pHLIP (Fig. 4g), indicating that anti-miR-33 treatment by pHLIP peptides induces macrophages towards a less in ammatory and more stable phenotype in the atherosclerotic lesions.
In conclusion, we demonstrated the remarkable selectivity of pHLIP peptides to target macrophages in atherosclerotic lesions based on their a nity for acidic microenvironments. The speci c delivery of anti-miR-33 to atherosclerotic plaques using pHLIP peptides promotes a more stable phenotype by induction of ABCA1-mediated cholesterol e ux and extracellular matrix deposition. These ndings highlight the therapeutic potential of anti-miR-33 pHLIP constructs for the regression of atherosclerosis, while avoiding the potential deleterious effects in other organs. The pHLIP technology can also be applied to the selective delivery of other protective miRNAs to the macrophages from atherosclerotic plaques for the therapy of atherosclerosis-associated cardiovascular diseases. One of the most unique aspects of miRNAs is their ability to target many different mRNAs, which allows them exert both very nuanced and extremely pronounced effects in different situations. However, this promiscuity has also raised important concerns for both research on and clinical applications, especially since the target preferences and impact of miRNAs can vary dramatically in different tissues and cell types. Considering the potential for disparate and possibly adverse effects in different organs, targeted delivery systems such as that described in this work may prove incredibly valuable for the development of safe and reliable miRNAbased therapies.

Synthesis of PNA
The synthesis of antimiR-33 and scramble PNA was performed as reported previously 27,28 . Boc-PNA regular monomers used for the synthesis were purchased from ASM Chemicals and Research (Germany). Solid phase synthesis of PNA was carried out on 4-methylbenzylhydramine (MBHA) resin and cysteine was conjugated on N-terminus of PNAs. After completion of synthesis, PNA was cleaved from the resin using cleavage cocktail (Tri uoromethanesulfonic acid: tri uoroacetic acid: dimethyl sul de: m-cresol, 2:6:1:1) and precipitated using diethyl ether. Further puri cation of PNA was performed using reverse phase high performance liquid chromatography (RP-HPLC). Mass spectroscopy (matrix assisted laser desorption/ ionization time of ight, MALDI) was used to con rm the molecular weight and concentration of PNAs was determined using UV-spectroscopy.

Assessment of pHLIP delivery
To determine the speci city of pHLIP delivery, animals were injected intravenously with pHLIP-A750-Var3 or the non-inserting control peptide pHLIP-A750-5K-Var3 (4 nmol). Animals were sacri ced after 4, 12 and 24 hours, and uptake in different tissues was determined by near-infrared uorescence imaging, performed on an IVIS Spectrum system (Caliper Life Science) with appropriate excitation (Ex) and emission (Em) lter sets (Ex/Em = 745/800 nm). For ow cytometry analysis, entire aortas (from the root to the iliac bifurcation) were harvested 4 hours after injection with pHLIP-A750-5K-Var3 or pHLIP-A750-Var3. The aortas were cut into small pieces and subjected to enzymatic digestion with 400 U/ml collagenase I, 125 U/ml collagenase XI, 60 U/ml DNAse I and 60 U/ml hyaluronidase (Sigma-Aldrich) for 1 h at 37ºC while shacking. Macrophages and monocytes were identi ed with the following antibodies Anti-miR-33 pHLIP treatment Ldlr −/− mice were fed a WD for 3 months to induce atherosclerosis and then transferred to chow diet (CD) for 1 month. anti-miR-33 pHLIP and Src pHLIP constructs were administered by intravenous injection at a dose of 1 mg/kg body weight in PBS with 5% DMSO. Injections were performed once every week (totally 5 times) during CD feeding before harvest. The mice fed with a WD for 3 months were harvested as Baseline for atherosclerosis analysis.

Plasma lipids, lipoprotein pro le and leukocytes analysis
After 3-month WD and 1-month CD feeding, mice were fasted for 12-14 h before blood samples were collected by retro-orbital venous plexus puncture. Plasma was separated by centrifugation and stored at -80 °C until analysis. Plasma total cholesterol (TC), high density lipoprotein-cholesterol (HDL-C) and triglyceride (TG) concentrations were determined by standard enzymatic methods (Wako Chemicals, USA). Plasma cholesterol fractions (VLDL, IDL/LDL and HDL) were detected by fast-performance liquid chromatography (FPLC) gel ltration on Superose 6 HR 10/30 size-exclusion column (Pharmacia) as described previously. 30 White blood cells (WBC) counting in circulation was determined from EDTAanticoagulated blood using a hemocytometer (Hemavet Counter HV950FS). Circulating aspartate aminotransferase (AST) and alanine aminotransferase (ALT) were analyzed using commercial enzymatic assays.

