- Email: [email protected]
Determinants of interferon-stimulated gene induction by RNAi vectors
Differentiation (2004) 72:103–111
r International Society of Differentiation 2004
O RI G INA L AR T I C L E
Stephanie Pebernard . Richard D. Iggo
Determinants of interferon-stimulated gene induction by RNAi vectors
Received October 20, 2003; accepted in revised form November 25, 2003
Abstract RNA interference is widely used to silence gene expression in mammalian cells. We recently reported that an shRNA expressed from the H1 promoter in a lentiviral vector could induce the expression of a large group of interferon-stimulated genes (ISGs). This response was unrelated to silencing of the gene targeted by the shRNA MORF4L1. In parallel, we constructed lentiviral vectors expressing shRNA from the U6 promoter and found that these too could induce expression of OAS1, a classic interferon target gene. The U6 vectors give a higher frequency of ISG induction than comparable lentiviral H1 vectors, suggesting that there might be a fundamental flaw in the vector design. We have characterized the U6 vectors in detail and report here that ISG induction is a consequence of the presence of an AA di-nucleotide near the transcription start site. A single nucleotide deletion in the siRNA sequence abolished OAS1 induction, suggesting that the mechanism underlying the response uses a sensor that can detect 19 bp RNA duplexes but not 14 bp duplexes. Adenoviral VA RNA I, which inhibits dsRNA-dependent protein kinase (PKR), was tested as a fusion partner to express shRNA on the grounds that it might prevent nonspecific off-target effects. Fusion of VA RNA I to a lamin shRNA was moderately effective in silencing lamin expression, but gave strong OAS1 induction by an shRNA that does not induce OAS1 when expressed from the U6 or H1 promoters. To avoid interferon induction by U6 vectors, we recommend preserving the wild-type sequence around the transcription start site, in particular a C/G sequence at positions 1/11, and we describe a simple cloning strategy using the Gateway recombination system that facilitates this task. . ) Stephanie Pebernard . Richard D. Iggo (* Oncogene Group Swiss Institute for Experimental Cancer Research (ISREC) Ch. des Boveresses 1066 Epalinges, Switzerland Tel: 141 21 692 58 89 Fax: 141 21 652 69 33 E-mail: [email protected] U.S. Copyright Clearance Center Code Statement:
Key words RNA interference shRNA U6 promoter lentiviral vector interferon-stimulated gene
Introduction RNA interference was first described in C. elegans (Fire et al., 1998). It was originally thought that no comparable silencing mechanism existed in mammalian cells, but in 2000 two groups reported that long dsRNA could silence gene expression in early mouse development (Svoboda et al., 2000; Wianny and ZernickaGoetz, 2000). Tuschl and colleagues then showed that transfection of short dsRNA oligonucleotides leads to specific gene silencing in cultured mammalian cells (Elbashir et al., 2001). To achieve prolonged gene silencing, several groups developed DNA-based vectors to express short hairpin RNAs (shRNAs) from RNA polymerase III promoters (Brummelkamp et al., 2002; Paddison et al., 2002). The small interfering RNA (siRNA), which confers target gene specificity on the RNA-induced silencing complex (RISC), is 21 nt long, and the dsRNA stem in shRNAs expressed from RNA pol-III promoters in DNA vectors is typically 19–26 bp in length. dsRNA molecules over 50 bp are known to induce a complex physiological response characterized by translation inhibition, mRNA degradation, and interferon secretion (Stark et al., 1998; reviewed by Sen, 2001). Although the length of the dsRNAs used for RNAi is below that which is typically required to induce an interferon-stimulated gene (ISG) response, we and others recently reported the activation of interferon signaling and ISG expression following transfection of siRNA or expression of shRNA from lentiviral vectors (Bridge et al., 2003; Sledz et al., 2003). We showed, using micro-array technology, that an shRNA targeting MORF4L1 was capable of inducing a large group of ISGs in human lung fibroblasts (Bridge et al., 2003). The MORF4L1 shRNA was expressed from the H1 promoter cloned into a lentiviral vector. The ISG response was unrelated to the silencing of MORF4L1,
0301–4681/2004/7202/3–103 $ 15.00/0
104
because it was not suppressed by reintroduction of a noncleavable MORF4L1 cDNA. In parallel, we constructed lentiviruses expressing shRNAs from the U6 promoter (Bridge et al., 2003). The U6 vectors give a higher frequency of ISG induction than the H1 vectors. We have characterized the U6 vectors in detail and report here that the ISG induction is a consequence of the presence of an AA di-nucleotide near the transcription start site.
Methods Plasmids, cells, and viruses The HAT-targeting vectors (pSP134–140, 142–143, and 146–158; see Fig. 1), the pSP-81 U6 promoter vector, and the pSP-161 Gateway donor lentiviral vectors are described by Bridge et al. (2003). 293 T and HCT116 cells were obtained from ATCC (Manassas, VA). Human embryonic lung fibroblasts (HLFs) were supplied by Dr. M. Nabholz. Lentiviruses were produced as described by Bridge et al. (2003), using pMD2-VSVG and pCMVDR8.91 (Naldini et al., 1996).
Gateway cloning strategy To prepare the RNA pol-III promoter-shRNA cassette for Gateway cloning, two consecutive PCR reactions (PCR1 and PCR2) were performed using Pfu Turbo polymerase (Stratagene, La Jolla, CA). The number of PCR cycles for each reaction was kept to a minimum to avoid introducing spurious mutations. PCR2 was inserted into pSP-161 using BP clonase (Invitrogen, Carlsbad, CA) according to the manufacturer’s instructions. To target a new mRNA, the following primers are required: an invariant attB1 forward primer in both reactions; a reverse primer in the first reaction that contains the hairpin sequence; and an invariant attB2 reverse primer in the second reaction. The invariant attB1 forward primer anneals to the upstream end of the pol-III promoter. The invariant attB2 reverse primer anneals immediately after the pol-III transcription termination signal in the hairpin oligonucleotide. By performing two steps, with invariant attB1/attB2 primers in the second step, the length of the hairpin oligonucleotide can be reduced and the final PCR product is much cleaner than is possible in a single step. For the sake of clarity, we first give the generic sequences that should be used to create new U6 vectors, taking into account the results of this study: invariant attB1-U6 forward primer, 5 0 -GGGGACAAGTTTGTACAAAAAAGCAGGCTCAAGGTCGGGCAGGAAG-3 0 ; invariant attB2 reverse primer, 5 0 -GGGGACCACTTTGTACAAGAAAGCTGGGTCCAAAAAGCTTCCA-3 0 ; and hairpin reverse primer, 5 0 -GGGTCCAAAAAGCTTCCAAAAAAA G (N18 sense) TCTCTTGAA (N18 antisense) C GGTGTTTCGTCCTTTCCACAAGATAT-3 0 . The siRNA target sequence ‘‘G (N18 sense)’’ should start with a G (the underlined nucleotide in the above sequence). The same pSP-161/Gateway cloning strategy can be used to create lentiviral vectors incorporating the H1 promoter fragment from pSUPER (Brummelkamp et al., 2002) using the following primers: invariant attB1-H1 forward primer, 5 0 -GGGGACAAGTTTGTACAAAAAAGCAGGCTGGAATTCGAACGCTGACG3 0 ; invariant attB2 reverse primer, as for U6; and hairpin reverse primer, 5 0 -GGGTCCAAAAAGCTTCCAAAAAAA (N19 sense) TCTCTTGAA (N19 antisense) GGGGATCTGTGGTCTCAT-3 0 .
Modified U6 vectors The construction of the vectors in Figure 3 is described below. The invariant attB1-U6 forward primer was used in all cases. Where two
sequences are given, the first is for the pSP-147 PCAF shRNA and the second is for the pSP-149 GCN5 shRNA. For the vectors in Figure 3B, a single PCR was performed on pSP-147 and pSP-149 with reverse primer 5 0 -GGGGACCACTTTGTACAAGAAAGCTGGGTGATCCTCTAGAGCTTCG-3 0 . The vectors in Figures 3C–3G were made by two-step PCR using the invariant attB2 reverse primer for PCR2. For the vectors in Figure 3C, PCR1 was performed on pSP-147 and pSP-149 with the reverse primers G5 0 -GGTCCAAAAAGCTTCCAAAAAAAGAAATTATTCATGGCAGACTCTCTTGA-3 0 and 5 0 -GGGTCCAAAAAGCTTCCAAAAAAAGAGATCATCAAGAAGCTGATCTCTTGA-3 0 . For the vector in Figure 3D, PCR1 was performed on pSP-149 with the same reverse primer as in Figure 3C, but the resulting vector was found by sequencing to have a single nucleotide deletion in the antisense siRNA strand. For the vectors in Figure 3E, PCR1 was performed on pSP-81 with the reverse primers 5 0 -GGGTCCAAAAAGCTTCCAAAAAAAGAAATTATTCATGGCAGACTCTCTTGAAGTCTGCCATGAATAATTTCTTCGAATCGTCCTTTCCACAAGATAT-3 0 and 5 0 -GGGTCCAAAAAGCTTCCAAAAAAAGAGATCATCAAGAAGCTGATCTCTTGAATCAGCTTCTTGATGATCTCTTCGAATCGTCCTTTCCACAAGATAT3 0 . For the vectors in Figure 3F, PCR1 was performed on pSP-81 with the reverse primers 5 0 -GGGTCCAAAAAGCTTCCAAAAAAAGAAATTATTCATGGCAGACTCTCTTGAAGTCTGCCATGAATAATTTCGGTGTTTCGTCCTTTCCACAAGATAT-3 0 and 5 0 -GGGTCCAAAAAGCTTCCAAAAAAAGAGATCATCAAGAAGCTGATCTCTTGAATCAGCTTCTTGATGATCTCGGTGTTTCGTCCTTTCCACAAGATAT-3 0 . For the vectors in Figure 3G, PCR1 was performed on pSP-81 with the reverse primers 5 0 -GGGTCCAAAAAGCTTCCAAAAAAAGAAATTATTCATGGCAGACTCTCTTGAAGTCTGCCATGAATAATTTCTTGGTGTTTCGTCCTTTCCACAAGATAT-3 0 and 5 0 -GGGTCCAAAAAGCTTCCAAAAAAAGAGATCATCAAGAAGCTGATCTCTTGAATCAGCTTCTTGATGATCTCTTGGTGTTTCGTCCTTTCCACAAGATAT-3 0 .
