Lentiviral strategies for RNAi gene knockdown


Among the 20-30 thousand protein-encoding genes in the human genome, the function of only a fraction are known. A new tool for systematically deciphering the functions and interactions of these thousands of genes has been developed: short hairpin RNAs. These hairpin RNAs are precursors to the short interfering RNAs (siRNAs) that are the powerful mediators of RNA interference (RNAi) (1, 2). In RNAi, genes homologous in sequence to the siRNA are silenced at the post-transcriptional state.

The original reports on the design and use of short hairpin RNAs revealed that there are a variety of different hairpin structures that may give rise to effective siRNAs (3-9). These reports addressed two major limitations of short interfering RNA (siRNA) technology: the efficiency of transfection and the longevity of the silenced state. These original reports used polymerase III (pol III) promoters to transcribe the siRNAs or RNA hairpins, and later reports have shown efficacy of siRNA generation from pol II promoters (10).

Lentiviruses, such as the human immunodeficiency virus (HIV) are capable of infecting non-dividing cells, including differentiated neurons of the brain. Short hairpin RNAs can be expressed from lentiviruses, allowing for high efficiency transfection of a variety of cell types. Approaches toward lentiviral hairpin-mediated RNAi are presented, and methodologies for deriving RNAi-transgenic cells and animals are discussed.


Choice of lentiviral vector

There are a number of lentiviral vectors available for performing short hairpin-directed RNAi in mice (11-14). Some vectors encode green fluorescent protein (GFP) as a marker for infectivity and to track the RNAi knockdown cells in real time. Other vectors are available which contain antibiotic selection markers, such as puromycin, which are useful when selection is imposed in an experiment. In addition, there are conditional versions of lentiviral RNAi vectors available, where the user may opt to 'turn on' or 'turn off' RNAi during the course of experimentation (11, 15). Although most RNAi lentiviral vectors are freely available from academic sources, several commercial sources also provide vectors.


Design of the short RNA hairpin

One of the first steps towards using the lentiviral RNAi method is to design an effective RNA hairpin construct. Although not every hairpin construct will produce an effective RNAi response, rules have been developed that enrich for successful constructs. These rules are based on the examination of large numbers of effective constructs and thermodynamic analysis of microRNAs and effective siRNAs(16-18). Prediction algorithms for designing hairpin RNAs (and siRNAs) are available on the web where investigators can input their gene sequence into an online-window, and a program will output the DNA oligos that need to be synthesized to make the constructs. Several institutions have undertaken large-scale efforts to create lentiviral RNAi libraries, ie. a vector for every human and mouse gene.


Cloning the short hairpin sequence into the lentivirus plasmid

Creating a short hairpin RNAi vector can be accomplished by several different methods. In one method, two long complementary DNA oligos are annealed and directionally cloned into the expression vector. Since only two DNA oligos need to be made, this approach is both intuitive and easy to design. In another strategy, the entire hairpin sequence may be included as part of one of the oligos. In this strategy, the ~300 nt U6 promoter may be amplified by PCR, and the resulting PCR product would contain the hairpin entity. Because these two approaches require long oligos, the final constructs may be more subject to sequence errors frequently found in long synthetic DNA oligos. An additional method relies on the synthesis of four different DNA oligos, whereby DNA oligos are annealed and directly cloned into the vector (in a manner similar to that mentioned above). However, this approach has the additional advantage in that short oligos are usually easier and faster to synthesize. There are additional strategies that may be used to clone hairpins into DNA vectors-in short, the creation of short hairpin lentiviral vectors is relatively easy. Since off-targeting may generate a phenotype, at least two different hairpin vectors should be made for each targeted gene, each of which should mitigate identical phenotypes. If the 3'-UTR of the gene is targeted by the RNA hairpin, rescue of the phenotype may be attempted by ectopic expression of the targeted gene that contains, for example, a SV40 UTR.

Safety note- Be aware of the local safety policies regarding construction of human-targeted genes in lentiviruses. For example, higher level BLS lab status may be required if cloning hairpins against human tumor suppressors.


