There are several methods for preparing siRNA, such as chemical synthesis, in vitro transcription, siRNA expression vectors, and PCR expression cassettes. Irrespective of which method one uses, the first step in designing a siRNA is to choose the siRNA target site. The guidelines below for choosing siRNA target sites are based on both the current literature, and on empirical observations by scientists at Ambion. Using these guidelines, approximately half of all siRNAs yield >50% reduction in target mRNA levels.
For the Best Results, Let Us Design Your siRNAs
Ambion has already designed siRNAs to >35,000 human, mouse, and rat targets using a proprietary siRNA design process. For more information on these highly effective siRNAs, please visit our
Silencer Select siRNAs Information Page. To purchase Ambion
Silencer Select Pre-designed or Validated siRNAs, start by searching our siRNA Database.
Silencer Select Pre-designed and Validated siRNAs are guaranteed to silence and available exclusively from Ambion/Applied Biosystems.
General Design Guidelines
If you prefer to design your own siRNAs, you can choose siRNA target sites in a variety of different organisms based on the following guidelines. Corresponding siRNAs can then be chemically synthesized, created by in vitro transcription, or expressed from a vector or PCR product.
1. Find 21 nt sequences in the target mRNA that begin with an AA dinucleotide.
2. Select 2-4 target sequences.
3. Design appropriate controls.
1. Find 21 nt sequences in the target mRNA that begin with an AA dinucleotide.
Beginning with the AUG start codon of your transcript, scan for AA dinucleotide sequences. Record each AA and the 3' adjacent 19 nucleotides as potential siRNA target sites.
This strategy for choosing siRNA target sites is based on the observation by Elbashir et al. (1) that siRNAs with 3' overhanging UU dinucleotides are the most effective. This is also compatible with using RNA pol III to transcribe hairpin siRNAs because RNA pol III terminates transcription at 4-6 nucleotide poly(T) tracts creating RNA molecules with a short poly(U) tail.
In Elbashir's and subsequent publications, siRNAs with other 3' terminal dinucleotide overhangs have been shown to effectively induce RNAi. If desired, you may modify this target site selection strategy to design siRNAs with other dinucleotide overhangs, but it is recommended that you avoid G residues in the overhang because of the potential for the siRNA to be cleaved by RNase at single-stranded G residues.
2. Select 2-4 target sequences.
Research at Ambion has found that typically more than half of randomly designed siRNAs provide at least a 50% reduction in target mRNA levels and approximately 1 of 4 siRNAs provide a 75-95% reduction. Choose target sites from among the sequences identified in Step 1 based on the following guidelines:
- Ambion researchers find that siRNAs with 30-50% GC content are more active than those with a higher G/C content.
- Since a 4-6 nucleotide poly(T) tract acts as a termination signal for RNA pol III, avoid stretches of > 4 T's or A's in the target sequence when designing sequences to be expressed from an RNA pol III promoter.
- Since some regions of mRNA may be either highly structured or bound by regulatory proteins, we generally select siRNA target sites at different positions along the length of the gene sequence. We have not seen any correlation between the position of target sites on the mRNA and siRNA potency.
- Compare the potential target sites to the appropriate genome database (human, mouse, rat, etc.) and eliminate from consideration any target sequences with more than 16-17 contiguous base pairs of homology to other coding sequences. We suggest using BLAST, which can be found on the NCBI server at: www.ncbi.nlm.nih.gov/BLAST.
3. Design appropriate controls.
A complete siRNA experiment should include a number of controls to ensure the validity of the data. The editors of Nature Cell Biology have recommended several controls (2). Two of these controls are:
- A negative control siRNA with the same nucleotide composition as your siRNA but which lacks significant sequence homology to the genome. To design a negative control siRNA, scramble the nucleotide sequence of the gene-specific siRNA and conduct a search to make sure it lacks homology to any other gene.
- Additional siRNA sequences targeting the same mRNA. Perhaps the best way to ensure confidence in RNAi data is to perform experiments, using a single siRNA at a time, with two or more different siRNAs targeting the same gene. Prior to these experiments, each siRNA should be tested to ensure that it reduces target gene expression by comparable levels.
