DNA Sequence Specificity Reveals a Role of the HLTF HIRAN Domain in the Recognition of Trinucleotide Repeats

Abstract

DNA damage tolerance (DDT) pathways enable cells to cope with a variety of replication blocks that threaten their ability to complete DNA replication. Helicase-like transcription factor (HLTF) plays a central role in the error-free DDT pathway, template switching (TS), by serving as a ubiquitin ligase to polyubiquitinate the DNA sliding clamp PCNA, which promotes TS initiation. HLTF also serves as an ATP-dependent DNA translocase facilitating replication fork remodeling. The HIP116, Rad5p N-terminal (HIRAN) domain of HLTF specifically recognizes the unmodified 3′-end of single-stranded DNA (ssDNA) at stalled replication forks to promote fork regression. Several crystal structures of the HIRAN domain in complex with ssDNA have been reported; however, optimal ssDNA sequences for high-affinity binding with the domain have not been described. Here we elucidated DNA sequence preferences of HLTF HIRAN through systematic studies of its binding to ssDNA substrates using fluorescence polarization assays and a computational analysis of the ssDNA:HIRAN interaction. These studies reveal that the HLTF HIRAN domain preferentially recognizes a (T/C)TG sequence at the 3′-hydroxyl ssDNA end, which occurs in the CTG trinucleotide repeat (TNR) regions that are susceptible to expansion and deletion mutations identified in neuromuscular and neurodegenerative disorders. These findings support a role for HLTF in maintaining the stability of difficult to replicate TNR microsatellite regions.

Graphical Abstract

Replication of cellular DNA is impeded by various events that cause replication stress, including DNA lesions, nicks and gaps, repetitive sequences, transcription−replication conflicts, dNTP depletion, histone irregularities, and aberrant assembly of replication machinery.^1,2 These events may lead to stalled replication forks, replication fork collapse, formation of vulnerable single-stranded DNA (ssDNA), and detrimental DNA damage such as double-stranded DNA (dsDNA) breaks that threaten genome integrity and cell viability. If replication stress persists, these events can hinder the ability of the cell to complete DNA replication.

To rescue stalled forks, cells employ DNA damage tolerance (DDT) pathways that alleviate replication stress by enabling replicative bypass of a fork blockage without immediate repair.³⁻⁸ The error-prone DDT pathway, translesion synthesis (TLS),⁶⁻⁸ is initiated by monoubiquitination of the sliding clamp PCNA at K164 by the Rad6/Rad18 (E2/E3) enzyme pair, which promotes recruitment of low-fidelity TLS DNA polymerases to copy over aberrant DNA structures.⁹⁻¹¹ The error-free DDT pathway, template switching (TS), is triggered by subsequent K63 polyubiquitination of PCNA by Ubc13-Mms2 (E2) and the two orthologs of yeast Rad5, helicase-like transcription factor (HLTF) and SNF2 histone linker PHD-finger RING-finger helicase (SHPRH) (E3).¹²⁻¹⁵ In addition to their function as E3 ubiquitin ligases, HLTF and Rad5 act as ATP-dependent DNA translocases capable of replication fork reversal that results in Holliday junction intermediate (“chicken foot” structure) whereby the two daughter strands are annealed against one another.¹⁶⁻¹⁸ By contrast, the fork reversal activity of SHPRH has not been confirmed. The replication fork reversal during TS enables the replisome to switch templates to a homologous sister chromatid and copy over replication blocks in a high-fidelity manner.³⁻⁵

HLTF, SHPRH, and Rad5 belong to the SWI2/SNF2 family of ATP-dependent chromatin remodeling enzymes, featuring a core helicase-like domain defined by seven conserved sequence motifs.^19,20 Embedded in the core domain are the two insert regions (“major” and “minor”) harboring additional modular domains.¹⁹ The RING-finger domain embedded in the “major” insert region confers HLTF, SHPRH, and Rad5 with the E3 ubiquitin-ligase activity necessary for PCNA polyubiquitination,¹²⁻¹⁵ while the “minor” insert region of SHPRH also includes the plant homeodomain (PHD) and the linker histone H1 and H5 (H1.5) domains.^19,21 In addition, HLTF and Rad5, but not SHPRH, contain the HIP116, Rad5p N-terminal (HIRAN) domain responsible for recognition of DNA features at stalled replication forks, which is essential for replication fork reversal.²²⁻²⁷

Both HLTF and SHPRH are considered tumor suppressors because depleted expression of these genes is correlated with increased genomic instability.^3,15,28 Despite their functional overlap, cellular knockouts of HLTF, SHPRH, or both genes exhibit distinct phenotypes during genotoxic stress, suggesting each protein has specificity that is not yet completely understood.^29,30 The functional divergence of SHPRH and HLTF likely stems from their preferred chromatin substrates. SHPRH binds nucleosomes and dsDNA,³¹ but its PHD domain surprisingly lacks the ability to recognize lysine methylation of the histone H3 tail.²¹ HLTF interacts with the 3′-end of ssDNA at replication forks through its HIRAN domain (residues 58−174), which positions the core helicase-like domain for fork remodeling.²²⁻²⁷ Interestingly, HLTF HIRAN also binds the free 3′-OH ends of duplexed DNA, highlighting HIRAN’s essential role in identifying the 3′-end of DNA at a stalled replication fork and displacing the nascent strand from template DNA during TS.²⁵

Biochemical and structural studies revealed that the HIRAN domain adopts a modified oligonucleotide/oligosaccharide fold that clamps around the first two 3′-terminal nucleotides of ssDNA and forms a binding pocket for the free 3′-hydroxyl group. The third nucleotide sits outside of this binding pocket but also forms stabilizing interactions with the domain.²³⁻²⁵ Further clarification of the optimal ssDNA sequence for HIRAN binding will lead to a better understanding of HLTF specificity and shed new light on its function(s) in TS. In this study, we sought to elucidate the DNA sequence preference of the HLTF HIRAN domain trough systematic analysis of its binding to ssDNA substrates using biochemical fluorescence polarization (FP) assays and extensive computational studies of the ssDNA:HIRAN interaction. This analysis reveals that the HIRAN domain has a higher affinity for ssDNA substrates with a 3′-purine (G preferred) following a pyrimidine (T preferred), which is consistent with an established role of HLTF in suppressing expansion and deletion mutations of CAG/CTG trinucleotide repeats³²⁻³⁵ and suggestive of a broader significance of HLTF for maintaining difficult to replicate microsatellite regions and preventing genetic instability.

■ MATERIALS AND METHODS

HIRAN Domain Expression and Purification.

BL21-(DE3) Escherichia coli cells were transformed with a pET28a-LIC-based plasmid containing the coding sequence for a His-tagged HLTF-HIRAN domain (residues 51−171) (Addgene plasmid 28152).²⁶ Cells were grown in LB medium at 37 °C until they reached an OD₆₀₀ of 0.8−1.0. Protein expression was induced by 1 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) at 20 °C overnight. Cells were harvested by centrifugation, resuspended in 10 mM imidazole, 20 mM NaHPO₄, 300 mM NaCl, and 6 M guanidine hydrochloride buffer, and lysed by sonication. The cell lysate was mixed with a Co²⁺ resin and extensively washed with the lysis buffer containing 6 M guanidine hydrochloride. Protein was refolded on the column by washing in a lysis buffer without guanidine hydrochloride. Protein was further purified by size exclusion chromatography on a Superdex 75 column in 20 mM sodium phosphate (pH 6.8), 100 mM NaCl, and 2 mM dithiothreitol.

Fluorescence Polarization Assay Protocols.

