miR-33 links SREBP-2 induction to repression of sterol transporters

miR33 Is Evolutionarily Conserved in Intron 16 of SREBP-2.

In an attempt to identify miRNAs that might affect cholesterol homeostasis, we performed an in silico analysis of human genes encoding nuclear receptors and transcription factors known to affect lipid homeostasis. This analysis identified sequences corresponding to miR-33 within intron 16 of human SREBP-2 (Fig. 1A). Importantly, these sequences are conserved across multiple animal species (Fig. 1B). To test the hypothesis that Srebp-2 and miR-33 are coexpressed and regulated by the same metabolic/sterol stimuli, we incubated mouse and human primary macrophages in media containing low or high sterols (see Materials and Methods), conditions known to regulate Srebp-2, SREBP-2 target genes, and LXR target genes (7–16). As expected, excessive exogenous cholesterol resulted in increased expression of the LXR target gene Abca1, with concomitant repression of the SREBP-2 target Ldl-r, whereas the opposite pattern of gene expression was found in cells incubated in media lacking cholesterol (Fig. 1 C and D). Importantly, and in agreement with our proposal, miR-33 levels paralleled those of Srebp-2; they were repressed in cells incubated in high sterols and induced in cells incubated in media devoid of cholesterol (Fig. 1 C and D).

Human and Murine ABCA1 and Murine ABCG1 Have Functional miR-33 Responsive Elements in Their 3′ UTR.

Because SREBP-2 and LXR control antagonistic aspects of cellular sterol homeostasis, we hypothesized that miR-33 might be involved in repression of LXR target genes, such as ABCA1 and ABCG1, which are known to promote the efflux of cholesterol from cells (17–25). Analysis of the 3′ UTR regions of ABCA1 and ABCG1 identified sequences in both genes that are partially complementary to miR-33 sequences (Fig. 2 A–C). In the case of human ABCA1, we identified a proximal element (nucleotides 120–172 downstream of the stop codon, which we have termed box 1) that contains three overlapping putative miR-33–responsive elements, and a distal element (nucleotides 1465–1481 after the stop codon; box 2) (Fig. 2 A and C). For human ABCG1, a single putative miR-33–responsive element was identified at nucleotides 516–541 after the stop codon (Fig. 2 B and C). The same putative sequences are found in nucleotides 120–171 and 1426–1442 in the mouse ABCA1 gene, and in nucleotides 715–733 in the mouse ABCG1 gene. Importantly, these sequences are evolutionarily conserved across animal species (Fig. 2 A and B). To assess the functionality of these sequences, we transfected HEK293 cells with luciferase reporters containing the different miR-33 putative elements inserted after the stop codon. Cotransfection of a plasmid expressing miR-33 resulted in a 50–60% decrease in luciferase activity when the reporter plasmid contained the human/mouse box 1 sequence from ABCA1 or the mouse ABCG1 sequence (Fig. 2D, lanes 5, 6, 9, and 10). The specificity of this effect is supported by the finding of no repression when the reporter gene contained either sequences corresponding to box 2 of ABCA1 (Fig. 2D), or mutant box 1 or mutant mouse ABCG1 sequences (Fig. S1). Intriguingly, the miR-33–responsive element in the human ABCG1 gene is degenerate compared with the mouse and rat sequences and does not confer miR-33 responsiveness (Fig. 2 B–D, lanes 11 and 12). In comparison, a perfect miR-33 complementary sequence resulted in 95% repression of luciferase activity (Fig. 2D, lanes 3 and 4).

We next measured the ability of miR-33 to modulate the expression of endogenous ABCA1 and ABCG1 mRNAs. To accomplish this, we transduced Hep3B cells with an empty adenovirus or an adenovirus encoding miR-33, and then incubated them in the absence or presence of agonists for LXR and RXR (T0901317 and 9-cis-retinoic acid, respectively) (Fig. 2E). In agreement with the foregoing data, overexpression of miR-33 resulted in reduced expression of ABCA1, but not ABCG1, mRNA (Fig. 2E, lanes 1–4), as well as reduced ABCA1 protein expression in these human hepatoma cells (Fig. 2F); other lipid metabolism genes remained unaffected by miR-33 expression (Fig. 2E, lanes 5–8). Collectively, these results identify human and murine ABCA1 and murine ABCG1 as bona fide targets for miR-33.

Manipulation of miR-33 Expression Alters Cholesterol Metabolism Both in Vitro and in Vivo.

