Divalent cations promote TALE DNA-binding specificity

Abstract

Recent advances in gene editing have been enabled by programmable nucleases such as transcription activator-like effector nucleases (TALENs) and CRISPR–Cas9. However, several open questions remain regarding the molecular machinery in these systems, including fundamental search and binding behavior as well as role of off-target binding and specificity. In order to achieve efficient and specific cleavage at target sites, a high degree of target site discrimination must be demonstrated for gene editing applications. In this work, we studied the binding affinity and specificity for a series of TALE proteins under a variety of solution conditions using in vitro fluorescence methods and molecular dynamics (MD) simulations. Remarkably, we identified that TALEs demonstrate high sequence specificity only upon addition of small amounts of certain divalent cations (Mg²⁺, Ca²⁺). However, under purely monovalent salt conditions (K⁺, Na⁺), TALEs bind to specific and non-specific DNA with nearly equal affinity. Divalent cations preferentially bind to DNA over monovalent cations, which attenuates non-specific interactions between TALEs and DNA and further stabilizes specific interactions. Overall, these results uncover new mechanistic insights into the binding action of TALEs and further provide potential avenues for engineering and application of TALE- or TALEN-based systems for genome editing and regulation.

INTRODUCTION

Transcription activator-like effector (TALE) proteins are a unique class of DNA-binding proteins that have found broad use in genomic engineering (1). TALEs are naturally secreted by Xanthomonas bacteria as a means of facilitating infection of plant cells (2). In recent years, TALEs have garnered a keen interest due to their modular, programmable DNA-binding domain that allows for targeted binding to DNA (1,3,4). Natural TALE proteins consist of an N-terminal region (NTR) that facilitates DNA binding (5,6), a central repeat domain (CRD) that confers sequence specificity (7,8), and a C-terminal region (CTR) that contains nuclear localization signals (9,10) (Figure 1A, B). The CRD is comprised of a series of tandem repeats of conserved modules consisting of 33–35 residues. Each repeat in the CRD contains two residues located at positions 12 and 13, known as the repeat variable diresidue (RVD), which specify the nucleotide binding partner of the repeat (1). In 2009, the binding code of RVDs was elucidated, thereby enabling the facile engineering of TALEs for sequence-specific binding of DNA (7,8). For direct gene editing applications, TALE proteins can be fused to a nuclease domain such as FokI, thereby generating a powerful molecular tool known as the TALEN system (7).

Figure 1.

Crystal structure of TALE protein bound to target DNA and design of fluorescence anisotropy experiments. (A) General schematic of the 21.5 repeat TALE protein construct used in our work and an expanded view of one repeat sequence in the central repeat domain (CRD). The repeat variable diresidues in all repeats are shown in the boxes. The lengths of the N-terminal, the central and the C-terminal domains are specified. (B) TALE PthXo1 bound to a 36 bp target DNA sequence (PDB ID: 3UGM (23)), with the NTR shaded in orange and the CRD shaded in green. (C) DNA templates used in fluorescence anisotropy experiments. The binding site for the 21.5 repeat TALE construct used in this work is shown in red and the non-specific elements are shown in black. The green star represents the fluorescein dye molecule conjugated to the 5’ terminus of DNA.

Despite the widespread use in a variety of gene editing and regulation applications (1,4,11), a complete understanding of TALE DNA-binding specificity has not yet been achieved. In prior work, there are a few studies focused on comparing the binding affinity of TALEs to specific versus non-specific DNA (5,12–16). In some cases, these studies have provided conflicting results, with some showing two or more orders of magnitude difference in specific versus non-specific binding affinity (12–15), while others show less than an order of magnitude difference (5,16). Only in those studies showing the greatest difference between specific and non-specific binding does the biological activity and gene engineering utility of TALEs appear plausible. Similar to TALEs, there have been reports of off-target binding effects for the other cutting-edge gene editing tools, including the CRISPR–Cas9 system (17–20) and zinc finger nucleases (21,22). For precise gene editing, the optimal tool must introduce specific and intended modifications exclusively at the target site while leaving the rest of genome unmodified. From this perspective, it is essential to understand the factors that influence TALE binding affinity and specificity.

Prior experimental work has probed a few key factors contributing to TALE DNA-binding specificity. Recent work by Gao et al. showed that the NTR is critical for TALE binding, and that the NTR is capable of binding to DNA on its own in a non-specific manner (5). Schreiber et al. corroborated these results by demonstrating the decline in TALE activity as the NTR was truncated to shorter and shorter lengths (6). Several studies have shown that TALEs with longer CRDs have higher overall DNA-binding affinity and specificity in a genomic context (24–26). Meckler et al. highlighted that the choice of RVDs in the CRD can affect TALE specific binding affinity by several orders of magnitude (27). Recent single molecule fluorescence microscopy studies of TALE binding and non-specific search dynamics by Cuculis et al. confirmed both the necessity of the NTR for nucleation of TALE binding events, while also demonstrating the non-specific one-dimensional search abilities of NTR-only truncation mutants (28,29). Furthermore, it was found that TALEs utilize a rotationally decoupled mechanism for non-specific search along the DNA double helix, wherein the TALE CRD repeats engage DNA in a fully wrapped conformation during search, yet remain loosely associated with the DNA backbone during the process (29,30). Rinaldi et al. investigated the effect of increasing number of repeats on TALE affinity for target and non-target DNA in monovalent salt solutions, which showed that TALE specificity increases then diminishes with increasing TALE length (15). Geiger-Schuller et al. reported that TALEs undergo partial unfolding prior to DNA binding and increasing number of repeats slows down conformational change rate thereby affecting binding kinetics (31). Finally, the requirement of a thymine at the 5’ end of the binding site was well characterized and found to significantly impact the binding of TALEs to target sequences (2,3,7,23).

A recent computational study on the energetics of TALE target binding suggested that TALE specificity arises mostly from negative discrimination by the RVDs (i.e. the ‘least bad’ binding site) (32). In this study, the major contributors to TALE binding to target sites are determined to be non-specific interactions between the lysine and glutamine residues at positions 16 and 17 within each repeat and the DNA backbone (32). The second largest contributor to binding energy is the NTR via an entirely nonspecific interaction. Specific interactions between the RVD (largely influenced by position 13 in each repeat, as the residue at position 12 is mainly a modulator of repeat structure) are suggested to be only a distant third contributor to overall TALE binding affinity (32). The authors postulated that TALE specificity is thus largely a result of steric clashes involving residue 13 of repeats in the RVD, and therefore described TALE specificity as arising from negative discrimination (32). Despite this understanding, however, several open questions still surround TALE specificity. Given the large amount of binding energy attributed to non-specific interactions, how do ionic strength and ionic species present in solution affect TALE specificity? How does the low contribution of specific binding to overall binding energy permit successful target search and localization? Given a negative discrimination model for TALE binding, how can experimental results showing similar binding activities for specific and non-specific binding be reconciled?

