SREBF1/MicroRNA-33b Axis Exhibits Potent Effect on Unstable Atherosclerotic Plaque Formation In Vivo

We analyzed the miR-33a/b expression levels in atherosclerotic plaques and their adjacent relatively healthy intima (plaque edge) from carotid endarterectomy specimens. These were obtained from 14 patients who were diagnosed with severe carotid artery stenosis and were underwent carotid endarterectomy at the Department of Neurosurgery, Kyoto University Hospital from 2012 to 2015.

All the in vivo experiments were performed in C57BL/6J background mice. MiR-33b^+/+ mice were generated as reported previously.¹² To obtain miR-33b^+/+Apoe^−/− mice, we mated miR-33b^+/+ mice with Apoe^−/− mice. MiR-33b^−/− or miR-33b^−/−Apoe^−/− littermates were used as the controls. To assess changes in miR-33a/b expression levels in the liver, miR-33b^+/+ and miR-33b^+/+Apoe^−/− mice at the age of 8 weeks were fed on normal chow containing 4.5% fat (Oriental Yeast, Tokyo, Japan) or a Western-type diet (WTD) containing 0.15% cholesterol and 40% fat (Oriental Yeast) for 2 weeks. To investigate the plaque formation, after being weaned at 4 weeks of age, miR-33b^+/+Apoe^−/− mice and their control mice were fed normal chow until 6 weeks of age and then switched to WTD for the next 12 weeks. We used male mice for our experiments unless otherwise stated.¹³

Total RNA was isolated using TRIzol reagent (Invitrogen). MiR-33a/b were measured in accordance with the TaqMan MicroRNA Assays (Applied Biosystems) protocol, and the products were analyzed using a thermal cycler (Applied Biosystems StepOnePlus real-time polymerase chain reaction [PCR] system). Samples were normalized by U6 small nuclear RNA expression.

Atherosclerotic plaque lesions were quantified by cross-sectional analysis of the proximal aorta.^14–16 For the cross-sectional analysis of the aorta, Tissue-Tek OCT compound (Sakura Finetek Japan)–embedded aortas were sectioned using a cryostat, and the 6-μm sections were obtained sequentially beginning at the aortic valve. Eight sections obtained every 24 μm from the aortic sinus were stained with oil red O, hematoxylin-eosin, and picrosirius red stain. The lesion areas of each aorta were measured using ImageJ. The average of the 8 sections from 1 mouse was taken as a value that represented the mouse.

Peritoneal macrophages (PEMs) were obtained from the peritoneal cavity of mice 4 days after the intraperitoneal injection of 3 mL of 3% thioglycollate. The obtained cells were washed, centrifuged at 1000 rpm for 5 minutes, and plated at a density of 8×10⁵ cells/mL with RPMI 1640 medium (Nacalai Tesque, Japan) containing 10% fetal bovine serum (FBS). THP-1 cells were cultured with RPMI 1640 supplemented with 10% FBS. To differentiate THP-1 monocytes into macrophages, cells were treated with 100 nmol/L PMA (phorbol 12-myristate 13-acetate) for 3 days. For cholesterol depletion experiments, cells were precultured in RPMI 1640 with or without 10 mmol/L methyl-β-cyclodextrin for 30 minutes and washed twice with RPMI 1640, followed culturing in RPMI 1640 containing 10% FBS with 10 ng/mL lipopolysaccharides (LPS). The antibodies used were a polyclonal anti–ATP-binding cassette transporter (ABC) A1 antibody (NB400-105), a polyclonal anti-ABCG1 antibody (NB400-132), a polyclonal anti–SR-BI (SCARB1) antibody (NB400-104; Novus Biologicals, Littleton, CO), an anti–single strand DNA antibody (no. 18731; IBL, Gunma, Japan), an anti-CPT1α (carnitine palmitoyltransferase 1A) antibody (ab128568), an anti-CROT (carnitine O-octanoyltransferase) antibody (ab103448), an anti–LDL receptor antibody (ab52818; Abcam, Cambridge, United Kingdom), an anti–β-actin antibody (AC-15; A5441, Sigma-Aldrich, St. Louis, MO), a PE (phycoerythrin)-conjugated anti-CD11b antibody (no. 12-0112; eBioscience, Santa Clara, CA), an APC (allophycocyanin)-conjugated anti-F4/80 antibody (no. 123115; BioLegend, San Diego, CA), and an anti-CD68 antibody (FA-11, Serotec, Kidlington, United Kingdom). Human acetylated LDL (AcLDL) and human HDL-C (high-density lipoprotein cholesterol) were purchased from Biomedical Technologies Inc. (Stoughton, MA). Anti-rabbit and anti-mouse IgG horseradish peroxidase (HRP)–linked antibody were purchased from GE Healthcare (Amersham, United Kingdom). Human apoA-I (apolipoprotein A1), polyethylene glycol, acyl-CoA: cholesterol acyl-transferase inhibitor, and oil red O were purchased from Sigma-Aldrich. [1, 2-3H (N)]-Cholesterol was purchased from Perkin Elmer (Boston, MA).

