Recently, it was reported that adeno-associated viral (AAV)-expressed short hairpin RNA (shRNA), when administered to mice, can result in profound toxicity, presumably by saturation of the cellular miRNA pathway⁴. AAV expression of shRNA seemed to elicit toxicological effects by interference of cellular pathways for miRNA biogenesis, including transport of nuclear miRNA precursors to the cytoplasm by the nuclear karyphorin exportin-5. shRNA-mediated toxicological effects were evident by marked reductions in cytoplasmic levels of mature miRNA, and occurred in a manner independent of both ‘on-target’ silencing and shRNA sequence. The findings of ref. ⁴ can be explained by viral vector expression of shRNA at high levels, because lower-level shRNA expression seems to be tolerated^4,5. However, the implications for cellular delivery of synthetic siRNA that acts downstream of miRNA biogenesis have remained undetermined.

We interrogated the biochemical and functional status of the miRNA pathway after in vivo administration of synthetic siRNA formulated in liposomal nanoparticles such as those recently described^2,6. Formulated, synthetic siRNA targeting two hepatocyte-expressed gene transcripts (si-Apob targeting apolipoprotein B (Apob) and si-FVII targeting factor VII (F7)) and an irrelevant gene (si-Luc targeting luciferase) were administered by a single intravenous bolus injection at different dose levels in groups of three mice. At both 2 and 30 days, silencing of on-target gene transcripts in liver homogenates was measured by the branched DNA assay², and levels of endogenous miRNA (miR-122, miR-16 and let-7a) were assessed by both northern blot and nuclease protection assays.

Administration of high-dose (5 mg kg⁻¹) si-Apob and si-FVII resulted in marked silencing of the hepatocyte-expressed genes to 22 ± 3% and 17 ± 2% relative to placebo-treated mice at the 2-day time point, respectively (Supplementary Fig. 1). The siRNA effects were dose-dependent, with intermediate silencing effects observed in the animals treated with a low dose (2 mg kg⁻¹) at the same time point. Consistent with previous findings^1,2, the RNA interference (RNAi)-mediated silencing effect was selective, with no measurable effect on Apob mRNA levels with si-FVII or si-Luc treatment. Furthermore, there was no significant effect on F7 mRNA levels in mice treated with si-Apob or si-Luc siRNA. At the 30-day time point, Apob and F7 mRNA returned to near normal levels. Treatment with siRNA was not associated with any observed toxicities over the 30-day time period.

To determine whether either ~80% silencing of two different hepatocyte-expressed genes or hepatocyte delivery of three distinct siRNAs results in dysregulation of miRNA biogenesis, levels of the liver-specific miRNA, miR-122, and the broadly expressed miRNAs miR-16 and let-7a were measured in liver tissue samples at both time points for all treatment groups at all dose levels. As shown in Fig. 1, there was no significant reduction in miRNA levels for individual-siRNA-treated animals, as assessed by northern blot assay. To quantify more accurately miRNA levels, a nuclease protection assay was used to measure simultaneously each miRNA together with the small nuclear RNA U6. Inclusion of the U6 control allowed for normalization of the total amount of RNA in each assay. As summarized in Supplementary Table 1, no significant differences were measured in miRNA levels in siRNA-treated animals. Expression levels ranged from 80% to 110% of the levels measured in PBS-treated animals, and were not associated with any trend related to the siRNA dose level or degree of gene silencing. Certainly, no reduction of miRNA levels to match the >80% level previously reported in ref. ⁴ was observed with administration of synthetic siRNA.

Open in a separate window

Figure 1

Northern blot analysis of mature miRNAs reveals no detectable differences among siRNA treatment groups

Northern blots of total RNA from mouse livers isolated two days (d 2) and 30 days (d 30) after intravenous injections of PBS, si-Apob (5 mg kg⁻¹ and 2 mg kg⁻¹), si-FVII (5 mg kg⁻¹ and 2 mg kg⁻¹) or si-Luc (5 mg kg⁻¹) hybridized to radiolabelled probes specific for miR-122, miR-16 and let-7a (one animal per lane). Hybridization to a radiolabelled probe specific for U6 was used to confirm equal loading of RNA.