All FP measurements were performed in triplicate in PerkinElmer ProxiPlate 384 F Plus black low-volume 384-well polystyrene plates. For the displacement assay, 5 μL of recombinant HIRAN protein at a concentration of 4 μM and 5 μL of 5′FAM-T₅ at a concentration of 1 μM were mixed in each well and complexed by being shaken for 15 min at room temperature. Nine concentrations of each oligonucleotide (final concentrations of 0.05−200 μM) were utilized to generate the IC₅₀ value. Ten microliters of the nucleotide was added to the complex (final volume of 20 μL), and the complex and nucleotides were mixed by being shaken for 30 min at room temperature. Twenty microliters of 5′-FAM-T₅ ssDNA at a concentration of 1 μM, phosphate-buffered saline (PBS), and 20 μL of HLTF protein at 4 μM were used as controls. The plates were read in a BioTek Synergy H1 hybrid reader with a filter cube with an excitation wavelength of 485 ± 20 nm and an emission wavelength of 528 ± 20 nm to ascertain the FP signal in millipolarization (mP) units using Gen5 software (version 1.11.5). Millipolarization data were analyzed using GraphPad Prism 5 (version 5.01) to ascertain the IC₅₀ for each unlabeled oligonucleotide evaluated. IC₅₀ values reported in Tables 1 and 2 are averages of at least three separate experiments performed in triplicate.

Table 1.

Binding Affinities of the HIRAN Domain for ssDNA Oligonucleotides

oligonucleotide	Kd (μM)a
5′-FAM-T5	0.14 ± 0.03
5′-FAM-A5	0.16 ± 0.04
5′-FAM-TATAT	0.25 ± 0.1
5′-FAM-ATATA	0.17 ± 0.06
T5-3′-FAM	>100
5′-FAM-T5-3′-PO4 (0.1 μM)	>100

^aK_d values represent the average ± the standard error of the mean at least three separate experiments performed in triplicate. Oligonucleotide concentrations in all binding assays were 0.1 μM.

Table 2.

ssDNA:HIRAN Interactions Evaluated by FP Displacement and Binding Assays

oligonucleotide	group(s)	IC50 (μM)a	Kd (μM)a
T5		2.4 ± 0.6	0.14 ± 0.05
TG	1	31.7 ± 3.4	−
AG	1	>100	−
CG	1	32.6 ± 7.9	−
TT	1	>100	−
TA	1	>100	−
TC	1	>100	−
CC	1	>100	−
CT	1	>100	−
TTT	2−4	4.3 ± 0.9	−
GGG	2	20.2 ± 5.1	−
CCC	2	>100	−
AAA	2	>100	−
TTG	3 and 5	2.8 ± 0.5	0.03 ± 0.01
TTA	3	4.3 ± 0.4	−
TTC	3	13.5 ± 0.6	−
TCT	4	6.8 ± 0.7	−
TGT	4	15.5 ± 0.9	−
TAT	4	34.9 ± 9.2	−
ATG	5	2.9 ± 0.4	0.09 ± 0.01
CTG	5	2.5 ± 0.2	0.06 ± 0.01
GTG	5	2.2 ± 0.07	0.18 ± 0.01
CAG	5	31.7 ± 3.9	−
TTTG	6	2.0 ± 0.2	0.10 ± 0.02
ATTG	6	1.8 ± 0.2	−
CTTG	6	1.6 ± 0.2	−
GTTG	6	1.7 ± 0.1	0.22 ± 0.09
GCC	7	31.8 ± 8.5	−
TCC	7	17.1 ± 4.8	0.21 ± 0.07
CTT	7	10.0 ± 1.5	−
CCT	7	6.9 ± 0.07	−
TGC	7	>100	−
GGT	7	5.4 ± 0.2	−
TCG	7	2.9 ± 0.3	−
CCG	7	3.6 ± 0.2	−

^aValues represent the averages ± the standard error of the mean of at least three separate experiments performed in triplicate.

FP binding affinity measurements were performed in triplicate in PerkinElmer ProxiPlate 384 F Plus black low-volume 384-well polystyrene plates. Nine concentrations of HIRAN were made with a concentration range of 0.005−10 μM (final assay concentrations of 0.0025−5 μM) or 0.05−30 μM (final assay concentrations of 0.025−15 μM). For each FAM-labeled ssDNA oligonucleotide, a stock solution at 0.2 μM was utilized. To each well were added 10 μL of the FAM-labeled ssDNA and 10 μL of the varying concentrations of the HIRAN domain, and solutions were mixed by being shaken for 30 min at room temperature. Controls used for this assay were 20 μL of 0.2 μM FAM-labeled ssDNA alone and 20 μL of HIRAN domain at a concentration of 10 or 30 μM alone. The plates were read in the same manner as described above to determine the mP signal value. The FP data were analyzed in GraphPad Prism 5 (version 5.01) using their standard one-site total binding nonlinear regression equation (eq 1) to determine the K_d values for each FAM-labeled oligonucleotide.

$$Y=B_{\max }\times X/\left(K_{\text{d}}+X\right)+\text{NS}\times X+\text{background}$$ (1)

Structure Preparation and Molecular Dynamics Simulations.

The crystal structures of the HLTF HIRAN domain complexed with ssDNA (ACCGCCGGGTGCC, PDB entry 4XZF;²⁵ TTTTTT, PDB entry 4S0N;²³ GTG, PDB entry 5BNH;²⁴ TTG, PDB entry 5BNH²⁴) as well as the apo HIRAN domain (PDB entry 4XZG)²⁵ were downloaded from the Protein Data Bank (PDB) and used for the molecular dynamics (MD) studies. All water molecules and other ions were removed from the complexes. The missing amino acid side chains and ssDNA nucleotides were restored and refined by using the OPLS3e force field in the Protein Preparation Wizard module of Schrodinger suite 2020. For the five additional trinucleotides studied, each ssDNA nucleotide was built and optimized using the LigPrep panel of Schrödinger maestro 12.0. Docking was performed in XP mode in Glide.³⁶ The nucleotide conformations that provided the highest docking scores and optimal root-mean-square deviation (RMSD) when superimposed with the ssDNA from the crystal structures were utilized as the initial structures for the MD simulations described below. MD simulations were performed with the AMBER 18 suite.³⁷ Amber14SB³⁸ and DNA.OL15³⁹ force fields were used for parametrization of the HIRAN domain and ssDNA from the complexes, respectively. Simulations employed periodic boundary conditions, and electrostatic interactions were treated using the particle-mesh Ewald algorithm⁴⁰ with a real space cutoff of 9 Å. Lennard−Jones interactions were truncated at 9 Å. A pair list was built with a buffer region, and a list update was triggered whenever a particle moved by >0.5 Å with respect to the previous update. The initial conformations of the ssDNA:HIRAN complexes were solvated with SPC/E water molecules. Periodic boundary conditions were imposed using a truncated octahedral box, ensuring a solvent shell of ≥8 Å around the solute. The solute was neutralized with sodium ions, and sufficient Na⁺/Cl⁻ ion pairs were added to reach a concentration of 150 mM. The ions were initially placed at random intervals, but at least 5 Å from ssDNA and 3.5 Å from one another.

Each system was initially subjected to energy minimization with harmonic restraints of 25 kcal mol⁻¹ Å⁻² on the solute atoms. The system was then heated to 298 K at a constant volume during 100 ps. Constraints were then relaxed from 5 to 1 kcal mol⁻¹ Å⁻² during a series of 1000 steps of energy minimization (500 steps of steepest descent and 500 steps of conjugate gradient) followed by 50 ps of equilibration with restraints of 0.5 mol⁻¹ Å⁻² and 50 ps without solute restraints. The 100 ns production simulations were carried out at a constant temperature (298 K) and a constant pressure (1 bar) with a 2 fs time step. During these simulations, the pressure and temperature were maintained using the Berendsen algorithm with a coupling constant of 5 ps and SHAKE constraints were applied to all bonds involving hydrogens. Conformational snapshots were saved every picosecond for further analysis. For the purpose of comparison, the free form of the HIRAN domain was also simulated using an identical protocol, creating a second set of 100 ns trajectories.