Eight sections of the aortic root per mouse were stained with an anti-CD68 antibody (5 μg/mL) and anti–single strand DNA antibody (0.083 μg/mL). The lesion positively stained areas or cells of each aortic were measured using ImageJ. The average of the 8 sections from 1 mouse was taken as a value that represented the mouse.

Total RNA was isolated and purified using TRIzol reagent (Invitrogen), and cDNA was synthesized from 1 μg of total RNA using Verso cDNA synthesis kit (Thermo Fisher SCIENTIFIC) in accordance with the manufacturer’s instructions. For quantitative real-time PCR, specific genes were amplified by 40 cycles using THUNDERBIRD SYBR qPCR MIX (TOYOBO). Expression was normalized to the housekeeping gene β-actin. Gene-specific primers are summarized in the Table in the online-only Data Supplement.

Western blotting was performed using standard procedures as described previously.¹⁷ A total of 20 μg of protein was fractionated using NuPAGE 4% to 12% Bis-Tris (Invitrogen) gels and transferred to a Protran nitrocellulose transfer membrane (Whatman). The membrane was blocked using 1× PBS containing 5% nonfat milk for 1 hour and incubated with a primary antibody (anti-ABCA1, 1 μg/mL; anti-ABCG1, 1 μg/mL; anti-CPT1a, 1 μg/mL; anti-CROT, 8 μg/m; anti-βactin, 1 μg/mL; anti-SCARB1, 0.5 μg/mL; and anti–LDL receptor, 0.4 μg/mL) overnight at 4°C. After a washing step in PBS-0.05% Tween 20 (0.05% T-PBS), the membrane was incubated with the secondary antibody (anti-rabbit IgG HRP-linked, 1:2000; anti-mouse IgG HRP-linked, 1:2000) for 1 hour at 4°C. The membrane was then washed in 0.05% T-PBS and detected using ECL Western Blotting Detection Reagent (GE Healthcare), using an LAS-4000 system (Fuji Film).

Blood was obtained from the inferior vena cava of anesthetized mice in the early morning without fasting or after a 4 to 6 hours fasting period. Serum was separated by centrifugation at 4°C and stored at −80°C. Biochemical data were measured using standard methods with a Hitachi 7180 Auto Analyzer (Nagahama Life Science Laboratory, Nagahama, Japan). Serum lipoproteins were analyzed using high-performance liquid chromatography (HPLC) methods^18,19 (Sky Light Biotech, Akita, Japan).

Cholesterol efflux via mouse apoB (apolipoprotein B)–depleted serum was measured as described previously.²⁰ Briefly, J774 mouse macrophage cells were plated in 24-well dish (7×10⁴ cells/well) and labeled with 2 μCi/mL ³H-cholesterol for 24 hours in RPMI 1640 with 1% FBS. Cells were incubated in RPMI 1640 containing Cpt-cAMP (0.3 mmol/L) and 0.2% BSA for an additional 16 hours to upregulate ABCA1 in J774 cells. Cells were washed 4× with RPMI 1640 and incubated for 4 hours in MEM-HEPES containing 2.8% apoB-depleted serum (equivalent to 2% serum), which was obtained after apoB lipoproteins were removed with polyethylene glycol. All steps were performed in the presence of acyl-CoA: cholesterol acyl-transferase inhibitor (2 μg/mL). Cholesterol efflux was expressed as the percentage of radioactivity released from the cells in the medium relative to the total radioactivity in cells plus medium.

