Inhibition of miR-33 a / b in non-human primates raises plasma HDL and lowers VLDL triglycerides

Cardiovascular disease remains the leading cause of mortality in westernized countries, despite optimum medical therapy to reduce the levels of low-density lipoprotein (LDL)-associated cholesterol. The pursuit of novel therapies to target the residual risk has focused on raising the levels of high-density lipoprotein (HDL)associated cholesterol in order to exploit its atheroprotective effects. MicroRNAs (miRNAs) have emerged as important posttranscriptional regulators of lipid metabolism and are thus a new class of target for therapeutic intervention. MicroRNA-33a and microRNA-33b (miR-33a/b) are intronic miRNAs whose encoding regions are embedded in the sterol-response-element-binding protein genes SREBF2 and SREBF1 (refs 3–5), respectively. These miRNAs repress expression of the cholesterol transporter ABCA1, which is a key regulator of HDL biogenesis. Recent studies in mice suggest that antagonizing miR-33a may be an effective strategy for raising plasma HDL levels and providing protection against atherosclerosis; however, extrapolating these findings to humans is complicated by the fact that mice lack miR-33b, which is present only in the SREBF1 gene of medium and large mammals. Here we show in African green monkeys that systemic delivery of an antimiRNA oligonucleotide that targets both miR-33a and miR-33b increased hepatic expression of ABCA1 and induced a sustained increase in plasma HDL levels over 12 weeks. Notably, miR-33 antagonism in this non-human primate model also increased the expression of miR-33 target genes involved in fatty acid oxidation (CROT, CPT1A, HADHB and PRKAA1) and reduced the expression of genes involved in fatty acid synthesis (SREBF1, FASN, ACLY and ACACA), resulting in a marked suppression of the plasma levels of very-low-density lipoprotein (VLDL)-associated triglycerides, a finding that has not previously been observed in mice. These data establish, in a model that is highly relevant to humans, that pharmacological inhibition of miR-33a and miR33b is a promising therapeutic strategy to raise plasma HDL and lower VLDL triglyceride levels for the treatment of dyslipidaemias that increase cardiovascular disease risk. Recent advances in the understanding of lipid metabolism have revealed that the genetic loci encoding the transcription factors SREBP1 and SREBP2 (known as SREBF1 and SREBF2) also encode the miRNAs miR-33b and miR-33a, respectively, which regulate cholesterol and fatty acid homeostasis together with their host genes. Although miR-33a and miR-33b differ by two nucleotides in their mature form, they are identical in their seed sequence and thus are predicted to repress the same subset of genes. Notably, miR-33a has been highly conserved throughout evolution, whereas miR-33b is encoded only by the SREBF1 gene of medium and large mammals. This difference between mice and humans may be particularly relevant under conditions in which the transcription of SREBF1, and thus miR-33b, is highly upregulated, such as insulin resistance. Recently, we and others have reported that the silencing of mature miR-33a in mice, by using modified antisense oligonucleotides, by viral delivery of hairpin inhibitors or by targeted deletion of the miR-33-encoding locus, increased the levels of hepatic ABCA1 and circulating HDL by as much as 40%. Although these studies highlight the therapeutic promise of miR-33 inhibitors for raising plasma HDL levels, the absence of miR-33b in mice limits the translational relevance of these findings. Thus, to gain a comprehensive understanding of the effects of inhibiting both miR-33a and miR-33b in a model highly related to humans, we treated African green monkeys (Chlorocebus aethiops) with a 29-fluoro/methoxyethyl (29-F/MOE)-modified, phosphorothioate-backbone-modified, antisense miR33 (denoted anti-miR33), which we showed was equally effective at inhibiting both miR-33a and miR-33b in vitro (Supplementary Fig. 1a). Six animals per group were subcutaneously administered a clinically relevant dose of antimiR-33 (5 mg kg) or a mismatch control twice weekly for the first two weeks and then weekly for the remainder of the study (Fig. 1a). Quantification of hepatic anti-miRNA levels by ion-pairing highperformance liquid chromatography (HPLC) coupled to electrospray mass spectrometry (ES/MS) after 4 and 12 weeks of treatment showed equivalent delivery of anti-miR-33 and control (mismatch) oligonucleotides (Supplementary Fig. 1b). No toxicity seemed to be associated with the anti-miRNA treatment, as shown by the clinical chemistries, blood counts, coagulation markers, body weights and serum cytokine profiles of the monkeys (Fig. 1b and Supplementary Fig. 1c, d), which remained within normal limits throughout the study. Microarray profiling of messenger RNA obtained from liver biopsies after 4 weeks of treatment revealed that anti-miR-33 selectively increased the expression of miR-33 heptamer-matched genes in monkeys fed a normal chow diet (Supplementary Table 1). Of these, the gene encoding the cholesterol transporter ABCA1 was the most highly derepressed miR-33 target gene. Reverse transcription followed by quantitative PCR (RT–qPCR) analysis confirmed the increase in ABCA1 expression, as well as that of other known miR-33 target genes (target site alignment is shown in Supplementary Fig. 2), including the genes encoding two enzymes involved in fatty acid oxidation, CROT and HADHB, and the insulin signalling gene IRS2 (Fig. 1c and Supplementary Fig. 3). To assess the effects of miR-33 inhibition under different metabolic conditions, monkeys were switched after 4 weeks to a high carbohydrate, moderate cholesterol diet for 8 weeks, thereby totalling 12 weeks of treatment. After 8 weeks of the high carbohydrate, moderate cholesterol diet, SREBF1 mRNA levels increased by 5-fold in the control animals, and a corresponding 2.2-fold increase in miR-33b was observed, making its expression more than 7-fold higher than miR-33a (Fig. 1d and Supplementary Fig. 3). Microarray and RT–qPCR analysis showed that