VA RNA I vectors For VA RNA I expression, a DNA fragment containing the whole VA RNA I gene plus 60 bp upstream was isolated by PCR from Adenovirus 5-infected cells using the primers 5 0 -TACGGATCCTGGCCGGTCAGGCGCGCGCAATCGT-3 0 and 5 0 -CTAGAATTCAAAAGGAAGCCCGGGAGCACTCCCCCGTTGTCTGACGTC-3 0 . This PCR product was cloned into pUC19 using BamHI and EcoRI restriction enzymes, to give pSP-93, which was used as the promoter template for PCR1. The VA RNA I lentiviral vectors were created by BP cloning of VA RNA I or VA RNA I-lamin shRNA cassettes into pSP-161. PCR1 was performed with the invariant attB1-VA forward primer 5 0 -GGGGACAAGTTTGTACAAAAAAGCAGGCTCCTGGCCGGTCAGGC-3 0 and the VA RNA I-specific reverse primer 5 0 -GGGTGCAAAAAGCTTCCAAAAAAAGGAGGAGTCCCCCGTTGTCTGA-3 0 or the VA RNA I-shRNA-specific reverse primer 5 0 -GGGTGCAAAAAGCTTCCAAAAAAAGGACTTCCAGAAGAACATCTCTCTTGAAGATGTTCTTCTGGAAGTCCGGAGGAGTCCCCCGTTGTCTGA-3 0 . PCR2 was performed using the invariant attB1-VA forward primer, and the invariant attB2 reverse primer (as for U6 above).
RNA extraction and quantitative RT-PCR Human embryonic lung fibroblasts (HLFs) were infected with the indicated amount of virus for 24 hr and then selected for 48 hr in puromycin. Total RNA was extracted with Qiagen (Hilden, Germany) RNeasy mini-columns, according to the manufacturer’s instructions. Reverse transcription was performed using random hexamers and Superscript II polymerase (Invitrogen). Quantitative
105 real-time PCR was performed on a PE 5700 PCR machine, using SyBR green PCR master mix (Perkin Elmer n14312704, Boston, MA) for OAS1, PCAF, GCN5, and lamin A/C mRNA, or Taqman PCR master mix (Perkin Elmer n14305719) for 18S ribosomal RNA (Perkin Elmer n14310875E). All mRNA values were normalized to 18S. The primers used for mRNA amplification were: for OAS1, 5 0 -AGGTGGTAAAGGGTGGCTCC-3 0 and 5 0 -ACAACCAGGTCAGCGTCAGAT-3 0 ; for PCAF, 5 0 -AACCCTAACCCCTCACCCAC-3 0 and 5 0 -TGGCTACAACTCCGACAGGAT-3 0 ; for GCN5, 5 0 -AATATTGAGCAGGGTGTGCTG-3 0 and 5 0 -AACATCGTCTGCCGCTCC-3 0 ; and for lamin A/C, 5 0 -CAAGCTTGAGGCAGCCCTAG-3 0 and 5 0 -CTCACGCAGCTCCTCACTGTA-3 0 .
Results Induction of an ISG response by shRNAs expressed from the U6 promoter The structure of the lentiviral vectors is shown in Figure 1A. To detect ISG induction, we used quantitative RTPCR for 2 0 ,5 0 -oligoadenylate synthetase (OAS1), a classic ISG. Human lung fibroblasts (HLFs) were infected with 22 lentiviruses targeting different histone acetyltransferases (HATs). Seven out of 22 strongly induced OAS1 expression (Fig. 1B). OAS1 induction was dependent on the amount of vector used, with no induction in cells infected at a multiplicity of infection (MOI) of 1, but 1000-fold induction at an MOI of 9 (Fig. 1C). Interferon signaling pathways are functional in normal cells, but often defective in tumor cells. HCT116 colon tumor cells are permissive for vesicular stomatitis virus (VSV) replication in the presence of interferon-a (Stojdl et al., 2000), and show impaired ISG induction by interferon-g as a result of mutant ras expression (Klampfer et al., 2003). To test whether our U6 vectors could induce ISG expression in tumor cells, HCT116 cells were infected with pSP-135, 138, and 147 at an MOI that induces ISGs in HLFs. None of the vectors induced OAS1 (Fig. 1D). We conclude that shRNAs expressed from the U6 promoter in our lentiviral construct can induce an ISG response in normal human cells.
A simple strategy to clone hairpin oligonucleotides into lentiviruses The high frequency of ISG induction by the U6 constructs suggests a fundamental problem with the vector design, which might be responsible for a recurrent error in transcription of the shRNA sequences. To perform mutagenesis of the U6 constructs and identify the cause of ISG induction, a simple strategy to modify the shRNA expression cassette in lentiviral vectors was developed (Fig. 2). In the first step, the RNA pol-III promoter is amplified by PCR using an oligonucleotide containing the shRNA sequence. The PCR product is re-amplified with primers
Fig. 1 OAS1 induction by lentiviruses expressing shRNA from the U6 promoter. (A) Schematic diagram of the U6 lentiviral vectors. The U6 promoter was cloned in the reverse orientation relative to the viral LTRs. (B–D) Quantitative RT-PCR for OAS1. Black bars: control virus infection (no shRNA insert). (B) HLFs were infected at an MOI of 10 with lentiviruses expressing shRNAs targeting different HATs. The shRNA numbering refers to pSP plasmid names. The sequence of the shRNA inserts is shown below the figure; the inserts are flanked in the vector by BstBI sites (ttcgaa; see Fig. 3). (C) HLFs were infected at increasing MOIs with the pSP149 virus. (D) HCT116 colon cancer cells were infected at an MOI of 10 with the pSP-135, 138, and 147 viruses.
106
Fig. 2 A simple strategy to clone shRNA oligonucleotides into lentiviral vectors. A two-step PCR is used to add attB sites to the ends of a pol-III/shRNA expression cassette. The attB sites recombine with attP sites in the lentiviral vector in the presence of BP clonase (Gateway cloning). The resulting lentiviral vector contains attL sites flanking the pol-III/shRNA expression cassette.
containing attB1 and attB2 sites, which permits recombinase-mediated cloning (Gateway cloning system, Invitrogen). The final PCR product is cloned by BP recombination to create the shRNA-expressing lentiviral vector.
