Gene targeting of human pluripotent cells by homologous recombination is inefficient, which has impeded the use of human ES and iPS cells (hESCs and hiPSCs) in disease models. To overcome this limitation, we and others have shown that zinc finger nucleases can be used to modify the genomes of hESCs and iPSCs1–3
. ZFNs can be engineered to induce a double strand break (DSB) precisely at a predetermined position in the genome4
. This DSB can be repaired by end-joining to drive targeted gene disruption, or via the homology-directed DNA repair pathway using an ectopically provided donor plasmid as a template. Depending on the donor design, this repair reaction can be used to generate large-scale deletions, gene disruptions, DNA addition4
or single nucleotide changes5
Recent work on transcription activator-like effectors (TALEs) suggests6
suggests an alternative approach to the design of site-specific nucleases. Natural TALEs are transcription factors used by plant-pathogens to subvert host genome regulatory networks6
. The DNA binding domain of TALEs is unusual: multiple ~34 amino acid units (”TALE repeats”) are arranged in tandem, their sequence nearly identical except for two highly variable amino acids that establish the base recognition specificity for each unit7, 8
. Each individual domain determines the specificity of binding to one DNA base pair in the TALE recognition sequence and therefore four different repeat units are sufficient to specify a novel site7, 8
. Nucleases based on such engineered TALE domains can target endogenous genes in transformed human cells9, 10
. Here, we assess whether TALENs can be used to genetically engineer endogenous loci in hESCs and iPSCs.
We designed TALENs targeting PPP1R12C (the AAVS1 locus), OCT4, PITX3 genes at precisely the same positions as targeted earlier by ZFNs2
. TALEN expression constructs and corresponding homology bearing donor plasmids were electroporated into hESCs (line WIBR#3)11
and iPSCs (line C1)12
(see FigureS 1, TableS 1, 2
). Southern blot analysis was used to identify correctly targeted clones.
We targeted the PPP1R12C gene with a gene trap approach (expressing puromycin (Puro) from the endogenous gene; ; ) or with an autonomous selection cassette (Puro expressed from the PGK-promoter (, ; Figure S1, 2, 3
). Targeting efficiency was high and comparable to that with ZFNs2
with 50% of the clones being targeted in one or both alleles carrying no randomly integrated transgenes (, ; Figure S2
). Similarly, a SA-Puro-eGFPeGFP transgene was highly expressed from this locus (; Figure S3A,B
). Importantly, such targeted cells remained pluripotent based on analysis of marker expression and teratomas. Cells of all germ layers expressed eGFP indicating that TALEN as well as ZFN-mediated targeting of the PPP1R12C gene results in robust transgene expression in pluripotent as well as in differentiated cells (Figure S3C–E
Genetic engineering of hESCs and iPSCs using TALENs
Summary of targeting experiments using TALENs OCT 4
The OCT4 gene was targeted using three different donor plasmids resulting in expression of Puro and an OCT4 exon1-eGFP fusion protein under control of the endogenous OCT4 promoter. The first two donor plasmids were designed to integrate a splice acceptor-eGFP-2A-Puro cassette into the first intron of OCT4 whereas the third donor generated an in-frame fusion of exon 1 with the eGFP-2A-puromycin cassette (Figure S4
). Targeting efficiency in ESCs and iPSCs was between 70% and 100% as determined by Southern blot analysis and DNA sequencing of single-cell-derived clones (, Figure S1,4
We also targeted the first exon of the non-expressed PITX3 gene and found that ~6% of drug resistant clones carried the transgene solely at the PITX3 locus as evaluated by Southern blot analysis (, Figure S1, 5
). Notably, in one clone the transgene had integrated into both alleles of the non-expressed PITX3 gene.
To target C’ terminus in the OCT4 gene, not previously targeted by ZFNs, we generated TALENs directed against sequences flanking the stop codon of the OCT4 gene using two donor plasmids: the last OCT4 codon was either fused in frame with a GFP-PGK-Puro construct or a with an eGFP preceded by a 2A sequence (2A-GFP-PGK-Puro; , ). Following excision of the LoxP flanked PGK-Puro cassettes either a C-terminal OCT4-eGFP fusion protein or a separately translated eGFP proteins will be expressed under the control of the endogenous OCT4 gene13
. Southern blot analysis showed that 2–10% of single cell derived hESC and hiPSC clones carried the transgene cassette at the OCT4 locus (; ). Cre-mediated excision of the PGK-Puro cassette from OCT4-eGFP targeted hESCs resulted in nuclear eGFP fluorescence and from those targeted with OCT4-2A-eGFP -- in pan-cellular eGFP-fluorescence (). Fluorescence was higher in OCT4-2A-eGFP than in OCT4-eGFP targeted clones as determined by FACS analysis suggesting differential protein stabilities of the OCT4-eGFP fusion protein and eGFP (Figure S6
). A gene trap vector fusing an eGFP-2A-Puro cassette with the last OCT4 codon had a 50% targeting efficiency (; ; Figure S6
) similar to that of gene trap-vectors designed to the first intron. As expected, eGFP expression became undetectable after differentiation into fibroblast-like cells (Figure S7
) validating eGFP expression as a faithful reporter for OCT4 expression.
To illustrate the general utility of TALENs to generate such C-terminal fusion proteins, we designed TALENs to cut at the last coding exon of PITX3 and generated hESCs and iPSCs with an in-frame fusion of 2A-eGFP or eGFP. Targeting was highly efficient resulting in some clones carrying the transgene on both alleles (Figure S8
To assess the frequency of off-target modification not detected by Southern blot analysis we determined the binding specificity of the PPP1R12C TALENs using SELEX. When genotyping a panel of 19 maximal-likelihood potential off-target sites unintended cleavage was found only at low-frequency (See Figure S9, 10; Table S3, 4 and Supplemental material
for detailed description). A strategy to minimize potential off-target events is to design TALENs to function as obligatory heterodimers14
. As shown in , such heterodimeric nucleases in combination with the TALE DNA binding domain can yield high efficiency targeting of the PPP1R12C locus.
Individual TALE repeats can be joined to produce DNA binding domains capable of recognizing endogenous sequences in mammalian cells9, 10, 15
. In the present work, we have built on the recent development of an efficient TALEN architecture9
, and evaluated the utility of TALENs to drive targeted gene modifications in human ESCs and iPSCs. At all five genomic sites tested, we obtained clones carrying transgenes solely at the TALEN-specified locus at a frequency comparable to that observed with ZFNs2
. As this approach couples a simple DNA recognition code with robust activity in human pluripotent stem cells, our data suggest that TALENs are a useful tool for investigator-specified targeting and genetic modification in human pluripotent cells with efficiencies similar to those we have previously reported for ZFNs.