PEMs from miR-33b^−/−Apoe^−/− mice were seeded at a density of 1.2×10⁶/mL in a 100 mm dish, incubated for 1 hour, and washed with PBS. After 4-hour incubation, cells were washed twice with PBS and incubated for another 24 hours with labeling RPMI 1640 medium containing 0.1 mg/mL AcLDL and 5 μCi/mL ³H-cholesterol. Cells were dissociated using Accutase (M&S TechnoSystems, Japan) followed by suspension in cold PBS. Each mouse was administered 500 μL of cell suspension intraperitoneally. Serum samples were collected at 6 and 24 hours after administration via the orbital sinus. Forty-eight hours after administration, we dissected the mice and collected serum, liver, bile, and all feces during the experiment. Liver samples were cut and then crushed using a BioMasher I (Nippi, Incorporated, Japan). After adding 500 μL hexane-isopropanol solution, crushed liver samples were centrifuged at 15 000 rpm for 3 minutes. Supernatants were moved to new tube and then a 30% volume of 0.47 mol/L Na₂SO₄ was applied. After shaking intensely for 1 minute, they were centrifuged and only upper layer was collected for measurement.

Feces samples were homogenized with 10 mL/g-feces water and incubated at 4°C over night. Then, an equal volume of ethanol was added and mixed.

All serum, liver, bile, and feces samples were counted their radioactivity by liquid scintillation counting. Part of ³H-cholesterol–labeled macrophage suspension in each experiment was used to measure the input value. Raw serum, liver, and feces radioactivity values were adjusted for the total individual values based on the body weight, the ratio of sampling liver weight to entire liver weight, and the whole amount of feces during experimental period, respectively. Full bile samples from each mouse were used for counting their raw radioactivity values. The percentages of macrophage-derived 3H-cholesterol in each specimen were calculated by dividing each estimated (serum, liver, and feces) or raw (bile) total individual value by the input value.

Mice aged 6 to 8 weeks old were fed WTD for 2 weeks. After a 4-hour fasting period, mice were injected with 500 mg/kg Triton WR-1339 (Tyloxapol, Sigma) via the postorbital vein to block lipolysis and hepatic VLDL (very-low-density lipoprotein) uptake. Mice were challenged with a bolus of 200 μL of olive oil by intragastric gavage 15 minutes after Triton WR-1339 injection. Blood samples for measuring triglyceride (TG) concentration were obtained from the opposite postorbital vein at 0, 1, 2, and 3 hours after olive oil administration. Serum TG levels were measured as described above. Intestinal chylomicron production rates were estimated using linear regression for serial TG data from each mouse.

Mice aged 6 to 8 weeks old were fed WTD for 2 weeks. After a 4-hour fasting period, mice were injected with 500 mg/kg Triton WR-1339 (Tyloxapol, Sigma) via the postorbital vein to block lipolysis and hepatic VLDL uptake. No food was available during the time course. Blood samples for measuring TG concentration were obtained from the opposite postorbital vein at 0, 30, 60, 90, and 120 minutes after Triton WR-1339 administration. Serum TG levels were measured as described above. Hepatic TG production rates were estimated using linear regression for serial TG data from each mouse.

Mice aged 6 to 8 weeks old were fed WTD for 3 weeks. After a 4-hour fasting period, mice were injected the 0.2 U/g heparin intravenously. Blood was obtained before and 10 minutes after heparin injection for the measurement of serum lipoprotein lipase (LPL) activity using an LPL activity assay kit (Cell Biolab no. STA-610) in accordance with the manufacturer’s instructions.

Cellular cholesterol efflux via apoA-1 and HDL was determined as described previously.^3,21 Briefly, thioglycollate-elicited mouse PEMs were plated in 24-well dishes (5×10⁶ cells/mL) and were cultured for 24 hours in RPMI 1640 containing 3H-labeled AcLDL (1.0 μCi/mL ³H-cholesterol and 25 μg/mL of AcLDL). Cells were washed 4× with RPMI 1640 and then incubated for 6 hours in RPMI 1640 with or without 10 μg/mL apoA-1 and 100 μg/mL HDL. Cholesterol efflux was expressed as a percentage of radioactivity released from the cells into the medium relative to the total radioactivity in the cells and medium.

