Construction of expression-ready vectors and a TALE repeat plasmid library

Plasmids containing the monomeric TALE repeat sequence of RVD variant: HD, NN, NG and NI were obtained from Zhang et al. (19). Monomeric repeats of each position were amplified with position-specific primers carrying corresponding linkers and BsaI cutting sites by polymerase chain reaction (PCR) (Agilent) (Supplementary Table S1 and S2). Purified PCR products were then digested by BsaI (New England BioLabs Inc?.; NEB) at 37°C for 3 h and cleaned up by PCR purification spin column (QIAGEN) before ligation by T7 ligase (Enzymatics) at 21°C for 3 h. Bands (300 bp) were then cut out and purified after gel electrophoresis. Secondary PCR with position-specific primers, TAL-F/R-assem were performed. They were then cloned into kanamycin resistant vectors by pCR-BluntII-TOPO^® PCR cloning kits (Invitrogen) and transformed by TOP10 One Shot^® chemically competent Escherichia coli (Invitrogen). Multiple colonies were randomly picked into 96-well plates for each position and sequenced. Sequencing traces were genotyped manually and correct clones were picked for expansion and archiving to create the master log. Missing triplet combinations not covered by the random assembly approach were assembled by targeted manual synthesis. Activator vectors were constructed based on the TALE architecture by Zhang et al. (19) and cloned into a piggyBac transposable backbone carrying a tetracycline-responsive element. The BsaI restriction site in the KRAB domain was removed and synthesized by GeneArt^® (Life Technologies). The VP64 transactivation domain was swapped with a KRAB repressor domain for the repressor construct. The Enhanced Green Fluorescent Protein (EGFP) reporter was also changed to a mCherry reporter. To assemble the TALE activator, bacterial clones carrying the triplet permutation at specific position and expression backbone were expanded in low salt overnight kanamycin (triplet)/ ampicillin (backbone) LB media followed by Miniprep spin column purification (QIAGEN). Eluted DNAs (50 µl) were digested for 3 h by BsaI (at 37°C, backbone)/ BsmBI (at 55°C, triplet). Digested products were then analysed by gel electrophoresis and expected bands (triplet: ∼300 bp and backbone: 7 kb) were cut out and purified with spin column (QIAGEN). Purified triplets and backbone were then ligated by T7 ligase for 1 h at 23°C. Ligated products were chemically transformed into OneShot TOP10 competent cells (Invitrogen). Clones were picked the day after and genotyped by XhoI-MluI digestion (1.7 and 2.2 kb band) and MfeI (multiple 102 bp band) digestion.

Activation of the Oct4 and Nanog loci by TALE and the dCas9 activators

We began by comparing the ability of activating the Oct4 locus through enhancer activation by TALE activator (TALE-A) and CRISPR/dCas9 activators (dCas9-As) in an Oct4-GFP reporter MEFs system. The transgenic Oct4-GFP MEFs contained the 18 kb fragment upstream of the Oct4 TSS and were previously shown to faithfully report the transcription status of the endogenous Oct4 locus (37). We previously showed that TALE-As targeting the distal enhancer rapidly reactivates transcription at the Oct4 locus and is able to replace exogenous Oct4 in reprogramming MEFs to iPS cells (6). In this study, we modified the dCas9 protein (31) and made three versions of activators in which the VP64 activation domain was fused to either the N-terminal, the C-terminal or both termini of the protein. They were named as PL-A1, PL-A2 and PL-A3, respectively (Supplementary Figure S1A). We also acquired three published dCas9 activator constructs (32,33,38). They are different in the deactivating mutations of the Cas9 nuclease, in the number of activation domain repeat units and in sequential arrangement of the domains. We termed these dCas9 activators as JW-A (32), GC-A (38) and RJ-A (33), according to their origins (Figure 1A). All the dCas9 activators were cloned into a piggyBac transposable vector. We also linked a BFP to the dCas9 protein via the T2A peptide, and mCherry to the gRNA expression vector in order to track the expression of the system in cells (Figure 1A). The DNA targeting sequences for TALE-A (three inside and one outside the Oct4 distal enhancer) were previously described (6). To ensure comparability of TALE-As and gRNA/dCas9-As targeting sites, we constructed a hemagglutinin-tagged dCas9 vector and designed multiple gRNAs to target sequences at close proximity to the TALE-A targeting sites in the distal Oct4 enhancer and compared their binding affinity by Chip-quantitative PCR (ChIP-qPCR). We then selected four gRNA targeting sites with high binding affinity for further analysis (Figure 1B and Supplementary Figure S1B and S1C). Although the two gRNAs (g3–2 and g3–3) targeting sites overlapped with more than 12 bp with the previously validated TALE-3 targeting region, they failed to induce dCas9 binding (Supplementary Figure S1C and Supplementary Table S4).

