HLTF's Ancient HIRAN Domain Binds 3' DNA Ends to Drive Replication Fork Reversal

Abstract

Stalled replication forks are a critical problem for the cell because they can lead to complex genome rearrangements that underlie cell death and disease. Processes such as DNA damage tolerance and replication fork reversal protect stalled forks from these events. A central mediator of these DNA damage responses in humans is the Rad5-related DNA translocase, HLTF. Here, we present biochemical and structural evidence that the HIRAN domain, an ancient and conserved domain found in HLTF and other DNA processing proteins, is a modified oligonucleotide/oligosaccharide (OB) fold that binds to 3' ssDNA ends. We demonstrate that the HIRAN domain promotes HLTF-dependent fork reversal in vitro through its interaction with 3' ssDNA ends found at forks. Finally, we show that HLTF restrains replication fork progression in cells in a HIRAN-dependent manner. These findings establish a mechanism of HLTF-mediated fork reversal and provide insight into the requirement for distinct fork remodeling activities in the cell.

Figure 1. HLTF and Rad18 are enriched at the replication fork

293T cells were pulsed with EdU for 10 min and chased with thymidine for the time shown. Cells were then fixed with 1% formaldehyde and collected. Nascent DNA-protein complexes were purified by iPOND, and EdU-associated proteins were analyzed by western blotting. The -clk condition represents cells pulsed with EdU, and processed without biotin-azide in the click reaction step.

Figure 2. The HLTF HIRAN domain binds the 3′-end of ssDNA

(A) Domain schematic of HLTF. The sequence alignment shows conservation within HIRAN domains from seven eukaryotic HLTF orthologs and two bacterial proteins of unknown function. Hs, Homo sapiens; Pt, Pan troglodytes; Mm, Mus musculus; Xl, Xenopus laevis; Dr, Danio Rerio; At, Arabidopsis thaliana; Sc, Saccharomyces cerevisiae; Lp, Lactobacillus plantarum; Ec, Escherichia coli. Arrows above the alignment indicate conserved residues that contact DNA (see Fig. 4 and S1). (B) EMSA of HIRAN binding to 5′-FAM labeled 40-mer ssDNA and dsDNA. Quantitation of the gel is shown on the right. (C,D,E) Capture of purified proteins by biotinylated 20mer ssDNA. The biotin position at the 5′- or 3′-end is indicated by the • symbol. Control, no DNA. (F) EMSA of HIRAN binding to 5′-FAM-labeled dT₁₀ oligonucleotides modified as shown at the 3′-ends. Quantification of the gel is shown on the right.

Figure 3. Structure of the HLTF-HIRAN domain bound to ssDNA

(A) The crystal structure of the HLTF HIRAN domain bound to dT₁₀ ssDNA (gold). (B) Sequence conservation from 150 HIRAN orthologs mapped onto the solvent accessible surface of the HIRAN crystal structure using the ConSurf web server (http://consurf.tau.ac.il/). (C) Comparison of HLTF HIRAN-ssDNA and SmpB-tRNA (PDB:2CZJ) structures. Homologous protein secondary structural elements are shown in green, OB-folds in yellow, and nucleic acid segments contacting the OB-fold in orange. Topology diagrams are shown below and colored according to the structures. See also Figure S2.

Figure 4. HIRAN residues responsible for DNA binding

(A) Details of the HIRAN-DNA interactions observed in the crystal structure. Hydrogen bonds are shown as dashed lines. (B) NMR chemical shift perturbation of ¹⁵N-labeled HIRAN in the absence (blue) and presence (red) of ssDNA (top) or dsDNA (bottom). The full ¹H-¹⁵N HSQC spectra are shown in Fig. S3A. In the absence of DNA, residue E114 gives rise to multiple weak NMR resonances that collapse into a single strong peak in the DNA-bound state. Similarly, the amide peaks of V68 and K113 are only observed in the spectrum of the HIRAN-ssDNA complex. (C) NMR chemical shift changes plotted by protein residue number. Resonances that are not detected in the free protein but appear in the complex were arbitrarily set to >200 Hz. (D) HIRAN solvent accessible surface colored according to NMR chemical shift perturbation from panel C, with the degree of magenta representing increased chemical shift changes in response to dT₁₀ binding. (E) EMSA of FAM-ssDNA in the presence of increasing concentrations of wild-type (WT) and mutant HIRAN proteins. Concentrations of protein were chosen to show the full range of binding for all mutants. (F) Quantitation of the data shown in panel E. Zero [HIRAN] is not plotted. (G) Dissociation constants (K_d) extracted from the binding isotherms shown in panel F. Absolute K_d values for WT and F142A are approximations since the transition range was not defined, but are consistent with K_d values in Figure 2B. Values represent the average ± S.D. of three independent measurements. See also Figure S3.

Figure 5. HIRAN is necessary for efficient DNA fork regression by HLTF

(A) Left, representative ATPase activity of 50 nM WT vs. ATPase-dead (DEAA) or HIRAN-mutant HLTF proteins incubated with the indicated amount of splayed-arm DNA. Right, average values ± SEM from triplicate experiments. (B) Left, representative fork regression experiment of model DNA forks by the indicated amount of WT, ATPase-dead (DEAA), or HIRAN-mutant HLTF proteins. The * represents the position of the 5′-FAM-labeled 75mer oligonucleotide in the fork structure and product. Right, quantification of the mean values ± SEM from triplicate experiments. See also Figure S4.

Figure 6. Fork regression by HLTF is promoted by the 3′-hydroxyl end of the leading strand

(A) Representative fork regression activity for HLTF on a model fork substrate with either a 3′-OH or a 3′-PO₄ end on the leading strand, with quantification on the right. A graph of mean values ± SEM for three replicate experiments is shown. (B) Fork regression experiments as in A, but with the indicated amounts of SMARCAL1 and quantification on the right. A graph of mean values ± SEM for three replicate experiments is shown. See also Figure S5.

Figure 7. Loss of HLTF leads to longer DNA replication tracks upon depletion of nucleotide pools

(A) Expression of HLTF in U2OS cells or CRISPR-generated HLTF knockout U2OS cells. (B) Experimental setup. Cells were pulsed with IdU (30 min), then incubated with CldU and 50 µM hydroxyurea (HU) for 30 min. (C) Representative IdU and CldU replication tracks in WT U2OS and HLTF-knockout clones. (D) Left, dot-plot of CldU replication track lengths in the indicated cell lines. Right, dot-plot of CldU replication tracks in the indicated cell lines after treatment with 50 µM HU as in B. In both experiments the line represents mean. **** = p < 0.0001, by two-tailed nonparametric Mann-Whitney test. (E) Left, western blots of lysates from HLTF KO #3 cells transfected with empty vector (vec), WT or mutant forms of HLTF. Middle, dot plot of CldU replication track lengths in HLTF KO #3 cells transfected as on the left and treated with HU as in B. Line represents mean. **** = P < 0.0001 were calculated using two-tailed nonparametic Mann-Whitney test. Right, Relative CldU track lengths for n≥3 replicates of the middle panel plotted ± SEM. 100% = CldU +HU track length in vector-transfected cells. (F) Model of fork regression by HLTF, which utilizes the HIRAN domain to drive replication fork reversal. See also Figure S6.