Thioglycollate-elicited PEMs from miR-33b^−/−Apoe^−/− and miR-33b^+/+Apoe^−/− mice were incubated with RPMI 1640 containing 10% FBS supplemented with or without 300 μg/mL AcLDL plus 10 μg/mL acyl-CoA: cholesterol acyl-transferase inhibitor (58-035) for 24 hours. To assess early-to-mid-stage apoptosis, cells were stained with Alexa Fluor 488–conjugated Annexin V (green) and DAPI (4’,6-diamidino-2-phenylindole; blue), as described previously,²² using a Vybrant Apoptosis Assay Kit (Molecular Probes). We counted the number of Annexin V–positive cells from representative fields (8–9 fields containing ≈650–1000 cells) and expressed as a percentage of the total number of cells.

To investigate an amount of lipid rafts, we used fluorescein isothiocyanate (FITC)-cholera toxin subunit B (CTB; Sigma, C1655) for labeling. Thioglycollate-elicited PEMs were prestained by PE-conjugated anti-CD11b antibody (1.25 μg/mL) and APC-conjugated anti-F4/80 antibody (2.5 μg/mL) to confirm cells as macrophages and followed by being stained with FITC-CTB at 4°C for 15 minutes. To quantify lipid rafts, we measured the signal intensity of FITC-CTB using a BD FACSAriaII (Becton Dickinson). THP-1 macrophages infected with miR-control or miR-33b by lentivirus vector were collected with or without 10 mmol/L methyl-β-cyclodextrin pre-treatment for 2 hours. Afterward, cells were stained with FITC-CTB and measured their FITC signals using BD FACSAriaII.

We produced lentiviral stocks in 293FT cells in accordance with the manufacturer’s protocol (Invitrogen). In brief, virus-containing medium was collected 48 hours post-transfection and filtered through a 0.45-μm filter. One round of lentiviral infection was performed by replacing the medium with virus-containing medium (containing 8 μg/mL of Polybrene), followed by centrifugation at 2500 rpm for 30 minutes at 32°C. Cells were used for analysis 3 days after DNA transduction.

BMT experiments were performed using the same protocol as reported previously.⁵ Male, 8-week-old age mice with genotypes of miR-33b^−/−Apoe^−/− and miR-33b^+/+Apoe^−/− were used as bone marrow (BM) donors. BM recipients were 8-week-old female miR-33b^−/−Apoe^−/− mice and miR-33b^+/+Apoe^−/− mice. All mice used for BMT had an Apoe^−/− background. BM donors were euthanized by cervical dislocation, and BM cells were collected by flushing femurs and tibias with PBS supplemented with 3% FBS. The suspension was passed through 40 μm nylon mesh cell strainer (BD Biosciences). Red blood cells were lysed using ACK (ammonium-chloride-potassium) lysing buffer (Lonza). BM cells were then washed twice with PBS supplemented with 3% FBS. To induce BM aplasia, recipients were irradiated with twice of 6 Gy within an interval of 3 hours (Cs¹³⁷; Gammacell 40 Exactor) and injected intravenously with 5×10⁶ BM cells 6 hours after irradiation.^23,24 After BMT, mice were fed normal chow diet for 4 weeks and then switched to a WTD for 10 weeks. At 22 weeks old, mice were euthanized and analyzed. Successful hematopoietic reconstitution after BMT was confirmed by PCR amplification of the whole-blood genome and tail genome at the time of euthanization.

We obtained the data about target genes for miR-33a/b-5p from TargetScan Human/Mouse release 7.1²⁵ and subsequent division of these genes into 3 distinctive groups: target genes that only exist in mouse, those only in human, and those in both mouse and human. Afterward, we analyzed the enriched KEGG pathways in each of the 3 groups using R clusterProfiler package.²⁶

Measurements are presented as means±SEM. Statistical comparisons were performed using paired or unpaired 2-tailed Student t-tests, Mann-Whitney test, a 1-way ANOVA with the Tukey post hoc test, or Kurskal-Wallis test with post hoc Dunn test where appropriate, with a P<0.05 taken to indicate significance. Parametric tests (such as t test) and nonparametric tests (such as Mann-Whitney test) were performed based on whether data passed the normality and equal variance tests. Statistical analyses were performed using GraphPad Prism 6 (GraphPad Software, Inc).

When we collect human carotid artery samples, informed consent was given by all patients or their families in accordance with the Declaration of Helsinki. Participants are identified by number, not by name. The Institutional Review Board of Kyoto University Graduate School and Faculty of Medicine approved this study.