Specific Guidelines for Designing siRNA Hairpins Encoded by siRNA Expression Vectors and siRNA Expression Cassettes
Researchers who initially reported the use of siRNA expression vectors to induce RNAi had different design criteria for their inserts encoding the expressed siRNA. Most of the designs had two inverted repeats separated by a short spacer sequence and ended with a string of T's that served as a transcription termination site. These designs produce an RNA transcript that is predicted to fold into a short hairpin siRNA as shown in Figure 1. The selection of siRNA target sequence, the length of the inverted repeats that encode the stem of a putative hairpin, the order of the inverted repeats, the length and composition of the spacer sequence that encodes the loop of the hairpin, and the presence or absence of 5'-overhangs, vary among different reports (3-11).
Figure 1 . Schematic of a Typical Hairpin siRNA Produced by an siRNA Expression Vector or an siRNA Expression Cassette and Its Relationship to the RNA Target Sequence.
Ambion's Recommended Procedure for siRNA Hairpin Design
The following recommendations for siRNA hairpin design and cloning strategy are made based on research by Ambion scientists. The first step in designing an appropriate insert is to choose the siRNA target site by following the steps described under "General Design Guidelines" above.
For screening, we typically test four siRNA sequences per target, spacing the siRNA sequences down the length of the gene sequence to reduce the chances of targeting a region of the mRNA that is either highly structured or bound by regulatory proteins. Because constructing and testing four siRNA expression plasmids per target is time-consuming, we find it much easier to screen potential siRNA sequences using PCR-derived siRNA expression cassettes (SECs). SECs are PCR products that include promoter and terminator sequences flanking a hairpin siRNA template. This screening strategy also permits the rapid identification of the best combination of promoter and siRNA sequence in the experimental system. SECs found to effectively elicit gene silencing can be readily cloned into a vector for long term studies. Ambion scientists have determined that sequences that function well as transfected siRNAs also function well as siRNAs that are expressed in vivo. The only exception is that siRNA sequences to be expressed in vivo should not contain a run of 4 or 5 A's or T's, as these can act as termination sites for Polymerase III.
For traditional cloning into pSilencer vectors, two DNA oligonucleotides that encode the chosen siRNA sequence are designed for insertion into the vector (Figures 2 and 3). In general, the DNA oligonucleotides consist of a 19-nucleotide sense siRNA sequence linked to its reverse complementary antisense siRNA sequence by a short spacer. Ambion scientists have successfully used a 9-nucleotide spacer (TTCAAGAGA), although other spacers can be designed. 5-6 T's are added to the 3' end of the oligonucleotide. In addition, for cloning into the pSilencer 1.0-U6 vector, nucleotide overhangs to the EcoR I and Apa I restriction sites are added to the 5' and 3' end of the DNA oligonucleotides, respectively (Figure 2). In contrast, for cloning into the pSilencer 2.0-U6, 2.1-U6, 3.0-H1, or 3.1-H1 vectors, nucleotide overhangs with BamH I and Hind III restriction sites are added to the 5' and 3' end of the DNA oligonucleotides, respectively (Figure 3). The resulting RNA transcript is expected to fold back and form a stem-loop structure comprising a 19 bp stem and 9 nt loop with 2-3 U's at the 3' end (Figure 1).
Figure 2. Insert Design for pSilencer 1.0-U6. This insert is specific for the pSilencer 1.0-U6 Vector and contains the appropriate 3' overhangs for directional cloning into this vector. The loop sequence and length can be varied as desired.
Figure 3 . Insert Design for pSilencer 2.0-U6 and pSilencer 3.0-H1. The insert design is specific for the pSilencer 2.0-U6, 2.1-U6, 3.0-H1 and 3.1-H1 Expression Vectors and contains the appropriate overhanging 5' ends for directional cloning into these plasmids. As with p Silencer 1.0-U6 shown in Figure 2, early indications suggest that a great deal of latitude is possible in the design of the loop; here we provide one loop sequence that we find works well.
For cloning into the p Silencer adeno 1.0-CMV vector, DNA oligonucleotides with stem-loop structures are created similar to those of p Silencer 2.0 and 3.0 vectors described above. However, one notable exception is the absence of 5-6 T's from the 3'-end of the oligonucleotides for the CMV-based vector system since the transcription termination signal for the CMV-based vector system is provided by the SV40 polyA terminator. In addition, for cloning into the pSilencer adeno 1.0-CMV vector, nucleotide overhangs containing the Xho I and Spe I restriction sites are added to the 5' and 3' end of the DNA oligonucleotides, respectively (Figure 4). However, for cloning into the pSilencer 4.1-CMV vector, nucleotide overhangs containing the Bam H1 and Hind III restriction sites are added to the 5' and 3' end of the DNA oligonucleotides, respectively (Figure 5).