Conformational and Environmental Analysis.

The average DNA conformation, DNA conformational fluctuations, and ion distributions around the ssDNA:HIRAN complexes during the MD simulations were analyzed with the cpptraj module of AMBER18.⁴¹ Hydrogen bond and salt bridges were analyzed using AMBER Tools. A distance cutoff of ≤3.5 Å between the relevant heavy atoms and an angle cutoff of ≥135° at the intervening hydrogen atom were used. These interactions were characterized by the percentage of the trajectory during which they are observed (% presence) and by their average lifetimes, which were calculated by ignoring interruptions in the interaction that last <1 ps. Indirect interactions mediated by water molecules were excluded from the analysis. All of the hydrogen bonds and their occupancy values were plotted using the MDplot library in R programming language.

Binding Free Energy and Per-Residue Interaction Calculations.

The binding free energies were calculated for the ssDNA:HIRAN complexes using the molecular mechanics/Poisson−Boltzmann surface area (MM/PBSA) approach developed previously.^42,43 Using the MM/PBSA method (eqs 2 and 3), binding free energies, ΔG_binding, were calculated as the sum of gas-phase energies ΔH^gas (including the energy of electrostatic interactions, ΔE_ele, and van der Waals energy, ΔE_vdw, as determined by a Lennard-Jones potential), solvation free energies ΔG^solvtion [including a nonpolar part, ΔG_SA, determined by eq 4 and an electrostatic part, ΔG_GB, obtained from the generalized Born (GB) calculations⁴⁴], and entropic terms TΔS, including translational, rotational, and vibrational contributions (estimated using normal mode analysis):

$$\begin{gathered} \text{Δ}{G}_{\text{binding}}=\text{Δ}{H}^{\text{gas}}+\text{Δ}{G}^{\text{solvation}}-T\text{Δ}S=\text{Δ}{E}_{\text{vdw}}+\text{Δ}{E}_{\text{ele}} \\ +\text{Δ}{G}_{\text{GB}}+\text{Δ}{G}_{\text{SA}}-T\text{Δ}S \end{gathered}$$ (2)

All energy terms in eq 1 were calculated as

$$\text{Δ}G=\langle G_{\text{complex}}\rangle -\langle G_{\text{protein}}\rangle -\langle G_{\text{ssDNA}}\rangle$$ (3)

where $\langle G_{\text{x}}\rangle$ are absolute free energies averaged over snapshots taken from MD trajectories.

The polar contribution (ΔG_GB) was calculated by solving the GB approximation;⁴⁴ the charges used in GB calculations were taken from the AMBER parameter set. The nonpolar contribution was determined on the basis of the solvent-accessible surface area (SASA) determined with the LCPO method (eq 4).⁴⁵

$$\text{Δ}{G}_{\text{SA}}=0.0072\left(\frac{\text{kcal}}{\text{mol}{\text{Å}}^2}\right)\text{ΔSASA}\left({\text{Å}}^2\right)$$ (4)

The entropy (TΔS) calculations were performed using normal mode analysis (NMA) in the gas phase. The distance-dependent dielectric constant was set to 1.0, and the convergence criteria for the minimized energy gradient were set at 0.0001 kcal mol⁻¹ Å⁻¹ with 10 000 minimization cycles per snapshot. Due to limited computational resources, we used only 50 frames, which were evenly extracted from 0 to 5 ns of the MD trajectories, for entropy calculations.

Per-residue and per-nucleotide decomposition analysis was performed to obtain the energetic contributions of individual HIRAN residues and ssDNA nucleotides to the stability of the ssDNA:HIRAN complexes. The binding energy contribution, ΔG_i, of nucleotide or amino acid i includes three terms: van der Waals (ΔG_vdw,i) and electrostatic (ΔG_ele,i) contributions computed using the sander program in AMBER14.0 and a solvation contribution $\left(\text{Δ}{G}_{\text{solvation,i}}=\text{Δ}{G}_{\text{GB},\text{i}}+\text{Δ}{G}_{\text{SA},\text{i}}\right)$

$$\text{Δ}{G}_{\text{i}}=\text{Δ}{G}_{\text{vdw},\text{i}}+\text{Δ}{G}_{\text{ele,i}}+\text{Δ}{G}_{\text{GB},\text{i}}+\text{Δ}{G}_{\text{SA},\text{i}}$$ (5)

The solvation contributions ΔG_GB,i and ΔG_SA,i were computed using the approach described above. The per-residue and per-nucleotide binding energy calculations used a total of 160 snapshots from the MD trajectories (20−100 ns) extracted evenly every 500 ps.

RESULTS

Initial Analysis of ssDNA:HIRAN Interactions.

Several structures of the HLTF HIRAN domain have been recently reported, including nuclear magnetic resonance (PDB entry 5KF5²⁶) and X-ray crystal (PDB entry 4XZG²⁵) structures of the apo domain, as well as multiple co-crystal structures of HLTF HIRAN with distinct DNA oligonucleotides (PDB entries 4S0N,²³ 5BNH,²⁴ and 4XZF²⁵). We utilized each of these structures to probe the specific intermolecular interactions that govern high-affinity binding of the HIRAN domain and the 3′-end of ssDNA. Specifically, we analyzed HLTF HIRAN complexes with the following 3′-ssDNA fragments: ACCGCCGGGTGCC (PDB entry 4XZF, 1.38 Å resolution²⁵), TTTTTT (PDB entry 4S0N, 1.50 Ų³), GTG:HIRAN (PDB entry 5BNH, 1.7 Å resolution²⁴), and TTG:HIRAN (PDB entry 5BNH, 1.7 Å resolution²⁴). Note that the 13-nucleotide ssDNA fragment in 4XZF²⁵ self-anneals and forms a partial duplex DNA with the two copies of HIRAN bound to short 3′-ssDNA overhangs on both sides of the duplex; however, for our computational studies, we used a single copy of HIRAN bound to the 3′-ssDNA fragment derived from this structure. In 4S0N,²³ the HIRAN domain was crystallized with the (dT)₁₀ ssDNA; however, because of the missing electron density for the four 5′-nucleotides, here we used the structure of HIRAN bound to (dT)₆ ssDNA. In 5BNH,²⁴ the two copies of HLTF HIRAN in an asymmetric unit cell were bound to GTG and CTG trinucleotides, and our computational studies utilized both of these complexes.

Figure 1 shows the superposition of the four HLTF HIRAN structures bound to distinct ssDNA fragments.²³⁻²⁵ The HIRAN domain, which is comprised of 125 amino acids (residues 55−180 of HLTF), adopts a modified OB fold featuring a β-barrel formed by six β-sheets (β₁−β₆) and two α-helices (α₁ and α₂). The surface of the α₁ helix and the loops between β₁ and β₂ and between β₂ and β₃ strands form a binding pocket for the two nucleotides at the 3′-end of ssDNA. The structural alignment reveals that the two 3′-nucleotides of all ssDNA fragments (hereafter termed NT₁ and NT₂) penetrate inside the binding pocket on the HIRAN domain and adopt a similar well-defined conformation, while the subsequent nucleotide NT₃ interacts with the domain outside of the binding pocket and exhibits more conformational variability (Figure 1). Several key intermolecular interactions between ssDNA and HIRAN are shared across each structure and contribute to high-affinity interaction. First, the 3′-hydroxyl moiety of ssDNA forms a hydrogen bond with the side chain carboxyl of D94 (Figure 1). The NT₁ and NT₂ nucleobases are stacked between the aromatic side chains of Y72 and Y93. The NT₁ and NT₂ phosphates form electrostatic interactions with R71 and K113 and a hydrogen bond with the side chain of Y72, while N91 forms a hydrogen bond with the NT₁ base. Point mutations of all of the residues mentioned above significantly weakened or abrogated ssDNA:HIRAN interaction in electrophoretic mobility shift assays (EMSAs),^23,25 while R71E, Y92A/Y93A,²³ and N90A/N91A²⁵ mutations also rendered HLTF defective in replication fork reversal. Remarkably, comparison of HLTF HIRAN structures bound to the four distinct nucleotides also revealed differences in key intermolecular interactions responsible for stabilization of each complex (Figure 1C and Figure S1). For example, the presence of a purine nucleobase at NT₁ (guanine) improves the affinity and stability of the ssDNA inside the binding site of the HIRAN protein (PDB entry 5BNH²⁴). The bicyclic structure of the purine nucleobase stabilizes a conformation of NT₁ and NT₂ that provides a network of inter- and intramolecular π−π stacking interaction with the side chain phenol of Y93 and the aromatic ring of the NT₂ nucleobase, respectively. These π−π stacking interactions orient the NT₂ nucleobase (thymine) such that an additional hydrogen bond is formed between the nitrogen in the nucleobase and the side chain phenol of Y72, which is not present when NT₁ contains a pyrimidine (thymine) nucleobase (PDB entry 4S0N²³). This difference suggests that HLTF HIRAN may preferentially recognize specific ssDNA sequences that we investigate below through systematic binding studies using an FP assay and computational analysis of ssDNA:HIRAN interaction.

