Design Rules for Dicer-Substrate RNAi - RNAi

Design Rules for Dicer-Substrate RNAi
2005
Mark A. Behlke, M.D., Ph.D.
Genetic Engineering News
RNA interference is a biochemical process common to most eukaryotes where double-stranded RNA (dsRNA) directs suppression of gene expression in a sequence-specific manner1-3. Long dsRNAs have been used for over a decade as tools to alter gene expression in plants, yeast, and C. elegans.
RNAs longer than 30-35 bases, however, usually trigger interferon responses in higher organisms and therefore cannot routinely be employed in studies done in mammals.
Long dsRNAs are not the actual mediators of RNAi. The RNase-III class endoribonuclease Dicer cleaves long dsRNAs into 21-23 base duplexes having 2-base 3'-overhangs and 5'-phosphate4,5. These species, called small interfering RNAs (siRNAs), enter the RNA Induced Silencing Complex (RISC) and direct sequence-specific degradation of mRNA6.
In Drosophila, siRNAs cannot directly enter RISC but must first associate with Dicer and a second RNA-binding protein R2D27,8. In mammals, Dicer is not required for siRNA to enter RISC9,10. The human ortholog of R2D2, TRBP, has recently been identified11.
TRBP also forms a heterodimer with Dicer and may similarly facilitate entry of the siRNA into RISC, even if the presence of this complex is not absolutely required for functional RISC assembly. The basic biochemical pathway of degradative RNA interference was recently reviewed by Sontheimer
The discovery of siRNAs permitted RNAi to be used as an experimental tool in mammalian cells and has met with amazing success over the past four years to study gene function in experiments that range from knockdown of a single gene to ambitious whole-genome surveys.
The reagents most commonly used today are made as single-stranded oligonucleotides annealed to produce 21-mer duplexes with 2-base 3'-overhangs. These duplexes can be transfected into cells and directly mimic the products made by normal Dicer processing. In Drosophila, siRNAs are limited to 21-23mers and duplexes 25-39 bp length are inactive6. In mammals, the biochemistry is slightly different and duplexes 25-30mer length can trigger RNAi responses9,13.
Increased Potency of Longer Duplexes
In fact, the use of longer duplexes can offer some advantages. It was recently shown that synthetic RNA duplexes of 25-30 base length can have as much as a 100-fold increase in potency compared with 21mers at the same location14.
In studies reported by Kim and colleagues, a series of RNA duplexes specific for the same site in Enhanced Green Fluorescent Protein (EGFP) were made that ranged from 21-30 bases in length. These duplexes were compared for relative potency in reducing EGFP fluorescence in a synthetic reporter assay system14.
The highest potency was observed for a blunt 27-mer duplex. The relative reduction in EGFP expression seen using 21-mer, 25-mer, and 27-mer duplexes is shown in Figure 2, which is adapted from Kim et al14.
At high dose (50 nM), all of the duplexes suppressed EGFP fluorescence. When dose was decreased to sub-nanomolar levels (50-200 pM), the longer duplexes continued to suppress EGFP while the 21mer was inactive. Improved potency has also been described for 29mer stem short hairpin RNAs (shRNAs) when compared with 19mer stem hairpins15, suggesting that length-related effects may reflect some aspect of RNAi biochemistry that is common to both the siRNA and shRNA pathways.
The basis for this increased potency was explored in a series of experiments. RNA duplexes 23 bases and longer were found to be substrates for Dicer and were processed to 21-22mer size duplexes, both in vitro using recombinant human Dicer and in cells. Thus, even though the 27mer showed improved potency, 21mer duplexes were still the final active compounds.
Why did the 27mer in this case show such improved potency? Shifting a 21mer siRNA by even a single base along a target mRNA sequence can produce a 10-fold or greater change in potency. Kim tested every possible 21mer that could be "diced" from the parent 27mer to see if any of these species could account for its improved performance. None of the 21mers showed sub-nanomolar potency like the 27mer, indicating that some specific benefit resulted from using a sequence of this length.
However, several of the 21mers tested did show potency that was as much as 10-fold better than the original 21mer studied at this site (effective in the low nanomolar range).
Electrospray ionization mass spectrometry (ESI-MS) was used to identify the specific products that result from Dicer digestion of this 27mer in vitro; two distinct 21mers were found to result from dicing, one of which was significantly more active than the original 21mer tested at that site in EGFP14.