Figure 4. Insert Design for pSilencer™ adeno 1.0-CMV Vector. This insert design is specific for the p Silencer adeno 1.0-CMV vector and contains the appropriate overhangs for directional cloning into this vector. The loop sequence and length can be varied as desired.
Figure 5. Insert Design for pSilencer™ 4.1-CMV Vector. This insert design is specific for the p Silencer 4.1-CMV vector and contains the appropriate overhangs for directional cloning into this vector. The loop sequence and length can be varied as desired.
Selection of siRNA Targets
In addition to the our own proprietary algorithm and our suggested procedure for selecting siRNA targets by scanning a mRNA sequence for AA dinucleotides and recording the 19 nucleotides immediately downstream of the AA, two other methods have been employed by other researchers. In the first method, the selection of the siRNA target sequence is purely empirically determined (4), as long as the target sequence starts with GG and does not share significant sequence homology with other genes as analyzed by BLAST search.
In the second report, a more elaborate method is employed to select the siRNA target sequences. This procedure exploits an observation that any accessible site in endogenous mRNA can be targeted for degradation by the synthetic oligodeoxyribonucleotide/RNase H method (5). Any accessible site identified in this fashion is then used as insert sequence in the U6 promoter-driven siRNA constructs.
Order of the Sense and Antisense Strands within the Hairpin siRNAs
A hairpin siRNA expression cassette is usually constructed to contain the sense strand of the target, followed by a short spacer, then the antisense strand of the target, in that order. One group of researchers has found that reversal of the order of sense and antisense strands within the siRNA expression constructs did not affect the gene silencing activities of the hairpin siRNA (6). In contrast, another group of researchers has found that similar reversal of order in another siRNA expression cassette caused partial reduction in the gene silencing activities of the hairpin siRNA (7). It is not clear what is responsible for this difference in observation. At the present time, it is still advisable to construct the siRNA expression cassette in the order of sense strand, short spacer, and antisense strand.
Length of the siRNA Stem
There appears to be some degree of variation in the length of nucleotide sequence being used as the stem of siRNA expression cassette. Several research groups including Ambion have used 19 nucleotides-long sequences as the stem of siRNA expression cassette (6-10). In contrast, other research groups have used siRNA stems ranging from 21 nucleotides-long (4-5) to 25-29 nucleotides-long (11). It is found that hairpin siRNAs with these various stem lengths all function well in gene silencing studies.
Length and Sequence of the Loop Linking Sense and Antisense Strands of Hairpin siRNA
Various research groups have reported successful gene silencing results using hairpin siRNAs with loop size ranging between 3 to 23 nucleotides (4, 6-9, 11). The following is a summary of loop size and specific loop sequences used by various research groups:
Presence of 5' Overhangs in the Hairpin siRNAs
Most research groups did not use a 5' overhang in their hairpin siRNA construct (4-8, 10-11). However, one research group included a 6 nucleotide 5' overhang in the hairpin siRNA constructs (9). These hairpin siRNAs with 5' overhangs were shown to be functional in gene silencing.
Figure 1 . Schematic of a Typical Hairpin siRNA Produced by an siRNA Expression Vector or an siRNA Expression Cassette and Its Relationship to the RNA Target Sequence.
Ambion's Recommended Procedure for siRNA Hairpin Design
The following recommendations for siRNA hairpin design and cloning strategy are made based on research by Ambion scientists. The first step in designing an appropriate insert is to choose the siRNA target site by following the steps described under "General Design Guidelines" above.
For screening, we typically test four siRNA sequences per target, spacing the siRNA sequences down the length of the gene sequence to reduce the chances of targeting a region of the mRNA that is either highly structured or bound by regulatory proteins. Because constructing and testing four siRNA expression plasmids per target is time-consuming, we find it much easier to screen potential siRNA sequences using PCR-derived siRNA expression cassettes (SECs). SECs are PCR products that include promoter and terminator sequences flanking a hairpin siRNA template. This screening strategy also permits the rapid identification of the best combination of promoter and siRNA sequence in the experimental system. SECs found to effectively elicit gene silencing can be readily cloned into a vector for long term studies. Ambion scientists have determined that sequences that function well as transfected siRNAs also function well as siRNAs that are expressed in vivo. The only exception is that siRNA sequences to be expressed in vivo should not contain a run of 4 or 5 A's or T's, as these can act as termination sites for Polymerase III.