Production of short hairpin RNA virus

Packaging refers to the preparation of competent virus from DNA vectors. Packaging an RNAi lentivirus is essentially the same as packaging a lentivirus carrying a cDNA. In essence, DNA vectors are transiently transfected into a packaging cell line- such as human 293 cells, and after 2-3 days the supernatant will contain the virus.

For the most part, lentiviral vector production systems are based on a 'split' system, where the natural viral genome has been split into individual helper plasmid constructs. This splitting of the different viral elements into three or four separate vectors diminishes the risk of creating a replication-capable virus by adventitious recombination of the lentiviral genome.

When choosing a lentiviral production system, the user has the option to prepare viruses that have a restricted host range (ie. virus that may infect only rodents) vs. a broad host range (virus that may infect mouse, birds, human, etc). For the most part, the viral surface coat protein determines the species specificity. Because the lentiviral production systems are split, this coat protein can be switched by using, for example, the vesicular stomatitis virus (VSV/G) glycoprotein (which display a wide host range tropism) vs. an ecotropic maltose binding surface glycoprotein (which displays a limited specificity).

There are two major points to consider when choosing which coat (pseudotype) to use in the laboratory. First, because VSV/G pseudotyped virus may infect human tissue, it may not be preferable to use if one is working strictly in a mouse system. Working with human infectious agents will likely require a heightened BLS lab safety level. A second point of consideration is the limitations imposed by the ecotropic coat: it is difficult to concentrate the virus due to the inherent instability of the ecotropic coat. Virus ultracentrifugation concentration is often necessary when infecting certain cell types or when generating transgenic mice, and for this reason, many choose to use VSV/G pseudotyped virus because the particles are stable to the high g-forces of ultracentrifugation. Even with highly concentrated virus, there is a possibility that the VSV/G or ecotropic pseudotypes may still display differing cell-type specificities for infection, an important consideration when planning an experiment. For example, in the CNS, VSV/G mediates vector entry primarily in neurons, while astrocytes appear to be better infected using other envelopes.

Safety note- VSV/G-pseudotyped lentiviruses are capable of infecting nondiving human primary cells, and appropriate laboratory safety considerations must be observed. The appropriate local biosafety officers should be aware of all experiments involving lentiviruses. In general, the use of lentiviruses will require review and approval of relevant biosafety protocols, training in the use of infectious virus, and at least a BL2 level laboratory status.

The transfer vector construct that express the shRNA contains several optimized genetic elements in order to fulfill biosafety criteria and to increase its transduction efficiency: deletion of the U3 sequence within the LTR creates a self-inactivating (SIN) vector, which reduces the number of HIV-1 viral sequences, prevent transcriptional interference between the LTR and internal promoters, and reduces the possibility of transcriptional activation of genes adjacent to the provirus (19, 20). To increase the nuclear translocation efficiency of lentiviral pre-integration complexes, a central polypurine tract (cPPT) associated with its central terminating sequence is integrated into the viral backbone vector(21, 22). Furthermore, the addition of the woodchuck hepatitis B post-transcriptional responsive element (WPRE) increases the messenger RNA stability, thereby enhancing the viral transgene expression level(23, 24).


Viral Packaging

Getting a high titer virus preparation depends on the efficiency of packaging. In general, the higher the transfection of the DNA vectors, the higher the viral titer. As a rough guideline, a 90-100% transfection efficiency of 293 packaging cells should be easy to achieve using the protocols described herein. The quality of the DNA, cells, and the transfection conditions are all potential variables that need to be considered when optimizing:

DNA: In terms of plasmid preparation, we have not observed a need to use E. coli cells that are highly defective for recombination, and lentiviruses carrying the short hairpins appear to keep the inverted DNA in a stable form. DH5_ or a similar strain should be completely suitable. High DNA quality usually means high transfection efficiency. We generally recommend that all DNA preparations be performed using either by Cesium prep or endotoxin-free ion exchange plasmid purification methods. If poor transfection is consistently observed, it may be worth performing a additional clean-up of the DNA. The transfection protocols described here are sensitive to the amount of DNA. It is important to optimize the DNA:Fugene ratios by following the manufacturer's recommendations.