We first investigated activation of the Oct4 enhancer by luciferase reporter assay 48 h after transfection of TALE-A and dCas9-A/gRNA in MEFs. These luciferase constructs contain the 2.4 kb region covering all three upstream regulatory elements of the Oct4 locus (6). Out of the three dCas9-As we constructed de novo, PL-A3, which has a VP64 domain at both N- and C- termini of dCas9 protein, produced the highest luciferase activities (Supplementary Figure S1D), we thus used PL-A3 in all subsequent CRISPR/dCas9 experiments. The luciferase activities of dCas9-As/gRNAs were comparable to that of TALE-A targeting the same region (Figure 1C). We next examined how dCas9-As affected expression of the endogenous Oct4 locus in MEFs by RT-qPCR. We observed the similar pattern of activation as in luciferase assay, but none of the dCas9-As activated Oct4 mRNA expression to the levels by the TALE-A (Figure 1D).

We also designed gRNA constructs and TALE-As to target the Nanog 5 kb upstream enhancer (Supplementary Figure 1E). Similar to the Oct4 locus, dCas9-As could bind their targeted regions and effectively activate the luciferase reporter carrying the 5 kb upstream enhancer (Figure 1E and Supplementary Figure S1F), but again failed to achieve the same level of mRNA expression from the Nanog locus as the TALE-A (Figure 1F). Interestingly, we did not observe significant additive effect on Oct4 and Nanog enhancer luciferase activation when we co-transfected multiple gRNAs with dCas9-A in MEFs (Data not shown).

dCas9 activators in reprogramming somatic cells to iPS cells

We previously detected GFP^Bright cells in Oct4-GFP transgene reporter MEFs after 5-day expression of ectopic reprogramming factors Myc, Klf4 and Sox2 (CKS) plus TALE-As targeting the Oct4 distal enhancer (6). We thus evaluated the ability of dCas9-A/gRNAs targeting the same enhancer to reprogramme the same MEFs to iPS cells. We constructed a vector that co-expresses the rtTA and gRNA cassette so that the number of vectors transfected in both the TALE-As and dCas9-A/gRNAs experiments is the same (Supplementary Figure S2A). Surprisingly, none of the dCas9-As/gRNAs produced GFP⁺ cells before day 8 (Figure 2A). Flow cytometry analysis of cells expressing TALE-As (mCherry⁺) showed that all three TALE-As targeting inside the distal enhancer (Site 2–4) produced bright GFP⁺ cells, in particular for the TALE-A targeting at the Site 3 where up to 50% of mCherry⁺ cells were GFP positive. In contrast, in cells expressing both dCas9-As/gRNAs, only GFP^dim cells were detected and at substantially lower percentages (Figure 2B). The discrepancy of GFP⁺ cells between dCas9-As and TALE-As was also reflected in endogenous transcription activity. RT-qPCR analysis of endogenous Oct4 mRNA levels in the GFP⁺ cells confirmed the less effective activation of the locus by the dCas9-As at day 8 of induction (Figure 2C). After 3–4 weeks of induction, dCas9-As produced much lower numbers of alkaline phosphatase-positive (AP⁺) colonies (Figure 2D). Nevertheless, the iPSC colonies generated by the dCas9-As expressed similar levels of key pluripotency genes (Supplementary Figure S2B). To further test the reprogramming potential of the two systems in a homogenous experimental setting, we linked a GFP cassette to the CKS reading frame by 2A peptide (termed GCKS) in a piggyBac vector and transfected wild-type MEFs with GCKS plus either TALE-As or dCas9-As/gRNAs (Figure 2E). Two days after transfection, cells of GFP⁺/mCherry⁺ (GCKS plus TALE-As), or of GFP⁺/mCherry⁺/BFP⁺ (GCKS plus dCas9-As/gRNAs) were sorted out for RT-qPCR analysis and for subsequent reprogramming induction. We found that the Oct4 mRNA levels in TALE-As transfected MEFs were 3–4 folds higher than in cells expressing dCas9-As/gRNAs (Figure 2F), whereas the expression levels of GCKS were similar (Supplementary Figure S2C). Consistently, TALE-As targeting sites 2–4 produced 4–10 folds more colonies than their dCas9/gRNAs counterparts after 3–4 weeks continuous induction (Figure 2G).