Mapping of ISG-inducing elements in the U6 constructs Two of the U6 vectors inducing an ISG response were selected for detailed characterization: pSP-147 and pSP149, which target PCAF and GCN5, respectively (Fig. 1B). The lentiviral vectors all have a deletion in the 3 0 LTR, which is copied to the 5 0 -LTR during reverse transcription and should abolish the transcriptional activity of the LTRs (long terminal repeats) in integrated proviruses (so-called self-inactivating, or
SIN vectors; Zufferey et al., 1998). Because the original U6 vectors contained the U6 promoter in the reverse orientation relative to the LTRs, residual LTR transcriptional activity could lead to the formation of long dsRNA molecules that would explain the ISG response. To test this possibility, the orientation of the U6 cassette was inverted in the Gateway vector (Fig. 3B). Both shRNA vectors tested still induced OAS1 expression, indicating that the reverse orientation of the U6 promoter relative to the LTRs in the original vectors was not responsible for the ISG response. Termination of RNA pol-III-dependent transcription occurs when the polymerase encounters a run of 4 or 5 Ts (Paule and White, 2000). The original U6 vector terminator contained 4 Ts. This should be sufficient for proper termination, but analysis of the termination signals of VA RNA I, an adenoviral RNA pol-III transcript, revealed that around 15% of transcripts continue past the first 4 T signal for an extra 35 nt to a second termination signal containing 5 Ts (Gunnery et al., 1999). To test whether ISG induction by the U6 vectors was caused by leaky termination, the stop signal length was increased to 7 Ts and a small downstream sequence from the H1 terminator was added, as this has been suggested to improve RNA pol-III termination (Gunnery et al., 1999). As shown in Figure 3C, both constructs still induced OAS1 expression despite modification of the terminator. This indicates that leaky termination is not responsible for the ISG induction by the U6 constructs. dsRNA-dependent protein kinase (PKR) is activated by dsRNA longer than 30 bp (Manche et al., 1992). Because the double-stranded stem in the shRNA is only 19 bp long, it is not expected to activate PKR. To test whether the 19 bp dsRNA stem in the U6 vectors was required for the ISG response, a construct with a single base deletion in the antisense GCN5 siRNA strand was tested (Fig. 3D). This construct is expected only to make a 14 bp dsRNA stem. The mutant was unable to trigger OAS1 mRNA expression (Fig. 3D), suggesting that a dsRNA recognition step may indeed be required for the ISG activation by our vectors. Proper U6 transcription initiation requires the presence of a purine at position 11 of transcription (Lobo et al., 1990). The three main sequence elements required for transcription initiation, the DSE, PSE, and TATA box, as well as the region immediately downstream of the PSE, are essential for start site localization (Lobo et al., 1990). The 3 0 end of the promoter in our original U6 vectors all contained a BstBI restriction site that was used to clone the shRNA-encoding oligonucleotides. The Gateway cloning strategy imposes no such restrictions on the possible sequences around the start site, because the shRNA sequence is fused to the promoter in a PCR reaction (Fig. 2). Modified U6 vectors were constructed to test whether changes around the start site might be responsible for the ISG
107
Fig. 3 Mapping the determinants of ISG induction. The PCAF and GCN5 shRNA vectors pSP-147 and 149 were mutagenized to identify the cause of the OAS1 induction. (A) Schematic diagram of the lentiviral vectors. (B–G) U6 transcription start site and shRNA sequence. Green boxes: siRNA sense and antisense strands. The
bent arrow marks the transcription start site. Right panels: quantitative RT-PCR for OAS1 in HLFs infected at an MOI of 10 with the modified vectors. Black bars: control virus infection (no shRNA insert).
induction by our original U6 vectors. In addition to the BstBI site, the vectors contain an AA di-nucleotide at the start site. First, we removed the AA di-nucleotide at the beginning of the transcript but retained the BstBI site upstream of the start site (Fig. 3E). In these vectors,
the PCAF and GCN5 shRNA transcripts should begin with the first G of the 19 nt siRNA sense strand, but the 1/ 2 residues remain AA instead of CC in the native promoter. Both constructs tested still induced OAS1 (Fig. 3E). Next, we restored the native U6 promoter
108
sequence (Fig. 3F). In these vectors, the PCAF and GCN5 shRNAs should begin with the first G of the 19 nt siRNA sense strand, and the 1/ 2 positions are CC as in the native promoter. As shown in Figure 3F, HLFs infected with these two vectors did not induce OAS1. Hence, a minor change in the sequence between the TATA box and 11 resulted in a major difference in OAS1 induction. To determine whether the presence of AA at the transcription initiation site was responsible for the OAS1 induction, these residues were added back at positions 11/12 (Fig. 3G). In these vectors, the PCAF and GCN5 shRNAs are expressed from the native U6 promoter, but the shRNA should start with AA. HLFs infected with the resulting lentiviruses induced strong OAS1 mRNA expression (Fig. 3G). We conclude that modification of the original C/G sequence at position 1/11 was responsible for the OAS1 induction by the defective U6 vectors.
RNAi efficiency of modified U6 constructs The reduction in PCAF and GCN5 mRNA levels using the original pSP-147 and pSP-149 U6 vectors was maximally 40% in HCT116 cells and HLFs (data not shown). Inefficient silencing is a recurrent problem with shRNA vectors, which is generally attributed to poor sequence selection. We considered the possibility that the explanation in this case might be defective transcription initiation from the modified start site. To determine whether ISG induction and target mRNA silencing were inversely correlated, the levels of PCAF and GCN5 mRNA were measured by quantitative RTPCR in HLFs infected with the modified lentiviruses described above (Fig. 4). The white bars in the figure show the results for the constructs that did not induce OAS1. There was no correlation between OAS1 induction and target gene silencing, suggesting either that the amount of shRNA was not reduced with the constructs that induced OAS1 or that the amount of shRNA was not limiting for silencing.
Fig. 4 Targeting efficiency of modified U6 vectors. Quantitative RT-PCR for PCAF and GCN5 in HLFs infected at an MOI of 10 with the modified vectors shown in Figure 3. Black bars: control virus infection (no shRNA insert). Gray bars: vectors that induce OAS1. White bars: vectors that do not induce OAS1. The corresponding panels in Figures 3B–3G are indicated below each column. U6-BstBI: U6 promoter with BstBI site. U6-native: U6 promoter without BstBI site. AA 11/12: transcripts starting with AA. 4 T, 7T: termination signal length.
shRNA vectors based on VA RNA I PKR activation is a major arm of the cellular response to dsRNA, in particular during viral infections. VA RNA I is an adenoviral RNA pol-III transcript abundantly expressed late during infection to prevent PKR activation by viral dsRNA. It contains an apical dsRNA stem, which binds to the PKR dsRNA binding site, and a central domain, which inhibits the kinase (Mathews and Shenk, 1991; Ma and Mathews, 1993; Mori et al., 1996). VA RNA I contains all the major transcription regulatory elements within the transcribed region. Because VA RNA I exists to suppress the cellular response to dsRNA during adenovirus infections,
we tested whether VA RNA I could be used to express shRNA. An oligonucleotide encoding a lamin A/C shRNA (Elbashir et al., 2001; Bridge et al., 2003) was fused to the 3 0 end of the VA RNA I-transcribed region, and the cassette was cloned into pSP-161 (Fig. 5A). HLFs transduced with this construct showed a dosedependent reduction in lamin A/C mRNA levels of 40% at an MOI of 1, and 60% at an MOI of 10 (Fig. 5B). In comparison, lamin A/C shRNA expressed from the U6 or H1 promoters reduced lamin A/C mRNA levels to 10% at an MOI of 10 (data not shown). To test whether VA RNA I could block OAS1 induction by shRNA vectors, we infected HLFs with lentiviruses
109
Fig. 5 VA RNA I-mediated RNAi. (A) Schematic diagram of the VA RNA I lentiviral vectors. (B) Quantitative RT-PCR for lamin A/C in HLFs infected at an MOI of 1 and 10 with VA RNA I vectors. Black bar: VA RNA I alone (no shRNA insert). Gray bars: VA RNA I-lamin shRNA vector. (C) Quantitative RT-PCR for OAS1 in HLFs infected at an MOI of 10 with RNAi vectors. White bar: control virus infection (H1 vector with no shRNA insert). Black bar: VA RNA I alone (no shRNA insert). Gray bar: VA RNA I-lamin shRNA vector.
expressing VA RNA I and VA RNA I-lamin shRNA (Fig. 5C). Compared to a blank lentiviral vector, OAS1 was induced 12-fold by VA RNA I alone, and 47-fold by the VA RNA I-lamin shRNA fusion. We conclude that VA RNA I fusions can be used to silence gene expression in mammalian cells, but additional modification of the construct will be required to achieve the same level of silencing as can be achieved with standard shRNA vectors, and this approach is unlikely to provide a means to circumvent the ISG induction by shRNA vectors.
Discussion In this study, we have identified the design flaw in a set of U6 promoter-based shRNA vectors previously reported to induce an interferon response (Bridge et al., 2003). The vectors contain a U6 promoter shRNA cassette in a self-inactivating lentiviral backbone. The use of lentiviruses to express shRNAs is attractive because they can be used to stably silence gene expression in a wide range of cells, including cells
that are difficult to transfect with plasmid vectors. The ISG induction is unrelated to the fact that a viral vector was used. Indeed, we previously showed that the lentiviral approach was less likely than plasmid vectors to induce OAS1 expression (Bridge et al., 2003). The cell contains at least three types of sensor for dsRNA: toll-like receptor 3 (TLR3), PKR, and members of the OAS enzyme family. TLR3 is a possible candidate for sensing the presence of shRNA but it is a cell membrane-bound receptor, which may not have access to cytosolic dsRNA (Alexopoulou et al., 2001). PKR is known to activate IRF-3, which would readily explain the ISG expression (Sen, 2001). The observation that VA RNA I, which inhibits PKR, itself activates OAS1 mRNA expression suggests that ISG expression can be induced in the absence of PKR activation. The increased OAS1 expression by a VA RNA I-shRNA fusion compared to VA RNA I alone further suggests that PKR is not the sensor that mediates shRNAdependent ISG induction. This is supported by the observation that VA RNA I also fails to suppress OAS1 induction by the MORF4L1 shRNA vector (A. Bridge and R.D.I., unpublished data). To completely exclude a role for PKR, we would have to demonstrate that our lentiviral VA RNA vector was biologically active, for example by showing that it could block eIF2a phosphorylation in response to dsRNA, or complement the growth defect of a VA RNA-deleted adenovirus (Cascallo et al., 2003). This is an important issue because Sledz et al. (2003) showed that PKR is required for the induction of interferon expression following transfection of siRNA oligonucleotides. The 19 bp dsRNA stem in our shRNAs is generally considered too short to induce an ISG response. However, some members of the OAS family of enzymes, the third type of sensor for dsRNA, can be activated by dsRNA as short as 15 bp (Sarkar et al., 1999). To define the precise length of dsRNA stem able to induce ISG expression in vivo would require extensive mutagenesis of our constructs, with compensatory mutations in each strand. Deletion of one base in the antisense strand of one of our vectors, which would be expected to reduce the length of the dsRNA stem to 14 bp, abolished the ISG response, suggesting that the response uses sensors that can detect 19 but not 14 bp dsRNA molecules. Because only 7 out of 22 shRNAs induced OAS1, the length of the stem alone is not sufficient to explain the ISG induction. To identify the additional causes of the ISG response, we performed mutational analysis of two ISG-inducing shRNA constructs. First, we determined that neither the orientation of the U6 promoter relative to the LTRs nor the length of transcription termination signal was responsible for the ISG response. Mattaj and colleagues have shown that Xenopus U6 transcription is extremely sensitive to changes near the start site. In particular, efficient transcription requires the presence of a G residue at the 11 position (Mattaj et al., 1988).