The Ethics Committee for Animal Experiments of Kyoto University approved all animal experimental protocols.

SREBF2/1 are expressed ubiquitously and transcriptionally regulate cholesterol and fatty acid homeostasis, respectively.²⁷ MiR-33a/b are transcribed from and with SREBF2/1, respectively.^1–3 We measured miR-33a/b expression levels in human plaque and adjacent relatively healthy intima using carotid endarterectomy specimens (Table 1). As a result, miR-33a expression levels were significantly repressed in the plaque area compared with those in the adjacent area. MiR-33b expression levels also tended to decrease in plaque area, but this was not significant (Figure 1A). Next, to elucidate the expression levels of miR-33a/b and their host genes in the livers with an atheroprone diet, we used miR-33b KI humanized mice. Expression levels of miR-33b and Srebf1 were significantly increased with a WTD, however, those of miR-33a and Srebf2 were decreased in the same conditions (Figure 1B). At the same time, the expression levels of miR-33b in liver is much higher than those of miR-33a (Figure IA in the online-only Data Supplement). MiR-33b is a much more prominent member of the miR-33 family in atheroprone conditions.

To clarify the role of miR-33b in atherogenesis, we generated miR-33b^+/+Apoe^−/− mice and fed them a WTD for 12 weeks. There are no significant differences in hepatic miR-33a/b and their host genes expression levels between miR-33b^+/+ mice and miR-33b^+/+Apoe^−/− mice (Figure IA and IB in the online-only Data Supplement). MiR-33b KI has little effect on expression levels of Srebf2/miR-33a under the background of Apoe^−/− mice (Figure IC in the online-only Data Supplement). MiR-33b^+/+Apoe^−/− mice were born with the expected Mendelian ratios and were viable and fertile. MiR-33b^+/+Apoe^−/− mice gained less weight compared with miR-33b^−/−Apoe^−/− mice (miR-33b^−/−Apoe^−/− 40.23 g versus miR-33b^+/+Apoe^−/− 34.63 g; P<0.0001; n=25–32). Although we previously reported on the diet consumption of miR-33^−/− mice fed a high-fat diet,¹⁰ miR-33b^+/+Apoe^−/− mice showed a decrease in food intake with WTD feeding (miR-33b^−/−Apoe^−/− 20.92 g/week per mouse versus miR-33b^+/+Apoe^−/− 19.40 g/week per mouse; P=0.001; n=13–23). The resultant atherosclerotic plaque area was significantly larger in miR-33b^+/+Apoe^−/− mice (Figure 2A and 2B). We found a greater hematoxylin-eosin–free area in plaques from miR-33b^+/+Apoe^−/− mice, which indicated that plaques in miR-33b^+/+Apoe^−/− mice contained larger necrotic core areas and more cholesterol crystals compared with those in miR-33b^−/−Apoe^−/− mice (Figure 2C and 2D). Next, we analyzed the infiltration of macrophages, the degree of fibrosis, and the number of apoptotic cells. As a result, a larger number of macrophages infiltrated into plaques in miR-33b^+/+Apoe^−/− mice than in miR-33b^−/−Apoe^−/− mice (Figure 2E and 2F). Fibrotic areas in plaques were smaller in miR-33b^+/+Apoe^−/− mice than those in miR-33b^−/−Apoe^−/− mice (Figure 2G and 2H). In addition, the number of apoptotic cells in plaques of miR-33b^+/+Apoe^−/− mice was significantly larger compared with miR-33b^−/−Apoe^−/− mice (Figure 2I and 2J). These data indicated that Srebf1-miR-33b axis promote the development of a so-called unstable plaque.