The effects of solution conditions (ionic strength and ion type) have been extensively investigated for several other sequence-specific DNA-binding proteins, while the effects of these factors on TALE specificity are not fully understood. One recurring theme in modulation of DNA-binding protein specificity is the suppression or enhancement of non-specific binding, which is mostly contributed by electrostatic interactions (33–39). Several sequence-specific DNA-binding proteins demonstrate minimal target site discrimination under low ionic strength conditions, when electrostatic interactions are not significantly screened (33–36). Indeed, these proteins display substantial target site specificity under elevated ionic strength conditions when non-specific binding affinity is suppressed. In some cases, sufficient electrostatic screening can be satisfied by higher concentrations of monovalent salts (33,34). For example, a fluorescence anisotropy study of tumor suppressor p53 binding to half and full target sites, as well as non-specific substrates, showed that a total ionic strength of 175 mM (NaCl and buffer agents) is required for apparent target discrimination (33). Several studies on basic leucine zipper proteins (B-ZIPs), including CREB and Fos-Jun (35), as well as the prokaryotic transcription factor NikR(36), also highlighted the roles of divalent cations in their binding specificity. In the case of B-ZIP proteins, no discernible sequence specificity is achieved even under elevated monovalent salt concentrations, and addition of a small amount of divalent salt nearly abolishes non-specific binding while only slightly impacting target binding (35,36). Several restriction enzymes, including BamHI (37), EcoRV (38) and EcoRI (39), have also been shown to require Mg²⁺ or Ca²⁺ for sequence-specific binding, though such divalent cations are also required for their enzymatic activity.

In this work, we presented an in-depth experimental and computational study of the effects of ionic strength and ionic species on TALE binding to DNA. Using an in vitro fluorescence anisotropy assay, we examined the binding affinities of a series of TALE proteins to both specific and non-specific DNA sequences under a variety of solution conditions. Under purely monovalent salt conditions, we found that TALE specificity is prohibitively minimal, especially for longer TALE proteins (here, a 21.5 repeat TALE construct). Strikingly, we observed a substantial degree of binding specificity upon addition of small amounts of certain divalent cations, such that non-specific binding is nearly abolished. The effects of divalent cations on binding specificity are enhanced in TALE constructs with larger CRDs. For the 21.5 repeat TALE protein, we also observed a slight enhancement of target binding, leading to an even greater target discrimination. Moreover, we found that the non-specific binding of NTR-only truncation mutants is significantly attenuated by divalent salts but not monovalent salts. To directly complement experimental results, we performed molecular dynamics (MD) simulations to quantify the preferential interactions between TALE and DNA in the presence of monovalent and divalent salts. Using MD simulations, we quantified the change in TALE binding free energy due to the presence of salts. Our results suggest that divalent cations preferentially bind to DNA, whereas monovalent cations are preferentially excluded from DNA. The significant attenuation of NTR binding by divalent cations is caused by the DNA-binding competition between the NTR and divalent cations, and the non-specific contacts between TALE CRD and DNA backbone are attenuated by divalent cations while some specific contacts may be stabilized by divalent cations. Taken together, these results elucidate the effects of divalent cations on TALE binding affinity and specificity, which can potentially be useful for future application of TALENs for gene editing and regulation.

MATERIALS AND METHODS

Preparation and cloning of tSCA21.5 and Naldt-tSCA21.5

In this work, we studied a series of AvrXa10 (40)-based TALE proteins, which were redesigned for editing the human β-globin gene containing a mutation associated with sickle cell disease. (41) We used this parent protein (containing 21.5 repeats) to design several TALEs with variable numbers of repeats in the CRD. The gene encoding for the untagged TALE protein (tSCA21.5) was assembled using the Golden Gate cloning method (Addgene TALEN Kit 1000000024, as described in our prior work (33)). For this purpose, repeats 1-10 of the TALE were assembled into the pFUSA30A vector, repeats 11-20 into the pFUSA30B vector, and repeat 21 was inserted into the pFUSB1 vector using BsaI-HF (New England Biolas (NEB)) and T4 DNA ligase (NEB). In this reaction, the temperature was cycled 10 times between 37°C (5 minutes) and 16°C (10 minutes), after which the reaction was treated with Plasmid-Safe nuclease (Epicentre Biotechnologies). DH5α (Cell Media Facility at the University of Illinois at Urbana-Champaign) transformants were assayed for proper assembly using restriction digestion of the purified (Qiagen) plasmids. Next, together with the plasmid coding for the last half-repeat pLR(NG), the repeats were further assembled into a specifically engineered destination vector pET28-GG-TALE using BsmBI (Fisher Scientific) and T4 DNA ligase. The destination vector contained an N-terminal His-tag and flanking N- (208 aa) and C- (63 aa) terminal regions of the TALE, as well as the BsmBI sites corresponding to the kit BsmBI sites (see sequence below). For fluorescent labeling of TALEs, we modified the original plasmid pET-tSCA21.5 with an oligonucleotide inserted encoding an N-terminal LCPTSR hexapeptide (aldehyde tag (34,42)) upstream of the His-tag. To this end, the plasmid was amplified in fragments containing the insert and assembled using the Gibson Assembly Kit (NEB), thereby generating the construct pET-Naldt-tSCA21.5. Similar to prior single molecule experiments, TALEs used in fluorescence anisotropy experiments contained a genetically encoded aldehyde tag, which showed no difference in binding affinity compared to wild type TALEs (28).

Protein sequence of Naldt-tSCA21.5

MGPLCTPSRSSHHHHHHSSGLVPRGSHMLDTSLLDSMPAVGTPHTAAAPAECDEVQSGLRAADDPPPTVRVAVTAARPPRAKPAPRRRAAQPSDASPAAQVDLRTLGYSQQQQEKIKPKVRSTVAQHHEALVGHGFTHAHIVALSQHPAALGTVAVTYQDIIRALPEATHEDIVGVGKQWSGARALEALLTEAGELRGPPLQLDTGQLLKIAKRGGVTAVEAVHAWRNALTGAPLNLTPDQVVAIASNIGGKQALETVQRLLPVLCQDHG  LTPDQVVAIASNNGGKQALETVQRLLPVLCQDHG  LTPDQVVAIASHDGGKQALETVQRLLPVLCQDHG  LTPDQVVAIASNIGGKQALETVQRLLPVLCQDHG  LTPDQVVAIASNIGGKQALETVQRLLPVLCQDHG  LTPDQVVAIASHDGGKQALETVQRLLPVLCQDHG  LTPDQVVAIASHDGGKQALETVQRLLPVLCQDHG  LTPDQVVAIASNGGGKQALETVQRLLPVLCQDHG  LTPDQVVAIASHDGGKQALETVQRLLPVLCQDHG  LTPDQVVAIASNIGGKQALETVQRLLPVLCQDHG  LTPDQVVAIASNIGGKQALETVQRLLPVLCQDHG  LTPDQVVAIASNIGGKQALETVQRLLPVLCQDHG  LTPDQVVAIASHDGGKQALETVQRLLPVLCQDHG  LTPDQVVAIASNIGGKQALETVQRLLPVLCQDHG  LTPDQVVAIASNNGGKQALETVQRLLPVLCQDHG  LTPDQVVAIASNIGGKQALETVQRLLPVLCQDHG  LTPDQVVAIASHDGGKQALETVQRLLPVLCQDHG  LTPDQVVAIASNIGGKQALETVQRLLPVLCQDHG  LTPDQVVAIASHDGGKQALETVQRLLPVLCQDHG  LTPDQVVAIASHDGGKQALETVQRLLPVLCQDHG  LTPDQVVAIASNIGGKQALETVQRLLPVLCQDHG  LTPDQVVAIASNG  GGKQALESIVAQLSRPDPALAALTNDHLVALACLGGRPALDAVKKGLPHAPELIRRINRRIPERTSHRVA

Preparation and cloning of Naldt-tSCA15.5 and Naldt-tSCA11.5

In order to construct TALEs containing only the first 15.5 and 11.5 repeats of the parent TALE with 21.5 repeats, we first engineered a destination vector pET28-Naldt-GG-TALE containing the aldehyde tag upstream of the His-tag using the Gibson Assembly Master Mix (NEB). Naldt-tSCA15.5 and Nald-tSCA11.5 were assembled using the Golden Gate cloning method (33) into the destination vector pET28-Naldt-GG-TALE. To this end, repeats 1–10 were assembled into vector pFUSA, repeats 11–15 were assembled into pFUSB5, and repeat 11 was inserted into vector pFUSB1 using the BsaI-HF and T4 DNA ligase by cycling the temperature 10 times between 37°C (5 min) and 16°C (10 min), after which the reactions were treated with Plasmid-Safe nuclease. DH5α transformants were assayed for proper assembly using restriction digestion of the purified plasmids. pFUSA containing the first 10 repeats was assembled with either pFUS-B5 (for TALE15.5) or pFUS-B1 (for TALE11.5) together with the plasmid coding for the last half-repeat into the pET28-Naldt-GG-TALE vector using BsmBI and T4 DNA ligase.

