Supplementary Information for:
Fatality in mice due to oversaturation of cellular micro/short hairpin RNA pathways
Dirk Grimm, Konrad L. Streetz†, Catherine L. Jopling*, Theresa A. Storm, Kusum Pandey, Corrine R. Davis#, Patricia Marion$, Felix Salazar$, and Mark A. Kay
Supplementary Methods
AAV vectors. We derived a vector from adeno-associated virus serotype 8 (AAV8), based on the superior efficiency of this isolate in liver5. Work with AAV2 suggested minimal doses could be used by encapsidating "double-stranded" (ds) vector DNA6 (wild-type AAV is a single-stranded DNA virus7). This led us to engineer an optimized vector consisting of a dsAAV2 genome pseudotyped with AAV8 capsids. We replaced one AAV2 DNA packaging signal with one from AAV4, yielding a novel "stabilized double-stranded" (sds) vector with significantly enhanced properties over conventional dsAAV (D.G. & M.A.K., manuscript in preparation). Details of the novel sdsAAV vector backbone, containing one intact AAV2 and one mutated AAV4 DNA packaging element, will be reported (D.G. & M.A.K., manuscript in preparation). Viral particles were generated/purified via a triple-transfection method and titered by dot blot29, considering that sdsAAV virions carried two shRNA copies. Typical particle yields from ~2x109 transfected 293 cells were at least 1x1013 per ml (total volume ~4 ml) and thus identical to conventional AAV vectors29. AAV vector genome copy numbers in total liver DNA were quantified as described5.
Design and cloning of shRNAs. All shRNAs used in this study were expressed from a human U6 promoter27, except for the final five shRNAs in Supp. Table 1 (target FATP or NONE) which were expressed from a human H1 promoter (kind gift from H. Doege and A. Stahl). To facilitate shRNA cloning, we engineered the sdsAAV vector plasmid to contain the U6 promoter followed by unique Bbs I restriction enzyme sites, allowing insertion of shRNAs as annealed oligonucleotides with appropriate overhangs (details will be reported, D.G. & M.A.K.). All hairpins were designed such that the sense came before the antisense strand (separated by a 7 to 9 nucleotide loop, Supp. Table 1) and verified by DNA sequencing.
Cell culture studies. Human Huh7 hepatoma cells in 6 cm dishes were transiently transfected with a plasmid encoding a destabilized gfp gene fused with a perfectly complementary miR-122 target site, shown to mediate miR-122 cleavage of the fusion transcript12. Selected dishes were co-transfected with equimolar ratios of shRNA-expressing sdsAAV plasmids, using Superfect transfection reagent (QIAGEN) according to the manufacturer’s protocol. Gfp expression was analysed and documented using a Zeiss Axiovert 200 fluorescence microscope and provided software.
Mice. Mice of the FVB strain transgenic for the hAAT gene (under a hepatocyte-specific promoter), or carrying an integrated copy of the HBV genome (STC lineage), were reported19,30. Animals were selected such that all groups had comparable average initial hAAT or HBsAg levels. Intravenous sdsAAV8 injections (total volume 300 ml in PBS) or high pressure delivery of plasmid DNA (in 1.8 ml 0.9% NaCl) were performed at an age of six to ten weeks, using established methods5. Blood was collected via retro-orbital bleeding, and plasma (hAAT) or serum (HBV) were prepared as described27,30. To induce liver damage, mice were given an intraperitoneal injection of 8 mg per 20 g mouse in mineral oil, twice a week. For characterization of shRNA toxicity, wildtype FVB or C57/BL6 mice (Jackson Laboratories) were injected at an age of six weeks. All procedures were approved by the Animal Care Committee at Stanford University.