Packaging cells: Human 293 cells are usually used for packaging, since they can be transfected with efficiencies in the range of 90-100%. If the cells do not transfect well in control experiments using a non-viral GFP vector, it may be worth discarding them and trying a different lot of cells. On occasion, we have observed cells to become refractive to transfection. Consequently, we prefer to use low passage cells for all transfections. Although 293 cells are adherent, they can easily detach from the plate during pipetting, or even by jostling a plate that has reached confluency. For this reason, we recommend performing the transfection when the cells are approximately 30% confluent. After 36-48 hours, the cells should reach confluency and be producing the maximum amount of virus. If the cells are fed fresh media, several harvests of virus can be made after the cells reach confluency. One final note: if human cells are used for packaging a RNA hairpin that targets a human sequence, lower viral titers may be experienced if silencing of the targeted gene is detrimental to the cell or packaging process.

Transfection methods: Protocols for transfection the viral plasmids are essentially identical to transfection of any DNA vector. A less expensive alternative is method that uses calcium phosphate. The most important tactic of a transfection is optimization! Obviously there are many ways to optimize, but one easy approach towards monitoring the efficiency of transfection is to use a non-viral GFP vector, and quantitate the number and brightness of the packaging cells using flow cytometry. Viral packaging could then be performed using the optimal transfection conditions. If using a viral vector that contains GFP, such as lentilox, it is tempting to directly analyze transfection efficiency using flow cytometry, but this is not advisable. A major safety concern is that since the transfected cells would be producing virus, aerosol formation of the supernatant could occur, presenting a breach of safety. For this reason, it is advisable to evaluate the transfection efficiency by visual examination of the transfected packaging cells using an inverted fluorescent microscope. Since this is noninvasive, examination can be done during the course of transfection. Be certain that the sensitivity of the microscope is sufficient to properly detect the transfected cells, although this is usually not an issue since CMV-GFP is quite bright in 293T cells.

Using the virus

Once optimal transfection of the cells have been achieved, it is necessary to separate the viral-containing supernatant from the cells. This is easily done by filtration of the supernatant through a 0.45 uM syringe-filter. Although the filtered supernatant may be directly and immediately used in experiments, it is worth getting in the habit of titering the supernatant. Titering will give the viral concentration of the supernatant, which can be a useful indicator to compare transfections done at separate times. In addition, titering will allow one to know how many infectious particles are added in a given experiment-which is important for getting good reproducibility over the course of multiple viral preparations and experiments. Although a single RNAi viral integration may be sufficient to reduce the level of gene expression to negligible amounts, in some experimental settings, multiple integrations may be preferred.

Different cell types differ in their ability to be infected. For cell types that are difficult to transfect, adding in a charge-neutralizing agent, such as polybrene may help. This is not effective in all cell types, in which case concentrating the supernatant may be necessary. Only VSV/G pseudotyped virus may be concentrated in high speed ultracentrifugation, although it may be possible to concentrate ecotropic psuedotyped virus in low speed centrifugation concentrators that are used to concentrate MMLV/MSCV virus. For some cell types, performing a 'spin infection', where virus and cells are spun together at a very low speed in a centrifuge, may increase infection rates.

Safety note- When preparing virus, it is imperative to monitor for the adventitious emergence of a replication-competent virus. This involves periodically testing for the presence of infectious virus in supernatants of cells that have been previously infected by virus. Refer to your local biosafety office for protocols and guidelines.

Examining the knockdown

Once cells have been infected, it will be necessary to remove any contaminating uninfected cells. In general, there are two ways 'to purify' and then to analyze RNAi-mediated gene knockdown in cells: one in which the whole population of infected cells are examined, and in the second approach a selected number of individual clonal cell lines are examined. Determining which strategy to perform depends on the nature of the experiment. If the entire population is to be analyzed, either flow cytometric sorting (when GFP-expressing virus is used) or drug selection (when the virus contains an antibiotic-resistance marker) may be used. If a constitutively-expressing small hairpin RNAi vector is used, it will be important to monitor the viability/growth of the cells throughout the procedure. Some gene knockdowns produce slow-growing or lethal phenotypes. This can be difficult to assess when drug selection of the infected cells is used, in which case a GFP marker may be preferred over antibiotic-resistance markers. Alternatively, it may be advisable to use inducible RNAi vectors.