Cloning of Naldt-NTR

To generate the NTR truncation mutant, we digested pET-Naldt-tSCA21.5 with SacII and HindIII and inserted the sequence: CAAAGCGTGGTGGCGTGACCGCGGTGGAAGCGGTCCATGCCTGGCGTAATGCGTTGACGGGCGCCCCCCTGAACTAAGTCAGATAACCGGATACAGACAAGCTTGCGGCCGCACTCGAGCACCAC. This sequence was synthesized as a gBlock Gene Fragment (IDT), using the Gibson Assembly Master Mix (NEB).

Protein sequence of Naldt-NTR

MGPLCTPSRSSHHHHHHSSGLVPRGSHMLDTSLLDSMPAVGTPHTAAAPAECDEVQSGLRAADDPPPTVRVAVTAARPPRAKPAPRRRAAQPSDASPAAQVDLRTLGYSQQQQEKIKPKVRSTVAQHHEALVGHGFTHAHIVALSQHPAALGTVAVTYQDIIRALPEATHEDIVGVGKQWSGARALEALLTEAGELRGPPLQLDTGQLLKIAKRGGVTAVEAVHAWRNALTGAPLN

Protein expression and purification

BL21 (DE3) electrocompetent Escherichia coli cells were co-transformed with plasmids encoding TALE constructs and the pBAD-FGE plasmid (generous gift of Dr Taekjip Ha, University of Illinois at Urbana-Champaign). A single colony was grown in 5 ml lysogeny broth (LB) supplemented with 100 μg/ml ampicillin and 25 μg/ml kanamycin as a seeding culture until saturation, and subsequently cultured in 500 ml of Terrific Broth at 37°C and 250 revolutions per minute (RPM) with the corresponding antibiotics until the cell culture reached an optical density OD₆₀₀ of 0.3–0.4. FGE expression was induced with 0.2l-arabinose (Sigma), and the culture was grown further at 37°C 250 RPM until reaching OD₆₀₀ 0.7–0.8. TAL effector expression was induced with 0.4 mM isopropyl β-d-1-thiogalactopyranoside (IPTG) (Sigma). The proteins were expressed overnight at 16°C in an incubated shaker at 250 RPM. Following expression, cell cultures were centrifuged at 4000 RPM at 4°C for 15 min, and the pellets were resuspended in 10–20 ml of lysis buffer (25 mM Tris–HCl (Fisher) pH 7.5, 300 mM NaCl (Fisher), 0.5 Triton-X100 (Sigma), 5 glycerol (Sigma), 4 U/ml DNase I (NEB), 0.3 mM phenylmethanesulfonylfluoride (Sigma), 1 mg/ml lysozyme (Sigma)). The cells were lysed by sonication for 20 min total with alternating 5 s pulse/5 s rest, and cell debris were centrifuged at 13 000 RPM for 20 min at 4°C. The His-tagged TALEs were purified using AKTA pure chromatography system (GE Healthcare) with a 1 ml HisTrap column (GE Healthcare). The cleared lysate was loaded on the column, washed with 20 mM Tris–HCl pH 7.5, 500 mM NaCl and 20 mM imidazole (Sigma), and eluted with 20 mM Tris–HCl pH 7.5, 500 mM NaCl and 250 mM imidazole. The purified protein was dialysed in 2 L 50 mM phosphate buffer pH 7-8.4 (depending on the predicted protein isoelectric point), 500 mM NaCl at 4°C overnight. When necessary, the protein was further purified using a 16/600 Superdex 200 pg gel filtration column (GE Healthcare) using 50 mM phosphate buffer pH 7–8.4 (depending on the predicted protein isoelectric point), and 500 mM NaCl.

DNA templates

DNA templates were prepared using the following method. Single stranded oligonucleotides (Integrated DNA Technologies) were first prepared in stock solutions in low salt buffer (20 mM Tris–HCl, pH 7.5, 50 mM KCl). DNA concentrations were determined via UV/vis spectrophotometry (NanoDrop, Thermo Fisher). Equivalent concentrations of a 5,6-FAM (fluorescein derivative) labeled oligonucleotides and their corresponding complimentary oligonucleotides were combined into solution in low salt buffer, placed in a covered heating block at 90°C for 2 min, followed by slow cooling to room temperature.

Fluorescence anisotropy experiments

Mixtures of 1 nM double stranded (ds) oligonucleotide templates and TALE protein were prepared in a standard buffer (20 mM Tris–HCl, pH 7.5) with varying amounts of monovalent salt (KCl or NaCl) and/or divalent salt (MgCl₂, MgSO₄, ZnCl₂, MnSO₄, SrCl₂, CaCl₂, where specified). Mixtures of TALE protein and DNA template (200 μl) were incubated for 10 min at room temperature (25°C) and then assayed in black 96-well plates (Corning) in duplicates. Fluorescence polarization measurements were performed on an Infinite 200 Pro microplate reader (Tecan) using excitation and emission wavelengths of 485 and 535 nm, respectively. Fluorescence polarization values were converted to fluorescence anisotropy values using the following relation:

(1)

where A is anisotropy and P is polarization. Fluorescence anisotropy values were determined by averaging over two repeated experiments and were plotted relative to anisotropy values in the absence of protein. The dissociation constant (K_d) was calculated using a non-linear least squares algorithm (Origin 8.5) using the following expression:

(2)

where A is experimental anisotropy value, A_f is anisotropy of DNA without protein, A_b is anisotropy of protein-bound DNA, L_T is total ligand (DNA) concentration, and R_T is total receptor (protein) concentration. For experiments with unlabeled competitor DNA, 100 nM of DNA templates lacking the fluorescein dye (5,6-FAM) was added to solutions containing 1 nM labeled probe DNA.

Molecular dynamics simulation

We simulated the NTR, the CRD (containing 21 repeats), and DNA molecules (a 31 bp random DNA for the NTR and a 36 bp target DNA for the CRD) in their dissociated states, in addition to also simulating the DNA-bound NTR and CRD complexes in aqueous solutions with varying concentrations of KCl and/or MgCl₂. The crystal structure of TAL effector PthXo1 (PDB ID: 3UGM, Figure 1B) (23) from Xanthomonas was used for the molecular dynamics (MD) simulation studies. Missing parts of the NTR in the TALE crystal structure were added via SWISS-MODEL homology-modelling server (43). The NTR structure included the 1-240 residues. We aligned the modeled NTR structure and a 31 bp DNA (5’GACACCATTAGCAACCTCAAACAGACACCAT3’) to the TALE crystal structure to get an initial structure of the NTR-DNA complex. The 1–21 repeats (the 289–1001 residues) in the CRD and the 36 bp DNA of the TALE crystal structure were used in MD simulations. The 36 bp DNA sequence in CRD-DNA complex was 5’TAGATATGCATCTCCCCCTACTGTACACCACCAAAA3’.