Measurement of HBV protein, DNA and RNA. HBsAg was quantified in mouse sera (pre-diluted between 1:50 and 1:500 in PBS with 10% fetal calf serum) using a commercial ELISA kit (AUSZYME® Monoclonal Diagnostic Kit, Abbott Laboratories, protocol 'C'). Serum HBV DNA was assessed by quantitative real-time PCR19. For quantification of HBV RNA, 20 mg of total liver RNA (isolated from ~100 mg tissue using TRIZOL reagent (Invitrogen)) were electrophoresed on 0.8% agarose-formaldehyde gels, transferred onto Hybond N+ membranes (Amersham Pharmacia Biotech) and probed with a 32P-labeled full-length HBV genome. HBV mRNAs were detected and quantified using a phosphoimager (Personal Molecular Imager FX, BIO-RAD).
Mouse pathology. Complete pathology analyses (necropsy, histology, chemistry and electrolyte panels, complete blood counts) were performed at the Department of Comparative Medicine at Stanford University. Briefly, tissues were fixed in 10% neutral buffered formalin, processed routinely into paraffin, sectioned at 5 microns and stained with hematoxylin and eosin for light microscopic examination. The automated systems used by the lab are: electrolytes: Ciba-Corning 644 Electrolyte analyzer; serum chemistries: Chiron Diagnostics Express-Plus; and complete blood counts: Cell-Dyn 3500R (Abbott). Cytokines were quantified in undiluted mouse serum, using commercial ELISA kits and provided protocols (Mouse IFN Alpha or Beta ELISA KIT, both PBL Biomedical Laboratories; OPTEIATM Mouse IFN-g Set, TNF Kit II or IL-6 Set, all BD Biosciences). Total and phosphorylated (Ser-51) eIF2a as well as p21 and p53 were detected via Western blotting, using total liver protein and specific antibodies with provided protocols (Cell Signaling, Novocastra and Abcam).
Analysis of Luciferase and hAAT expression. Luciferase expression in living mice and plasma hAAT levels were analysed as described28,,30. For histology, livers were embedded in OCT compound (Tissue-Tek, Sakura Finetek), cryosectioned and immunostained using antibodies against hAAT (RDI) or incorporated BrdU (BD Dickinson). BrdU labeling was performed by intraperitoneal injection of selected mice with 4 mg BrdU (Sigma) two hours prior to sacrificing.
Small RNA Northern blot analyses. Total RNA was extracted using TRIZOL and electrophoresed as reported recently12, along with probes to detect pre- and mature miR-122 and methods for their generation. Additional probes were (all 5'-3') : let-7a AACTATACAACCTACTACCTCA, hAAT-sense GAAGCGTTTAGGCATGTTT, hAAT-anti AAACATGCCTAAACGCTTC, luc-sense TCCCGCTGAATTGGAATCC, luc-anti GGATTCCAATTCAGCGGGA, sAg-sense TTACTAGTGCCATTTGTTC, sAg-anti GAACAAATGGCACTAGTAA, cAg-sense AGAAGAACTCCCTCGCCTC, cAg-anti GAGGCGAGGGAGTTCTTCT. To control for equal loading, the same total RNAs were analysed via conventional Northern blotting. Probes to detect mouse b-actin, GAPDH and cyclophilin mRNAs, or 28S rRNA, were generated using the Mouse Internal Standard Screening and the MAXIscript T7 Kits (both Ambion). In Fig. 2a, the 'sense' signals for sAg-25', cAg-25' and luc-29 are underrepresented because of limited probe binding due to three nucleotide mismatches in each shRNA (Supp. Table 1). Also note that the sizes of the shRNA precursors could vary, for reasons unknown. In Fig. 2b and Supp. Fig 5, probing for 28S rRNA and cyclophilin (CP) mRNA showed equal RNA loading, while b-actin and GAPDH mRNAs were elevated in ailing mice, typical for proliferating hepatocytes. Simultaneous probing against pre- and mature miR-122 or let-7a verified global miRNA repression.
