TALE (transcription activator-like effector) proteins can be tailored to bind to any DNA sequence of choice and thus are of immense utility for genome editing and the specific delivery of transcription activators. However, to perform these functions, they need to occupy their sites in chromatin. In the present study, we have systematically assessed TALE binding to chromatin substrates and find that in vitro TALEs bind to their target site on nucleosomes at the more accessible entry/exit sites, but not at the nucleosome dyad. We show further that in vivo TALEs bind to transcriptionally repressed chromatin and that transcription increases binding by only 2-fold. These data therefore imply that TALEs are likely to bind to their target in vivo even at inactive loci.
The packaging of eukaryotic DNA into chromatin provides a formidable obstacle to protein binding and thus to the initiation of essential nuclear processes, such as transcription, recombination and repair. Even the first level of chromatin packaging, the nucleosome, restricts protein binding, but the extent of this inhibition depends on the protein in question . For example, the affinity of some transcription factors, such as heat-shock factor, is decreased >100-fold, essentially eliminating physiological binding . In contrast, the glucocorticoid receptor and Sp1 (specificity protein 1) bind to their target sites on nucleosomes, but with affinities decreased between 2–5-fold and 10–20-fold respectively compared with free DNA [3–5].
Complementary studies using restriction enzymes to probe nucleosome accessibility have shown that DNA at the nucleosome dyad is two to three orders of magnitude less accessible than at the more ‘breathable’ entry/exit points . Binding close to the nucleosome dyad can be achieved, however, by synergism between transcription factors where binding of the first factor aids subsequent factor binding . In contrast with these detailed studies of nucleosomes, relatively little is known about the influence of higher-order chromatin structure on protein accessibility, although one study suggested that the effect is relatively small .
The ability of proteins to target their sites in chromatin is particularly important for TALE (transcription activator-like effector) proteins. These proteins were identified in the plant pathogen Xanthomonas and are extremely powerful since they can be tailored to bind to any DNA sequence of choice. Moreover, provided the target is greater than 16 bp, TALEs can pinpoint unique locations within the mammalian genome. TALEs have already been used to great effect to regulate gene expression by linking them to activation or repression domains [9,10]. In addition, TALEs linked to the Fok I endonuclease are highly effective in genome editing . However, their ability to perform these functions absolutely requires TALE binding to target sites in chromatin.
Initial studies showed that TALE binding to the repressed Oct4 (octamer-binding protein 4) promoter in neuronal stem cells is enhanced by pre-treatment with either the histone deacetylase inhibitor valproic acid or the DNA methyltransferase inhibitor 5-azacytidine . This implies that repressive chromatin modifications restrict TALE binding. Consistent with this, genome editing was found to occur with variable efficiency, possibly due to different chromatin environments influencing TALE binding [13,14].
Although these studies imply that chromatin reduces TALE binding, they do not determine how it specifically mediates its inhibitory effects. One possibility is that packaging the TALE-binding site into a nucleosome prevents binding. A second possibility is that repressive chromatin marks additionally, or alternatively, play a role. In the present study, we have systematically examined the ability of TALEs to bind to chromatin templates and found that they cannot bind to nucleo-somes in vitro when their binding site is close to the dyad axis; however, they can bind at more accessible sites closer to the entry/exit points. We complement these studies with an in vivo analysis of TALE binding and show that TALEs bind strongly to an actively transcribed region. This is decreased only approximately 2-fold in cells where the gene is not transcribed. By measuring the nucleosome repeat length and the change in chromatin accessibility upon transcription, we propose that a key determinant of TALE binding is the availability of linker DNA between nucleosomes. Together, these data imply that TALEs have the potential to bind and perform effective genome editing and gene regulation at most sites in the genome.
Further details are given in the Experimental section of the Supplementary Online Data (at http://www.biochemj.org/bj/458/bj4580153add.htm).
Cloning of TALE A2
TALE A2 was cloned using Golden Gate ligation  into the pTALEN(HD) vector to generate the plasmid pTALEN-TALEA2. The DNA-binding domain was subcloned into: (i) a mammalian expression vector containing an HA (haemagglutinin) tag, and (ii) an Escherichia coli expression vector with a biotin tag.
Purification of TALE A2
The full-length TALE protein was expressed in E. coli BL21(DE3) cells and biotinylated in vivo by co-expression of the biotin ligase BirA using the pBirAcm plasmid (Avidity). It was purified using Softlink Avidin Resin (Promega).