As shown in Figure 3A, the expression levels of miR-33b were significantly increased, and on the contrary, those of miR-33a were decreased in the livers of miR-33b^+/+Apoe^−/− mice fed on WTD as with miR-33b^+/+ mice (Figure 1B). These data indicated that miR-33a/b are differentially regulated in the livers after WTD feeding and suggested that miR-33b could have stronger repressive effects on target genes than miR-33a in such an atheroprone context. To determine whether miR-33b has additional repressive effects on hepatic target genes of miR-33a or not, we measured hepatic ABCA1, CPT1a, and CROT mRNA and protein levels as representative target genes after 12 weeks WTD feeding. We found further repression of ABCA1 and CROT at both the mRNA and protein levels in the livers of miR-33b^+/−Apoe^−/− and miR-33b^+/+Apoe^−/− mice, except for CPT1A (Figure 3B and 3C). The additive repressive effect of miR-33b on ABCA1 protein levels was also confirmed in small intestines (Figure IIIA and IIIB in the online-only Data Supplement). As a result, serum HDL-C levels were significantly decreased in miR-33b^+/+Apoe^−/− mice compared with miR-33b^−/−Apoe^−/− mice (Figure 3D). Although we reported that miR-33^−/− mice showed fatty liver when they were fed a high-fat diet or were old,¹⁰ there were no significant differences in liver histology (Figure IIIC in the online-only Data Supplement) or serum liver enzymes (data not shown) between miR-33b^−/−Apoe^−/− and miR-33b^+/+Apoe^−/− mice at the age of 6 to 8 weeks. Next, we examined the cholesterol efflux capacity of serum HDL and found that the cholesterol efflux capacity of apoB-depleted serum was significantly attenuated in miR-33b^+/+Apoe^−/− mice compared with miR-33b^−/−Apoe^−/− mice (Figure 3E). In addition to in vitro cholesterol efflux experiments, systemic reverse cholesterol transport (RCT) measured by tracing ³H-labeled cholesterol was significantly repressed in miR-33b^+/+Apoe^−/− mice (Figure 3F). There are no significant differences in both mRNA and protein levels of hepatic SCARB1 which have potential contribution to RCT (Figure IVA and IVB in the online-only Data Supplement). These results indicated that Srebf1-miR-33b axis significantly reduces serum HDL-C levels via repressing both liver and small intestine ABCA1 levels and represses systemic RCT.

We found serum TG levels in miR-33b^+/+Apoe^−/− mice were lower than those in miR-33b^−/−Apoe^−/− mice. Moreover, we found a higher ratio of serum LDL cholesterol (LDL-C) to HDL-C in miR-33b^+/+Apoe^−/− mice (Figure 3D). To further investigate serum lipoprotein profiles, we performed analyses using HPLC. HPLC analysis for cholesterol content showed a decreased peak in the HDL-C fraction in miR-33b^+/+Apoe^−/− mice and decreased cholesterol levels in the chylomicron fraction and a peak shift from the large VLDL fraction to small LDL fraction. In addition, HPLC analysis of TG content showed that TG levels were reduced especially in the chylomicron and VLDL fractions in miR-33b^+/+Apoe^−/− mice (Figure 4A; Table 2).

Serum TG levels reflected the balance among production by the small intestine and the liver and clearance by peripheral organs and tissues.^28,29 Using Triton WR-1339, which inhibits LPL activity, we measured the lipid absorption after chylomicron-TG production from the small intestine and hepatic VLDL-TG production. As a result, lipid absorption and subsequent chylomicron-TG production rates were significantly repressed in miR-33b^+/+Apoe^−/− mice (Figure 4B), and hepatic VLDL-TG production rates were similar between miR-33b^+/+Apoe^−/− and miR-33b^−/−Apoe^−/− mice (Figure 4C). Hepatic lipogenic genes expression levels were also similar (Figure VA in the online-only Data Supplement). Post-heparin administration, serum LPL activity was also similar between them (Figure 4D). Furthermore, there were no significant differences in hepatic expression of genes involved in modulating LPL activity (Figure VB in the online-only Data Supplement). These results were consistent with HPLC data that showed a reduction in serum TGs, especially in the chylomicron fraction in miR-33b^+/+Apoe^−/− mice (Figure 4A). The reduction rate of serum TGs from the fed state to 4 to 6 hours fast was much higher in miR-33b^+/+Apoe^−/− mice than in miR-33b^−/−Apoe^−/− mice (−54.7% versus −26.1%, respectively; Figure VC in the online-only Data Supplement), which also supports the reduction in lipid absorption and subsequent chylomicron-TG production in the small intestine. Next, we checked gene expression profiles related to serum LDL-C clearance because HPLC data showed a cholesterol peak shift from the large VLDL fraction to small LDL fraction. There were no significant changes in hepatic expression levels of genes related to LDL-C uptake and hepatic LDL receptor protein levels (Figure VD and VE in the online-only Data Supplement). These results indicated that Srebf1-miR-33b axis remodels lipoprotein metabolism broadly, as well as the reduction in serum HDL-C with attenuated cholesterol efflux capacity.