Supplementary Table 1 | List of shRNAs expressed from sdsAAV8 in vivo
hAAT / hAAT-19 / 19 / GAAGCGTTTAGGCATGTTT / AAACATGCCTAAACGCTTC / A / -
hAAT-21 / 21 / GAAGCGTTTAGGCATGTTTAA / TTAAACATGCCTAAACGCTTC / A / +
hAAT-23 / 23 / GAAGCGTTTAGGCATGTTTAACA / TGTTAAACATGCCTAAACGCTTC / A / +
hAAT-25 / 25 / GAAGCGTTTAGGCATGTTTAACATC / GATGTTAAACATGCCTAAACGCTTC / A / ++
A-25 / 25 / GAAGCGTTTAGGCATGTTTAACATC / GATGTTAAACATGCCTAAACGCTTC / B / ++
R-25 / 25 / GATGTTAAACATGCCTAAACGCTTC / GAAGCGTTTAGGCATGTTTAACATC / A / ++
A-3 / 21 / AAGCGTTTAGGCATGTTTAAC / GTTAAACATGCCTAAACGCTT / A / +
A-6 / 19 / AAGCGTTTAGGCATGTTTA / TAAACATGCCTAAACGCTT / A / -
R-19 / 19 / AAACATGCCTAAACGCTTC / GAAGCGTTTAGGCATGTTT / A / -
Luc / Luc-19 / 19 / TCCCGCTGAATTGGAATCC / GGATTCCAATTCAGCGGGA / A / ++
Luc-21 / 21 / GCTCCCGCTGAATTGGAATCC / GGATTCCAATTCAGCGGGAGC / A / ++
Luc-23 / 23 / TGGCTCCCGCTGAATTGGAATCC / GGATTCCAATTCAGCGGGAGCCA / A / +
Luc-25 / 25 / GGTGGCTCCCGCTGAATTGGAATCC / GGATTCCAATTCAGCGGGAGCCACC / A / -
Luc-29 / 29 / ATCGGGCGGCTCTCGCTGAGTTGGAATCC / GGATTCCAATTCAGCGGGAGCCACCTGAT / B / +
L19.1 / 19 / GGTGGCTCCCGCTGAATTG / CAATTCAGCGGGAGCCACC / A / ++
L19.2 / 19 / GCTCCCGCTGAATTGGAAT / ATTCCAATTCAGCGGGAGC / A / -
HBsAg / sAg-19 / 19 / TTACTAGTGCCATTTGTTC / GAACAAATGGCACTAGTAA / A / -
sAg-21 / 21 / GTTTACTAGTGCCATTTGTTC / GAACAAATGGCACTAGTAAAC / A / +
sAg-23 / 23 / CAGTTTACTAGTGCCATTTGTTC / GAACAAATGGCACTAGTAAACTG / A / +
sAg-25 / 25 / CTCAGTTTACTAGTGCCATTTGTTC / GAACAAATGGCACTAGTAAACTGAG / A / ++
sAg-25' / 25 / CTCGGTTTATTAGTGCCGTTTGTTC / GAACAAATGGCACTAGTAAACTGAG / B / ++
FA1 / 25 / CTCAGTTTACTAGTGCCATTTGTTC / GAACAAATGGCACTAGTAAACTGAG / B / ++
FA2 / 25 / CTCGGTTTATTAGTGCCGTTTGTTC / GAACAAACGGCACTAATAAACCGAG / B / ++
FA7 / 25 / CTCGGTTTATTAGTGCCGTTTGTTC / GAACAAATGGCACTAGTAAACTGAG / A / +
hbv22 / 19 / GGCTCAGTTTACTAGTGCC / GGCACTAGTAAACTGAGCC / A / -
M3 / 21 / ATTGTGAGGATTCTTGTCAAC / GTTGACAAGAATCCTCACAAT / A / +
M4 / 21 / ATACAGGTGCAATTTCCGTCC / GGACGGAAATTGCACCTGTAT / A / ++
M5 / 21 / TGTAACACGAGAAGGGGTCCT / AGGACCCCTTCTCGTGTTACA / A / +
M6 / 21 / ACAAGTTGGAGGACAGGAGGT / ACCTCCTGTCCTCCAACTTGT / A / +
M7 / 21 / TGGTACAGCAACAGGAGGGAT / ATCCCTCCTGTTGCTGTACCA / A / +
HBcAg / cAg-19 / 19 / AGAAGAACTCCCTCGCCTC / gaggcgagggagttcttct / A / -
cAg-21 / 21 / GAAGAAGAACTCCCTCGCCTC / gaggcgagggagttcttcttc / A / ++
cAg-23 / 23 / TAGAAGAAGAACTCCCTCGCCTC / gaggcgagggagttcttcttcta / A / ++
cAg-25 / 25 / CCTAGAAGAAGAACTCCCTCGCCTC / gaggcgagggagttcttcttctagg / A / ++
cAg-25' / 25 / CCTAGGAGAAGGACTCCCTTGCCTC / gaggcgagggagttcttcttctagg / B / ++