The northern blot and nuclease protection results established that synthetic siRNAs do not inhibit the synthesis or processing of cellular miRNAs. We extended these findings to investigate whether systemically administered siRNAs interfere with endogenous miRNA function. Although in vivo data on miRNA function are limited to only a few published reports^7–9, it has been shown that inhibiting mouse miR-122 with antagomirs significantly increases the mRNA levels of a number of liver-expressed genes including Aldoa, Hfe2, Tmed3, Lass6, Slc35a4, Tmem50b and Gpx7 (ref. ⁹, see Supplementary Fig. 2). Accordingly, liver mRNA levels for these seven putative miR-122 targets were quantified using branched DNA assays (normalized to Gapdh) in mice treated with si-Apob, si-FVII and si-Luc (two days after injection), and compared to PBS-treated animals. Moreover, protein levels of total Aldolase and another miR-122 target, Iqgap1 (ref. ⁹), were assayed by western analysis of liver lysates. Consistent with the lack of effect on mature miR-122 levels, there was no increase in the mRNA or protein levels of any of these genes that could be correlated with the dose of siRNA administered or the degree of gene silencing observed (Fig. 2 and Supplementary Fig. 3).

Open in a separate window

Figure 2

Liver mRNA levels of miR-122 targets are unaffected by siRNA treatment

Mouse liver Aldoa, Gpx7, Hfe2, Slc35a4, Tmed3, Tmem50b and Lass6 mRNA levels normalized to Gapdh mRNA measured 2 days after intravenous injections of PBS or different doses (5 mg kg⁻¹ and 2 mg kg⁻¹) of formulated siRNA targeting Apob (si-Apob), F7 (si-FVII) or luciferase (si-Luc) (n = 3 per group). Data are shown as group mean ± s.d. normalized to the PBS group for each gene.

These studies in mice demonstrate that acute gene silencing of ~80% by single administration of siRNA is not associated with any changes in biosynthesis of miR-122, miR-16 or let-7a, nor with any increase in miR-122 target-gene expression. To explore whether prolonged gene silencing by multiple doses of siRNA would alter miRNA levels, and to extend our studies to another species, we targeted the hepatocyte-expressed gene SREBP cleavage-activating protein (Scap) in hamsters. To silence Scap, hamsters were given three, weekly injections of a formulated siRNA targeting Scap (si-Scap) or a mismatch control siRNA (si-Scapmm) at a dose of 2.5 mg kg⁻¹. Repeated treatment with si-Scap resulted in significant (~75%) silencing of Scap mRNA, as well as a very significant reduction in Scap protein on day 21 relative to saline-treated animals (Supplementary Fig. 4). As shown in Fig. 3, silencing of Scap for three weeks by multiple doses of si-Scap had no effect on levels of miR-122, nor did repeat-dosing of the mismatch control, si-Scapmm.

Open in a separate window

Figure 3

Long-term silencing of Scap in hamster liver by repeat administration of siRNA does not affect miR-122 levels

Hamster liver miR-122 levels measured by nuclease protection (normalized to U6 RNA levels) 7 days after three, weekly intravenous injections of PBS or 2.5 mg kg⁻¹ formulated siRNA targeting Scap (si-Scap) or a mismatch control siRNA (si-Scapmm). Data are shown as mean ±s.d.

Our findings demonstrate that robust gene silencing, both acute and longer-term, can be achieved in the liver without any detectable alteration of cellular miRNA biogenesis or function. Certainly, additional studies are warranted such as those with longer ‘dicer substrate’ siRNAs (>21 nucleotides in length) that act upstream of 21-nucleotide siRNA and could interfere with miRNA biogenesis¹⁰. Meanwhile, effective gene silencing with siRNA, in which RNA interference is mediated downstream of miRNA biogenesis, can be dissociated from toxicities that may be intrinsic to DNA-expressed shRNA.