Structure of the human RECQ1 helicase reveals a putative strand-separation pin

Abstract

RecQ-like helicases, which include 5 members in the human genome, are important in maintaining genome integrity. We present a crystal structure of a truncated form of the human RECQ1 protein with Mg-ADP. The truncated protein is active in DNA fork unwinding but lacks other activities of the full-length enzyme: disruption of Holliday junctions and DNA strand annealing. The structure of human RECQ1 resembles that of Escherichia coli RecQ, with some important differences. All structural domains are conserved, including the 2 RecA-like domains and the RecQ-specific zinc-binding and winged-helix (WH) domains. However, the WH domain is positioned at a different orientation from that of the E. coli enzyme. We identify a prominent β-hairpin of the WH domain as essential for DNA strand separation, which may be analogous to DNA strand-separation features of other DNA helicases. This hairpin is significantly shorter in the E. coli enzyme and is not required for its helicase activity, suggesting that there are significant differences between the modes of action of RecQ family members.

The RecQ helicases are a family of DNA-unwinding enzymes conserved from prokaryotes to mammals that play a key role in the maintenance of genome stability. The RecQ helicase family has 5 representatives in the human genome (1–3): RECQ1 (also known as RECQL or RECQL1), BLM, WRN, RECQ4, and RECQ5. Although these 5 enzymes are similar in their catalytic core, they probably have distinct functions, as indicated by the genetic disorders associated to mutations in the genes of BLM, WRN, and RECQ4. In particular, mutations in the gene encoding for BLM (4) are associated with the Bloom's syndrome (BS), which is manifested as an increased incidence of a wide spectrum of cancers. Werner's syndrome (WS), which is linked to mutations in the WRN (5) gene, involves many signs of premature aging, as well as a predisposition to a more limited spectrum of cancers. Mutations in the gene of RECQ4 are the cause of more varied genetic disease phenotypes, including Rothmund–Thomson (RTS) (6, 7), RAPADILINO (8), and Baller–Gerold (9) syndromes. No disease phenotypes have been associated with mutations in the genes of the other 2 family members, RECQ1 and RECQ5 yet, although they may be responsible for additional cancer predisposition disorders that are distinct from RTS, BS, and WS. In this regard, interesting candidates are patients with a phenotype similar to that of RTS individuals who do not carry any mutations in the RECQ4 gene (7). A possible role of RECQ1 in genome maintenance is suggested by several observations (reviewed in ref 10). Biochemical purification from human embryonic kidney cells recovered RECQ1 as the major Holliday junction (HJ) branch migration activity (11). Knockout of the RECQ1 gene in mice (12) or suppression of its expression in HeLa cells (11) resulted in cellular phenotypes that include chromosomal instability, increased sister chromatid exchange, and heightened sensitivity to ionizing radiation (12, 13). However, the knockout mice seemed healthy and fertile (12). Two recent reports show that RECQ1 depletion using inhibitory RNA has anticancer effects in cells (14) and xenografts (15); this indicates that RECQ1 may be an important target for cancer therapy.

RecQ helicases are ATP- and Mg²⁺-dependent enzymes that unwind DNA with a 3′ to 5′ polarity and, although with some differences, are capable of unwinding a variety of DNA structures other than standard B-form DNA duplexes such as forked duplexes, D-loops, triple helices, 3- or 4-way junctions, and G-quadruplex DNA (16–21). Even when acting on the same substrates, they may do so by using different mechanisms (22). In addition to unwinding DNA, some RecQ helicases are also able to promote the annealing of complementary ssDNA molecules in an ATP-independent fashion (23–28). Interestingly, a recent study indicated that different activities of RECQ1 are associated with distinct oligomeric forms: A form identified as monomers or dimers is responsible for DNA unwinding activity, whereas higher-order oligomers (hexamers or pentamers) possess strand-annealing activity (23). Moreover, ATP binding is the key that controls the equilibrium between these 2 assembly states, favoring the smaller form.

Several aspects of the mechanisms of RecQ helicase action remain unsolved. Prominent among them are the mechanisms of coupling nucleotide hydrolysis to DNA tracking and unwinding, and the determinants of substrate specificity. High-resolution structures have provided crucial information on the conformation and structural organization of various DNA helicases in their nucleotide-bound and unbound forms, revealing both common and distinct features. In this study, we present the first high-resolution structural data on the human RECQ1 helicase in its DNA-unbound form. Although similar in to Escherichia coli RecQ, there are significant structural differences.