Figure 1.

Structures of ssDNA complexed with the HLTF HIRAN domain. (A) Aligned structures of the HIRAN domain in complex with TTT (4S0N, deep teal), GCC (4XZF, blue), GTG (5BNH, magenta), and TTG (5BNH, orange) fragments at 3′-ssDNA. The ssDNA binding site on the HIRAN domain (total surface area of 1060 Ų in 5BNH) is colored pink. (B) Three-dimensional structure of the TTT fragment (4S0N) in the binding site on the HIRAN domain. (C) Key intermolecular interactions between the 3′-end of the TTT ssDNA and the HIRAN domain (from 4S0N).

Fluorescence Polarization (FP) Assay Development.

To explore ssDNA sequence preferences of the HIRAN domain, we sought to develop a rapid, robust binding assay that that could be utilized to probe the ssDNA:HIRAN interaction. For our assay, we focused on an FP protocol in which a fluorescein amidite label (FAM) was appended to the 5′-hydroxyl of the ssDNA oligonucleotide. The high affinity reported previously for the binding of HIRAN to the (dT)₁₀ oligonucleotide (K_d = 0.013 μM) and a crystal structure of the HIRAN:(dT)₁₀ complex (PDB entry 5BNH) in which only a few 3′-nucleotides interact with the domain, while the electron density of the four 5′-nucleotides is missing,²³ informed our decision to initially evaluate the binding of HIRAN to a 5′FAM-labeled (dT)₅ oligonucleotide (5′-FAM-T₅). Titration of increasing concentrations of the HIRAN domain into 0.1 μM 5′-FAM-T₅ resulted in a sensitive sigmoidal FP curve indicative of binding (Table 1 and Figure S2). The data analysis yielded a K_d of 0.14 ± 0.03 μM for the 5′-FAM-T₅:HIRAN complex under these assay conditions, while the use of higher 5′-FAMT₅ concentrations (0.5 and 1.0 μM) resulted in a weaker affinity (data not shown). We also used the FP assay to probe the interactions of HLTF HIRAN with several more FAM-labeled 3′-OH pentanucleotides, which demonstrated comparable binding affinities for the HIRAN domain under our assay conditions (Table 1). As a negative control, we evaluated the two (dT)₅ oligonucleotides with the 3′-end modified with either a FAM label (T₅-3′-FAM) or phosphate (5′-FAM-T₅-3′PO₄) that, in agreement with previous results,²³⁻²⁵ exhibited no detectable binding to HLTF HIRAN, validating our FP approach as a suitable strategy for evaluating ssDNA:HIRAN interaction.

Next, we evaluated the ability of unlabeled (dT)₅ to displace 5′-FAM-T₅ from the HIRAN domain across a range of concentrations of both the oligonucleotide and the protein. For the majority of parameters evaluated, a robust, concentration-dependent displacement of 5′-FAM-T₅ from the HIRAN domain was observed. The most reproducible assay and the largest change in FP (mP values) were achieved with 1 μM HIRAN, 0.25 μM 5′-FAM-T₅, and a 30 min incubation time. Under these assay conditions, unlabeled (dT)₅ disrupted the ssDNA:HIRAN interaction with an IC₅₀ of 2.4 ± 0.6 μM (Figure 2). Note that the close correspondence between IC₅₀ and K_d is expected only when [L] ≪ K_d, where [L] is the concentration of the labeled ligand,⁴⁶ which is not the case under our assay conditions. While other assay conditions utilized during the assay optimization provided an IC₅₀ that more closely correlated with the K_d we initially calculated for 5′-FAM-T₅, these conditions did not result in the most robust and reproducible assay.

Figure 2.

Disruption of the 5′-FAM-T₅:HIRAN interaction by unlabeled T₅ in the FP displacement assay. Data on this graph are from a single representative experiment performed in triplicate.

DNA Sequence Preferences of HLTF HIRAN Evaluated by FP Assays.

Having optimized the FP binding and displacement assay protocols, we sought to probe the DNA sequence requirements for the ssDNA:HIRAN binding by using these assays to systematically evaluate various short oligonucleotides (Table 2). First, we tested whether single nucleotides (A, T, G, or C) could displace the 5′-FAM-T₅ from the HIRAN domain using the FP displacement assay. Not surprisingly, none of single nucleotides was capable of disrupting the ssDNA:HIRAN interaction (Figure S3), which is consistent with the structural basis of ssDNA:HIRAN complex formation. ²³⁻²⁵ Then, we used the FP displacement assay to evaluate 3′-OH dinucleotides, including TX (X = A, T, G, or C) and several other sequences (Table 2, group 1, and Figure.S4). Among the considered nucleotide pairs, only TG and CG significantly disrupted the ssDNA:HIRAN interaction at concentrations below 100 μM (IC₅₀ values of 31.7 ± 3.4 and 32.6 ± 7.9 μM, respectively). Interestingly, a short oligonucleotide ending with TG was co-purified with HLTF HIRAN expressed in bacterial cells in one of the structural studies of the domain,²⁴ suggesting this might be one of HIRAN’s preferentially recognized sequences. Overall, our analysis of dinucleotide sequences suggests that longer ssDNA fragments are needed to effectively disrupt the ssDNA:HIRAN interaction.

Our exploration of trinucleotides with the FP displacement assay initially focused on homogeneous sequences (TTT, GGG, CCC, and AAA) (Table 2, group 2), of which the TTT and GGG trinucleotides demonstrated disruption of the ssDNA:HIRAN interaction (IC₅₀ values of 4.3 ± 0.9 and 20.2 ± 5.5 μM, respectively). This observation is consistent with our results for dinucleotides, suggesting that XG might be the optimal 3′-sequence for HIRAN binding. By contrast, both CCC and AAA trinucleotides demonstrated no disruption of ssDNA:HIRAN interaction at concentrations below 100 μM. The loss of activity for the CCC result was unexpected, as it is so structurally similar to the related pyrimidine trinucleotide TTT, which yielded robust disruption of the complex.