One alternative is to simply plate the cells at low density and pick individual colonies for analysis. If the virus expresses GFP, the cell colonies can be picked under a fluorescent microscope, allowing only for green cells to be carried forward to analysis. Each colony should be considered an individual line, and it should be verified that the lines are pure. One scenario we have observed that individual colonies may present themselves as 'chimeric' colonies in which only a portion of the cells in the colony are green. This may occur if the cells are not completely broken up into individual cells during trypsinization. In this case, it may be necessary to repeat the colony selection until an individual line is obtained. The number of individual lines to obtain depends on the efficiency of infection; we find it convenient to start by picking 12 or 24 lines, which can be amplified into 12-well plates for analysis.

If possible, the analysis of gene knockdown should be done at the protein level. This may be done using antibodies (Western blots, etc.) and by functional analysis. It should be expected that not all of the individual lines will display the same degree of gene knockdown. One reason is that the integration site may modulate the degree of expression. Once good RNAi-knockdown lines are obtained, it is important that at least 2-3 individual lines are carried through the experimental analysis.


Creation of transgenic RNAi mice by lentiviral transduction

Since lentiviruses are capable of infecting nondividing primary cells, they may be introduced directly into the whole animal by injection, or by ex vivo infection of cells and transplantation into irradiated mice. In a powerful application, transgenic whole animals may be derived by injection of lentiviral-infected embryonic stem (ES) cells into blasts using conventional methodologies, or direct lentiviral infection of single cell embryos. Alternatively, transgenic RNAi mice may be obtained by tetraploid complementation strategies, allow the investigator to bypass the need to generate chimeric animals and obtaining germline transmissible in the first generation.

Blast injection of RNAi-trangenic ES cells is highly recommendable. First, most transgenic mouse facilities are skilled in the art of performing ES-blast injections. Second, it gives the investigator the opportunity to characterize the degree of knockdown and phenotypes in the ES cell prior to injection. Selection of optimally knocked-down cell lines may save a lot of time in the long run.

Derivation of RNAi transgenic mice by injection of lentiviral-infected ES cells into blasts begins with the successful infection of ES cells. This usually requires concentrated virus, and typically infected cell lines are obtained by either GFP or antibiotic colony selection. Although a single integration is sufficient to cause an effective RNAi response, it is not absolutely necessary to focus on ES cells that contain only a single integrant- in fact, it may be advisable to use ES cells that contain multiple integrants. First, because integration site may affect transcriptional activity of the RNA hairpin, not all integrations will be active. One puzzling observation is that not all animals derived from lentiviral transgenesis will exhibit RNAi activity. This is presumably due to the effects of integration site, and perhaps collaborated by recognition of the inverted DNA sequence by cellular silencing factors and subsequent spreading of silencing to adjacent regions. Thus multiple integrations may enhance the odds of getting a good integration site and and thus an effective hairpin expression and RNAi response. Second, animals derived from mating of germline-transmissive chimeras may yield litters of pups that contain different integrations (depending on the proximity of the integrations to each other). This is advantageous because a large number of animals containing individual integrations can be generated quickly. Each animal should be considered an independent line, and will hopefully yield consistent phenotypes independent of the integration.

Direct lentiviral infection of mouse zygotes may be performed by direct injection of the single cell embryo with virus, or by digestion of the zona pellucida and subsequent incubation in virus. Again, this requires concentration of virus, and the quality of the viral prep is important since blasts can be exceptionally sensitive to any impurities that may be present. In general, the highest success rates are obtained from high titer viral preparations.