For traditional cloning into pSilencer vectors, two DNA oligonucleotides that encode the chosen siRNA sequence are designed for insertion into the vector (Figures 2 and 3). In general, the DNA oligonucleotides consist of a 19-nucleotide sense siRNA sequence linked to its reverse complementary antisense siRNA sequence by a short spacer. Ambion scientists have successfully used a 9-nucleotide spacer (TTCAAGAGA), although other spacers can be designed. 5-6 T's are added to the 3' end of the oligonucleotide. In addition, for cloning into the pSilencer 1.0-U6 vector, nucleotide overhangs to the EcoR I and Apa I restriction sites are added to the 5' and 3' end of the DNA oligonucleotides, respectively (Figure 2). In contrast, for cloning into the pSilencer 2.0-U6, 2.1-U6, 3.0-H1, or 3.1-H1 vectors, nucleotide overhangs with BamH I and Hind III restriction sites are added to the 5' and 3' end of the DNA oligonucleotides, respectively (Figure 3). The resulting RNA transcript is expected to fold back and form a stem-loop structure comprising a 19 bp stem and 9 nt loop with 2-3 U's at the 3' end (Figure 1).
Figure 2. Insert Design for pSilencer 1.0-U6. This insert is specific for the pSilencer 1.0-U6 Vector and contains the appropriate 3' overhangs for directional cloning into this vector. The loop sequence and length can be varied as desired.
Figure 3 . Insert Design for pSilencer 2.0-U6 and pSilencer 3.0-H1. The insert design is specific for the pSilencer 2.0-U6, 2.1-U6, 3.0-H1 and 3.1-H1 Expression Vectors and contains the appropriate overhanging 5' ends for directional cloning into these plasmids. As with p Silencer 1.0-U6 shown in Figure 2, early indications suggest that a great deal of latitude is possible in the design of the loop; here we provide one loop sequence that we find works well.
For cloning into the p Silencer adeno 1.0-CMV vector, DNA oligonucleotides with stem-loop structures are created similar to those of p Silencer 2.0 and 3.0 vectors described above. However, one notable exception is the absence of 5-6 T's from the 3'-end of the oligonucleotides for the CMV-based vector system since the transcription termination signal for the CMV-based vector system is provided by the SV40 polyA terminator. In addition, for cloning into the pSilencer adeno 1.0-CMV vector, nucleotide overhangs containing the Xho I and Spe I restriction sites are added to the 5' and 3' end of the DNA oligonucleotides, respectively (Figure 4). However, for cloning into the pSilencer 4.1-CMV vector, nucleotide overhangs containing the Bam H1 and Hind III restriction sites are added to the 5' and 3' end of the DNA oligonucleotides, respectively (Figure 5).
Figure 4. Insert Design for pSilencer™ adeno 1.0-CMV Vector. This insert design is specific for the p Silencer adeno 1.0-CMV vector and contains the appropriate overhangs for directional cloning into this vector. The loop sequence and length can be varied as desired.
Figure 5. Insert Design for pSilencer™ 4.1-CMV Vector. This insert design is specific for the p Silencer 4.1-CMV vector and contains the appropriate overhangs for directional cloning into this vector. The loop sequence and length can be varied as desired.
Selection of siRNA Targets
In addition to the our own proprietary algorithm and our suggested procedure for selecting siRNA targets by scanning a mRNA sequence for AA dinucleotides and recording the 19 nucleotides immediately downstream of the AA, two other methods have been employed by other researchers. In the first method, the selection of the siRNA target sequence is purely empirically determined (4), as long as the target sequence starts with GG and does not share significant sequence homology with other genes as analyzed by BLAST search.
In the second report, a more elaborate method is employed to select the siRNA target sequences. This procedure exploits an observation that any accessible site in endogenous mRNA can be targeted for degradation by the synthetic oligodeoxyribonucleotide/RNase H method (5). Any accessible site identified in this fashion is then used as insert sequence in the U6 promoter-driven siRNA constructs.