Inspired by differential activities displayed by dinucleotides and homotrinucleotides in our FP displacement assay, we further explored the nucleotide preferences of the HIRAN domain at NT₁, NT₂, and NT₃ positions of ssDNA by evaluating a series of TTX, TXT, and XTG trinucleotides substituted with X = A, T, G, or C by the FP displacement assay (Table 2, groups 3−5, and Figure S5). As expected, these studies reveal a strong preference for G at the NT₁ position (Table 2, group 3) with the TTG trinucleotide displacing (dT)₅ ssDNA from its complex with HIRAN with an IC₅₀ of 2.8 ± 0.5 μM, which was more active than the TTA, TTT, and TTC trinucleotides. Studies of TTT, TCT, TAT, and TGT trinucleotides to probe the preferred nucleotide at the NT₂ position (Table 2, group 4) revealed that replacing the T with another pyrimidine (C) resulted in a slight decrease in activity, while incorporating either bicyclic purine (G or A) in this location reduced activity by severalfold to an order of magnitude. Finally, appending each individual nucleotide to the 5′-end of TG (Table 2, group 5) revealed that each of the resulting trinucleotides significantly disrupted the ssDNA:HIRAN interaction with IC₅₀ values of 2.2−2.9 μM that are similar to the IC₅₀ value of unlabeled (dT)₅. Finally, we evaluated trinucleotide CAG, which is a common trinucleotide repeat expanded in Huntington’s disease.⁴⁷ As expected, CAG (IC₅₀ = 31.7 ± 3.9 μM) was significantly less active than the CTG trinucleotide repeat, consistent with a preference for thymine in the NT₂ position. Overall, the results of our FP displacement assay for trinucleotides suggest that the HLTF HIRAN domain preferentially recognizes the XTG-3′-OH sequence.

The next series of oligonucleotides that we evaluated in the FP displacement assay substituted X = A, T, G, or C in the XTTG sequence to determine whether the fourth NT₄ nucleotide improved the displacement activity (Table 2, group 6, and Figure S6). Each of these oligonucleotides disrupted the ssDNA:HIRAN interaction with IC₅₀ values of 1.6−2.0 μM comparable to TTG, suggesting that the NT₄ nucleotide has a minimal effect on ssDNA:HIRAN interaction.

Finally, to further establish structure−activity relationships (SARs) and validate our computational results (see below), we evaluated several trinucleotides based on TTT and CTG in which one or two cytosines were incorporated in place of the related pyrimidine thymidine (Table 2, group 7). Trinucleotides TCC, CTT, and CCT all demonstrated reduced activity compared to that of TTT, regardless of which thymidine was substituted, while both TCG and CCG demonstrated activity comparable to that of TTG in the displacement assay. Taken together, these results further highlight the enhanced activity of trinucleotides containing guanosine at NT₁ followed by a pyrimidine at the NT₂ position.

Disruption of the (dT)₅:HIRAN complex with oligonucleotides displaying IC₅₀ values in the low micromolar range (<5 μM) is at the detection limit of the FP displacement assay. Therefore, to cross-validate our results, we performed FP binding studies of HLTF HIRAN with 5′-FAM-labeled analogues of several trinucleotides that were most potent in the FP displacement assay (Table 2, column 4, and Figure S7). Not surprisingly, the XTG (X = A, T, G, or C) trinucleotides exhibited high binding affinities for the HIRAN domain (K_d = 0.03−0.18 μM). While we were not able to establish a clear nucleotide preference at the NT₃ position based on IC₅₀ values from the FP displacement assay, in the FP binding studies 5′FAM-labeled TTG and CTG demonstrated affinities for HIRAN significantly higher than those of GTG and ATG (Table 2, column 4), suggesting a preference for pyrimidine at the NT₃ position. As a control, we also performed the FP binding assay for 5′-FAM-TCC, which demonstrated a weaker binding affinity for the HIRAN domain consistent with the decreased activity of this trinucleotide in the FP displacement assay. Finally, we measured K_d values of HIRAN for 5′-FAMTTTG and 5′-FAM-GTTG oligonucleotides, which surprisingly were weaker than K_d values obtained for the corresponding trinucleotides (Table 2, column 4, and Figure S8). Taken together, our FP displacement and binding studies established (T/C)TG-3′-OH as the ssDNA sequence preferentially recognized by the HLTF HIRAN domain.

Molecular Dynamics Simulations.

To rationalize the results of our FP assays (Table 2) and provide the atomic level description of intermolecular interactions that stabilize the ssDNA:HIRAN complexes, we performed MD simulations of the HIRAN bond to various ssDNA fragments. Initial series of MD trajectories (100 ns, 298 K) included simulations of the four ssDNA:HIRAN structures determined by X-ray crystal-lography (ACCGCCGGGTGCC:HIRAN, PDB entry 4XZF;²⁵ TTTTTT:HIRAN, PDB entry 4S0N;²³ GTG:HIRAN, PDB entry 5BNH;²⁴ TTG:HIRAN, PDB entry 5BNH²⁴), as well as the structure of the apo HIRAN domain (PDB entry 4XZG²⁵). These MD trajectories were subsequently used to assess energetics of ssDNA:HIRAN interaction and probe specific intermolecular contacts that confer binding affinity and specificity. To validate our computational protocol for the prediction of binding energies (see below), we also modeled the structures of HLTF HIRAN complexes with the five additional 3′-OH trinucleotides (ATG, CTG, TCC, TAT, and GGG) and performed MD simulations for these complexes (100 ns, 298 K) with the protocol similar to that used for experimental structures. The backbone atoms of the HIRAN domain in each trajectory (1.76−2.75 Å) indicate that the overall structure of the protein was well-maintained throughout the MD simulations (Table 3 and Figures S9−S13). By contrast, ssDNA from each of the complexes demonstrated a range of fluctuations, with a RMSD of >6 Å obtained for 13-nucleotide ssDNA for 4XZF, which is significantly higher than the RMSD of 2.1−3.1 Å obtained for hexa- and trinucleotides for other structures (Table 3). For this reason, ssDNA in the 4XZF trajectory did not reach equilibrium until 40 ns, while the shorter oligonucleotides equilibrated after MD simulation for ~20 ns. However, the 3′-nucleotides NT₁−NT₃ maintained interactions with the HIRAN domain through all of the MD trajectories, as indicated by RMSD values of 1.44−3.02 Å for the NT₁−NT₃ ssDNA region (Table 3). Finally, the comparison of isotropic temperature factors (B-factors) calculated from each MD trajectory with the crystallographic B-factors²³⁻²⁵ (Figures S9−S13) suggests that our MD trajectories correctly captured rigid and flexible regions of the HIRAN domain, providing us with confidence that the MD-generated conformational ensembles can be utilized to further dissect the details of ssDNA:HIRAN interaction.

Table 3.

RMSD Values for the MD Simulations on the Four ssDNA:HIRAN Complexes^a

structure	HIRAN domain	ssDNA NT1−NT3 (full-length ssDNA)b
4XZG	1.83 ± 0.2	−
4XZF	2.04 ± 0.2	2.57 ± 0.6 (6.01 ± 1.2)
4S0N	2.45 ± 0.3	1.44 ± 0.3 (2.7 ± 0.7)
5BNH/GTG	1.98 ± 0.1	1.65 ± 0.6
5BNH/TTG	1.76 ± 0.2	2.24 ± 0.5
ATG	2.02 ± 0.2	1.17 ± 0.5
CTG	2.11 ± 0.3	1.07 ± 0.2
GGG	1.89 ± 0.4	3.02 ± 0.4
TAT	2.75 ± 0.5	2.69 ± 0.4
TCC	2.07 ± 0.4	1.45 ± 0.2

^aAll values represent average RMSD ± SD in angstroms.

^bRMSD values for NT₁−NT₃ of the ssDNA in each complex. RMSD values in parentheses are for the full-length ssDNA oligonucleotide when the ssDNA in the complex is longer than three nucleotides.

Analysis of ssDNA:HIRAN Binding Energies from MD Trajectories.