The advent of lentiviral RNAi technologies has provided biological research the ability to quickly generate transgenic animals. The time to generate lentiviral transgenic mice can be very short once the procedure is set up and optimized. It can typically take 4-6 weeks, starting from ordering the hairpin oligos to generation of the first litter of transgenic pups. That is much more rapid than the typical 9-12 months it may take to generate a knockout mouse using conventional homologous recombination technologies. Since RNAi is based on a reduction of gene expression, and not always a complete ablation of gene expression, it is rapidly finding its place as a complement to knockout technologies.

Lentiviral vectors are continuing to generate great interest as tools for gene therapy in vivo. Over the past two years, considerable progress has been made in demonstrating short RNAi hairpin lentiviral delivery to a wide range of tissues, and in a wide variety of experimental settings. The basic science of lentiviral RNAi use is expanding and the large number of laboratories reporting successful gene-knockdown reflects their relative ease of use. Indeed, the use of viral RNAi approaches in both cells and whole animals is transforming how basic science is approached and performed. Although none have been used in vivo in human clinical trials, it may be only a matter of time before this occurs.



Literature Cited

1. D. M. Dykxhoorn, C. D. Novina, P. A. Sharp, Nat Rev Mol Cell Biol 4, 457 (Jun, 2003).

2. M. T. McManus, P. A. Sharp, Nat Rev Genet 3, 737 (Oct, 2002).

3. C. P. Paul, P. D. Good, I. Winer, D. R. Engelke, Nat Biotechnol 20, 505 (May, 2002).

4. M. Miyagishi, K. Taira, Nat Biotechnol 20, 497 (May, 2002).

5. M. T. McManus, C. P. Petersen, B. B. Haines, J. Chen, P. A. Sharp, RNA 8, 842 (Jun, 2002).

6. P. J. Paddison, A. A. Caudy, E. Bernstein, G. J. Hannon, D. S. Conklin, Genes Dev 16, 948 (Apr 15, 2002).

7. T. R. Brummelkamp, R. Bernards, R. Agami, Science 296, 550 (Apr 19, 2002).

8. Y. Zeng, E. J. Wagner, B. R. Cullen, Mol Cell 9, 1327 (Jun, 2002).

9. J. Y. Yu, S. L. DeRuiter, D. L. Turner, Proc Natl Acad Sci U S A 99, 6047 (Apr 30, 2002).

10. H. Xia, Q. Mao, H. L. Paulson, B. L. Davidson, Nat Biotechnol 20, 1006 (Oct, 2002).

11. A. Ventura et al., PNAS 101, 10380 (July, 2004).

12. D. A. Rubinson et al., Nat Genet 33, 401 (Mar, 2003).

13. T. Kunath et al., Nat Biotechnol 21, 559 (May, 2003).

14. M. A. Carmell, L. Zhang, D. S. Conklin, G. J. Hannon, T. A. Rosenquist, Nat Struct Biol 10, 91 (Feb, 2003).

15. G. Tiscornia, V. Tergaonkar, F. Galimi, I. M. Verma, Proc Natl Acad Sci U S A 101, 7347 (May 11, 2004).

16. D. S. Schwarz et al., Cell 115, 199 (Oct 17, 2003).

17. A. Khvorova, A. Reynolds, S. D. Jayasena, Cell 115, 209 (Oct 17, 2003).

18. A. Reynolds et al., Nat Biotechnol 22, 326 (Mar, 2004).

19. R. Zufferey et al., J Virol 72, 9873 (Dec, 1998).

20. T. Iwakuma, Y. Cui, L. J. Chang, Virology 261, 120 (Aug 15, 1999).

21. A. Follenzi, L. E. Ailles, S. Bakovic, M. Geuna, L. Naldini, Nat Genet 25, 217 (Jun, 2000).

22. V. Zennou et al., Cell 101, 173 (Apr 14, 2000).

23. J. E. Donello, J. E. Loeb, T. J. Hope, J Virol 72, 5085 (Jun, 1998).

24. R. Zufferey, J. E. Donello, D. Trono, T. J. Hope, J Virol 73, 2886 (Apr, 1999).