699. shRNA Sense Strand Neutralization Reduces Off-Targeting, Ameliorates Toxicity and Enhances Efficacy of RNAi-Based HBV Gene Therapy

Abstract

A decade after first reports of successful inhibition of hepatitis B virus (HBV) replication in vivo using shRNAs, only a single RNAi therapeutic is under phase II efficacy investigation in Hepatitis B patients. This slow clinical translation is puzzling, as current standard-of-care requires life-long and daily drug treatment which is often accompanied by side effects and occurrence of resistant HBV mutants. Also, while preventing the generation of infectious HBV virions, reverse-transcriptase inhibitors fail to suppress HBV surface and e-antigen, which enable viral persistence by modulating the host immune system and contribute to HBV-associated hepatocellular carcinoma. These problems may be solved using RNAi as anti-HBV therapeutic since (i) long-term suppression can be achieved by a single dose of an shRNA-encoding vector, and (ii) all four HBV transcripts share a common 3 end, allowing concurrent inhibition of HBV pre-genomic RNA and all viral proteins with a single RNAi molecule, and potentially restoring antiviral immunity and preventing carcinogenesis. Still, clinical translation remains hampered by safety concerns as shRNA over-expression in the mouse liver causes elevated liver transaminases, jaundice and weight loss. One explanation is that ectopic RNAi triggers overload the endogenous miRNA pathway, perturbing miRNA biogenesis and/or activity, and causing cytotoxicity. Evidence for this saturation model is that (i) shRNAs circumvent Drosha processing, a gatekeeper for miRNA biogenesis, (ii) embedding shRNAs in a recombinant miRNA context yields lower and safer levels of mature RNAi triggers, and (iii) overexpression of Argonaute-2 (Ago2) ameliorates toxicity and enhances RNAi efficacy. Moreover, RNAi triggers can perturb cell physiology through unwanted inhibition of off-target genes with partial complementarities to one of the two strands of the RNAi molecule. To alleviate such unintended gene silencing, we have developed a novel bi-cistronic AAV vector that expresses, in addition to the shRNA, a second RNA hairpin called tough decoy or TuD. In cell culture, the TuD effectively sequestered and inactivated shRNA sense strands, and thereby improved RNAi specificity. These remarkable features translated well into an HBV-transgenic mouse model, where sense strand off-target activity was likewise prevented, as validated by transcriptome analysis of liver RNA, coinciding with ameliorated toxicity. Enhanced in vivo safety was also noted for two alternative AAV vectors, either co-expressing the shRNA and Ago2, or embedding the shRNA in a miR-122 backbone. Notably, the new shRNA/TuD vector outperformed the other expression strategies regarding efficacy with stable HBV reduction of up to 98% over 3 months. We speculate that its enhanced antiviral efficacy results from increased loading of the desired antisense strand into RISC in absence of the sense strand, possibly also further amending toxicity by attenuating RISC saturation. The simple TuD design and the versatility of our new AAV vector pave the way for adaptation of our strategy to other applications and should facilitate clinical translation.