Next, we used conformational ensembles generated in MD simulations of the four experimental ssDNA:HIRAN structures²³⁻²⁵ to perform the binding free energy analysis using the molecular mechanics/Poisson−Boltzmann surface area (MM/PBSA) approach,⁴⁸ which has been successfully used to rationalize affinities of protein interactions with small molecule ligands.⁴⁹ In this method, the total binding energy is divided into the change in total solvation energy ΔG_SA + ΔG_GB (where ΔG_SA and ΔG_GB are nonpolar and polar contributions, respectively), the changes in van der Waals and electrostatic energies (ΔE_vdw + ΔE_ele), and the entropic contribution (−TΔS). The calculated binding free energy of the ssDNA:HIRAN interaction ranged from −24.2 kcal/mol for the complex with the TTG trinucleotide in 5BNH (the strongest interaction)²⁴ to −5.7 kcal/mol for the 13-nucleotide ssDNA fragment ending with GCC in 4XZF²⁵ (Table 4, top). The total predicted binding free energy for all complexes included large mutually compensating changes in polar solvation energy ΔG_GB and electrostatic energy ΔE_ele (Table 4, top), reflecting the loss of favorable interactions of protein and DNA polar groups with the solvent compensated by the formation of favorable electrostatic interactions in the protein:DNA complex. In addition to electrostatic interactions, the ssDNA:HIRAN complexes are stabilized by the formation of favorable van der Waals interactions between protein and DNA groups [large negative ΔE_vdw (Table 4)]. The higher (less negative) predicted total binding free energies ΔG for ssDNA:HIRAN complexes correlated with the higher IC₅₀ values (indicative of weaker binding) obtained for the corresponding 3′-OH trinucleotides in the FP displacement assay (Table 2). A good correlation was also observed between the experimental IC₅₀ values and the predicted binding enthalpy (ΔH, a sum of ΔG_SA, ΔG_GB, ΔE_vdw, and ΔE_ele terms), suggesting that the total free energy (ΔG) and enthalpy (ΔH) changes can be used as predictors of the affinity of the ssDNA:HIRAN interaction. As only the first two nucleotides of ssDNA are well-organized inside the binding pocket on the HINAN domain and account for the majority of intermolecular interactions,²³⁻²⁵ the predicted enthalpic and entropic contributions to the binding free energy exhibit little correlation with the length of ssDNA fragments, albeit the entropy term is somewhat more unfavorable for the complexes containing longer oligonucleotides (PDB entries 4S0N and 4XZF) (Table 4).

Table 4.

Total Binding Free Energies and Decompositions for the ssDNA:HIRAN Complexes

ssDNA	ΔGa	ΔGGBb	ΔGSAb	ΔEvdwc	ΔEelec	ΔHd	TΔSd
ACCGCCGGGTGCC (4XZF)	−5.7	897 ± 13	−6.3 ± 0.2	−44.8 ± 5.5	−890 ± 16	−43.9 ± 11	−38.2 ± 9.7
TTTTTT (4S0N)	−10.9	670 ± 25	−7.7 ± 0.4	−57.7 ± 4.9	−655 ± 27	−49.6 ± 7.0	−38.7 ± 7.8
GTG (5BNH)	−21.2	326 ± 25	−6.8 ± 0.3	−59.8 ± 6.3	−311 ± 25	−52.0 ± 7.0	−30.9 ± 4.1
TTG (5BNH)	−24.2	341 ± 22	−7.7 ± 0.3	−64.5 ± 22	−326 ± 26	−57.5 ± 7.8	−33.3 ± 3.8
TAT	−4.1	353 ± 38	−6.7 ± 0.6	−54.4 ± 6.5	−340 ± 45	−47.8 ± 10	−31.5 ± 4.1
TCC	−5.9	353 ± 21	−7.3 ± 0.3	−58.6 ± 4.3	−339 ± 25	−51.7 ± 7.6	−33.2 ± 5.8
GGG	−6.8	363 ± 30	−6.3 ± 0.4	−51.7 ± 4.5	−347 ± 34	−40.9 ± 7.3	−32.0 ± 6.8
CTG	−9.8	385 ± 32	−7.2 ± 0.6	−57.7 ± 5.6	−372 ± 34	−52.0 ± 5.8	−32.8 ± 9.7
ATG	−15.0	340 ± 21	−6.6 ± 0.3	−54.7 ± 4.3	−332 ± 21	−54.1 ± 7.5	−30.3 ± 4.9

^aΔG is the total binding free energy in kilocalories per mole.

^bΔG_SA and ΔG_GB are nonpolar and polar contributions, respectively, to ΔG in kilocalories per mole.

^cΔE_vdw and ΔE_ele are van der Waals and electrostatic energy changes, respectively, in kilocalories per mole.

^dΔH and TΔS are enthalpy and entropy changes, respectively, in kilocalories per mole.

To gain further insights into energetics of the ssDHA:HIRAN interaction and validate our computational approach through the comparison with the results of binding studies (Table 2), we modeled the structures of HLTF HIRAN complexes with the five additional 3′-OH trinucleotides (ATG, CTG, TCC, TAT, and GGG) and performed MD simulations for these complexes (100 ns, 298 K) through a protocol similar to that used for experimental structures. The MD-generated conformational ensembles of the HIRAN:trinucleotide complexes were used to predict the binding free energies (ΔG) using the MM/PBSA approach as described above (Table 4, bottom). Not surprisingly, ATG and CTG demonstrated the most favorable negative binding free energies, while higher (less favorable) binding energies were calculated for trinucleotides that were less potent in the FP displacement assay (TCC, TAT, and GGG) (Table 2), further validating the robustness of our computational approach. Overall, the binding free energies (ΔG) calculated using the MM/PBSA methodology for the nine experimental and modeled ssDNA:HIRAN complexes (Table 4) exhibit good correlation with the experimentally determined log(IC₅₀) values for the corresponding trinucleotide sequences [r² = 71% (Figure 3)], demonstrating that our MD-based method correctly captures the energetics of the ssDNA:HIRAN interaction and can be used to predict optimal ssDNA sequences for HLTF HIRAN domain binding.

Figure 3.

Correlation between the predicted binding free energies for the ssDNA:HIRAN complexes (Table 4) and experimental IC₅₀ values obtained in the FP displacement assays with the corresponding ssDNA oligonucleotides (Table 2).

Per-Nucleotide Contributions to the ssDNA:HIRAN Complex Stability.

Having demonstrated that the binding energies for ssDNA:HIRAN complexes predicted from MD trajectories are in agreement with the experimental binding affinities, we sought to gain additional insights into HLTF HIRAN specificity for 3′-ssDNA by evaluating contributions of individual ssDNA nucleotides and HIRAN residues to the stability of the four experimentally determined complexes (Table 5 and Tables S1−S4). The per-nucleotide and per-residue contributions to the binding free energies were divided into the same energy terms as described above for the total binding energies of the entire complex calculated using the MM/PBSA approach (Table 4).

Table 5.

Per-Nucleotide Decomposition of the Binding Free Energy for ssDNA Nucleotides in Complex with the HIRAN Domain^a,b

	ΔEvdwa	ΔEelea	ΔGGBa	ΔGSAa	ΔGa
NT1
4XZF/-GCC	−11.7 ± 1.6	−78.6 ± 6.0	83.0 ± 2.8	−2.3 ± 0.1	−9.5 ± 3.3
4S0N/-TTT	−13.0 ± 2.0	−69.7 ± 11	77.6 ± 8.5	−2.1 ± 0.1	−7.3 ± 2.6
5BNH/GTG	−15.8 ± 1.4	−58.8 ± 8.4	71.9 ± 3.3	−1.2 ± 0.1	−5.4 ± 2.0
5BNH/TTG	−15.2 ± 1.3	−60.6 ± 7.3	72.2 ± 6.5	−2.1 ± 0.1	−5.8 ± 1.7
NT2
4XZF/-GCC	−6.1 ± 1.1	−65.5 ± 1.1	71.2 ± 2.2	−0.85 ± 0.1	−1.2 ± 2.1
4S0N/-TTT	−7.6 ± 1.3	−69.5 ± 3.4	76.0 ± 4.0	−1.1 ± 0.2	−2.2 ± 1.0
5BNH/GTG	−9.3 ± 1.1	−77.2 ± 9.1	82.9 ± 8.7	−1.2 ± 0.1	−5.4 ± 2.0
5BNH/TTG	−9.9 ± 1.2	−78.4 ± 6.3	83.5 ± 5.8	−1.3 ± 0.1	−6.2 ± 1.5
NT3
4XZF/-GCC	−3.3 ± 1.9	−53.2 ± 7.7	57.8 ± 9.5	−0.54 ± 0.3	0.70 ± 2.4
4S0N/-TTT	−1.2 ± 0.4	−61.1 ± 8.8	63.9 ± 8.8	−0.16 ± 0.1	1.5 ± 0.6
5BNH/GTG	−4.8 ± 2.1	−18.8 ± 2.7	22.9 ± 3.3	−0.67 ± 0.3	−1.2 ± 1.2
5BNH/TTG	−7.1 ± 1.2	−24.2 ± 3.6	28.4 ± 2.9	−1.13 ± 0.2	−4.0 ± 2.0

^aAll values are in kilocalories per mole, and terms are as defined in Table 4.