Histology and morphometric analysis
Following anesthesia (100 mg/kg ketamine; 10 mg/kg xylazine), thoracic cavity was exposed immediately and in situ perfusion xation through the left cardiac ventricle was performed by thorough perfusion with PBS and 4% paraformaldehyde (PFA). Subsequently, hearts and aortas were harvested and xed in 10% formaldehyde solution overnight. Hearts were embedded in OCT after dehydration with 30% sucrose and serial sections were cut at 6 µm thickness using a cryostat. Every third slide from the serial sections was stained with haematoxylin and eosin (H&E) and each consecutive slide was stained with oil red O (ORO) for quanti cation of lesion area. Aortic lesion size of each animal was obtained by averaging the lesion areas in four sections from the same mouse. Collagen content was assessed by Picro Sirus Red staining of consecutive slides from serial sections and quanti ed as a percentage of the total plaque area.

Immuno uorescence staining
The atherosclerotic sections were xed with 4% PFA and incubated overnight with primary antibodies for CD68 (Serotec; #MCA1957) and Actin, α-smooth muscle-Cy3 (Sigma, #C6198) after blocking with blocker buffer (5% Donkey Serum, 0.5% BSA, 0.3% Triton X-100 in PBS) for 1 hour at RT, followed by incubation with Alexa Fluor secondary antibody (Invitrogen, Carlsbad, CA) for 1 hour at RT. The stained sections were captured using a Carl Zeiss scanning microscope Axiovert 200M imaging system and images were digitized under constant exposure time, gain, and offset. Results are expressed as the percent of the total plaque area stained measured with the Image J software (ImageJ version 1.51, Yale software library, Yale University).

Aortic cells isolation from atherosclerotic lesions
To obtain the whole cells in the atherosclerotic lesions, the aorta was digested (from the root to the diaphragm) with 1 mg/ml Collagenase A (Roche, Cat 11088785103) for 7 min at 37 °C to remove the adventitia under microscope. Aortic tissue was cut into small pieces and subjected to enzymatic digestion with 1.5 mg/ml Collagenase A and 0.5 mg/ml Elastase (Worthington, Cat LS006365) for 40 min at 37 °C while shaking. The digested aortas were passed through a 70 µm Cell Strainer to obtain single cell suspensions followed by incubation of 10 minutes at 4 °C with 10 µg/ml of puri ed rat anti-mouse FcgRII/III (Biolegend) to block non-speci c binding of antibodies to Fc Receptors. Total cell viability was obtained using live/dead viability dye eFluor 450 (Thermo Fisher Scienti c). Viable cells were sorted by FACS Aria III (BD Biosciences) and immediately processed for single-cell RNA-seq.
Droplet-based scRNA-seq library construction and sequencing The sorted viable cells were encapsulated into droplets and processed following manufacturer's speci cations using 10X Genomics GemCode Technology. Equal numbers of cells per sample were loaded on a 10x Genomics Chromium controller instrument to generate single-cell Gel Beads in emulsion (GEMs) at Yale Center for Genome Analysis. Lysis and barcoded reverse transcription of polyadenylated mRNA from single cells were performed inside each GEM followed by cDNA generation using the Single Cell 3' Reagent Kits v3 (10X Genomics). Libraries were sequenced on an Illumina NovaSeq 6000 as 2 × 100 paired-end reads.
Single cell RNA-seq data analysis Sample demultiplexing, aligning reads to the mouse genome (mouse UCSC mm10 reference genome) with STAR and unique molecular identi er (UMI) processing were processed using CellRanger software (version 4.0.0) as previously described. 31

In vivo foam cell formation
In vivo foam cell formation was performed as previously described. 33

qPCR analysis
Total RNA from cells was isolated using TRIzol reagent. One microgram of total RNA was reversetranscribed using the iScript RT Supermix (Bio-Rad, Hercules, CA, USA), following the manufacturer's protocol. Quantitative real-time PCR was performed in triplicate using iQ SYBR green Supermix (Bio-Rad) on a Real-Time Detection System (BioRad). The mRNA level was normalized to GAPDH as a housekeeping gene. Real-time PCR was conducted with gene expression levels with oligonucleotides speci c for each of the genes.

Statistical analysis
All data are expressed as mean ± s.e.m. Statistical differences were calculated with either unpaired twosided Student's t-test or one-way analysis of variance (ANOVA, followed by the Bonferroni post-test). A value of P ≤ 0.05 was considered statistically signi cant. Data analysis was performed using GraphPad Prism Software Version 7.0 (GraphPad, San Diego, CA).

Declarations
COMPETING FINANCIAL INTEREST O.A.A. and Y.K.R. are founders of pHLIP, Inc. They have shares in the company, but the company did not fund any part of the work reported in this paper, which was carried out in their academic laboratories.