Order of the Sense and Antisense Strands within the Hairpin siRNAs
A hairpin siRNA expression cassette is usually constructed to contain the sense strand of the target, followed by a short spacer, then the antisense strand of the target, in that order. One group of researchers has found that reversal of the order of sense and antisense strands within the siRNA expression constructs did not affect the gene silencing activities of the hairpin siRNA (6). In contrast, another group of researchers has found that similar reversal of order in another siRNA expression cassette caused partial reduction in the gene silencing activities of the hairpin siRNA (7). It is not clear what is responsible for this difference in observation. At the present time, it is still advisable to construct the siRNA expression cassette in the order of sense strand, short spacer, and antisense strand.
Length of the siRNA Stem
There appears to be some degree of variation in the length of nucleotide sequence being used as the stem of siRNA expression cassette. Several research groups including Ambion have used 19 nucleotides-long sequences as the stem of siRNA expression cassette (6-10). In contrast, other research groups have used siRNA stems ranging from 21 nucleotides-long (4-5) to 25-29 nucleotides-long (11). It is found that hairpin siRNAs with these various stem lengths all function well in gene silencing studies.
Length and Sequence of the Loop Linking Sense and Antisense Strands of Hairpin siRNA
Various research groups have reported successful gene silencing results using hairpin siRNAs with loop size ranging between 3 to 23 nucleotides (4, 6-9, 11). The following is a summary of loop size and specific loop sequences used by various research groups:
| Loop Size (# of Nucleotides) | Specific Loop Sequence | Reference |
|---|---|---|
| 3 | AUG | 4 |
| 3 | CCC | 7 |
| 4 | UUCG | 5 |
| 5 | CCACC | 7 |
| 6 | AAGCUU | 2 |
| 7 | CCACACC | 7 |
| 9 | UUCAAGAGA | 6 |
| 23 | Not Reported | 9 |
Presence of 5' Overhangs in the Hairpin siRNAs
Most research groups did not use a 5' overhang in their hairpin siRNA construct (4-8, 10-11). However, one research group included a 6 nucleotide 5' overhang in the hairpin siRNA constructs (9). These hairpin siRNAs with 5' overhangs were shown to be functional in gene silencing.
Chemical Synthesis of siRNA
Ambion synthesizes both customer designed siRNAs and siRNAs pre-designed using the Cenix algorithm.
To order a chemically synthesized siRNA for which you already have the design, you can either provide:
OR
Ambion will synthesize a complementary pair of siRNA oligonucleotides according to your sequence. By default, siRNAs for which you provide only the mRNA target sequence will be synthesized with dTdT 3' overhangs. If you wish, you can choose UU or other overhangs. Our scientists observe no functional difference in the potency of siRNA made with dTdT or UU overhangs. (Note: the 3' dTdT of the sense strand does not have to be complementary to the target gene.)
Currently, Ambion Pre-designed siRNAs are available for >98% of all human, mouse, and rat genes in the RefSeq database maintained by NCBI. To order a pre-designed siRNA, search our siRNA database for your gene of interest, choose the design(s) you'd like to purchase, add them to your cart, and transfer the relevant information about each to our online oligo order form. See Designing a Better siRNA for information on the design algorithm used.
To order a chemically synthesized siRNA for which you already have the design, you can either provide:
- the ~21 base mRNA sequence (starting with the AA dinucleotide) to which the siRNA will be directed
OR
- the sequence of each siRNA strand (This option is recommended if you wish your siRNA to have 3' termini other than dTdT or UU.)
Ambion will synthesize a complementary pair of siRNA oligonucleotides according to your sequence. By default, siRNAs for which you provide only the mRNA target sequence will be synthesized with dTdT 3' overhangs. If you wish, you can choose UU or other overhangs. Our scientists observe no functional difference in the potency of siRNA made with dTdT or UU overhangs. (Note: the 3' dTdT of the sense strand does not have to be complementary to the target gene.)
Currently, Ambion Pre-designed siRNAs are available for >98% of all human, mouse, and rat genes in the RefSeq database maintained by NCBI. To order a pre-designed siRNA, search our siRNA database for your gene of interest, choose the design(s) you'd like to purchase, add them to your cart, and transfer the relevant information about each to our online oligo order form. See Designing a Better siRNA for information on the design algorithm used.
Other Methods of siRNA Preparation
To prepare siRNA by in vitro transcription, siRNA expression vector, or PCR-generated siRNA expression cassette, appropriate templates must be prepared. Web-based tools for designing these templates are available for the following Ambion kits/products:
These tools are also accessible from the siRNA Target Finder described above.
