A pipeline for the generation of shRNA transgenic mice

Single-copy shRNA transgenic mice can be produced in about 14 weeks by subcloning a tested miR30-based shRNAmir into the cTGM construct (ColA-TRE-GFP-miR30), which places the shRNAmir in the 3′ UTR of a TRE-driven GFP, and integrating it by Flp-mediated recombinase-mediated cassette exchange downstream of the Col1a1 locus in KH2 ES cells. KH2 cells also carry Rosa26-driven rtTA-M2, so doxycycline-regulated GFP-shRNAmir expression can be validated immediately; targeted cells then yield mice by blastocyst injection or tetraploid complementation.

In the Col1a1-targeted cassette the TRE promoter gives shRNA expression equal to or better than all constitutive promoters tested, including CMV, the MSCV retroviral LTR and EF1α.

Including GFP as a spacer upstream of the shRNAmir increases gene silencing, whereas placing the shRNAmir in an intron of GFP moderately reduces shRNA potency. Changing the miRNA backbone downstream of GFP can alter GFP fluorescence, possibly through altered processing of the nascent transcript.

Only about 3% of all possible shRNA sequences trigger potent knockdown when expressed from a single genomic copy. A two-step in silico filter, DSIR guide predictions followed by seven Sensor exclusion criteria, raises the hit rate seven- to eightfold to about 25%, and testing 10–15 high-confidence candidates per gene is recommended.

To build a miR30 shRNAmir from a 21-mer DSIR guide, the first base of the 22-mer sense strand is set to C when the mRNA nucleotide 5′ of the target site is A or U, and to A when it is C or G. This creates a mismatch at the base of the stem-loop that mimics endogenous miR30 and enables more efficient processing.

Valid shRNA knockdown testing requires single genomic integration: 5–20% transduction of NIH 3T3 or other immortalized fibroblasts followed by antibiotic selection gives single-copy populations, while transduction above 30% suggests multiple integrations and overestimates potency. Transient co-transfection of shRNAmir and target cDNA overestimates silencing, and quantitative PCR does not always reflect protein reduction, so protein-level assays are used.

Single-copy miR30-based shRNAs expressed in cultured cells do not affect processing of endogenous miRNAs. A control transgenic strain such as a firefly or Renilla luciferase shRNA (Luc.1309, Ren.713) and two independent strains per target gene are still recommended to control for sequence-dependent off-target effects.

Doxycycline-regulated GFP-shRNAmir expression at Col1a1 depends on rtTA strength: Rosa26-rtTA and CMV-rtTA induce limited expression in liver, lung, kidney, pancreas and muscle, CAGs-rtTA3 induces it more widely, and the spleen shows mosaic induction with any rtTA strain tested. Dietary doxycycline or minocycline gives little to no induction in neurons, whereas ventricular injection of doxycycline or treatment of cultured brain slices induces strong GFP.

In Col1a1-targeted mice, shRNAs against the essential gene Rpa3 are silenced in vivo, especially in highly proliferative tissues such as bone marrow and intestine, by a mechanism not yet resolved that may involve suppression of the shRNA or of rtTA expression.

miR30 shRNAs are cloned by PCR amplification of 97-mer Ultramer templates followed by XhoI/EcoRI digestion of a 110-bp product, or by annealing complementary phosphorylated 110-bp oligonucleotides with XhoI/EcoRI sticky ends. 97- and 110-bp Ultramers from IDT give the lowest mutation rate, and shRNAs that fail PCR cloning show no correlation with eventual silencing.