Cloning of the TALE A2-binding site
A 122 bp fragment from the Jλ1 gene, containing the TALE-binding site, was amplified from mouse genomic DNA using the primers TALElamb_F (primer 1, 5′-CAGCCCGGGATTTTCTG-GAAAGACTTCCTATGAG-3′) and TALElamb_R (primer 2, 5′-CTGAAGCTTTGCATCTGTGATGTATGCAGCTC-3′) (Supp-lementary Table S1 at http://www.biochemj.org/bj/458/bj4580153add.htm). The PCR product was cloned into the XmaI and HindIII sites of pBluescriptSK to generate pBS-TALEA2bind.
Nucleosome reconstitution and binding assays
Two DNA fragments were excised from pBS-TALEA2bind: Fragment 1 (156 bp) using Acc65I and XmaI and Fragment 2 (146 bp) using NotI and HindIII. These were end-labelled and reconstituted with recombinant Xenopus histone octamers via salt dialysis . Nucleosomes were gel purified and their positions mapped via micrococcal nuclease digestion as described previously . Binding reactions contained 20 mM Tris/HCl (pH 8.0), 100 mM NaCl, 10% glycerol and 0.05 μg/μl poly(dI-dC).
Preparation of RNA and reverse transcription
Total RNA was isolated using TRIzol® (Invitrogen) and reverse transcribed with M-MLV (Moloney murine leukaemia virus) reverse transcriptase (Invitrogen) according to the manufacturer's instructions. cDNA was amplified with primers for the Hprt (hypoxanthine guanine phosphoribosyl transferase) and Jλ1 genes (Supplementary Table S1) via qPCR (quantitative PCR).
Transfection and ChIP analysis
Cells were transfected with pmRuby2-TALEA2-HA and an expression vector for GFP (pmaxGFP; Lonza) via electroporation using the relevant Amaxa nucleofector kit (Mouse B cell kit for 103/BCL-2 cells and Nucleofector R kit for NIH 3T3 cells). Transfection levels (25.3% and 35.1% respectively) were determined 48 h later using an LSRFortessa FACS analyser (Becton Dickinson). ChIP was performed with 5×106 cells per assay as described previously .
Bandshift analysis of TALE A2 binding
Extract was made from transfected cells  and the level of functional TALE protein was determined using a gel mobility-shift assay.
Mapping nucleosome positions in vivo
Nucleosome positioning at the Igλ locus was analysed by indirect end labelling as described previously .
Restriction enzyme accessibility
Nuclei were prepared as described previously  and incubated at 37°C with or without 50 units of enzyme for 20 min. DNA was prepared and normalized by qPCR using the IntIII primers (Supplementary Table S1, primers 11 and 12) and the level of accessibility was detected using the J11Sty primers (Supplementary Table S1, primers 13 and 14).
TALE binding to nucleosomes
To measure TALE binding to nucleosomes in vitro, a TALE protein was generated to bind to an 18 bp sequence within the Jλ1 region of the murine Igλ light-chain locus (Figure 1A). This TALE (A2) was expressed in E. coli cells and purified using a C-terminal biotin tag (Supplementary Figure S1A at http://www.biochemj.org/bj/458/bj4580153add.htm). Next, the TALE-binding site and surrounding DNA sequence, was amplified from mouse genomic DNA and cloned into pBluescript. A 156 bp DNA fragment was then excised and reconstituted into a nucleosome in vitro by salt dialysis. Notably, this fragment (Fragment 1) has the TALE-binding site located centrally and, following reconstitution, the binding site mapped close to the nucleosome dyad (Figure 1B and Supplementary Figure S1B, left-hand panel).
TALE binding in vitro
The ability of the purified TALE protein to bind to Fragment 1 was then examined, using both free DNA and nucleosome substrates. Although strong binding was observed to free DNA, no detectable binding was observed to the nucleosome (Figure 1B) or the hexasome (Supplementary Figure S2 at http://www.biochemj.org/bj/458/bj4580153add.htm). However, the TALE–DNA complex had an almost identical mobility with the nucleosome and a band of increased intensity was observed at the position of the nucleosome following the addition of TALE A2 to the nucleosome preparation. Free DNA was present in the nucleosome preparation and, to verify that the band was due to the formation of a TALE–DNA complex, streptavidin was added. The biotinylated TALE–DNA complex was super-shifted, whereas the intensity of the nucleosome and hexasome bands remain unchanged (Figure 1C). Together, these data imply that the TALE has negligible binding to nucleosomes (or hexasomes) reconstituted on to Fragment 1.