Because previous reports have shown that miR-33a regulates macrophage phenotypes,^4,5,30,31 we examined whether miR-33b altered macrophage characteristics or not. We measured ABCA1 and ABCG1 protein levels in PEMs. When PEMs were treated with acetylated LDL-C as the indicated time, induction of ABCA1/G1 was significantly repressed in PEMs from miR-33b^+/+Apoe^−/− mice (Figure 5A). The cholesterol efflux rate to apolipoprotein A-1 was significantly reduced to the same degree both in PEMs from miR-33b^+/−Apoe^−/− and miR-33b^+/+Apoe^−/− mice, whereas the cholesterol efflux rate to HDL was attenuated according to the number of miR-33b KI alleles (Figure 5B). To investigate whether this attenuated cholesterol efflux capacity affected PEMs survival or not, we analyzed FC-induced apoptosis in PEMs. Annexin V–positive apoptotic cells increased significantly in PEMs from miR-33b^+/+Apoe^−/− mice after FC loading (Figure 5C). Thus, miR-33b enhances the susceptibility of FC-induced apoptosis, which might lead to a plaque with increased number of apoptotic cells in vivo (Figure 2J).

We reported that a loss of miR-33a might have multiple effects on both pro- and anti-inflammatory process in PEMs.⁵ We checked the expression levels of proinflammatory genes in thioglycollate-elicited PEMs from miR-33b^+/+Apoe^−/− and miR-33b^−/−Apoe^−/− mice. PEMs from miR-33b^+/+Apoe^−/− mice showed increased expression levels of Il-6 and Ccl2 (Figure 5D). Because miR-33b represses cholesterol efflux capacity in PEMs (Figure 5B), we hypothesized that miR-33b alters the cellular membrane lipid raft compartment, which is enriched in FC and is important for regulating inflammatory signal transduction. The amount of lipid rafts labeled with FITC-CTB on the membrane of PEMs from miR-33b^+/+Apoe^−/− mice were significantly increased compared with those from miR-33b^−/−Apoe^−/− mice (Figure 5E). We could recapitulate these phenotypes in THP-1 macrophages transduced with miR-33b expression vectors (Figure 5F). To confirm whether increased lipid rafts is responsible for proinflammatory gene expression profile or not, we depleted lipid raft cholesterol using methyl-β-cyclodextrin treatment. After methyl-β-cyclodextrin treatment, the amount of lipid rafts was similar between miR-control and miR-33b overexpressing THP-1 macrophages. In these conditions, increased expression levels of Il-1β and Il-6 shown in miR-33b–expressing THP-1 macrophages after LPS treatment were rescued (Figure 5G). Furthermore, to access the possibility of controlling miR-33a/b expression levels separately and its efficacy on inflammatory status, we treated PEMs from miR-33b^+/+Apoe^−/− using locked ribonucleotides, which have complementary sequences to miR-33b selectively. We could significantly repress miR-33b expression levels without changing miR-33a expression levels (Figure VIA in the online-only Data Supplement). As a result, the expression levels of Abca1 were increased and those of proinflammatory genes were significantly repressed in a dose-dependent manner (Figure VIB in the online-only Data Supplement). These data indicated that miR-33b could reprogram macrophages to a more proinflammatory phenotype in a proinflammatory milieu via regulating lipid rafts.

To elucidate the contribution of miR-33b in macrophages to atherogenesis, we performed BMT experiments. As a result, regardless of the recipients’ genotype, the resultant atherosclerotic plaque areas and CD68-positive macrophage infiltration areas were significantly increased in mice transfused with miR-33b^+/+Apoe^−/− BM cells compared with miR-33b^−/−Apoe^−/− BM cells (Figure 6A–6C). Serum HDL-C levels and serum TG levels were decreased in the miR-33b^+/+Apoe^−/− recipient group regardless of the genotype of transfused BM cells (Figure 6D). However, we could not detect significant differences in the resultant plaque areas between miR-33b^−/−Apoe^−/− and miR-33b^+/+Apoe^−/− mice transfused with the same genotype BM cells (Figure 6A–6C).