At the Nanog locus, dCas9-As/gRNAs targeting the 5 kb upstream enhancer activated luciferase reporter and even increased the mRNA expression, yet they failed to produce any iPSC colonies from Oct4-GFP reporter EpiSCs (3) (Figure 2H). In contrast, expressing the TALE-A targeting at the Site 2, which is inside the enhancer, consistently produced iPSC colonies (Figure 2H). To confirm the inability of dCas9-As/gRNAs to reprogramme EpiSCs, we designed and tested additional four gRNAs with target sequences across the enhancer region (Supplementary Figure S1E and Supplementary Table S4). Again, these dCas9-As/gRNAs bound their target regions and induced substantial luciferase reporter activities (Supplementary Figure S1F and S2D), but no iPSC colonies were produced (data not shown).

We investigated the possibility of inadvertent off-target gene activation by dCas9-A/gRNAs, which may impede the reprogramming process. We computationally identified 58 genes that contained potential off-target binding sites 3 kb up/down stream of their TSSs (Supplementary Table S5) for all the gRNAs tested in this study. Among which, Snai1, a key mesenchymal gene, which may have blocked the mesenchymal-epithelial transition during reprogramming, was present. However, we found no induction of Snai1 mRNA expression by either Oct4 or Nanog dCas9-As/gRNAs (data not shown). Therefore, the failure of dCas9-As/gRNAs in reprogramming was unlikely to be caused by off-target gene activation.

Less efficient epigenetic changes caused by dCas9 activators at the enhancers

We next elected to investigate the epigenetic changes at the Oct4 distal enhancer induced by either the dCas9-A or TALE-A. It has been reported that the VP64 transactivation domain recruits activating complex component p300 and facilitates histone acetylation (39). To this end, we performed secondary reprogramming experiment using cells differentiated from iPSC clones obtained from doxycycline-inducible (Dox) CKS and dCas9-As or TALE-As (Figure 3A) to exclude the variation of transfection in primary reprogramming. These iPSCs cells contained all the reprogramming factors integrated in the genome which could be reactivated by addition of doxycycline after retinoic acid-induced differentiation. We determined the enrichment of p300 at the Site 3 after three days of Dox induction, and found that TALE-A induced significantly higher levels of p300 at the Oct4 distal enhancer than any of the dCas9-As, detected by ChIP-qPCR (Figure 3B). Furthermore, higher levels of active histone mark H3K27Ac were induced by the TALE-A at the enhancer (Figure 3C). Similar results of p300 and H3K27Ac induction by TALE-As and dCas9-As were also found at the Site 4 (Supplementary Figure S3A and B). In agreement with the primary reprogramming experiment (Figure 2D and G), reactivation of the endogenous Oct4 locus by dCas9-As in these differentiated cells was also significantly less efficient than the TALE-A (Figure 3D). Finally, TALE-A also outperformed all four dCas9-As (Figure 3E) in secondary reprogramming experiment.