All the simulations were setup in AmberTools16 (44) and performed using OpenMM 7.0 GPU platform (45). Each initial structure was solvated with TIP4P-EW water molecules in a cubic simulation box. The minimum distance from the boundary of the simulation box to the solute surface was at least 15 Å. Certain number of cations and anions were added to neutralize the system and match the specific salt solutions (Supplementary Table S1). Amber ff14SB force field parameters were used for the protein and DNA (with bsc1 modifications) (46). Monovalent ions parameters for TIP4P-EW water model by Joung et al. (47) were used for K⁺ and Cl⁻. 12–6–4 model with fine tuned C4 terms by Panteva et al. for divalent ions were used for Mg²⁺, which were optimized to improve the approximation of interactions between divalent ions and nucleic acids in TIP4P-EW water box (48). After energy minimization, each system was running in NPT ensemble (300 K, 1 bar) for at least 50 ns with an integration time step of 2 ps. Monte Carlo barostat algorithm (49) was employed to maintain the constant pressure and Langevin integrator (50) was used to control the system temperature with a collision frequency of 1 ps^-1. Periodic boundary conditions were employed and the particle-mesh Ewald method (51) was used to treat the electrostatic interactions with a 10 Å cutoff distance. All the covalent bonds involving hydrogen atoms were constrained.

Preferential interaction coefficient calculation

The theory of computation of preferential interaction coefficient (Γ₂₃, Scatchard notation (52): 1—water, 2—solute, 3—salt) from MD simulations based on a statistical mechanical method with no adjustable parameters was developed by Baynes and Trout (53). The MD simulation data was used to calculate Γ₂₃ using the following expression (53–58):

(3)

where n₁(r, t), n₃(r, t) are the number of water molecules and ions within distance r from solute surface at time step t, and n₁, n₃ are the total number of water molecules and ions in the simulation box. The average Γ₂₃ over all time frames (〈Γ₂₃〉) was determined as the final Γ₂₃ value. To quantify the error bars on 〈Γ₂₃〉, we divided the individual MD trajectory into three blocks, and calculated the block average values of 〈Γ₂₃〉. The standard deviation of the block averages was determined as the error bar (54). For KCl or MgCl₂, the preferential interaction coefficient calculations require an estimate of Γ₂₃ for both cation and anion, given by (54,59):

(4)

(5)

where |Z₂| is the net charge of the solute.

Theory for quantifying the effect of salt on free energy of TALE binding

The theory is based on the previous method that estimates the effect of cosolutes on free energy of protein-protein binding (60). The protein–DNA binding free energy is defined by

(6)

Then the change in protein–DNA binding free energy (ΔΔG) in the presence of salts in solution, compared to that in pure water, is given by

(7)

Also, ΔΔG can be calculated by the following equation:

(8)

where Δμ^tr is the transfer free energy that measures the change in interaction energy between the solute and water when the solute is transferred from pure water to a salt solution. The transfer free energy is equal to the change of chemical potential of the solute. Using Scatchard notation (1—water, 2—solute, 3—salt), the transfer free energy of protein is given by

(9)

where μ₂ is the chemical potential of the protein. Considering the process of adding a protein to a salt solution, the chemical potential of the salt is disturbed by the protein, , and the chemical potential of the protein is disturbed by the salt, . Here, m is molality, T and P are temperature and pressure, respectively. The total change in chemical potential of the protein due to the transfer to a salt solution is given by

(10)

(11)

(12)

where Γ₂₃ is the preferential interaction coefficient. From a thermodynamic view, Γ₂₃ is a measure of the number of salt molecules added to the solution per protein to keep the chemical potential of the salt constant. It is also a measure the excess number of salt molecules within the local domain around protein compared to the bulk. The derivative term is the variation in the chemical potential of the salt as a function of its molality, which can be obtained from the experimental activity or osmolality data. Using the expression for chemical potential as a function of activity and the Gibbs–Duhem relationship to approximate the change in the activity of the salt in terms of the change in the activity of water:

(13)

(14)

(15)

where Osm denotes the osmality of the salt solution. Further, assuming that Γ₂₃ is a linear function of cosolute molality, the transfer free energy can be approximated as

(16)

(17)

(18)

In the end, we can get the change in binding free energy ΔΔG_b:

(19)

(20)

Therefore, the experimental water activity or the osmolality data for the salt solution and the preferential interaction coefficients for the DNA, protein and complex are required to determine the ΔΔG_b. The water activities in unsaturated KCl and MgCl₂ solutions have been previously reported (61,62). The water activity in salt solution is given by the following equation (61):

(21)

where ϕ is practical osmotic coefficient, M₁ is molar mass of water, ν is stoichiometric parameter. ν is the number of moles of ions produced by complete dissociation of one mole of salt, and for KCl ν equals 2 and for MgCl₂ν equals 3.

The experimental osmotic coefficient and water activity data used in this study were summarized in Supplementary Table S2. The preferential interaction coefficients were calculated from MD simulation data.

RESULTS

TALEs show minimal target recognition in pure monovalent salt.

We utilized a fluorescence anisotropy assay in order to probe the role of monovalent and divalent salts on TALE binding specificity. To this end, we used both target and random DNA sequences to probe the affinity and specificity of TALEs containing 21.5, 15.5 and 11.5 repeats in the CRD (Figure 1C). We further compared these results to the binding data of an NTR-only TALE truncation mutant. In all cases, we characterized TALE binding affinities under a variety of solution conditions in terms of equilibrium dissociation constants, K_{d,non-specific} for non-target binding and K_d,specific for target binding (Supplementary Table S3), obtained from fluorescence anisotropy measurements. A small value of K_{d,non-specific} or K_d,specific suggests a strong association between TALEs and random or target DNA.

In pH-buffered aqueous solutions containing the monovalent salt KCl in the absence of divalent cations, TALE binding affinities to target and random DNA sequences appear to be similar (Figure 2). Strikingly, the apparent low degrees of specificity were observed for a series of different TALEs (21.5, 15.5 and 11.5 repeats) studied in this work and across a wide range of different concentrations of KCl and NaCl (Figure 2, Supplementary Figure S1). For the TALE containing 21.5 repeats in the CRD, we observed a significant lack of specificity between target and random DNA even under high concentrations of KCl (Figure 2A–C and Supplementary Figure S2). No difference in binding specificity was observed when the monovalent cation was exchanged for Na⁺ in place of K⁺ (Supplementary Figure S1). In the presence of monovalent salt, the 15.5 repeat TALE appears to show slightly greater discrimination between random and target sequences compared to both the 11.5 repeat and the 21.5 repeat TALEs (Figure 2). Our results are in agreement with the previous report that under pure monovalent salt solutions, TALE specificity increases and then decreases with increasing number of repeats, peaking between 15 and 19 repeats (15). Rinaldi et al. suggested that TALE binding affinities for both target and non-target DNA increases non-linearly as the number of repeats increases, with the gain decaying exponentially (15). However, the affinity gain for non-target DNA decays slower compared to that for target DNA (15). Nevertheless, there is only less than an order of magnitude preference for target DNA when comparing K_{d,non-specific} and K_d,specific, which is not enough for the specificity required for TALE and TALEN function in vivo (two or more orders of magnitude difference is required) (12–14).

Figure 2.