^bPer-residue energy decompositions for HIRAN bound to ssDNA are listed in Tables S1−S4.

The first three nucleotides (NT₁−NT₃) in all considered ssDNA:HIRAN structures were primary contributors to the overall complex stability; therefore, we mainly focused our analysis on these nucleotides (Table 5). Overall, per-nucleotide energy decompositions demonstrate that the 3′-NT₁ nucleotide contributes the majority of the binding energy to each complex. The van der Waals interactions (ΔE_vdw) of NT₁−NT₃ contribute significantly to the stabilization of all complexes, with each nucleotide forming vdW contacts with numerous residues in the HIRAN domain. Most notably, the NT₁ and NT₂ nucleotides are locked inside the binding pocket of the HIRAN domain by π−π interactions with Y72 and Y93. In agreement with the experimentally determined nucleotide preferences, the most favorable ΔE_vdw contributions are observed for G and T at the NT₁ and NT₂, positions, respectively, suggesting that the DNA sequence specificity of HIRAN (in part) stems from its vdW interactions with NT₁ and NT₂ bases. In addition, for the complexes containing ssDNA fragments longer than three nucleotides (PDB entries 4XZF and 4S0N), base and sugar moieties of NT₃ form favorable contacts with the domain outside of the binding pocket; however, water molecules loosely associated with the HIRAN surface prevent the formation of the optimal NT₃ interface with the domain, decreasing the ΔE_vdw contribution to binding energy. By contrast, NT₃ in HIRAN complexes with GTG and TTG trinucleotides demonstrates more favorable nonpolar interactions. The likely reason for this difference between the trinucleotides and longer ssDNA sequences is the formation of π−π interactions between NT₃ and NT₄, which compete with NT₃ binding to the domain surface. In HIRAN complexes with trinucleotides, the NT₃:NT₄ interaction is missing; NT₃ is oriented toward the domain surface and forms favorable vdW contacts with F142. This observation may explain weaker binding affinities of the HIRAN domain for the four-nucleotide TTTG and GTTG sequences than for the corresponding trinucleotides obtained in the FP binding assays (Table 2).

Per-residue binding energy decompositions for key amino acid residues in the HIRAN domain confirmed the importance of both van der Waals and electrostatic interactions for high-affinity ssDNA:HIRAN binding (Tables S1−S4). As expected, the two tyrosine residues (Y72 and Y93) that form vdW contacts with NT₁ and NT₂ make significant ΔE_vdw contributions to the binding free energy. In addition, positively charged (basic) amino acids R71, K112, and K113 contribute significantly to stabilization of the complex through electrostatic interactions with the negatively charged ssDNA back-bone, while several more polar and/or charged residues contribute to the complex stability through the formation of hydrogen bonds with ssDNA base and sugar moieties (see below).

Central to the formation of ssDNA:HIRAN complexes are π−π stacking interactions of the NT₁ and NT₂ bases locked between the aromatic side chains of Y72 and Y93, including both intermolecular (Y93−NT₁ and Y72−NT₂) and intramolecular (NT₁−NT₂) π−π stacking. The empirical potentials used in the classical MD approach lack accuracy to quantitatively describe the long-range electrostatic and dispersion interactions underlying π−π stacking, and while the more rigorous quantum mechanical (QM) methods can be utilized,^50,51 these are often too computationally demanding for biomolecular applications. Previous quantum chemical calculations of the interaction energies between oligonucleotides and aromatic amino acids demonstrated that guanine exhibits the strongest π−π stacking interactions with aromatic residues in proteins.^52,53 The bicyclic aromatic guanine in the NT₁ position of the 5BNH complexes provides a larger surface for the π electron distribution relative to the single pyrimidine rings in the 4S0N (NT₁ = T) and 4XZF (NT₁ = C) complexes, which allows the NT₁ guanine to form a more extensive π−π stacking interaction network with the side chain of Y93 and the NT₂ thymine (Figure S14). This can be clearly seen in Table 5 with regard to the per-nucleotide binding free energy values for the individual nucleotides in each ssDNA. The individual per-nucleotide contributions to binding energy of a cytosine at NT₁ are significantly greater than the contributions of a guanine at NT₁ (ΔG values of −9.5 and −5.8/−5.4 kcal/mol, respectively). However, the presence of the guanine at NT₁ improves the per-nucleotide contributions of NT₂ and NT₃, which results in an overall higher binding energy for the GTG/ TTG ssDNA trinucleotides.

Analysis of Intermolecular Hydrogen Bonds.

Important contributions to the affinity and specificity of ssDNA:HIRAN interactions result from the direct intermolecular hydrogen bonds between protein and DNA groups. In addition, water molecules can form transient hydrogen bond bridges between ssDNA and HIRAN that are more difficult to quantify. With this in mind, we sought to further define specific electrostatic interactions that determine the affinity and specificity of HIRAN for ssDNA sequences through the analysis of high-occupancy intermolecular hydrogen bonds in MD simulations of the four experimentally determined ssDNA:HIRAN complexes. This analysis revealed that both NT₁ and NT₂ are stabilized inside the binding pocket of the HIRAN domain by a network of hydrogen bonds between several key amino acid residues and the phosphate backbone of ssDNA. As shown in Figure 4, several hydrogen bonds between the ssDNA and the HIRAN domain in each complex persisted through the majority of the MD simulation (100 ns). In each complex, NT₁ forms two stable, high-occupancy hydrogen bonds: (i) between the backbone amide of K113 and a phosphate oxygen and (ii) between the amino group of N91 side chain amine and either a nitrogen or a carbonyl of the nucleotide base (Figure 4 and Figures S15−S19). In addition, the 3′-hydroxyl forms a high-occupancy bond in each complex, but the HIRAN residue participating in the hydrogen bond is different across the different complexes. In the 5BNH and 4S0N complexes featuring thymidine at the NT₂ position, the second nucleotide forms several high-occupancy intermolecular hydrogen bonds (i) between the amino group of R71 side chain and a phosphate oxygen and (ii) between the side chain OH of Y72 and a phosphate oxygen. The 4S0N complex is also stabilized by a third high-occupancy hydrogen bond between the side chain OH of Y73 and the pyrimidine carbonyl. By contrast, cytosine in the NT₂ position of the 4XZF complex did not demonstrate any hydrogen bonds with the HIRAN domain at an occupancy rate of >5% during the MD simulation, providing a possible reason why T is preferred over C at the NT₂ position. Taken together, the analysis of a hydrogen bond network of the ssDNA:HIRAN complex and the binding energy decompositions provided insights into the molecular mechanisms for 3′-hydroxyl ssDNA recognition by the HIRAN domain and helped rationalize the experimentally determined HIRAN preference for (T/C)TG trinucleotides.

Figure 4.