We next examined the enrichment of H3K27Ac and p300 at the Nanog 5 kb enhancer region in EpiSC reprogramming, BFP⁺/mCherry⁺ and mCherry⁺ EpiSCs were FACS sorted after transfection of dCas9-As/gRNAs and TALE-As respectively. Similar to the Oct4 distal enhancer, the TALE-A caused higher levels of p300 and H3K27Ac than dCas9-As at this enhancer. Nevertheless, even though dCas9-As failed to reprogramme EpiSCs, they were still able to induce substantial epigenetic changes at the enhancer (Figure 3F and G).

Effective gene repression by dCas9 repressors

To test dCas9 as repressors, we added the repressive KRAB domain to the C-terminal of the GC-dCas9 (38) and both termini of PL-dCas9 to make GC-R and PL-R. We also included a published dCas9 repressor, JW-R (32) for comparison. For all three repressors, a BFP cassette was co-expressed either through 2A peptide or direct fusion (Figure 4A). The same gRNAs that are specific to the Oct4 distal enhancer or the Nanog 5 kb enhancer described above were used to guide the dCas9-Rs to their respective target sites. To compare the repression function of the two systems, the dCas9-R/gRNAs and TALE-Rs were expressed separately in either Oct4-GFP or Nanog-GFP reporter mouse ESCs (40) so that repression of Oct4 and Nanog could be conveniently tracked by GFP intensity and the number of GFP⁺ cells.

We first examined the Oct4 locus 3 days after repressor expression. For the repressors targeting at the Site 1, which is upstream of the Oct4 distal enhancer, neither dCas9-Rs nor TALE-R substantially affected GFP intensity or the percentage of GFP⁺ cells, indicating that the repressive function is depending on the genomic context of the targeting sequence. By contrast, all the dCas9-Rs/gRNAs and the TALE-Rs targeting at the Site 2, 3 and 4, which are within the enhancer region, efficiently suppressed GFP expression at comparable levels (Figure 4B). To quantitate the repression, cells expressing the dCas9-R/gRNAs and the TALE-Rs targeting at the Site 1 and 3 were harvested and subjected to qRT-PCR analysis. The endogenous Oct4 mRNA was reduced to comparable levels by the dCas9-Rs/gRNAs and TALE-Rs (Figure 4C), indicating that both systems performed similarly at this enhancer.

We next tested whether the repression of the Oct4 locus by either the TALE-R or the dCas9-R would affect MEF reprogramming. Lrh1 was previously shown to replace exogenous Oct4 in reprogramming MEFs to iPSCs by direct binding and activating the Oct4 locus (41). We thus co-expressed TRE-Lrh1 with CKS (CKSL as the control) in MEFs, which eventually produced 79 AP⁺ colonies at day 25. Once the dCas9-R/gRNA or the TALE-R was co-expressed with CSKL, only fewer than 40 AP⁺ colonies were obtained (Figure 4D). Importantly, none of these colonies expressed the dCas9-R/gRNA or the TALE-R. To exclude the effects of transfection efficiency, we repeated this reprogramming experiment with GCKS and TRE-Lrh1 combined with PL-R/gRNAs or TALE-Rs targeting at the Site 2, 3 and 4. The transfected cells were sorted out (GFP⁺/mCherry⁺ for GCKSL plus TALE-As, or GFP⁺/mCherry⁺/BFP⁺ for GCKSL plus PL-R/gRNAs) 2 days after transfection. RT-qPCR analysis showed similar induction level of LRH1 and GCKS in different transfection (Supplemental Figure S4A and S4B). PL-R targeting at all three sites inside the Oct4 distal enhancer suppressed the CKSL-induced reprogramming as efficiently as TALE-R targeting at the same sites (Figure 4E). Therefore, suppression of the Oct4 enhancer and thus of reactivation of the locus by either dCas9-R/gRNA or the TALE-R effectively inhibited MFF reprogramming.

Besides the Oct4 locus, we also examined the two systems in suppressing the Nanog locus via the 5 kb enhancer. We expressed both repression systems in Nanog-GFP ESCs. Effective repression of the locus was evident when GFP⁺ cells were quantitated 3 days after expression of the repressor systems. Specifically, in cells expressing GC-R and PL-R targeting at the Nanog enhancer Site 2, 77 and 85% cells became GFP⁻/^dim, whereas in cells expressing the TALE-R, only 62% of them became GFP⁻/^dim (Figure 4F). The effective repression of the Nanog locus was also confirmed at the mRNA levels (Figure 4G).