We tested the hypothesis that miR-33–dependent repression of ABCA1 and/or ABCG1 could affect cellular sterol homeostasis in vitro and in vivo. To test the proposal that miR-33 modulates the efflux of cellular cholesterol, we transfected HEK293 cells with an empty plasmid or a plasmid encoding miR-33 and then incubated the cells with [³H]-cholesterol in the presence or absence of ligands for LXR and RXR to induce ABCA1 and ABCG1. Importantly, altering miR-33 levels produced no differences in total cellular cholesterol content or [³H]-cholesterol uptake (Fig. S2). The ability of the cells to efflux the radiolabeled sterol to BSA, ApoAI, and FBS was analyzed 6 h later (see Materials and Methods for details). The data shown in Fig. 3A demonstrate that after activation of LXR:RXR, miR-33 overexpression resulted in reduced cholesterol efflux to ApoAI or FBS (compare lanes 5 and 6 to lanes 7 and 8, and lanes 9 and 10 to lanes 11 and 12). Thus, miR-33 overexpression decreases the ability of cells to efflux cholesterol. Conversely, silencing miR-33 using antisense oligonucleotides resulted in enhanced LXR agonist–dependent cholesterol efflux in the same HEK293 cells (Fig. 3B; compare lanes 5 and 6 to lanes 7 and 8, and lanes 9 and 10 to lanes 11 and 12). These data offer a molecular explanation for previously reported results showing that treatment with statins, which induce Srebp-2, resulted in decreased expression of both ABCA1 and/or ABCG1 mRNAs and attenuated cholesterol efflux in macrophages and endothelial cells (26, 27).

To put these latter results in a more physiological context, we injected empty or miR-33–expressing adenovirus, or scrambled or anti–miR-33 antisense oligonucleotides, into the tail veins of C57BL/6 mice and measured hepatic ABCA1 mRNA and protein expression and plasma lipoprotein and cholesterol levels 5 d or 12 d after the infusion of the adenovirus or the antisense oligonucleotides, respectively. The data show that miR-33 adenoviral transduction resulted in reduced ABCA1 mRNA and protein (Fig. 3C), along with a 19% reduction in total plasma cholesterol (130.0 ± 15.0 vs. 159.8 ± 6.4 mg/dL, P = 0.00002) and a 29% reduction in HDL-cholesterol (66.8 ± 16.8 vs. 93.6 ± 23.6 mg/dL, P = 0.0086) , compared with mice transduced with the empty adenovirus (n = 6 for Ad-GFP; n = 8 for Ad-miR33) (Fig. 3D). Conversely, silencing miR-33 with antisense oligonucleotides resulted in a substantial increase in both Abca1 expression (Fig. 3E) and HDL-cholesterol levels as measured by FPLC (Fig. 3F). Importantly, these effects were specific for Abca1, because the expression of other lipid-related genes did not change following manipulation of hepatic miR-33 expression (Fig. S3). Collectively, these in vitro and in vivo studies strongly suggest that miR-33 modulates intracellular cholesterol homeostasis and lipoprotein metabolism, presumably by silencing the expression of ABCA1. Whether other miR-33 targets are also essential for this effect remains to be determined.

Plasmid Constructs and Production of Adenovirus.

A fragment containing miR-33 flanked by 150 bp upstream and 150 bp downstream of genomic sequence was amplified by PCR using Platinum Pfx enzyme (Invitrogen) from mouse genomic DNA obtained from tail biopsy specimens from C57BL/6 mice. The reverse primer used in the PCR contained the appropriate terminator sequence for RNA-pol-III (TTTTTCT). This fragment was cloned into the HpaI-XhoI sites of pSicoR-GFP (Addgene), which provides a U6 promoter to control the expression of the transgene (and a GFP cassette under control of a cytomegalovirus promoter), thus generating pSicoR-miR33. The integrity of the clones was analyzed by sequencing. Adenoviruses were produced using the AdEasy Adenoviral System (Stratagene) by cloning a XbaI-XhoI fragment containing the U6 promoter with or without the miR-33 sequences from the pSicoR vectors into pAdTrack (which also contains a GFP cassette). The pAdTrack and pAdTrack-miR33 vectors were electroporated into pAdeasy-1 cells to generate the final adenoviral vectors. Recombinant vectors were identified by the presence of a 4.5-kb PacI restriction fragment. These final vectors were transfected individually into AD293 cells (Stratagene) for the production and amplification of the adenoviral particles. The adenoviruses were purified by CsCl gradient ultracentrifugation, and the titers were calculated to be 1 × 10¹² plaque-forming units (pfu)/μL for Ad-GFP and 1 × 10¹¹ pfu/μL for Ad-miR33. Experiments that used the empty pSicoR or the empty adenoviral vector are referred to as “gfp.”

Cells.