110
Substitution of this nucleotide with a C reduced transcription by about 20-fold. Replacement of G by A gave only a 3-fold reduction (Mattaj et al., 1988), indicating that the main requirement is for a purine at this position. Analysis of tRNAleu3 gene transcription likewise revealed that a pyrimidine-purine sequence (with initiation at the purine residue) was necessary for proper transcription initiation (Fruscoloni et al., 1995). The ISG response by our shRNA vectors was abolished when a C/G sequence at the –1/11 position was respected. Insertion of AA residues at the transcription start site, giving a C/A sequence at 1/11, systematically resulted in OAS1 expression. While this may solve the problem in a technical sense, it does not address the mechanism. It is possible that by changing the sequences at the normal start site, we force the polymerase to initiate transcription at a new site. This could result in the formation of aberrant transcripts that do not fold correctly. If delayed processing of shRNA to siRNA is the critical factor, it is worth noting that the presence of AA residues near the start site of our constructs could lead to the formation of blunt-ended shRNA by base pairing with the UU tail. The efficiency of silencing showed no correlation with OAS1 induction, suggesting that the amount of shRNA and the efficiency of processing were not direct determinants of OAS1 induction. The overall silencing efficiency of the U6 vectors was disappointing, a result best explained by recent reports that 5 0 -end stability determines which strand is incorporated into RISC (Khvorova et al., 2003; Schwarz et al., 2003). Taken together with our data showing that an AA sequence at the 5 0 end of the shRNA triggers ISG expression, it is clearly much better to place the sense strand before the antisense strand in shRNA vectors. In this configuration, the 5 0 end of the antisense strand (the strand complementary to the target mRNA) can be destabilized by sequences near the hairpin loop rather than the transcription start site. The mechanism of ISG induction may not be the same in our U6 vectors as in the MORF4L1 H1 vector that induced interferon expression (Bridge et al., 2003), because it had a C/G sequence at the transcription start site. In conclusion, we have identified the proximate cause of ISG induction by some of our U6 shRNA constructs. The exact start site and structure of the RNA molecules triggering the response are unknown, but they may only be a small fraction of the total pool of transcripts, which would make their identification extremely difficult. Because small differences in vector design can have such large effects on the cellular response to shRNA vectors, the key message from this study is that offtarget effects should always be considered as a possible explanation for shRNA phenotypes. Acknowledgments We thank the Swiss National Foundation for financial support.
References Alexopoulou, L., Holt, A.C., Medzhitov, R. and Flavell, R.A. (2001) Recognition of double-stranded RNA and activation of NF-kB by Toll-like receptor 3. Nature 413:732–738. Bridge, A.J., Pebernard, S., Ducraux, A., Nicoulaz, A.L. and Iggo, R. (2003) Induction of an interferon response by RNAi vectors in mammalian cells. Nat Genet 34:263–264. Brummelkamp, T.R., Bernards, R. and Agami, R. (2002) A system for stable expression of short interfering RNAs in mammalian cells. Science 296:550–553. Cascallo, M., Capella, G., Mazo, A. and Alemany, R. (2003) Rasdependent oncolysis with an adenovirus VAI mutant. Cancer Res 63:5544–5550. Elbashir, S.M., Harborth, J., Lendeckel, W., Yalcin, A., Weber, K. and Tuschl, T. (2001) Duplexes of 21-nucleotide RNAs mediate RNA interference in cultured mammalian cells. Nature 411:494– 498. Fire, A., Xu, S., Montgomery, M.K., Kostas, S.A., Driver, S.E. and and Mello, C.C. (1998) Potent and specific genetic interference by double-stranded RNA in Caenorhabditis elegans. Nature 391:806–811. Fruscoloni, P., Zamboni, M., Panetta, G., De Paolis, A. and Tocchini-Valentini, G.P. (1995) Mutational analysis of the transcription start site of the yeast tRNA(Leu3) gene. Nucleic Acids Res 23:2914–2918. Gunnery, S., Ma, Y. and Mathews, M.B. (1999) Termination sequence requirements vary among genes transcribed by RNA polymerase III. J Mol Biol 286:745–757. Khvorova, A., Reynolds, A. and Jayasena, S.D. (2003) Functional siRNAs and miRNAs exhibit strand bias. Cell 115:505. Klampfer, L., Huang, J., Corner, G., Mariadason, J., Arango, D., Sasazuki, T., Shirasawa, S. and Augenlicht, L. (2003) Oncogenic Ki-Ras inhibits the expression of interferon-responsive genes through inhibition of STAT1 and STAT2 expression. J Biol Chem 278:46278–46287. Lobo, S.M., Ifill, S. and Hernandez, N. (1990) cis-acting elements required for RNA polymerase II and III transcription in the human U2 and U6 snRNA promoters. Nucleic Acids Res 18:2891–2899. Ma, Y. and Mathews, M.B. (1993) Comparative analysis of the structure and function of adenovirus virus-associated RNAs. J Virol 67:6605–6617. Manche, L., Green, S.R., Schmedt, C. and Mathews, M.B. (1992) Interactions between double-stranded RNA regulators and the protein kinase DAI. Mol Cell Biol 12:5238–5248. Mathews, M.B. and Shenk, T. (1991) Adenovirus virus-associated RNA and translation control. J Virol 65:5657–5662. Mattaj, I.W., Dathan, N.A., Parry, H.D., Carbon, P. and Krol, A. (1988) Changing the RNA polymerase specificity of U snRNA gene promoters. Cell 55:435–442. Mori, K., Juttermann, R., Wienhues, U., Kobayashi, K., Yagi, M., Sugimoto, T., Tjia, S.T., Doerfler, W. and Hosokawa, K. (1996) Anti-interferon activity of adenovirus-2-encoded VAI and VAII RNAs in translation in cultured human cells. Virus Res 42:53–63. Naldini, L., Blomer, U., Gallay, P., Ory, D., Mulligan, R., Gage, F.H., Verma, I.M. and Trono, D. (1996) In vivo gene delivery and stable transduction of nondividing cells by a lentiviral vector. Science 272:263–267. Paddison, P.J., Caudy, A.A., Bernstein, E., Hannon, G.J. and Conklin, D.S. (2002) Short hairpin RNAs (shRNAs) induce sequence-specific silencing in mammalian cells. Genes Dev 16:948–958. Paule, M.R. and White, R.J. (2000) Survey and summary: transcription by RNA polymerases I and III. Nucleic Acids Res 28:1283–1298. Sarkar, S.N., Bandyopadhyay, S., Ghosh, A. and Sen, G.C. (1999) Enzymatic characteristics of recombinant medium isozyme of 2 0 5 0 oligoadenylate synthetase. J Biol Chem 274:1848–1855.
111 Schwarz, D.S., Hutvagner, G., Du, T., Xu, Z., Aronin, N. and Zamore, P.D. (2003) Asymmetry in the assembly of the RNAi enzyme complex. Cell 115:199–208. Sen, G.C. (2001) Viruses and interferons. Annu Rev Microbiol 55:255–281. Sledz, C.A., Holko, M., De Veer, M.J., Silverman, R.H. and Williams, B.R. (2003) Activation of the interferon system by short-interfering RNAs. Nat Cell Biol 5:834–839. Stark, G.R., Kerr, I.M., Williams, B.R., Silverman, R.H. and Schreiber, R.D. (1998) How cells respond to interferons. Annu Rev Biochem 67:227–264. Stojdl, D.F., Lichty, B., Knowles, S., Marius, R., Atkins, H., Sonenberg, N. and Bell, J.C. (2000) Exploiting tumor-specific
defects in the interferon pathway with a previously unknown oncolytic virus. Nat Med 6:821–825. Svoboda, P., Stein, P., Hayashi, H. and Schultz, R.M. (2000) Selective reduction of dormant maternal mRNAs in mouse oocytes by RNA interference. Development 127:4147– 4156. Wianny, F. and Zernicka-Goetz, M. (2000) Specific interference with gene function by double-stranded RNA in early mouse development. Nat Cell Biol 2:70–75. Zufferey, R., Dull, T., Mandel, R.J., Bukovsky, A., Quiroz, D., Naldini, L. and Trono, D. (1998) Self-inactivating lentivirus vector for safe and efficient in vivo gene delivery. J Virol 72:9873– 9880.