FA3 / 25 / CCTAGAAGAAGAACTCCCTCGCCTC / gaggcgagggagttcttcttctagg / B / ++
FA4 / 25 / CCTAGGAGAAGGACTCCCTTGCCTC / gaggcAagggagtCcttctCctagg / B / ++
FA5 / 25 / CCTAGGAGAAGGACTCCCTTGCCTC / gaggcgagggagttcttcttctagg / A / ++
FA6 / 25 / CCTAGAAGAAGAACTCCCTCGCCTC / gaggcAagggagtCcttctCctagg / B / +
K19.1 / 19 / CCTAGAAGAAGAACTCCCT / agggagttcttcttctagg / A / ++
K19.2 / 19 / GAAGAAGAACTCCCTCGCC / ggcgagggagttcttcttc / A / -
HCV / HCV1 / 20 / GCGAAAGGCCTTGTGGTACT / AGTACCACAAGGCCTTTCGC / B / -
HCV2 / 20/21 / GTGCACGGTCTACGAGACCTC / GAGGTCTCGTAGACCGTGCA / B / ++
HCV3 / 19 / ATTGGAGTGAGTTTAAGCT / AGCTTAAACTCACTCCAAT / B / ++
FATP / F2-6 / 19 / GGTATGAGCTGATCAAGTA / TACTTGATCAGCTCATACC / C / ++
F2-7 / 19 / GGCGACATCTACTTCAACA / TGTTGAAGTAGATGTCGCC / C / -
F5-2 / 19 / GTGGAAATCTCCTGCCATA / TATGGCAGGAGATTTCCAC / C / ++
F5-3 / 19 / GTTCTCTGCCTCCCGATTC / GAATCGGGAGGCAGAGAAC / C / -
NONE / SCR / 19 / gatcgaatgtgtacttcga / tcgaagtacacattcgatc / C / -
'Target', gene targeted by the shRNAs (HCV, Hepatitis C Virus; FATP, fatty acid transporter protein; NONE, no target in the mouse genome); 'shRNA', name of shRNA with shown stem length ('Stem', in nucleotides), sequence ('Sense' and 'Antisense' strand (underlined nucleotides were mismatches between the strands) and loop ('Loop', sequences ranged from 7 to 9 nucleotides, 5' TCAAGAG 3' (A), 5' GAAGCTTG 3' (B), or 5' TTCAAGAGA 3' (C)); 'Toxic', degree of in vivo toxicity from injection of 1x1012 particles per mouse (measured by RNAi persistence and/or morbidity onset; '-', no/mild toxicity; '+', strong toxicity, causing rapid RNAi loss and ultimately morbidity; '++', severe toxicity, resulting in early lethality); SCR, scrambled control shRNA.
Supplementary Figure Legends
Supplementary Figure 1 | sdsAAV8-mediated in vivo anti-luciferase RNAi. FVB mice were injected with Firefly luciferase-expressing sdsAAV8 at particle doses of 2x1011 (controls, mice 1 and 2), 5x1011 (mouse 3) or 2x1012 (mouse 4). Animal 1 was co-injected with 1x1012 sdsAAV8 expressing a control anti-hAAT shRNA. Another control mouse remained non-injected (right). The top panels document dose-dependent and persistent Luciferase expression. Mice in lower panels were co-injected with 2x1011 luciferase expression vector and the shown doses of sdsAAV8 encoding anti-luciferase shRNAs (numbers are stem lengths, Supp. Table 1). Luciferase repression was rapid and effective with luc-29, while other shRNAs caused liver toxicity (evident from artificially high Luciferase levels) and ultimately death.
