siRNA and shRNA Design Guidelines – Review

June/July 2003

The design of siRNAs and short hairpin siRNAs (shRNAs) remains an empirical process since the molecular mechanisms underlying RNAi are not yet sufficiently understood to allow for the rational design of siRNAs. However, based on the research from various laboratories including our own, InvivoGen has been able to develop siRNA Wizard, an online tool accessible from our homepage, that will help you find the best siRNA sequences on your target gene. The siRNA Wizard tool will also design the pair of oligonucleotides needed to generate shRNAs using InvivoGen’s psiRNA plasmids. Below is the list of general rules, used by the siRNA Wizard, that have been revised to better suit the design of shRNAs.

The current guidelines recommend avoiding the first 50-100 nt located downstream of the Start codon and the 100 nt located upstream of the Stop codon, as well as 5′ and 3’UTRs. These regions contain binding sequences for regulatory proteins that may affect the accessibility of the RNA target sequence to the RISC complex. However, we and others have successfully silenced the expression of several genes by targeting the 5′ or 3’UTRs [1, 2, 3, 4]. Therefore, 5′ and 3’UTRs should also be considered when selecting a region on your target gene.

The first nucleotide of the siRNA sequence can either be an A or a G. Although we recommend choosing an A (see Selection criteria for Standard search), a G can also be used since in several examples siRNAs starting with a G and expressed from the human H1 promoter have worked [4, 5].
The pyrimidines C and T should be avoided because expression of RNAs from RNA polymerase III promoters is only efficient when the first transcribed nucleotide is a purine. In cases where your siRNA sequence starts with a C or T, we recommend adding an A as the first nucleotide.
This addition will not affect the activity of your siRNA since it will generate a T at the end of the antisense siRNA strand that will be included in the termination signal maintaining its complementarity with the target sequence. This point is important since according to current knowledge recognition of the specific gene target is achieved by the antisense siRNA strand.
It is usually recommended to choose sequences with low GC content (between 30-55%). There are also many examples of active siRNAs with high GC content [6, 7, 8].

siRNA-mediated RNAi is based on using dsRNA < 30 nt to avoid nonspecific silencing. According to Hannon et al. siRNA of 25-29 nt are generally more effective than shorter ones. However, we and others found that hairpin siRNAs with duplex length of 19-21 nt are as effective as longer hairpin siRNAs [6, 9, 10].

Several teams including ours have tested a variety of sequences for the loop between the two complementary regions of a shRNA, ranging from 3 to 9 nt in length. Similar effectiveness have been obtained for loops of 5, 7 or 9 nt. We use a 7 nt loop sequence (TCAAGAG) for the psiRNA vectors.

Despite the fact that this set of rules is still not well defined, sequences generated by the siRNA Wizard will likely work better than randomly selected sequences. However, because some candidate siRNAs are more active than others, it is recommended to vary the selection criteria and to compare a panel of three siRNAs to find the most efficient.

1- Yokota T. et al., 2003. EMBO reports AOP.
2- Yu JY. et al., 2002. PNAS 99(9):6047-6052
3- Rubinson DA. et al., 2003. Nature Genetics 33:401-406
4- Mcmanus MT. et al., 2002. RNA 8:842-850
5- Tiscornia G. et al., 2003. PNAS 100(4):1844-1848
6- Kim MH. et al., 2002. BBRC 296:1372-1377
7- Hasuwa H. et al., 2002. FEBS Letters 532:227-230
8- Bertrand JR. et al., 2002. BBRC 296:1000-1004
9- Yu JY. et al., 2003. Molecular Therapy 7(2):228-236
10- Song E. et al., 2003. Nature Medicine 9(3): 347-351

Reference from


RNAi in good Cell growing

Cells growing in log phase are the most suitable one for studying gene silencing because once the cells come to be in confluence, cell cycle will be blocked, senescence will start and a group of inhibins will be secreted. All these will affect the physiological status of the cells and in turn affect gene silencing.

I am assuming that the protein of interest is not density- or cell-cycle-dependent, otherwise the experimental needs are dictated by the exact question you are addressing. A “generic” protocol very much depends on your cells’ doubling time and the time-frame of the physiological effect you expect, starting with the stability of your protein of interest. A fast-growing cell line (say, doubling time of 16-30 hrs) is best transfected at ~60-70% and analysed 24 hrs post transfection. BUT if your protein is stable (e.g. structural component), 24 hours may not be sufficient for its degradation, so starting with a sparser culture (~40%) and waiting 48 hours might be better. GFP protein, for example has a half life of several days! In a slow-cycling cell line, 80-90% is a good density.
The most robust approach is to try the knock down in parallel in dishes with different density first, choose a set of conditions, and stick to them.

Its good to avoid complete confluency at the time you harvest your cells, best would be 70-80% at harvesting time. So,dependent on the time you have to give the siRNA to work (could be 24,48,72 or more h) you will have to seed the amount of cells that ends up with 70-80% of cells at the harvesting time. Depends on the division time of your cell line. Design a test experiment, (differnt KD times, cell# etc.. ) and give it a try.

Confluent monolayer is not a good thing when working with eukaryotic cells. If my cells reach confluence I am forced to eliminate them. So in theory, the cells on plastic should never reach confluence, especially if you are going to freeze them and use as stocks.

A more convenient and efficient way of transfecting cells is reverse-transfection. You detach the cells from plastic and transfect them in solution and then add medium and aliquot to the tissue culture plates. This is a more robust and high-throughput method compared to just adding your transfection mix to cells growing on plastic and embedded in extracellular matrix (I am not talking here about suspension cell cultures as these are a special case).

For siRNA knock-down, 24 hours is almost never enough to see 1-2 log difference. You need 48 hours, at least, to get a big difference.

In principle, cell confluency per se brings up stress to the cell population and is able to activate signaling pathways (like NF-kB) and, if the process is persistent, can certainly establish some selective pressure for the cells. Depending on the target gene you are focusing on, your gene silencing strategy might be affected when cells are under this kind of stress. Besides, compensatory effects frequently happen when stably shRNA expressing cells are cultured for long time (better to freeze stocks and culture cells for only 2-3 passages)

You no need to seed cells at 80 – 90% confluency, there is difference between knock-down and over-expression.

For silencing, transfection at 30 – 40 % is more efficient than higher confluency, We usually do knockdown for 72 hrs (for smaller proteins) and 90 hrs (for larger proteins).

Again, it depends on how efficient the siRNA is targeting the mRNA. In my experience, once the siRNA is taken up by the cells (usually within 8hrs post-trans) it is very stable inside.

PS: I did knock-down for a protein, 72 hours post-trans, I transfected its deletion mutant (only C-Terminus region) to see the effect (after 36 hours). But, the siRNA was still active and it silenced my deletion mutant also.

Good idea from