r International Society of Differentiation 2004
O RI G INA L AR T I C L E
Stephanie Pebernard . Richard D. Iggo
Determinants of interferon-stimulated gene induction by RNAi vectors
Received October 20, 2003; accepted in revised form November 25, 2003
Abstract RNA interference is widely used to silence gene expression in mammalian cells. We recently reported that an shRNA expressed from the H1 promoter in a lentiviral vector could induce the expression of a large group of interferon-stimulated genes (ISGs). This response was unrelated to silencing of the gene targeted by the shRNA MORF4L1. In parallel, we constructed lentiviral vectors expressing shRNA from the U6 promoter and found that these too could induce expression of OAS1, a classic interferon target gene. The U6 vectors give a higher frequency of ISG induction than comparable lentiviral H1 vectors, suggesting that there might be a fundamental flaw in the vector design. We have characterized the U6 vectors in detail and report here that ISG induction is a consequence of the presence of an AA di-nucleotide near the transcription start site. A single nucleotide deletion in the siRNA sequence abolished OAS1 induction, suggesting that the mechanism underlying the response uses a sensor that can detect 19 bp RNA duplexes but not 14 bp duplexes. Adenoviral VA RNA I, which inhibits dsRNA-dependent protein kinase (PKR), was tested as a fusion partner to express shRNA on the grounds that it might prevent nonspecific off-target effects. Fusion of VA RNA I to a lamin shRNA was moderately effective in silencing lamin expression, but gave strong OAS1 induction by an shRNA that does not induce OAS1 when expressed from the U6 or H1 promoters. To avoid interferon induction by U6 vectors, we recommend preserving the wild-type sequence around the transcription start site, in particular a C/G sequence at positions 1/11, and we describe a simple cloning strategy using the Gateway recombination system that facilitates this task. . ) Stephanie Pebernard . Richard D. Iggo (* Oncogene Group Swiss Institute for Experimental Cancer Research (ISREC) Ch. des Boveresses 1066 Epalinges, Switzerland Tel: 141 21 692 58 89 Fax: 141 21 652 69 33 E-mail: [email protected] U.S. Copyright Clearance Center Code Statement:
Key words RNA interference shRNA U6 promoter lentiviral vector interferon-stimulated gene
Introduction RNA interference was first described in C. elegans (Fire et al., 1998). It was originally thought that no comparable silencing mechanism existed in mammalian cells, but in 2000 two groups reported that long dsRNA could silence gene expression in early mouse development (Svoboda et al., 2000; Wianny and ZernickaGoetz, 2000). Tuschl and colleagues then showed that transfection of short dsRNA oligonucleotides leads to specific gene silencing in cultured mammalian cells (Elbashir et al., 2001). To achieve prolonged gene silencing, several groups developed DNA-based vectors to express short hairpin RNAs (shRNAs) from RNA polymerase III promoters (Brummelkamp et al., 2002; Paddison et al., 2002). The small interfering RNA (siRNA), which confers target gene specificity on the RNA-induced silencing complex (RISC), is 21 nt long, and the dsRNA stem in shRNAs expressed from RNA pol-III promoters in DNA vectors is typically 19–26 bp in length. dsRNA molecules over 50 bp are known to induce a complex physiological response characterized by translation inhibition, mRNA degradation, and interferon secretion (Stark et al., 1998; reviewed by Sen, 2001). Although the length of the dsRNAs used for RNAi is below that which is typically required to induce an interferon-stimulated gene (ISG) response, we and others recently reported the activation of interferon signaling and ISG expression following transfection of siRNA or expression of shRNA from lentiviral vectors (Bridge et al., 2003; Sledz et al., 2003). We showed, using micro-array technology, that an shRNA targeting MORF4L1 was capable of inducing a large group of ISGs in human lung fibroblasts (Bridge et al., 2003). The MORF4L1 shRNA was expressed from the H1 promoter cloned into a lentiviral vector. The ISG response was unrelated to the silencing of MORF4L1,
0301–4681/2004/7202/3–103 $ 15.00/0
104
because it was not suppressed by reintroduction of a noncleavable MORF4L1 cDNA. In parallel, we constructed lentiviruses expressing shRNAs from the U6 promoter (Bridge et al., 2003). The U6 vectors give a higher frequency of ISG induction than the H1 vectors. We have characterized the U6 vectors in detail and report here that the ISG induction is a consequence of the presence of an AA di-nucleotide near the transcription start site.
Methods Plasmids, cells, and viruses The HAT-targeting vectors (pSP134–140, 142–143, and 146–158; see Fig. 1), the pSP-81 U6 promoter vector, and the pSP-161 Gateway donor lentiviral vectors are described by Bridge et al. (2003). 293 T and HCT116 cells were obtained from ATCC (Manassas, VA). Human embryonic lung fibroblasts (HLFs) were supplied by Dr. M. Nabholz. Lentiviruses were produced as described by Bridge et al. (2003), using pMD2-VSVG and pCMVDR8.91 (Naldini et al., 1996).
Gateway cloning strategy To prepare the RNA pol-III promoter-shRNA cassette for Gateway cloning, two consecutive PCR reactions (PCR1 and PCR2) were performed using Pfu Turbo polymerase (Stratagene, La Jolla, CA). The number of PCR cycles for each reaction was kept to a minimum to avoid introducing spurious mutations. PCR2 was inserted into pSP-161 using BP clonase (Invitrogen, Carlsbad, CA) according to the manufacturer’s instructions. To target a new mRNA, the following primers are required: an invariant attB1 forward primer in both reactions; a reverse primer in the first reaction that contains the hairpin sequence; and an invariant attB2 reverse primer in the second reaction. The invariant attB1 forward primer anneals to the upstream end of the pol-III promoter. The invariant attB2 reverse primer anneals immediately after the pol-III transcription termination signal in the hairpin oligonucleotide. By performing two steps, with invariant attB1/attB2 primers in the second step, the length of the hairpin oligonucleotide can be reduced and the final PCR product is much cleaner than is possible in a single step. For the sake of clarity, we first give the generic sequences that should be used to create new U6 vectors, taking into account the results of this study: invariant attB1-U6 forward primer, 5 0 -GGGGACAAGTTTGTACAAAAAAGCAGGCTCAAGGTCGGGCAGGAAG-3 0 ; invariant attB2 reverse primer, 5 0 -GGGGACCACTTTGTACAAGAAAGCTGGGTCCAAAAAGCTTCCA-3 0 ; and hairpin reverse primer, 5 0 -GGGTCCAAAAAGCTTCCAAAAAAA G (N18 sense) TCTCTTGAA (N18 antisense) C GGTGTTTCGTCCTTTCCACAAGATAT-3 0 . The siRNA target sequence ‘‘G (N18 sense)’’ should start with a G (the underlined nucleotide in the above sequence). The same pSP-161/Gateway cloning strategy can be used to create lentiviral vectors incorporating the H1 promoter fragment from pSUPER (Brummelkamp et al., 2002) using the following primers: invariant attB1-H1 forward primer, 5 0 -GGGGACAAGTTTGTACAAAAAAGCAGGCTGGAATTCGAACGCTGACG3 0 ; invariant attB2 reverse primer, as for U6; and hairpin reverse primer, 5 0 -GGGTCCAAAAAGCTTCCAAAAAAA (N19 sense) TCTCTTGAA (N19 antisense) GGGGATCTGTGGTCTCAT-3 0 .
Modified U6 vectors The construction of the vectors in Figure 3 is described below. The invariant attB1-U6 forward primer was used in all cases. Where two
sequences are given, the first is for the pSP-147 PCAF shRNA and the second is for the pSP-149 GCN5 shRNA. For the vectors in Figure 3B, a single PCR was performed on pSP-147 and pSP-149 with reverse primer 5 0 -GGGGACCACTTTGTACAAGAAAGCTGGGTGATCCTCTAGAGCTTCG-3 0 . The vectors in Figures 3C–3G were made by two-step PCR using the invariant attB2 reverse primer for PCR2. For the vectors in Figure 3C, PCR1 was performed on pSP-147 and pSP-149 with the reverse primers G5 0 -GGTCCAAAAAGCTTCCAAAAAAAGAAATTATTCATGGCAGACTCTCTTGA-3 0 and 5 0 -GGGTCCAAAAAGCTTCCAAAAAAAGAGATCATCAAGAAGCTGATCTCTTGA-3 0 . For the vector in Figure 3D, PCR1 was performed on pSP-149 with the same reverse primer as in Figure 3C, but the resulting vector was found by sequencing to have a single nucleotide deletion in the antisense siRNA strand. For the vectors in Figure 3E, PCR1 was performed on pSP-81 with the reverse primers 5 0 -GGGTCCAAAAAGCTTCCAAAAAAAGAAATTATTCATGGCAGACTCTCTTGAAGTCTGCCATGAATAATTTCTTCGAATCGTCCTTTCCACAAGATAT-3 0 and 5 0 -GGGTCCAAAAAGCTTCCAAAAAAAGAGATCATCAAGAAGCTGATCTCTTGAATCAGCTTCTTGATGATCTCTTCGAATCGTCCTTTCCACAAGATAT3 0 . For the vectors in Figure 3F, PCR1 was performed on pSP-81 with the reverse primers 5 0 -GGGTCCAAAAAGCTTCCAAAAAAAGAAATTATTCATGGCAGACTCTCTTGAAGTCTGCCATGAATAATTTCGGTGTTTCGTCCTTTCCACAAGATAT-3 0 and 5 0 -GGGTCCAAAAAGCTTCCAAAAAAAGAGATCATCAAGAAGCTGATCTCTTGAATCAGCTTCTTGATGATCTCGGTGTTTCGTCCTTTCCACAAGATAT-3 0 . For the vectors in Figure 3G, PCR1 was performed on pSP-81 with the reverse primers 5 0 -GGGTCCAAAAAGCTTCCAAAAAAAGAAATTATTCATGGCAGACTCTCTTGAAGTCTGCCATGAATAATTTCTTGGTGTTTCGTCCTTTCCACAAGATAT-3 0 and 5 0 -GGGTCCAAAAAGCTTCCAAAAAAAGAGATCATCAAGAAGCTGATCTCTTGAATCAGCTTCTTGATGATCTCTTGGTGTTTCGTCCTTTCCACAAGATAT-3 0 .