Hydrogen bonds between ssDNA and the HIRAN domain. Key hydrogen bonds identified for the NT₁ and NT₂ nucleobases in MD simulations of the ssDNA:HIRAN complexes: (A) NT₁ (T)−4S0N, (B) NT₂ (T)−4S0N, (C) NT₁ (G)−5BNH/TTG, and (D) NT₂ (T)−5BNH/GTG. Additional hydrogen bond diagrams and occupancy graphs are shown in Figures S15−S19.

DISCUSSION

HLTF and SHPRH are the human orthologs of yeast Rad5 that function as E3 ligases to polyubiquitinate K63 of PCNA to initiate the error-free DDT pathway, TS.^{3−5,12−15} In addition, HLTF and Rad5 function as ATP-dependent DNA translocases that catalyze replication fork reversal, but this activity has not been reported for SHPRH.¹⁶⁻¹⁸ The N-terminal HIRAN domain, which is found in HLTF and Rad5 but not SHPRH, binds the 3′-hydroxyl of ssDNA at a stalled replication fork and enables the core helicase-like domain to perform strand separation, branch migration, and annealing of the daughter strands, ultimately resulting in a Holliday junction intermediate (“chicken foot” structure).^4,22−27 During TS, this intermediate allows error-free replication over DNA lesions using a newly synthesized daughter strand as a template. The HIRAN domain confers HLTF and Rad5 with the ability to recognize DNA features at stalled replication forks and conduct the fork reversal required for TS, while the lack of this domain in SHPRH is likely responsible for its functional divergence. Prior to our studies, it was not known whether the HIRAN domain exerts preferential binding to DNA sequences that allow HLTF and Rad5 to guard specific regions of the genome.

To elucidate the DNA sequence preferences of HLTF HIRAN, we used FP assays to probe HIRAN interactions with a variety of ssDNA oligonucleotides. As expected, ssDNA fragments with an obstructed 3′-end (T₅-3′-FAM and 5′-FAM-3′-PO₄) were unable to bind HIRAN, while the 5′-FAM-labeled 3′-hydroxyl pentanucleotides displayed affinities for HIRAN in the range of 0.14−0.25 μM (Table 1), confirming that a free 3′-hydroxyl is integral to ssDNA:HIRAN complex formation. To probe interactions of HIRAN with a broader range of ssDNA sequences, we have developed a robust FP displacement assay, which tested the ability of unlabeled oligonucleotides to displace the 5′-FAM-T₅ from its complex with HIRAN (Figure 2). We found that single nucleotides were unable to displace 5′-FAM-T₅ bound to HIRAN, while TG was the only dinucleotide that demonstrated concentration-dependent displacement of 5′-FAM-T₅ with an IC₅₀ of 31.7 μM (Table 2). Subsequent FP displacement studies of TTX and TXT (X = A, T, G, or C) trinucleotides revealed a strong preference of HLTF HIRAN for ssDNA sequences with a 3′-purine (G) following a pyrimidine (T) at NT₁ and NT₂ positions (Table 2), which is consistent with the results obtained for the TG dinucleotide and our computational studies. Although our FP displacement assay revealed no preference for the NT₃ nucleotide, more robust FP binding studies with 5′-FAM-XTG (X = A, T, G, or C) sequences suggest a slight specificity of HIRAN for a pyrimidine (T or C) at this position (Table 2). Addition of a fourth nucleotide did not improve the binding affinity, in agreement with our computational results, validating that the first three nucleotides contribute the majority of the binding free energy. Therefore, our studies identified the 5′-TTG-3′-OH and 5′-CTG-3′-OH sequences as the preferred ssDNA binding substrates of the HLTF HIRAN domain.

To rationalize the observed (T/C)TG-3′-OH nucleotide preference of HLTF HIRAN, we have performed extensive computational studies, including MD simulation of the four experimental ssDNA:HIRAN structures (PDB entries 4XZF, 4SON, and 5BNH)²³⁻²⁵ as well as models of the HIRAN complexes with five additional trinucleotides, per-nucleotide and per-residue decompositions of the binding free energies calculated using the MM/PBSA approach (Tables 4 and 5), and analysis of the ssDNA:HIRAN hydrogen bond network. Our computational studies revealed that the HLTF HIRAN domain binds to the free 3′-hydroxyl of ssDNA with the first three 3′-nucleotides contributing a majority of the binding free energy, while the following nucleotides exhibited increased flexibility in MD simulations (Table 3). The 3′-nucleotide, NT₁, contributes the most binding energy to the ssDNAHIRAN complex, while the contributions from the preceding nucleotides NT₂ and NT₃ are less pronounced (Table 5). We found that total binding energies of the ssDNA:HIRAN complexes (Table 4) exhibit good correlation with log(IC₅₀) obtained for the corresponding oligonucleotides in the FP displacement assays providing the measure of binding affinity (Figure 3). The TTG:HIRAN complex displayed the most favorable binding energy (Table 5), in agreement with our experimental results. The complexes are stabilized by multiple interactions of NT₁ and NT₂ with a well-organized binding pocket on the HIRAN domain, most notably by the π−π interaction of NT₁ and NT₂ bases with Y72 and Y93, and a network of hydrogen bonds, including those formed by the 3′-OH of NT₁. The detailed analysis of these interactions provided the high-resolution picture of the ssDNA:HIRAN recognition and helped explain HIRAN specificity for (T/C)TG-3′-OH sequences.

The consensus sequence of (T/C)TG for HLTF HIRAN binding is both biologically and clinically relevant. Improper replication of repetitive short DNA sequences (microsatellites) such as trinucleotide repeats (TNRs) has been implicated in various human diseases, including neurological, neurodegenerative, and muscular disorders.³²⁻³⁴ Specifically, the CTG repeat is one of a few TNRs expanded in human disorders linked to myotonic dystrophy type 1,^54,55 while expansion of the CCTG repeat is associated with myotonic dystrophy type 2.⁵⁶ The CTG repeats are difficult to replicate due to atypical secondary structures such as DNA hairpins formed in these TNR regions,^57,58 which can stall replication fork progression and cause DNA polymerase slippage leading to expansion or contraction mutations.³²⁻³⁴ Therefore, replication of these problematic DNA regions requires additional factors to unwind unwanted hairpins and prevent mutagenesis.

There is a growing body of evidence that TS helps maintain the stability of the difficult to replicate and susceptible to fork stalling CTG TNRs.^34,35 Thus, human HLTF, Rad18, and RTEL1 helicase inhibit expansions and the fragility of CAG/CTG,³⁵ which is in line with similar observations reported for Rad5, Rad18, and Srs2 helicase in budding yeast.⁵⁹ Furthermore, HLTF knockdown but not SHPRH knockdown in human cells displays an increased level of mutagenesis in CTG repeat regions.³⁵ This result is in agreement with HLTF, but not SHPRH, harboring the HIRAN domain that recognizes CTG sequences. Previous studies along with our findings suggest that when the replisome pauses in CTG repeat regions, PCNA is monoubiquitinated by the Rad6/Rad18 E2/E3 enzyme pair.^4,10,11 When the HLTF HIRAN domain recognizes and binds the 5′-CTG-3′-OH sequence, polyubiquitination of PCNA by Ubc13-Mms2 (E2) and HLTF (E3) commits DDT to the error-free pathway, TS. It is likely that, along with the subsequent fork reversal, HLTF takes part in resolution of replication blocking DNA hairpin structures in the CTG repeat regions, thus protecting microsatellites from expansion and deletion mutations and maintaining genome stability.

Collectively, our computational studies and FP assays revealed the HLTF HIRAN domain preferentially recognizes 3′-trinucleotide sequences occurring in TNR repeat regions susceptible to replication errors that cause expansion or deletion mutations found in neuromuscular and neurodegenerative disorders. Therefore, this study adds to the growing body of literature that supports a role for HLTF in maintaining the stability of difficult to replicate TNR microsatellite regions and safeguarding against TNR-associated diseases.