Although TALEs are unable to bind when their target sites are at the nucleosome dyad, it remained possible that they could bind to more accessible sites at the nucleosome entry/exit points. Therefore a second DNA fragment was prepared where the TALE-binding site is close to the 5′-end. Importantly, following nucleosome reconstitution and mapping, the TALE-binding site was located within the last 30–35 bp of nucleosomal DNA (Supplementary Figure S1B, right-hand panel). The addition of the TALE protein to this nucleosome resulted in a marked shift (Figure 1D), consistent with the idea that the TALE can indeed bind to nucleosome 2, albeit with an affinity approximately 3-fold lower than to free DNA (which has a Kd of 21.7 nM; Supplementary Figure S3 at http://www.biochemj.org/bj/458/bj4580153add.htm).
Together, these data imply that TALEs can bind to a nucleosome when the binding site is located close to the more accessible entry/exit points. However, if the site lies further towards the nucleosome dyad, TALE–nucleosome binding is impaired.
TALE binding in vivo
Next, we examined how chromatin packaging affects TALE binding in vivo. To this end, two cell types were used: the fibroblast cell line NIH 3T3, where Igλ is inactive, and the pro/pre-B-cell line 103/BCL-2 . In this latter cell line, Igλ is poised for activity when the cells are grown at 33°C; however, a temperature shift to 39.6°C strongly induces Igλ transcription.
First, we verified the transcription levels through the TALE-binding site in the two cell types. Negligible expression was observed in the NIH 3T3 and 103/BCL-2 cells grown at 33°C; in contrast, transcription of this region increased significantly in 103/BCL-2 cells following a temperature shift to 39.6°C (Figure 2A), consistent with previous observations .
TALE binding in vivo
To determine the ability of the TALEs to bind in vivo, we transfected the two cell types with an expression construct for TALE A2. TALE A2 has a C-terminal HA epitope and, after determining the number of transfected cells by flow cytometry, the binding of TALE A2 was determined by ChIP using an anti-HA antibody. Remarkably, in 103/BCL-2 cells following the temperature shift, binding was enriched over 100-fold (Figure 2B) compared with a non-transfected control. In contrast, in NIH 3T3 cells and 103/BCL-2 cells grown at 33°C the binding was enriched an average of 35- and 50-fold respectively. This implies that TALEs can bind to their sites in non-expressed chromatin, but that this is reduced compared with when the region is transcribed. Control experiments verified that the TALE was expressed in both cell types and that it was capable of binding to DNA (Figure 2C). Moreover, to eliminate differences due to transfection efficiency, 103/BCL-2 cells from the same transfection were used to compare binding at 33°C and 39.6°C.
Next, to investigate the molecular basis for increased TALE binding in the temperature-shifted cells, we added the transcription inhibitor α-amanitin. Consistent with the idea that transcription modulates TALE binding, α-amanitin reduced binding in temperature-shifted 103/BCL-2 cells to close to the levels observed in the non-temperature-shifted cells (Figure 2B).
Modulation of TALE binding in vivo
Transcription could increase TALE binding in a number of ways. It could alter histone modifications to increase the accessibility of the TALE-binding site or, alternatively or additionally, it could unravel the higher-order chromatin structure. To begin to differentiate between these possibilities, we first examined the level of the major active chromatin mark, histone acetylation. Remarkably, relatively little enrichment in acetylation was observed at the TALE-binding site even in the presence of ongoing transcription (Figure 3). This suggests that changes in histone acetylation do not significantly influence TALE A2 binding.
Levels of histone H3 acetylation
Therefore we next assessed how transcription alters the higher-order structure. Initially, we measured the accessibility of the TALE-binding site directly using the restriction enzyme SphI as a probe. This has a recognition site 30–40 bp from the TALE-binding site and, using an amount of enzyme that had been shown previously to give maximum digestion , nuclei were digested, followed by analysis of the level of cutting via qPCR. When transcription was ongoing, the average accessibility of the site was 28%; this is reduced by approximately 50% when Jλ1 is transcriptionally inactive in NIH 3T3 cells and 103/BCL-2 cells at 33°C (Figure 4A). This decrease correlates well with the observed reduction in TALE binding (Figure 2B).
Accessibility of the TALE-binding site in vivo
Next, we determined the expected level of accessibility. To this end, the nucleosome repeat length was measured by digesting nuclei with micrococcal nuclease and separating the resulting DNA by electrophoresis. This showed that the average distance between nucleosomes is 195 and 205 (±5 bp) in NIH 3T3 and 103/BCL-2 cells respectively (Figure 4B), implying that approximately 25% of DNA lies within the linker.