Effect of monovalent ions on TALE-DNA binding for 21.5 repeat, 15.5 repeat and 11.5 repeat TALEs. Binding of the 21.5 repeat TALE to (A) target (dots for raw data, solid lines for fitted curves) and (B) random (dots for raw data, dashed lines for fitted curves) DNA sequences measured by fluorescence anisotropy. (C) Equilibrium dissociation constants, K_{d,non-specific} and K_d,specific, determined for the 21.5 repeat TALE to random and target DNA sequences. Binding of the 15.5 repeat TALE to (D) target (dots for raw data, solid lines for fitted curves) and (E) random (dots for raw data, dashed lines for fitted curves) DNA sequences, and (F) K_{d,non-specific} and K_d,specific values. Binding of the 11.5 repeat TALE to (G) target (dots for raw data, solid lines for fitted curves) and (H) random (dots for raw data, dashed lines for fitted curves) DNA sequences, and (I) K_{d,non-specific} and K_d,specific values. TALE-DNA binding was measured in 20 mM Tris–Tris HCl buffer with different amounts of KCl. DNA templates were maintained at a constant concentration of 1 nM and TALE concentrations were adjusted. Duplicate experiments were performed to determine the mean and standard deviation of fluorescence anisotropy value.

Certain divalent cations enhance TALE binding specificity in an ionic strength dependent manner.

We further sought to understand the apparent lack of sequence specificity for TALE binding in monovalent salt by assessing the impact of ionic strength and solution conditions. Here, we first studied the 21.5 repeat TALE binding to target and random DNA in pH-buffered aqueous solutions containing 100 mM KCl and 10 mM various divalent salts (MgCl₂, MgSO₄, CaCl₂, MnSO₄, SrCl₂, ZnCl₂). Strikingly, specificity in TALE binding to target sites is restored in the presence of certain divalent cations (Figure 3A–E, Supplementary Figures S3 and S4). In particular, the divalent cations Mg²⁺ or Ca²⁺ substantially diminish non-target binding and simultaneously enhance the binding affinity to target sites (Figure 3E) compared to the binding in purely monovalent solutions containing 140 mM KCl (Figure 2C). We characterized TALE binding in the presence of a variety of divalent cations, and we observed the strongest effect with Mg²⁺ or Ca²⁺, a moderate effect with Mn²⁺, a slight effect with Sr²⁺, and essentially no discrimination effect with Zn²⁺ (Figure 3A–E and Supplementary Figure S3). No differences in TALE binding were observed upon changing the monovalent anions or divalent anions (Supplementary Figure S4). Interestingly, these effects appear to be sensitive to the chemical identity of divalent cations, with an apparent correlation with the abundance of common cellular cations. However, this phenomenon is not entirely driven by ionic radius or any apparent particular periodic trend.

Figure 3.

Effects of divalent cations on TALE-DNA binding for the 21.5 repeat TALE. Binding of the 21.5 repeat TALE to target (dots for raw data, solid lines for fitted curves) and random DNA (dots for raw data, dashed lines for fitted curves) measured by fluorescence anisotropy in 20 mM Tris–Tris HCl buffer containing 100 mM KCl and 10 mM (A) MgCl₂, (B) CaCl₂, (C) SrCl₂ or (D) ZnCl₂. (E) Equilibrium dissociation constants K_{d,non-specific} and K_d,specific for TALE binding in the presence of various divalent cations. Fluorescence anisotropy results for binding to (F) target DNA and (G) random DNA in 20 mM Tris–Tris HCl buffer with the total ionic strength of added salts held constant at 120 mM. (H) Equilibrium dissociation constants K_{d,non-specific} and K_d,specific for TALE–DNA binding in the presence of KCl and MgCl₂. Fluorescence anisotropy results for the binding to (I) target DNA and (J) random DNA in 20 mM Tris–Tris HCl buffer with increasing concentrations of MgCl₂ in the absence of any monovalent salt. (K) Equilibrium dissociation constants K_{d,non-specific} and K_d,specific for TALE–DNA binding in the presence of only MgCl₂. DNA concentration was held constant at 1 nM, and TALE concentrations were adjusted as noted. Duplicate experiments were performed to determine the mean and standard deviation of fluorescence anisotropy value.

Next, we studied the effect of increasing the overall ionic strength on the binding behavior of the 21.5 repeat TALE (Supplementary Figure S5). Here, we maintained KCl concentration at 100 mM and increased divalent cation concentration from 5 to 10 mM. Our results show that TALE non-specific binding affinity generally decreases upon increasing the concentration of divalent cations, which is similar to TALE binding behavior in the presence of monovalent salt. Strikingly, TALE specific binding affinity is essentially unaffected by the increase in divalent cation concentration (Supplementary Figure S5). However, the decrease in DNA-binding affinity to off-target sites is significantly larger when the increase in total ionic strength results from a higher proportion of divalent cations (as opposed to pure monovalent cations).

Based on these results, we next studied the effect of increasing the proportion of divalent cations in solution while maintaining the total ionic strength at a constant value (90 or 120 mM). For all full length TALEs studied in this work (TALEs with 11.5, 15.5 and 21.5 repeats), we observed that TALE binding specificity increases for target DNA (corresponding to an increase in the difference between K_{d,non-specific} and K_d,specific) with increasing proportion of Mg²⁺ (Figure 3F–H, Supplementary Figures S6 and S7). For shorter TALEs (11.5 and 15.5 repeats), differences in binding affinity between random and target sites are minimal until approximately 40 of the total ionic strength is contributed by Mg²⁺ (Supplementary Figures S6 and S7). In contrast, the 21.5 repeat TALE shows higher specificity for target sites when the proportion of Mg²⁺ reaches similar levels (where the difference between K_{d,non-specific} and K_d,specific is >500-fold) (Figure 3H). In addition, the specific binding affinity for the 21.5 repeat TALE increases with increasing amount of divalent cations, in contrast to the slight decreases for both the 15.5 repeat and the 11.5 repeat TALEs. At higher total ionic strength with moderate proportion of Mg²⁺ (similar to physiological conditions), the 21.5 repeat TALE shows extremely high specificity (Supplementary Figure S8), consistent with prior studies that utilize TALEs for specific genomic edits (12–14). Interestingly, in buffered solutions containing only Mg²⁺ as the primary cation (40 mM MgCl₂), the 21.5 repeat TALE binds to target DNA with extremely high specificity (the ratio between K_{d,non-specific} and K_d,specific is approaching infinity, effectively as K_{d,non-specific} is too large to measure) (Figure 3I–K). These results suggest that TALE binding specificity increases with higher concentrations of divalent cations, as opposed to proportional increases in ionic strength provided by monovalent cations only.

We further utilized a DNA competitor assay for fluorescence anisotropy in which a large molar excess (100-fold) of unlabelled random DNA was added to 1 nM of fluorescently labelled target probe DNA. We hypothesized that in the presence of a large excess of competitor (random) DNA, TALEs exhibiting relatively weak binding specificity would be sequestered away from target DNA, as they are unable to discriminate between target and random DNA. In purely monovalent salt conditions, the 21.5 repeat TALE shows substantially attenuated binding to both target and random DNA in the presence of random competitor DNA (Supplementary Figure S9). In the presence of only monovalent salt, therefore, there appears to be minimal differences in the behavior of TALE binding to target versus random DNA. Interestingly, the behavior of TALE binding in purely monovalent salt conditions is sharply contrasted by TALE binding to target DNA in the presence of 10 mM MgCl₂. Here, we note that TALE binding in either monovalent salt or monovalent and divalent salt conditions are compared under conditions of equivalent total solution ionic strength for consistency. Notably, TALE binding to target DNA is essentially unaffected by the large excess of random competitor DNA in the presence of Mg^{2 +} (Supplementary Figure S9). Overall, our results suggest that the presence of small amounts of certain divalent cations generally attenuates non-specific binding and stabilizes specific binding, thereby enhancing TALE binding specificity.

TALE NTR binding to DNA is impaired by divalent cations.