Thus, the 100-fold improvement in activity seen with the 27mer resulted from the additive effects of Dicer processing producing "a better 21mer" (10-fold boost) and some other effect that is dependent upon intact 27mer being provided to the cells (another 10-fold boost).
Design Rules for Dicer-substrate RNAi Duplexes
Unfortunately, not all 27mers show this effect. In a series of studies done at Integrated DNA Technologies (IDT) to establish design rules for Dicer-substrate RNAi duplexes, examples were identified where 27mers designed around a known effective 21mer showed improved, equivalent, or worse potency than the "parent" 21mer.
One example of this kind of experiment is shown in Figure 3, where two blunt 27mers were tested at the site of a previously identified effective 21mer siRNA in the human STAT1 gene. This 21mer is a potent duplex that results in 90% suppression of STAT1 mRNA levels at 1 nM concentration and 80% suppression at 100 pM concentration.
If "27mer conversion" is done at this site by adding bases to the left side of the 21mer, the resulting duplex (27a) has low potency (worse than the 21mer), whereas if bases are added to the right side of the 21mer, the resulting duplex (27b) has high potency (slightly better than the 21mer).
Why is this large variation in potency seen between 27mers "at the same site" and can design rules be established that ensure that the most potent 27mer is obtained?
The problem is that a given 27mer can produce a variety of different 21mers after dicing and sometimes these 21mers are potent and sometimes they are not. For example, the EGFP 27mer studied in Figure 2 was diced into a "good" 21mer that was more potent than the original parent 21mer at that site.
In the STAT1 examples shown in Figure 3, duplex 27a was functionally diced into "bad" 21mer(s), even though a known "good" 21mer was present within the 27mer sequence. To effectively use 27mer Dicer-substrates in RNAi, design rules must be available to obtain duplexes that consistently result in high-activity 21mer products after Dicer processing.
To better understand the biochemistry of Dicer processing, cleavage patterns for a series of different 27mer RNA duplexes were studied using ESI-MS16. Blunt duplexes generally give rise to multiple 21mer species, and the precise cleavage patterns are unpredictable.
Through testing a number of design variants, a strategy was developed that gives predictable results. Asymmetric duplexes that have a single 2-base 3?-overhang on one end and are blunt on the other end usually have a single cleavage event 21-22 bases from the overhang. Substitution of two DNA residues at the 3?-end of the blunt side of an asymmetric duplex further helps limit heterogeneity in dicing patterns.
Currently, the combination of asymmetric single 3?-overhang with a DNA-modified blunt end is the preferred design for Dicer-substrate RNA duplexes. It is likely that the single 3?-overhang provides a binding site for the PAZ domain in Dicer, which serves to anchor the duplex with cleavage occurring 21-22 bases away.
Conversely, the opposing blunt end with DNA residues offers an unfavorable site for PAZ binding, further helping to ensure the orientation of Dicer processing16.
Using this approach, two different 27mers can be designed that are each cleaved to produce the same active 21mer siRNA product after Dicer processing. An example of the relationship between these sequences is shown for a target site in hnRNPH in Figure 4A. The "L-form" 27mer has the 3?-overhang on the sense (top) strand with added bases (which will later be removed by Dicer) on the "left" side while the "R-form" 27mer has the 3?-overhang on the antisense (bottom) strand with added bases on the "right" side.
Given that both 27mers produce the same 21mer after processing (confirmed by ESI-MS), it would be predicted that they should have similar functional potency. Interestingly, this is not the case. The duplexes shown in Figure 4A were co-transfected into HEK293 cells with a plasmid that expresses an EGFP-hnRNPH fusion protein and cells were examined for EGFP fluorescence (Figure 4B, adapted from Rose et al)16.
The "L-form" duplex performed about the same as the 21mer while the "R-form" duplex was significantly more potent. This same pattern where functional potency was biased by the direction of Dicer processing was observed in several sites in different genes.

Conclusion
In summary, the position of the 3?-overhang ("R" vs. "L") influences potency and asymmetric duplexes with the 3?-overhang on the AS strand are generally more potent than those with the 3?-overhang on the S strand. Interestingly, the opposite pattern is seen when targeting the antisense transcript. These observations could result from asymmetrical strand loading into RISC.
Strand targeting experiments were performed and differential strand utilization was observed for the two 27mer forms, with the "R" form favoring S strand and the "L" form favoring AS strand silencing. Therefore, Dicer processing confers functional polarity within the RNAi pathway and this biochemistry can be exploited through use of Dicer-substrate RNAi duplexes16.
A free web-based design to