If nucleosomes are randomly distributed, then the probability of the TALE-binding site being within this linker DNA is the same as that predicted by the repeat length, i.e. 25%. If, however, nucleosomes are positioned, then the TALE-binding site might always be accessible or always inaccessible. To assess nucleosome positioning at the Jλ1 gene, nuclei were digested with micrococcal nuclease and used in an indirect end labelling experiment. A smear was observed in both NIH 3T3 and 103/BCL-2 cell digestions (Figure 4C), implying that the nucleosomes are not positioned regardless of Jλ1 transcription. Together with the results shown in Figure 4(B), these data suggest that approximately 25% of the TALE-binding sites lie in the linker DNA.
Notably, in the absence of transcription, the level of accessibility is lower (Figure 4A, 14%) than predicted by the nucleosome repeat length (25%). To test whether this is because transcription increases accessibility by disrupting higher-order chromatin structure, we added α-amanitin for 40 min to 103/BCL-2 cells grown at 39°C. The addition of α-amanitin reduced accessibility to close to that seen in cells where this locus was not being transcribed (Figure 4A). This suggests that a key function of transcription is to increase the accessibility of linker and nucleosomal DNA, which then facilitates TALE binding. Accessibility does not fall completely to repressed levels upon inhibition of transcription. This might be because changes in chromatin modifications make a small contribution to the accessibility. Alternatively, it might be due to incomplete re-folding of the higher-order chromatin structure during the relatively short α-amanitin treatment. It is also notable that the decrease in TALE binding upon α-amanitin treatment was slightly greater than the decrease in SphI accessibility. This might be because the TALE-binding site (18 bp) is longer than the SphI site (6 bp) and is thus more likely to be occluded.
TALE proteins are of immense utility in genome editing and gene regulation, but their ability to carry out these functions relies on their effective binding to target sites in chromatin. In the present study, we show that although the nucleosome partially inhibits binding, TALE binding is possible in vivo even at a transcriptionally repressed locus. Our data therefore contrast with early ideas that TALEs should be targeted to hypersensitive sites. Instead, we find that there is a relatively small difference in binding to active and repressed loci, implying that TALEs should function at inactive loci. Consistent with this, TALEs were shown recently to function at such loci .
We find that TALE A2 binds at the more accessible nucleosome entry/exit sites, albeit with approximately a 3-fold lower affinity than to free DNA. This, together with the recent reports that TALE proteins synergize to activate genes [23,24], begins to explain how TALEs can function on repressed chromatin substrates . The binding of a TALE to the linker DNA or to the more accessible regions of the nucleosome could increase binding of a second TALE or another transcription factor closer to the nucleosome dyad, a phenomenon first described for GAL4 .
Transcription increases the accessibility of the TALE-binding site and this correlates well with the disruption of higher-order chromatin structure and the release of linker DNA. Transcription is also known to dislodge histone H1, thereby removing a protein that competes with transcription factors for nucleosome binding . Recently, a third way in which transcription increases accessibility was described: transcription transiently evicts an histone H2A–H2B dimer from each nucleosome, which temporarily releases 30–40 bp of DNA from the nucleosome, providing a window of opportunity for protein binding . It is probable that all three of these processes co-operate to aid TALE binding to transcriptionally active chromatin. However, the strong correlation between the level of accessibility and amount of linker DNA suggests that the transcription-dependent release of linker DNA contributes significantly to increased protein binding.
The lack of TALE binding at the nucleosome dyad means that the target site for any given TALE is probably occluded in some cells in the population. Consistent with the idea that TALEs do not bind in all cells, genome editing was found to vary in efficiency between 2% and 100% . Moreover, some TALE proteins were found not to function at all . Although this could be due to weak interaction with the particular DNA sequence, it could also be due to the TALE-binding site falling close to the dyad of a positioned nucleosome. Although nucleosome occlusion will prevent the binding of some TALEs, the results of the present study suggest that, in the absence of such occlusion, TALEs can bind to repressed chromatin at least in some cells in the population. Together with the recent report that TALEs synergize to perform their function, our data reinforce the idea that using a series of TALEs with adjacent binding sites [23,24] is likely to increase the chances of TALE proteins being effective in vivo.
James Scott, Adam Kupinski, Christopher Kirkham and Joan Boyes designed and performed the experiments, analysed the data and wrote the paper. Roman Tuma analysed the data and wrote the paper.
We thank Professor Anthony Turner for helpful comments on the paper.
This work was funded by the Yorkshire Cancer Research [grant number LPP052] and a Lady Tata Memorial Trust studentship (to J.N.F.S.). C.M.K. is a recipient of an Engineering and Physical Sciences Research Council studentship.
These authors contributed equally to this work.