In order to further elucidate the effect of divalent cations on target site discrimination for TALEs, we designed a TALE truncation mutant containing only the NTR (known as the NTR-only construct), which was used to probe the role of divalent cations on TALE NTR binding. In prior work, it has been clearly shown that the NTR is critical for nucleation of TALE binding events and NTR binding to DNA occurs in a non-specific fashion, driven primarily by electrostatic interactions (5,6,28). Furthermore, we have previously shown that the TALE NTR is capable of binding to and diffusing one-dimensionally along DNA in reduced ionic strength conditions in the presence of monovalent salt (28).

In the present work, our results show that Mg^{2 +} entirely suppresses the ability of the NTR-only TALE mutant to bind to DNA, even in low concentrations of divalent cations (Figure 4). In particular, the binding affinity of the NTR-only TALE to DNA is significantly attenuated even at very low total ionic strength in the presence of 2.5 mM MgCl₂ compared to pure monovalent salt conditions (Figures 4D and E). As expected, a TALE construct lacking the NTR (CRD+CTR only) shows no binding affinity for target or random DNA even with addition of divalent cations (Supplementary Figure S10). Overall, our results suggest that addition of monovalent salt and/or an increase in overall ionic strength can weaken the NTR binding interactions, however, divalent cations greatly suppress NTR binding to DNA. Together, these data help to elucidate the effects of divalent cations on TALE binding. When the NTR binding is significantly attenuated by the presence of divalent cations, the overall non-specific binding affinity of TALEs decreases. TALEs with shorter CRDs have a greater proportion of their overall binding energy conferred via the NTR compared to longer TALEs. Thus, the overall binding affinities of shorter TALEs are expected to be more strongly impacted than the binding affinity of longer TALEs by the presence of divalent cations, which is consistent with the results presented in this study (Figure 3, Supplementary Figures S6 and S7).

Figure 4.

TALE NTR binding to DNA in a variety of solution conditions. Binding of the NTR-only TALE construct (without a CRD or CTR) to random DNA template was measured via fluorescence anisotropy in 20 mM Tris–Tris HCl buffer with increasing concentrations of (A) KCl only, (B) MgCl₂ only or (C) a combination of KCl and MgCl₂ with total ionic strength of either 15 or 25 mM (dots for raw data, dashed lines for fitted curves). Equilibrium dissociation constants for TALE NTR binding to DNA in the presence of (D) KCl, (E) MgCl₂ and (F) KCl and MgCl₂. Binding is rapidly diminished in the presence of MgCl₂, even when accounting for total ionic strength. DNA concentration was held constant at 1 nM. Duplicate experiments were performed to determine the mean and standard deviation of fluorescence anisotropy value.

Divalent cations introduce significant changes in TALE binding free energy compared to monovalent cations.

On the basis of our experimental results, we hypothesized that divalent cations significantly attenuate TALE non-specific binding energy more than monovalent cations, thereby enhancing TALE target discrimination. To validate this hypothesis, we further sought to use MD simulations to characterize the effects of monovalent and divalent cations on TALE target binding free energy. Specifically, we aimed to estimate the change in free energy of TALE target binding (ΔΔG_b) under salt conditions versus pure water environment. A positive ΔΔG_b therefore indicates destabilization of binding by the presence of salts, whereas a negative ΔΔG_b suggests a stabilizing effect. We tested our hypothesis by comparing ΔΔG_b under monovalent and divalent salt conditions at the same ionic strength.

We sought to evaluate ΔΔG_b for individual TALE components of the NTR and the CRD, respectively. Estimation of ΔΔG_b requires determination of the preferential interaction coefficients (Γ₂₃; 1—water, 2—protein or DNA (solute) and 3—ion (cosolute)) of the solutes (e.g. the NTR, the CRD, DNA and their complexes) in the presence of different ionic species (Materials and Methods) (55–58,60,63,64). Γ₂₃ was determined from MD simulation data (53–58), which measures the excess number of ions in the local domain of a solute compared to the bulk (53,54,59). Positive values of Γ₂₃ suggest that ions preferentially bind to the solute surface, whereas negative values of Γ₂₃ suggest that ions are preferentially excluded from the solute surface.

For all simulations, the time-averaged Γ₂₃ values converge after 50 ns (Supplementary Figures S11 and S12), and approach constant values for r >6 Å (Figure 5 and Supplementary Figure S13). Previous studies have also found that the extent of the local domain around the solute where Γ₂₃ varies is ∼6 Å. (53–58,60,64) We observed that the TALE CRD relaxes to an extended superhelical conformation at the dissociated state within 50 ns, which is consistent with the previously reported crystal structures of DNA-free and DNA-bound TALEs (65). Final Γ₂₃ values were determined at r = 8 Å and ΔΔG_b values were then estimated using the calculated Γ₂₃ (Table 1). Strikingly, our data show that divalent salts exhibit positive preferential interaction with all the solutes compared to monovalent salts. Moreover, divalent salts preferentially bind to all the solutes, and monovalent salts are preferentially excluded from all the solutes (except the CRD in 100 mM KCl).

Figure 5.

Variation in preferential interaction coefficients as a function of distance from the solute surface. For simulations involving DNA, TALE NTR and NTR-DNA complex in (A) 100 mM KCl and (B) 25 mM MgCl₂, and simulations involving DNA, TALE CRD and CRD–DNA complex in (C) 100 mM KCl and (D) 25 mM MgCl₂, the preferential interaction coefficients approach constant values beyond 6 Å from the molecule surface.

Table 1.

Change in free energy of binding (ΔΔG_b) between TALE NTR/CRD and DNA in the presence of monovalent and divalent salts, and the preferential interaction coefficient (Γ₂₃) for the simulated systems. The error bars on ΔΓ₂₃ and ΔΔG_b are given, and the error bars on Γ₂₃ for the simulated systems are ∼1.

System	Concentration (mM)	Γ23, complex	Γ23, DNA	Γ23, TALE	ΔΓ23	ΔΔGb (kcal/mol)
NTR, KCl	100	−13.5	−11.9	−1.0	−0.6 ± 0.6	0.7 ± 0.7
NTR, MgCl2	25	3.4	1.3	4.3	−2.2 ± 1.1	3.6 ± 1.8
NTR, KCl	200	−17.1	−12.5	−2.5	−2.1 ± 0.7	2.3 ± 0.8
NTR, MgCl2	50	3.6	2.0	5.6	−4.0 ± 1.4	6.5 ± 2.3
CRD, KCl	100	−14.6	−13.5	1.9	−3.0 ± 1.5	3.3 ± 1.6
CRD, MgCl2	25	2.7	0.1	14.0	−11.4 ± 1.3	18.6 ± 2.1
CRD, KCl	200	−20.7	−15.6	−3.4	−1.7 ± 1.1	1.8 ± 1.2
CRD, MgCl2	50	7.0	2.8	13.1	−8.9 ± 2.1	14.5 ± 3.4

Our results suggest that divalent salts introduce significant changes in TALE binding free energy compared to monovalent salts (Table 1). For both the NTR and the CRD, positive ΔΔG_b values suggest that the presence of either monovalent or divalent salts in solution destabilizes their DNA-bound complexes. For the TALE NTR, the presence of MgCl₂ results in an over 3-fold change in ΔΔG_b compared to KCl at the same ionic strength. Furthermore, ΔΔG_b is observed to increase upon increasing ionic strength. For example, the ΔΔG_b values in 200 mM KCl and 50 mM MgCl₂ are 2.3 and 6.5 kcal/mol, respectively. The large differences in ΔΔG_b suggest that the presence of divalent cations will strongly impact the non-specific NTR binding compared to monovalent cations. The predictions are in agreement with experimental observations that divalent cations greatly attenuate and even suppress the NTR binding at low concentrations. For the TALE CRD, the differences in ΔΔG_b caused by divalent and monovalent salts are even larger (more than 5-fold). Under conditions of 25 mM and 50 mM MgCl₂, ΔΔG_b values are found to be 18.6 and 14.5 kcal/mol, respectively, which could be large enough to suppress the TALE CRD binding to DNA. Again, these results are consistent with the experimental observations that the TALE CRD+CTR construct without the NTR cannot bind to DNA. Overall, these results demonstrate that divalent salts have clear effects on TALE binding free energy due to their strong preferential interactions with TALE or DNA compared to monovalent salts. Furthermore, the non-specific NTR binding affinity is significantly decreased by divalent salts compared to monovalent salts, which confirms our hypothesis.