HEK293 cells (American Type Culture Collection) and Hep3B cells (a kind gift from Dr. Richard Lee, Isis Pharmaceuticals) were maintained in DMEM plus 10% FBS. Human monocyte-derived macrophages from three healthy donors and mouse thioglycollate–elicited primary peritoneal macrophages were obtained and maintained as described previously (36). Murine macrophages were incubated for 48 h in media A [DMEM supplemented with 10% lipoprotein-deficient serum (LPDS) (Intracel), 100 μmol/L mevalonic acid (Sigma-Aldrich), and 50 μmol/L pravastatin (Cayman Chemicals)] in the presence or absence of 10 μg/mL of cholesterol (Sigma-Aldrich) and 1 μg/mL of 25-hydroxycholesterol (Sigma-Aldrich). Human macrophages were incubated in media A with or without supplementation with 40 μg/mL of oxidized LDL (Biomedical Technologies). Where indicated, cells were transfected with pSicoR-GFP or pSicor-miR33, or with miRIDIAN-scrambled or miR33-specific hairpin inhibitors (Dharmacon), using Lipofectamine 2000 (Invitrogen) or Dharmafect 1 (Dharmacon) reagents, respectively, or were transduced with Ad-GFP or Ad-miR-33 adenovirus. Where indicated, cells were incubated for 16 h in the absence or presence of agonists for LXR (1 μmol/L T0901317; Cayman Chemicals) and RXR (1 μmol/L 9-cis-retinouic acid; Sigma-Aldrich).

Cholesterol Efflux Assays.

HEK293 cells were seeded in 24-well plates (0.25 × 10⁶ cells per well) and transfected with pSicoR-GFP or pSicoR-miR33, or with scrambled or miR33-specific hairpin inhibitors. After 36 h in complete media, the cells were washed and incubated for an additional 16 h in DMEM supplemented with 10% LPDS and [³H]-cholesterol (1 μCi/mL). After 16 h, the cells were washed with PBS and then incubated in DMEM supplemented with 0.2% BSA for a 2-h equilibration period. When indicated, the cells were incubated with T0901317 (1 μmol/L) and 9-cis retinoic acid (1 μmol/L) during radiolabeling and equilibration. To determine cholesterol efflux, the cells were rinsed three times with PBS and then incubated for 6 h in DMEM supplemented with 0.2% BSA and, where indicated, either ApoAI (15 μg/mL) or FBS (10%). Subsequently, the media was removed, the cells washed with PBS, and the radioactive content of the media and cells was determined by scintillation. Cholesterol efflux was determined by dividing the radioactive content of the media by the sum of the radioactivity in the cells and media.

RNA Analysis.

RNA was isolated from cells with TRIzol reagent, using a slight modification of the manufacturer's protocol to preserve miRNAs. In brief, after homogenization in 2 mL of TRIzol, 400 μL of chloroform and 200 μL of 2 mol/L sodium acetate (pH 4.2) were added before separation of the organic and aqueous phases. The latter was combined with 500 μL of acid phenol:chloroform:isoamylic acid and then centrifuged again. RNAs in the aqueous phase were then precipitated with 3 mL of ethanol overnight at 4 °C. After centrifugation, the RNA pellet was resuspended in water and stored at −80 °C until use. Alternatively, RNAs were isolated using the mirVana miRNA Isolation Kit (Ambion). cDNAs were generated from 5 μg of DNase1-treated RNA using SuperScript III and Random hexamers (Invitrogen). Real-time PCR was done with SYBR Green reagent (Roche), using a LightCycler 480 real-time PCR detection system (Roche). Primer sets are available from the authors on request. Values were normalized to GAPDH and calculated using the comparative C_T method. The expression of miR-33 was normalized to U6 and analyzed using Taqman miRNA assays (Applied Biosystems).

Protein Analysis.

Protein extracts from cells or livers were obtained and analyzed as described previously (31). ABCA1 levels were determined using rabbit polyclonal anti-ABCA1 (Novus Biologicals, 1:1,000 dilution in TBS + 1% Tween 20) and normalized to α-tubulin (Santa Cruz Biotechnologies, 1:5,000 dilution in TBS + 1% Tween 20).

Murine Studies.

Male C57BL/6J mice were obtained from Jackson Laboratories and maintained in a 12-h/12-h light/dark cycle with unlimited access to chow and water. The mice (n = 6–8 per group) were infused via tail vein injection with adenoviral vectors (2 × 10⁹ pfu) or with control or anti–miR-33 antisense oligonucleotides (5 mg/kg/d for 3 consecutive d; miRagen Therapeutics). Blood and livers were collected 5 d (for adenovirus) or 12 d (for antisense oligonucleotides) postinfusion. Total cholesterol and HDL-cholesterol were assayed enzymatically with Waco Chemicals Cholesterol E and HDL-Cholesterol E Kits. Plasma lipoprotein profiles for cholesterol were determined in pooled plasma samples from five mice by FPLC as described previously (37).

Note Added in Proof.

While this manuscript was under revision, two independent laboratories reported similar findings (38, 39).