VA RNA I vectors For VA RNA I expression, a DNA fragment containing the whole VA RNA I gene plus 60 bp upstream was isolated by PCR from Adenovirus 5-infected cells using the primers 5 0 -TACGGATCCTGGCCGGTCAGGCGCGCGCAATCGT-3 0 and 5 0 -CTAGAATTCAAAAGGAAGCCCGGGAGCACTCCCCCGTTGTCTGACGTC-3 0 . This PCR product was cloned into pUC19 using BamHI and EcoRI restriction enzymes, to give pSP-93, which was used as the promoter template for PCR1. The VA RNA I lentiviral vectors were created by BP cloning of VA RNA I or VA RNA I-lamin shRNA cassettes into pSP-161. PCR1 was performed with the invariant attB1-VA forward primer 5 0 -GGGGACAAGTTTGTACAAAAAAGCAGGCTCCTGGCCGGTCAGGC-3 0 and the VA RNA I-specific reverse primer 5 0 -GGGTGCAAAAAGCTTCCAAAAAAAGGAGGAGTCCCCCGTTGTCTGA-3 0 or the VA RNA I-shRNA-specific reverse primer 5 0 -GGGTGCAAAAAGCTTCCAAAAAAAGGACTTCCAGAAGAACATCTCTCTTGAAGATGTTCTTCTGGAAGTCCGGAGGAGTCCCCCGTTGTCTGA-3 0 . PCR2 was performed using the invariant attB1-VA forward primer, and the invariant attB2 reverse primer (as for U6 above).
RNA extraction and quantitative RT-PCR Human embryonic lung fibroblasts (HLFs) were infected with the indicated amount of virus for 24 hr and then selected for 48 hr in puromycin. Total RNA was extracted with Qiagen (Hilden, Germany) RNeasy mini-columns, according to the manufacturer’s instructions. Reverse transcription was performed using random hexamers and Superscript II polymerase (Invitrogen). Quantitative
105 real-time PCR was performed on a PE 5700 PCR machine, using SyBR green PCR master mix (Perkin Elmer n14312704, Boston, MA) for OAS1, PCAF, GCN5, and lamin A/C mRNA, or Taqman PCR master mix (Perkin Elmer n14305719) for 18S ribosomal RNA (Perkin Elmer n14310875E). All mRNA values were normalized to 18S. The primers used for mRNA amplification were: for OAS1, 5 0 -AGGTGGTAAAGGGTGGCTCC-3 0 and 5 0 -ACAACCAGGTCAGCGTCAGAT-3 0 ; for PCAF, 5 0 -AACCCTAACCCCTCACCCAC-3 0 and 5 0 -TGGCTACAACTCCGACAGGAT-3 0 ; for GCN5, 5 0 -AATATTGAGCAGGGTGTGCTG-3 0 and 5 0 -AACATCGTCTGCCGCTCC-3 0 ; and for lamin A/C, 5 0 -CAAGCTTGAGGCAGCCCTAG-3 0 and 5 0 -CTCACGCAGCTCCTCACTGTA-3 0 .
Results Induction of an ISG response by shRNAs expressed from the U6 promoter The structure of the lentiviral vectors is shown in Figure 1A. To detect ISG induction, we used quantitative RTPCR for 2 0 ,5 0 -oligoadenylate synthetase (OAS1), a classic ISG. Human lung fibroblasts (HLFs) were infected with 22 lentiviruses targeting different histone acetyltransferases (HATs). Seven out of 22 strongly induced OAS1 expression (Fig. 1B). OAS1 induction was dependent on the amount of vector used, with no induction in cells infected at a multiplicity of infection (MOI) of 1, but 1000-fold induction at an MOI of 9 (Fig. 1C). Interferon signaling pathways are functional in normal cells, but often defective in tumor cells. HCT116 colon tumor cells are permissive for vesicular stomatitis virus (VSV) replication in the presence of interferon-a (Stojdl et al., 2000), and show impaired ISG induction by interferon-g as a result of mutant ras expression (Klampfer et al., 2003). To test whether our U6 vectors could induce ISG expression in tumor cells, HCT116 cells were infected with pSP-135, 138, and 147 at an MOI that induces ISGs in HLFs. None of the vectors induced OAS1 (Fig. 1D). We conclude that shRNAs expressed from the U6 promoter in our lentiviral construct can induce an ISG response in normal human cells.
A simple strategy to clone hairpin oligonucleotides into lentiviruses The high frequency of ISG induction by the U6 constructs suggests a fundamental problem with the vector design, which might be responsible for a recurrent error in transcription of the shRNA sequences. To perform mutagenesis of the U6 constructs and identify the cause of ISG induction, a simple strategy to modify the shRNA expression cassette in lentiviral vectors was developed (Fig. 2). In the first step, the RNA pol-III promoter is amplified by PCR using an oligonucleotide containing the shRNA sequence. The PCR product is re-amplified with primers
Fig. 1 OAS1 induction by lentiviruses expressing shRNA from the U6 promoter. (A) Schematic diagram of the U6 lentiviral vectors. The U6 promoter was cloned in the reverse orientation relative to the viral LTRs. (B–D) Quantitative RT-PCR for OAS1. Black bars: control virus infection (no shRNA insert). (B) HLFs were infected at an MOI of 10 with lentiviruses expressing shRNAs targeting different HATs. The shRNA numbering refers to pSP plasmid names. The sequence of the shRNA inserts is shown below the figure; the inserts are flanked in the vector by BstBI sites (ttcgaa; see Fig. 3). (C) HLFs were infected at increasing MOIs with the pSP149 virus. (D) HCT116 colon cancer cells were infected at an MOI of 10 with the pSP-135, 138, and 147 viruses.
106
Fig. 2 A simple strategy to clone shRNA oligonucleotides into lentiviral vectors. A two-step PCR is used to add attB sites to the ends of a pol-III/shRNA expression cassette. The attB sites recombine with attP sites in the lentiviral vector in the presence of BP clonase (Gateway cloning). The resulting lentiviral vector contains attL sites flanking the pol-III/shRNA expression cassette.
containing attB1 and attB2 sites, which permits recombinase-mediated cloning (Gateway cloning system, Invitrogen). The final PCR product is cloned by BP recombination to create the shRNA-expressing lentiviral vector.