Divalent cations attenuate non-specific binding and stabilize specific binding via preferential binding to DNA.

MD simulations provide detailed molecular descriptions of interactions between ionic species and TALE/DNA complexes. Using this approach, we further sought to identify the interactions of divalent cations with solutes (TALE or DNA) that are responsible for suppression of TALE non-specific binding, as well as possible stabilization of specific binding. DNA is a negatively charged polymer, and divalent cations will preferentially associate with DNA compared to monovalent cations due to stronger electrostatic interaction, as suggested by Γ₂₃ (Figure 5).

We characterized the interaction strength between DNA and different ions by determining a contact coefficient, which is the ratio of the local ion concentration around each DNA nucleotide (within 8 Å from nucleotide surface) to the bulk ion concentration. We observed a >3-fold DNA-binding preference for Mg²⁺ compared to K⁺, which is consistent with prior studies (Supplementary Figure S14) (66,67). A snapshot from MD simulations suggests that divalent cations can bind to the minor groove, the major groove, and the backbone of DNA (Supplementary Figure S14). As a result, when a TALE protein binds to DNA, divalent cations that are associated with DNA are excluded from its surface (Supplementary Figures S15 and S16), which leads to decreased binding affinity. Some divalent cations might still reside around the interface between TALE and DNA, which can influence TALE–DNA interactions. By comparing the distribution of ions near the TALE–DNA interface before and after a TALE protein binds to DNA, we can obtain structural insight into the effects of divalent cations on TALE binding. We therefore identified the TALE residues at the interfaces between the NTR/CRD and DNA and quantified their interactions with ions by determining the contact coefficient during a binding event.

Our results suggest that the suppression of NTR binding by divalent cations is caused by preferential binding of divalent cations to DNA (Figure 6). The residues in the NTR that interact with DNA were identified from the changes in solvent-accessible area (ΔSAA) of the NTR residues when the NTR is in the dissociated state compared to the associated state. Among the top ten residues with the largest ΔSAA values, seven are positively charged arginine or lysine residues (Figure 6A). This further highlights the significant contribution of electrostatic interaction to overall non-specific NTR binding energy. As expected, the NTR residues in contact with DNA do not have favorable interactions with divalent cations (Figure 6B, C). Therefore, divalent cations effectively abolish the electrostatic interactions between the NTR and DNA, which reduces the overall affinity of NTR to DNA backbones in the presence of divalent cations. In this way, divalent cations indirectly suppress NTR binding to DNA as opposed to acting by direct interactions with the NTR. In other words, for the NTR to bind DNA, the exclusion of divalent cations from the DNA interface results in an overall decreased binding affinity. In particular, in the presence of 25 mM MgCl₂, ∼4 Mg²⁺ are required to be excluded from the DNA surface to form a NTR–DNA complex (Supplementary Figure S15). Overall, these results show that divalent cations preferentially bind to DNA over monovalent cations, which in turn suppresses TALE NTR binding.

Figure 6.

Preferential interactions of the NTR with ions. (A) Change in the solvent-accessible area ΔSAA of the residues in the TALE NTR when the NTR is in the dissociated state compared to the associated state. Ten residues with the largest ΔSAA values are labeled, with seven positively charged residues (red) and three hydrophobic residues (blue). (B) Contact coefficients for the residues in the NTR with the different ionic species. Mg²⁺ shows stronger interactions with parts of the NTR compared to K⁺ and Cl⁻. However, (C) the residues in the NTR that interact with DNA have little interactions with Mg²⁺. The NTR surface is colored based on the contact coefficients of the NTR residues with Mg²⁺ at the dissociated state. Binding of the NTR to DNA requires the exclusion of cations from DNA. In the presence of Mg²⁺, the TALE NTR binding is attenuated due to the strong associations between DNA and Mg²⁺.

Our results also suggest that the presence of divalent cations attenuates the non-specific contacts between the CRD and DNA while stabilizing some specific contacts (Figure 7). From the co-crystal structure of a TALE–DNA complex, the interactions between the CRD and DNA include non-specific contacts mediated by a lysine and glutamine at positions 16 and 17 in each repeat, respectively, and specific contacts by the second residue (position 13) in RVDs (Figure 7A). When the CRD is dissociated from DNA, residues 16 and 17 in the CRD repeats do not have favorable interactions with monovalent and divalent cations; however, residue 13 in the RVD can have strong interactions with divalent cations compared to monovalent cations (Figure 7B). These strong interactions are mainly contributed by the negatively charged aspartic acid residue in the CRD repeats containing histidine/aspartic acid (HD) in the RVD (Figure 7C). When the CRD is bound to DNA, MD simulations suggest that the non-specific contacts between the CRD and DNA backbone are significantly attenuated by the presence of either monovalent or divalent cations (Supplementary Figure S17). Snapshots show that monovalent or divalent cations preferentially bind to DNA such that non-specific interactions are suppressed (Figure 7D, E). This is consistent with our previous single molecule observation and model for TALE non-specific binding in which the CRD only makes partial contacts with the DNA backbone (28). Strikingly, MD simulations suggest that Mg²⁺ can still bind to the major groove of DNA and interact with both DNA nitrogenous bases and the aspartic acid residues in RVDs containing the sequence HD (Figure 7F). We hypothesize that DNA-bound Mg²⁺ can stabilize the specific interactions between aspartic acid and cytosine, thereby contributing to overall specific interaction energy. The experimental observation that addition of small amounts of divalent cations enhances the specific binding affinity of the 21.5 repeat TALE can be explained by the stabilization of these specific contacts. Indeed, for the 21.5 repeat TALE construct used in our experiments, eight RVD pairs have the sequence HD and three of them are located in the last five CRD repeats. Compared to the 21.5 repeat TALE, the stabilizing effect is not as evident for TALEs with shorter CRDs (11.5 or 15.5 repeats), presumably due to the reduced number of RVDs bearing the sequence HD. As a result, for the 11.5 repeat and 15.5 repeat TALEs, the reduction in non-specific interaction can not be compensated by the enhancement in specific interaction. Therefore, no enhancement in overall specific binding affinity is observed for these shorter TALE constructs. Overall, these results suggest that the presence of divalent cations attenuates the non-specific binding between the CRD and DNA backbone and also may contribute to specific binding by stabilizing contacts between HD residues in the RVD and cytosine.

Figure 7.

Preferential interactions of the CRD with ions. (A) The average change in the solvent-accessible area ΔSAA of the residues at each location within a repeat when the CRD is in the dissociated state compared to the associated state. (B) The average contact coefficients for (B) the residues at each position of the repeat and (C) different type of residues in the CRD at the dissociated state. Divalent cations show stronger interaction with the CRD residues compared to monovalent cations, which is mostly driven by the negatively charged aspartic acid residue at position 4 of each repeat or position 13 of the repeats with RVD of HD. Snapshots of the simulated CRD–DNA complex structures in the presence of (D) 200 mM KCl and (E) 50 mM MgCl₂. K⁺ and Mg²⁺ preferentially bind to DNA that the CRD only partially contacts with DNA backbone. The 1–21 repeats in the CRD are shown in tube (cyan), and the lysine (pink) and glutamine (blue) residues are shown in stick. K⁺ (green) and Mg²⁺ (magenta) are shown in spheres. (F) Mg²⁺ binds to the major groove of DNA, and interacts with DNA bases and the aspartic acid residue (the second residue in RVD of HD) in repeat 9, which may stabilize the specific interaction.