Supplementary Figure 2 │ shRNA toxicity does not require a target and is not caused by general shutdown of liver protein synthesis. hAAT-transgenic mice were peripherally infused with a high dose of sdsAAV-8 (2x1012 particles per animal), expressing the anti-HBsAg 25-mer or the anti-hAAT 19-mer shRNA. Plasma hAAT levels were determined at various timepoints after the injection. Shown are two representative mice per group. All sAg-25-injected animals died within three weeks, but continued to express high hAAT levels in the liver until the timepoint of death, proving that toxicity did not correlate with a general block of liver protein synthesis. It moreover exemplifies that shRNA-induced lethality did typically not require the presence of an shRNA target. In contrast, mice treated with the anti-hAAT vector showed a rapid and strong, albeit transient (see main text for reasons), drop in hAAT levels, in the absence of morbidity. To rule out that the transient RNAi was due to hAAT escape mutants, the mice were re-injected (purple arrow) with 3x1011 particles of the same sdsAAV8, or the same shRNA pseudotyped with AAV1. hAAT levels dropped again in the AAV1-injected mouse, proving that the shRNA target was still present and accessible. They remained stable in the AAV8-re-injected mouse, albeit likely only due to anti-AAV8 antibodies from the first injection.
Supplementary Figure 3 | shRNA overexpression does not induce eIF2a phosphorylation or changes in p21/p53. Total protein was extracted from livers of sdsAAV8-treated mice at the onset of morbidity and analysed for total and phosphorylated (Ser-51) eIF2a, or p21 and p53 protein via Western blotting. Shown as examples are the anti-luciferase shRNA-treated mice (numbers indicate shRNA stem lengths; C is an untreated control mouse). shRNA-induced toxicity did not affect the phosphorylation status of eIF2a, and had no effect on levels of p21 or p53 proteins.
Supplementary Figure 4 | Induction of liver regeneration does not generally affect cellular RNA levels. The same total RNAs as shown in panels b and c of Figure 2 (main text) were probed for cyclophilin (CP) mRNA and 28S rRNA. In contrast to b-actin (and GAPDH) (Fig. 2b and c, main text), the levels of these endogenous RNAs remained relatively stable in regenerating liver, regardless of induction of hepatocyte proliferation by shRNA overexpression (a), or by surgical liver injury (b).
Supplementary Figure 5 | Model for competition of shRNA and miRNA pathways. miRNAs are generated from long primary molecules (pri-miRNAs), which are processed by the nuclear enzyme Drosha into shorter precursors (pre-miRNAs). As of yet, it is unclear whether shRNAs share this particular step. Both pre-miRNAs and shRNAs are exported to the cytoplasm via the karyopherin exportin-5. Our data suggest that three factors determine the extent of shRNA interaction with exportin-5, and thus the level of competition with miRNAs : shRNA length, sequence and dose. Once in the cytoplasm, pre-miRNAs and shRNAs share the enzyme Dicer, catalyzing their cleavage into effector molecules. The overall ratio of shRNAs to miRNAs will also determine the extent and outcome of competition for this enzyme : low amounts of shRNA will readily be tolerated by the cell, but may be insufficient to trigger effective target degradation (scenario 1). On the other hand, shRNA excess will saturate the processing machinery (exportin-5 and/or Dicer and co-factors) and thus inhibit miRNA expression/function. As shown, this can have detrimental effects on the cell and ultimately the organism (scenario 3). Ideally, shRNAs are expressed at levels not saturating the miRNA processing pathway, but sufficiently high to give strong and stable RNAi (scenario 2).