Mapping of ISG-inducing elements in the U6 constructs Two of the U6 vectors inducing an ISG response were selected for detailed characterization: pSP-147 and pSP149, which target PCAF and GCN5, respectively (Fig. 1B). The lentiviral vectors all have a deletion in the 3 0 LTR, which is copied to the 5 0 -LTR during reverse transcription and should abolish the transcriptional activity of the LTRs (long terminal repeats) in integrated proviruses (so-called self-inactivating, or
SIN vectors; Zufferey et al., 1998). Because the original U6 vectors contained the U6 promoter in the reverse orientation relative to the LTRs, residual LTR transcriptional activity could lead to the formation of long dsRNA molecules that would explain the ISG response. To test this possibility, the orientation of the U6 cassette was inverted in the Gateway vector (Fig. 3B). Both shRNA vectors tested still induced OAS1 expression, indicating that the reverse orientation of the U6 promoter relative to the LTRs in the original vectors was not responsible for the ISG response. Termination of RNA pol-III-dependent transcription occurs when the polymerase encounters a run of 4 or 5 Ts (Paule and White, 2000). The original U6 vector terminator contained 4 Ts. This should be sufficient for proper termination, but analysis of the termination signals of VA RNA I, an adenoviral RNA pol-III transcript, revealed that around 15% of transcripts continue past the first 4 T signal for an extra 35 nt to a second termination signal containing 5 Ts (Gunnery et al., 1999). To test whether ISG induction by the U6 vectors was caused by leaky termination, the stop signal length was increased to 7 Ts and a small downstream sequence from the H1 terminator was added, as this has been suggested to improve RNA pol-III termination (Gunnery et al., 1999). As shown in Figure 3C, both constructs still induced OAS1 expression despite modification of the terminator. This indicates that leaky termination is not responsible for the ISG induction by the U6 constructs. dsRNA-dependent protein kinase (PKR) is activated by dsRNA longer than 30 bp (Manche et al., 1992). Because the double-stranded stem in the shRNA is only 19 bp long, it is not expected to activate PKR. To test whether the 19 bp dsRNA stem in the U6 vectors was required for the ISG response, a construct with a single base deletion in the antisense GCN5 siRNA strand was tested (Fig. 3D). This construct is expected only to make a 14 bp dsRNA stem. The mutant was unable to trigger OAS1 mRNA expression (Fig. 3D), suggesting that a dsRNA recognition step may indeed be required for the ISG activation by our vectors. Proper U6 transcription initiation requires the presence of a purine at position 11 of transcription (Lobo et al., 1990). The three main sequence elements required for transcription initiation, the DSE, PSE, and TATA box, as well as the region immediately downstream of the PSE, are essential for start site localization (Lobo et al., 1990). The 3 0 end of the promoter in our original U6 vectors all contained a BstBI restriction site that was used to clone the shRNA-encoding oligonucleotides. The Gateway cloning strategy imposes no such restrictions on the possible sequences around the start site, because the shRNA sequence is fused to the promoter in a PCR reaction (Fig. 2). Modified U6 vectors were constructed to test whether changes around the start site might be responsible for the ISG
107
Fig. 3 Mapping the determinants of ISG induction. The PCAF and GCN5 shRNA vectors pSP-147 and 149 were mutagenized to identify the cause of the OAS1 induction. (A) Schematic diagram of the lentiviral vectors. (B–G) U6 transcription start site and shRNA sequence. Green boxes: siRNA sense and antisense strands. The
bent arrow marks the transcription start site. Right panels: quantitative RT-PCR for OAS1 in HLFs infected at an MOI of 10 with the modified vectors. Black bars: control virus infection (no shRNA insert).
induction by our original U6 vectors. In addition to the BstBI site, the vectors contain an AA di-nucleotide at the start site. First, we removed the AA di-nucleotide at the beginning of the transcript but retained the BstBI site upstream of the start site (Fig. 3E). In these vectors,
the PCAF and GCN5 shRNA transcripts should begin with the first G of the 19 nt siRNA sense strand, but the 1/ 2 residues remain AA instead of CC in the native promoter. Both constructs tested still induced OAS1 (Fig. 3E). Next, we restored the native U6 promoter
108
sequence (Fig. 3F). In these vectors, the PCAF and GCN5 shRNAs should begin with the first G of the 19 nt siRNA sense strand, and the 1/ 2 positions are CC as in the native promoter. As shown in Figure 3F, HLFs infected with these two vectors did not induce OAS1. Hence, a minor change in the sequence between the TATA box and 11 resulted in a major difference in OAS1 induction. To determine whether the presence of AA at the transcription initiation site was responsible for the OAS1 induction, these residues were added back at positions 11/12 (Fig. 3G). In these vectors, the PCAF and GCN5 shRNAs are expressed from the native U6 promoter, but the shRNA should start with AA. HLFs infected with the resulting lentiviruses induced strong OAS1 mRNA expression (Fig. 3G). We conclude that modification of the original C/G sequence at position 1/11 was responsible for the OAS1 induction by the defective U6 vectors.
RNAi efficiency of modified U6 constructs The reduction in PCAF and GCN5 mRNA levels using the original pSP-147 and pSP-149 U6 vectors was maximally 40% in HCT116 cells and HLFs (data not shown). Inefficient silencing is a recurrent problem with shRNA vectors, which is generally attributed to poor sequence selection. We considered the possibility that the explanation in this case might be defective transcription initiation from the modified start site. To determine whether ISG induction and target mRNA silencing were inversely correlated, the levels of PCAF and GCN5 mRNA were measured by quantitative RTPCR in HLFs infected with the modified lentiviruses described above (Fig. 4). The white bars in the figure show the results for the constructs that did not induce OAS1. There was no correlation between OAS1 induction and target gene silencing, suggesting either that the amount of shRNA was not reduced with the constructs that induced OAS1 or that the amount of shRNA was not limiting for silencing.
Fig. 4 Targeting efficiency of modified U6 vectors. Quantitative RT-PCR for PCAF and GCN5 in HLFs infected at an MOI of 10 with the modified vectors shown in Figure 3. Black bars: control virus infection (no shRNA insert). Gray bars: vectors that induce OAS1. White bars: vectors that do not induce OAS1. The corresponding panels in Figures 3B–3G are indicated below each column. U6-BstBI: U6 promoter with BstBI site. U6-native: U6 promoter without BstBI site. AA 11/12: transcripts starting with AA. 4 T, 7T: termination signal length.
shRNA vectors based on VA RNA I PKR activation is a major arm of the cellular response to dsRNA, in particular during viral infections. VA RNA I is an adenoviral RNA pol-III transcript abundantly expressed late during infection to prevent PKR activation by viral dsRNA. It contains an apical dsRNA stem, which binds to the PKR dsRNA binding site, and a central domain, which inhibits the kinase (Mathews and Shenk, 1991; Ma and Mathews, 1993; Mori et al., 1996). VA RNA I contains all the major transcription regulatory elements within the transcribed region. Because VA RNA I exists to suppress the cellular response to dsRNA during adenovirus infections,
we tested whether VA RNA I could be used to express shRNA. An oligonucleotide encoding a lamin A/C shRNA (Elbashir et al., 2001; Bridge et al., 2003) was fused to the 3 0 end of the VA RNA I-transcribed region, and the cassette was cloned into pSP-161 (Fig. 5A). HLFs transduced with this construct showed a dosedependent reduction in lamin A/C mRNA levels of 40% at an MOI of 1, and 60% at an MOI of 10 (Fig. 5B). In comparison, lamin A/C shRNA expressed from the U6 or H1 promoters reduced lamin A/C mRNA levels to 10% at an MOI of 10 (data not shown). To test whether VA RNA I could block OAS1 induction by shRNA vectors, we infected HLFs with lentiviruses
109
Fig. 5 VA RNA I-mediated RNAi. (A) Schematic diagram of the VA RNA I lentiviral vectors. (B) Quantitative RT-PCR for lamin A/C in HLFs infected at an MOI of 1 and 10 with VA RNA I vectors. Black bar: VA RNA I alone (no shRNA insert). Gray bars: VA RNA I-lamin shRNA vector. (C) Quantitative RT-PCR for OAS1 in HLFs infected at an MOI of 10 with RNAi vectors. White bar: control virus infection (H1 vector with no shRNA insert). Black bar: VA RNA I alone (no shRNA insert). Gray bar: VA RNA I-lamin shRNA vector.
expressing VA RNA I and VA RNA I-lamin shRNA (Fig. 5C). Compared to a blank lentiviral vector, OAS1 was induced 12-fold by VA RNA I alone, and 47-fold by the VA RNA I-lamin shRNA fusion. We conclude that VA RNA I fusions can be used to silence gene expression in mammalian cells, but additional modification of the construct will be required to achieve the same level of silencing as can be achieved with standard shRNA vectors, and this approach is unlikely to provide a means to circumvent the ISG induction by shRNA vectors.
Discussion In this study, we have identified the design flaw in a set of U6 promoter-based shRNA vectors previously reported to induce an interferon response (Bridge et al., 2003). The vectors contain a U6 promoter shRNA cassette in a self-inactivating lentiviral backbone. The use of lentiviruses to express shRNAs is attractive because they can be used to stably silence gene expression in a wide range of cells, including cells
that are difficult to transfect with plasmid vectors. The ISG induction is unrelated to the fact that a viral vector was used. Indeed, we previously showed that the lentiviral approach was less likely than plasmid vectors to induce OAS1 expression (Bridge et al., 2003). The cell contains at least three types of sensor for dsRNA: toll-like receptor 3 (TLR3), PKR, and members of the OAS enzyme family. TLR3 is a possible candidate for sensing the presence of shRNA but it is a cell membrane-bound receptor, which may not have access to cytosolic dsRNA (Alexopoulou et al., 2001). PKR is known to activate IRF-3, which would readily explain the ISG expression (Sen, 2001). The observation that VA RNA I, which inhibits PKR, itself activates OAS1 mRNA expression suggests that ISG expression can be induced in the absence of PKR activation. The increased OAS1 expression by a VA RNA I-shRNA fusion compared to VA RNA I alone further suggests that PKR is not the sensor that mediates shRNAdependent ISG induction. This is supported by the observation that VA RNA I also fails to suppress OAS1 induction by the MORF4L1 shRNA vector (A. Bridge and R.D.I., unpublished data). To completely exclude a role for PKR, we would have to demonstrate that our lentiviral VA RNA vector was biologically active, for example by showing that it could block eIF2a phosphorylation in response to dsRNA, or complement the growth defect of a VA RNA-deleted adenovirus (Cascallo et al., 2003). This is an important issue because Sledz et al. (2003) showed that PKR is required for the induction of interferon expression following transfection of siRNA oligonucleotides. The 19 bp dsRNA stem in our shRNAs is generally considered too short to induce an ISG response. However, some members of the OAS family of enzymes, the third type of sensor for dsRNA, can be activated by dsRNA as short as 15 bp (Sarkar et al., 1999). To define the precise length of dsRNA stem able to induce ISG expression in vivo would require extensive mutagenesis of our constructs, with compensatory mutations in each strand. Deletion of one base in the antisense strand of one of our vectors, which would be expected to reduce the length of the dsRNA stem to 14 bp, abolished the ISG response, suggesting that the response uses sensors that can detect 19 but not 14 bp dsRNA molecules. Because only 7 out of 22 shRNAs induced OAS1, the length of the stem alone is not sufficient to explain the ISG induction. To identify the additional causes of the ISG response, we performed mutational analysis of two ISG-inducing shRNA constructs. First, we determined that neither the orientation of the U6 promoter relative to the LTRs nor the length of transcription termination signal was responsible for the ISG response. Mattaj and colleagues have shown that Xenopus U6 transcription is extremely sensitive to changes near the start site. In particular, efficient transcription requires the presence of a G residue at the 11 position (Mattaj et al., 1988).