We also tested the biological consequence of suppressing the Nanog locus in EpiSC reprogramming. Overexpression of KLF4 efficiently reprogrammes EpiSCs to iPSCs (3). We found that once the Nanog locus was repressed by either the dCas9-R/gRNA or the TALE-R, only very few iPSC colonies could be obtained from KLF4-mediated EpiSC reprogramming, demonstrating effective repression of the locus and the essential role of Nanog in the reacquisition of naïve pluripotency (Figure 4H).

The dCas9/gRNA complex interferes with binding of transcription factors at enhancers

Comparing to the TALE proteins, the CRISPR/dCas9 system was less effective in activation but worked equally well, if not more effective, in repressing a locus. We showed in this study that one possible mechanism is the less efficient ability of dCas9-As to recruit epigenetic modifiers and co-regulator complexes. The CRISPR/dCas9 system requires a gRNA to form a complex with dCas9 at the target sites by guide RNA/targeting DNA paring, which requires local helix unwinding. This may interfere with enhancer function. In particular, it may have detrimental impact on neighbouring transcription factor binding. To address this possibility, we reviewed the ChIP-seq information of several pluripotency transcription factors, including KLF4, OCT4, NANOG and SOX2 at the Nanog 5kb upstream enhancer region(42) and found that the Site 2 (targeted by both TALE and dCas9) was surrounded by the predicted KLF4 and NANOG binding sites (Supplementary Figure S5). We then investigated the effect of expressing the TALE proteins or dCas9/gRNA (both without either the VP64 or KRAB domains) on the binding of NANOG and KLF4 at the Nanog enhancer in mouse ESCs. Expression of the TALE protein or the dCas9/gRNA that targets at the Site 1, which is outside the enhancer, did not significantly change KLF4 or NANOG binding at the enhancer detected by ChIP-qPCR (Figure 5A). On the other hand, the dCas9/gRNA targeting at the Site 2, which is within the enhancer, significantly interfered KLF4 and NANOG binding (Figure 5A). Indeed, expression of this gRNA with dCas9 (again without either VP64 or KRAB) in Nanog-GFP reporter ES cells substantially increased GFP^dim cell populations, from 33.8 to 48.3% (Figure 5B). Binding of this regulatory-domain-free dCas9/gRNA complex at the Nanog 5 kb enhancer also decreased luciferase reporter activities (Figure 5C). These functional consequences were not observed in ES cells expressing either the TALE protein targeting at the Site 2 or the dCas9/gRNA targeting at sequence outside the enhancer. These results showed that the dCas9/gRNA complex acted as a steric hindrance to native transcription factor binding at enhancers.

Construction of a TALE repeat plasmid library and expression-ready vectors

In view of the distinct advantages of TALE in enhancer activation, we sought to develop a one-step process to facilitate TALE activators and repressors construction. We simplified the Golden-Gate TALE repeat assembly cloning steps into a single-step ligation reaction by constructing a plasmid library of TALE triplet repeat. Residue variable di-residues: HD, NI, NG and NN were chosen to target nucleotide cytosine, adenosine, thymine and guanine for library construction. A final 18-mer-long TALE repeats assembly design was chosen for this library and it is constructed by ligating six triplet-inserts, each carrying position-specific linkers, to an expression-ready activator/repressor backbone. Our previous report showed that the length of the TALE repeats affects the modulation efficiency and an array of 18-mer appears to be an optimal balance of targeting specificity and assembly complexity (6). In total, 384 (i.e. 6 × 64) unique triplet plasmids were cloned to cover all the possible permutations of an 18-mer-long TALE repeat by randomly ligating a pool of RVD monomers amplified with primers carrying position-specific linkers. Multiple colonies were then picked and sequenced to confirm all the possible permutations. We evaluated the assembly efficiency by constructing 10 additional TALEs, and the average success rate is 73.8% (50–100%). Luciferase reporter activation and MEF reprogramming experiment validated the TALEs constructed by this platform (Supplementary Figure S6A–D).