These tools are also accessible from the siRNA Target Finder described above.
References
1. Elbashir, et al. (2001) Functional anatomy of siRNA for mediating efficient RNAi in Drosophila melanogaster embryo lysate. EMBO J 20: 6877-6888.
2. Editors of Nature Cell Biology (2003) Whither RNAi? Nat Cell Biol. 5:489-490.
3. Brown, D., Jarvis, R., Pallotta, V., Byrom, M., and Ford, L. (2002) RNA interference in mammalian cell culture: design, execution, and analysis of the siRNA effect. Ambion TechNotes 9(1): 3-5.
4. Sui, G., Soohoo, C., Affar, E.B., Gay, F., Shi, Y., Forrester, W.C., and Shi, Y. (2002) A DNA vector-based RNAi technology to suppress gene expression in mammalian cells. Proc. Natl. Acad. Sci. US A 99(8): 5515-5520.
5. Lee, N.S., Dohjima, T., Bauer, G., Li, H., Li, M.-J., Ehsani, A., Salvaterra, P., and Rossi, J. (2002) Expression of small interfering RNAs targeted against HIV-1 rev transcripts in human cells. Nature Biotechnology 20 : 500-505.
6. Yu, J.-Y., DeRuiter, S.L., and Turner, D.L. (2002) RNA interference by expression of short-interfering RNAs and hairpin RNAs in mammalian cells. Proc. Natl. Acad. Sci. USA 99(9) : 6047-6052.
7. Paul, C.P., Good, P.D., Winer, I., and Engelke, D.R. (2002) Effective expression of small interfering RNA in human cells. Nature Biotechnology 20 : 505-508.
8. 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.
9. Jacque, J.-M., Triques, K., and Stevenson, M. (2002) Modulation of HIV-1 replication by RNA interference. Nature 418 : 435-438.
10. Miyagishi, M., and Taira, K. (2002) U6 promoter-driven siRNAs with four uridine 3' overhangs effectively suppress targeted gene expression in mammalian cells. Nature Biotechnology 20 : 497-500.
11. Paddison, P.J., Caudy, A.A., Berstein, E., Hannon, G.J., and Conklin, D.S. (2002) Short hairpin RNAs (shRNAs) induce sequence-specific silencing in mammalian cells. Genes Devel. 16: 948-958.
2. Editors of Nature Cell Biology (2003) Whither RNAi? Nat Cell Biol. 5:489-490.
3. Brown, D., Jarvis, R., Pallotta, V., Byrom, M., and Ford, L. (2002) RNA interference in mammalian cell culture: design, execution, and analysis of the siRNA effect. Ambion TechNotes 9(1): 3-5.
4. Sui, G., Soohoo, C., Affar, E.B., Gay, F., Shi, Y., Forrester, W.C., and Shi, Y. (2002) A DNA vector-based RNAi technology to suppress gene expression in mammalian cells. Proc. Natl. Acad. Sci. US A 99(8): 5515-5520.
5. Lee, N.S., Dohjima, T., Bauer, G., Li, H., Li, M.-J., Ehsani, A., Salvaterra, P., and Rossi, J. (2002) Expression of small interfering RNAs targeted against HIV-1 rev transcripts in human cells. Nature Biotechnology 20 : 500-505.
6. Yu, J.-Y., DeRuiter, S.L., and Turner, D.L. (2002) RNA interference by expression of short-interfering RNAs and hairpin RNAs in mammalian cells. Proc. Natl. Acad. Sci. USA 99(9) : 6047-6052.
7. Paul, C.P., Good, P.D., Winer, I., and Engelke, D.R. (2002) Effective expression of small interfering RNA in human cells. Nature Biotechnology 20 : 505-508.
8. 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.
9. Jacque, J.-M., Triques, K., and Stevenson, M. (2002) Modulation of HIV-1 replication by RNA interference. Nature 418 : 435-438.
10. Miyagishi, M., and Taira, K. (2002) U6 promoter-driven siRNAs with four uridine 3' overhangs effectively suppress targeted gene expression in mammalian cells. Nature Biotechnology 20 : 497-500.
11. Paddison, P.J., Caudy, A.A., Berstein, E., Hannon, G.J., and Conklin, D.S. (2002) Short hairpin RNAs (shRNAs) induce sequence-specific silencing in mammalian cells. Genes Devel. 16: 948-958.