110
Substitution of this nucleotide with a C reduced transcription by about 20-fold. Replacement of G by A gave only a 3-fold reduction (Mattaj et al., 1988), indicating that the main requirement is for a purine at this position. Analysis of tRNAleu3 gene transcription likewise revealed that a pyrimidine-purine sequence (with initiation at the purine residue) was necessary for proper transcription initiation (Fruscoloni et al., 1995). The ISG response by our shRNA vectors was abolished when a C/G sequence at the –1/11 position was respected. Insertion of AA residues at the transcription start site, giving a C/A sequence at 1/11, systematically resulted in OAS1 expression. While this may solve the problem in a technical sense, it does not address the mechanism. It is possible that by changing the sequences at the normal start site, we force the polymerase to initiate transcription at a new site. This could result in the formation of aberrant transcripts that do not fold correctly. If delayed processing of shRNA to siRNA is the critical factor, it is worth noting that the presence of AA residues near the start site of our constructs could lead to the formation of blunt-ended shRNA by base pairing with the UU tail. The efficiency of silencing showed no correlation with OAS1 induction, suggesting that the amount of shRNA and the efficiency of processing were not direct determinants of OAS1 induction. The overall silencing efficiency of the U6 vectors was disappointing, a result best explained by recent reports that 5 0 -end stability determines which strand is incorporated into RISC (Khvorova et al., 2003; Schwarz et al., 2003). Taken together with our data showing that an AA sequence at the 5 0 end of the shRNA triggers ISG expression, it is clearly much better to place the sense strand before the antisense strand in shRNA vectors. In this configuration, the 5 0 end of the antisense strand (the strand complementary to the target mRNA) can be destabilized by sequences near the hairpin loop rather than the transcription start site. The mechanism of ISG induction may not be the same in our U6 vectors as in the MORF4L1 H1 vector that induced interferon expression (Bridge et al., 2003), because it had a C/G sequence at the transcription start site. In conclusion, we have identified the proximate cause of ISG induction by some of our U6 shRNA constructs. The exact start site and structure of the RNA molecules triggering the response are unknown, but they may only be a small fraction of the total pool of transcripts, which would make their identification extremely difficult. Because small differences in vector design can have such large effects on the cellular response to shRNA vectors, the key message from this study is that offtarget effects should always be considered as a possible explanation for shRNA phenotypes. Acknowledgments We thank the Swiss National Foundation for financial support.
References Alexopoulou, L., Holt, A.C., Medzhitov, R. and Flavell, R.A. (2001) Recognition of double-stranded RNA and activation of NF-kB by Toll-like receptor 3. Nature 413:732–738. Bridge, A.J., Pebernard, S., Ducraux, A., Nicoulaz, A.L. and Iggo, R. (2003) Induction of an interferon response by RNAi vectors in mammalian cells. Nat Genet 34:263–264. Brummelkamp, T.R., Bernards, R. and Agami, R. (2002) A system for stable expression of short interfering RNAs in mammalian cells. Science 296:550–553. Cascallo, M., Capella, G., Mazo, A. and Alemany, R. (2003) Rasdependent oncolysis with an adenovirus VAI mutant. Cancer Res 63:5544–5550. Elbashir, S.M., Harborth, J., Lendeckel, W., Yalcin, A., Weber, K. and Tuschl, T. (2001) Duplexes of 21-nucleotide RNAs mediate RNA interference in cultured mammalian cells. Nature 411:494– 498. Fire, A., Xu, S., Montgomery, M.K., Kostas, S.A., Driver, S.E. and and Mello, C.C. (1998) Potent and specific genetic interference by double-stranded RNA in Caenorhabditis elegans. Nature 391:806–811. Fruscoloni, P., Zamboni, M., Panetta, G., De Paolis, A. and Tocchini-Valentini, G.P. (1995) Mutational analysis of the transcription start site of the yeast tRNA(Leu3) gene. Nucleic Acids Res 23:2914–2918. Gunnery, S., Ma, Y. and Mathews, M.B. (1999) Termination sequence requirements vary among genes transcribed by RNA polymerase III. J Mol Biol 286:745–757. Khvorova, A., Reynolds, A. and Jayasena, S.D. (2003) Functional siRNAs and miRNAs exhibit strand bias. Cell 115:505. Klampfer, L., Huang, J., Corner, G., Mariadason, J., Arango, D., Sasazuki, T., Shirasawa, S. and Augenlicht, L. (2003) Oncogenic Ki-Ras inhibits the expression of interferon-responsive genes through inhibition of STAT1 and STAT2 expression. J Biol Chem 278:46278–46287. Lobo, S.M., Ifill, S. and Hernandez, N. (1990) cis-acting elements required for RNA polymerase II and III transcription in the human U2 and U6 snRNA promoters. Nucleic Acids Res 18:2891–2899. Ma, Y. and Mathews, M.B. (1993) Comparative analysis of the structure and function of adenovirus virus-associated RNAs. J Virol 67:6605–6617. Manche, L., Green, S.R., Schmedt, C. and Mathews, M.B. (1992) Interactions between double-stranded RNA regulators and the protein kinase DAI. Mol Cell Biol 12:5238–5248. Mathews, M.B. and Shenk, T. (1991) Adenovirus virus-associated RNA and translation control. J Virol 65:5657–5662. Mattaj, I.W., Dathan, N.A., Parry, H.D., Carbon, P. and Krol, A. (1988) Changing the RNA polymerase specificity of U snRNA gene promoters. Cell 55:435–442. Mori, K., Juttermann, R., Wienhues, U., Kobayashi, K., Yagi, M., Sugimoto, T., Tjia, S.T., Doerfler, W. and Hosokawa, K. (1996) Anti-interferon activity of adenovirus-2-encoded VAI and VAII RNAs in translation in cultured human cells. Virus Res 42:53–63. Naldini, L., Blomer, U., Gallay, P., Ory, D., Mulligan, R., Gage, F.H., Verma, I.M. and Trono, D. (1996) In vivo gene delivery and stable transduction of nondividing cells by a lentiviral vector. Science 272:263–267. Paddison, P.J., Caudy, A.A., Bernstein, E., Hannon, G.J. and Conklin, D.S. (2002) Short hairpin RNAs (shRNAs) induce sequence-specific silencing in mammalian cells. Genes Dev 16:948–958. Paule, M.R. and White, R.J. (2000) Survey and summary: transcription by RNA polymerases I and III. Nucleic Acids Res 28:1283–1298. Sarkar, S.N., Bandyopadhyay, S., Ghosh, A. and Sen, G.C. (1999) Enzymatic characteristics of recombinant medium isozyme of 2 0 5 0 oligoadenylate synthetase. J Biol Chem 274:1848–1855.
111 Schwarz, D.S., Hutvagner, G., Du, T., Xu, Z., Aronin, N. and Zamore, P.D. (2003) Asymmetry in the assembly of the RNAi enzyme complex. Cell 115:199–208. Sen, G.C. (2001) Viruses and interferons. Annu Rev Microbiol 55:255–281. Sledz, C.A., Holko, M., De Veer, M.J., Silverman, R.H. and Williams, B.R. (2003) Activation of the interferon system by short-interfering RNAs. Nat Cell Biol 5:834–839. Stark, G.R., Kerr, I.M., Williams, B.R., Silverman, R.H. and Schreiber, R.D. (1998) How cells respond to interferons. Annu Rev Biochem 67:227–264. Stojdl, D.F., Lichty, B., Knowles, S., Marius, R., Atkins, H., Sonenberg, N. and Bell, J.C. (2000) Exploiting tumor-specific
defects in the interferon pathway with a previously unknown oncolytic virus. Nat Med 6:821–825. Svoboda, P., Stein, P., Hayashi, H. and Schultz, R.M. (2000) Selective reduction of dormant maternal mRNAs in mouse oocytes by RNA interference. Development 127:4147– 4156. Wianny, F. and Zernicka-Goetz, M. (2000) Specific interference with gene function by double-stranded RNA in early mouse development. Nat Cell Biol 2:70–75. Zufferey, R., Dull, T., Mandel, R.J., Bukovsky, A., Quiroz, D., Naldini, L. and Trono, D. (1998) Self-inactivating lentivirus vector for safe and efficient in vivo gene delivery. J Virol 72:9873– 9880.