DISCUSSION

Despite a large body of prior work on TALE proteins, a comprehensive view of the binding mechanism of TALEs to specific and non-specific DNA is generally lacking. In this study, we systematically characterized the effects of different ionic species over a range of concentrations on TALE specificity using both in vitro fluorescence anisotropy assay and MD simulations. Our results show that the presence of low concentrations of certain divalent cations (Mg²⁺ or Ca²⁺) results in the ability of TALEs to discriminate between target and non-target sequences. This trend appears to be consistent with prior reports of TALEs binding to DNA; in particular, prior in vitro binding measurements carried out in the absence of Mg²⁺ (5,16) tend to show less sequence specificity when compared to those that include Mg²⁺ in binding buffers (12–14).

Our results show that even supra-physiological concentrations of monovalent cations are insufficient to account for the sequence specificity of TALEs. This behavior generally contrasts with many sequence-specific DNA-binding proteins, such as tumor suppressor p53 (33) and integration host factor (34), wherein non-specific protein binding to DNA can be adequately suppressed via sufficient concentrations of monovalent cations, thereby resulting in enhanced binding specificity. Surprisingly, our results suggest that the presence of Mg²⁺ or Ca²⁺ substantially suppresses off-target binding for all full length TALEs (11.5, 15.5, 21.5 repeats) and simultaneously enhances binding to target sites for the 21.5 repeat TALE, thereby enhancing TALE binding specificity. Furthermore, Mg²⁺ suppresses non-specific TALE NTR binding even at very low ionic strength. The divalent cation dependence for sequence-specific binding of TALEs is analogous to prior reports that Mg²⁺ is required to prevent non-specific binding of the B-ZIP proteins (CREB and Fos-Jun) (35) and the prokaryotic transcription factor NikR (36).

Our MD simulations provide energetic and structural insight into the effects of divalent cations on TALE binding. Compared to K⁺, Mg²⁺ introduces significant changes in the free energy of binding for both the NTR and the CRD. Non-specific TALE binding energy includes contributions from both the NTR and the CRD. NTR binding is dominantly driven by the electrostatic interactions between the positively charged residues in the NTR and the negatively charged DNA backbone. Such electrostatic interactions are inhibited by competition with cations. As DNA preferentially interacts with divalent cations over monovalent cations, the effect of divalent cations on NTR binding is more pronounced. Similar to the NTR, non-specific contacts between the CRD and the DNA backbone are suppressed by competition with monovalent and divalent cations. Overall, the presence of Mg²⁺ or Ca²⁺ significantly attenuates TALE non-specific binding.

TALE binding to target sites is driven by non-electrostatic interactions, including hydrogen bonding and van der Waals interactions, between the RVDs in the CRD and nitrogenous bases of target DNA sequences. Despite the exclusion of ions during TALE binding, MD simulations suggest that some Mg²⁺ specifically bind to the major groove of DNA and interact with both the aspartic acid in HD and nitrogenous bases. Prior studies have also reported that Mg²⁺ can interact favorably with π-systems of nucleic acid bases through the major groove (68). The diresidue HD is the most common RVD occurring in the 21.5 TALE construct used in our work, and the HD RVD is known to be one of the highest affinity RVDs (27). We conjecture that due to the presence of bound Mg²⁺ ions, the interactions between HD residues in the RVD and the cytosine base will be stabilized, thereby enhancing overall specific binding energy. This finding explains the experimental observation that divalent cations enhance the specific binding of the 21.5 repeat TALE construct to the target DNA sequence. For the 11.5 repeat and 15.5 repeat TALEs, the stabilization of specific binding is less pronounced likely due to the reduced number of HD RVD sequences. Taken together, these observations explain the enhanced overall non-electrostatic interactions between the 21.5 repeat TALE and target DNA sequences.

Compared to Mg²⁺ or Ca²⁺, our results show that Mn²⁺ only moderately attenuates the non-specific TALE binding, while Sr²⁺ and Zn²⁺ have little effect. This trend appears to be correlated with the binding preferences of divalents cations to DNA bases relative to phosphates: Zn²⁺>Mn²⁺>Ca²⁺>Mg²⁺ (69). The pronounced attenuation of non-specific TALE binding in the presence of Mg²⁺ or Ca²⁺ could be attributed to their relative preferential binding to the DNA backbone. However, the overall effects of divalent cations on TALE specificity do not appear to be correlated with the strength of divalent cation interaction with DNA bases (Mn²⁺>Ca²⁺>Mg²⁺, Sr²⁺) (69) or the divalent cation Hofmeister series at the protein carboxylate sites (Zn²⁺>Ca²⁺>Sr²⁺ >Mg²⁺) (70,71).

Similar to other DNA-binding proteins, the binding of TALEs to target DNA is governed by a balance of electrostatic and non-electrostatic interactions, which are in turn influenced by the presence of divalent cations and total ionic strength. In the presence of small amounts of certain divalent cations (Mg²⁺ or Ca²⁺), the electrostatic interactions between the NTR or CRD and DNA backbone will be attenuated, and the non-electrostatic interactions between the RVDs and specific DNA sequences will be enhanced. In this way, the non-specific binding affinity generally decreases upon addition of divalent cations, as it is primarily driven by electrostatic interactions. On the other hand, changes in the specific binding affinity will depend on the length of the CRD in TALEs and the presence of HD RVD. For longer TALEs, our results suggest that attenuation of electrostatic binding interactions can be outweighed by enhanced non-electrostatic binding interactions in the presence of divalent cations, and therefore the specific binding affinity can increase or remain unchanged under these conditions. For shorter TALEs, the attenuated electrostatic binding interactions cannot be compensated by the enhanced non-electrostatic binding interactions, and therefore the specific binding affinity is impacted by divalent cations. These results are consistent with our observations that the effects of divalent cations are largest for the 21.5 repeat TALE, and generally decrease as the TALE CRD is shortened. We find that the 21.5 repeat TALE exhibits extremely high specificity under physiologically relevant salt concentrations, which suggests that physiological salt conditions may provide a near optimal balance between specific and non-specific TALE binding to DNA.

Finally, we propose a general model for the divalent cation effect on TALE specificity. Divalent cations strongly affect NTR binding, and the NTR contributes a significant proportion of the overall binding affinity for shorter TALEs than longer for TALEs. Meanwhile, divalent cations can stabilize some specific interactions between the CRD and DNA bases. Longer TALEs are able to overcome the reduction in non-specific binding affinity by their larger CRD regions. Thus, it holds that the effects of divalent cations can be thought of as a balancing act; they increase specificity by decreasing affinity for non-specific DNA. Our findings suggest that divalent cations in the cellular milieu are a critical component to enhance the binding specificity of TALEs and possibly other sequence-specific DNA-binding proteins. Overall, this work provides new insights into the fundamental nature of TALE–DNA binding process, which could potentially be useful for gene editing applications.

Notes

Present address: Zhanar Abil, Department of Biochemistry, University of Texas at Austin, Austin, TX 78712, USA.

SUPPLEMENTARY DATA

Supplementary Data are available at NAR Online.

FUNDING

Carl R. Woese Institute for Genomic Biology (to H.Z.); FMC Fellowship (to L.W.C.); 3M Corporate Fellowship (to C.Z.); David and Lucile Packard Foundation (to C.M.S.). Funding for open access charge: University of Illinois.

Conflict of interest statement. None declared.