Our mutational analysis highlights the role of a prominent feature of the divergent winged-helix (WH) domain of RECQ1 and sheds light on the relationship between different activities of the protein.

Results

Properties of the Truncated RECQ1 Protein.

Previous biochemical investigations of human RECQ1 used a full-length recombinant protein produced in insect cells. The full-length protein (RECQ1^FL) appears as a mixture of different oligomeric forms (23, 29). In common with other proteins of the RecQ family, RECQ1^FL has DNA-dependent ATPase activity, unwinds forked DNA substrates, resolves HJs, and has a DNA strand-annealing activity. It is less active on tailed-duplex DNA and has no unwinding activity on blunt-ended duplex DNA. A previous study indicated that fork unwinding is performed by a monomer or a dimer, whereas DNA annealing requires a higher-order oligomeric form of the protein (23).

We initially attempted to crystallize nearly full-length versions of the protein. However, we were unsuccessful, possibly because of the mixture of oligomeric forms. In an attempt to produce a crystallizable form of RECQ1, we cloned multiple truncated versions, with modest deletions of both the N- and C-terminal regions of the protein. A construct encompassing amino acids 49–616 (of 649) of RECQ1, followed by a C-terminal tag of 22 aa, could be readily produced in E. coli and purified to >95% homogeneity [see supporting information (SI) Fig. S1]. This truncated protein (termed RECQ1^T1) appears as a single peak in gel filtration that elutes later than the smallest form of the full-length protein (Fig. 1A). The elution profile remains the same in the presence of ATPγS and/or ssDNA, indicating that the oligomeric state of the RECQ1^T1 is not affected by the interaction with the nucleotide or ssDNA. We then compared the biochemical activities of the truncated and full-length proteins. Both proteins have very similar DNA-unwinding activity on forked-DNA substrates (Fig. 1B). However, the RECQ1^T1 is not active on HJ DNA, and has no detectable strand-annealing activity, unlike the full-length protein (Fig. 1 C and D). This confirms the previous observations that higher-order oligomers are required for the strand-annealing activity of RECQ1 and suggests that these oligomers are also necessary for the HJ resolution activity of RECQ1 (22).

Fig. 1.

Characterization of the RECQ1^T1 mutant. (A) Size exclusion chromatography profiles of RECQ1^FL (black) RECQ1^T1 (red), RECQ1^T1 + ATPγS (green), RECQ1^T1 + ATPγS + ssDNA (blue), RECQ1^T1 +ssDNA (purple). (B) Plot of DNA unwinding activity as a function of protein concentration using a fork-duplex substrate of 20 bp with ssDNA tails of 30 nt. Concentration dependence experiments were performed by using 0–100 nM RECQ1^FL (▵) or RECQ1^T1 (•). All of the reactions were stopped after 15 min. Data points were the mean of 3 independent experiments with the standard deviation indicated by error bars. The Inset shows an example of the unwinding assay using 0–200 nM RECQ1^T1. (C) Plot of the strand annealing activity as a function of protein concentration using the same partially complementary synthetic oligonucleotides used to prepare the forked duplex substrate. Concentration dependence experiments were performed by using 0–200 nM full-length RECQ1 (▵) or RECQ1^T1 (•). All of the reactions were stopped after 20 min. Data points were the mean of 3 independent experiments with the standard deviation indicated by error bars. The Inset shows an example of the strand annealing assay using 0–200 nM RECQ1^T1. (D) Concentration dependence experiments using 0–200 nM RECQ1^FL or RECQ1^T1 and a 50-bp-long synthetic HJ substrate that contains a 12-bp homologous core. All of the reactions were stopped after 20 min.

Structure Overview: The RECQ1 Monomer.

The truncated RECQ1^T1 protein was crystallized in presence of MgCl₂, ATPγS and DNA. Although this is a poorly hydrolyzable ATP analogue, the crystal structure contains Mg-ADP in the binding site; the DNA is absent. The modeled structure encompasses residues S63–K592 of RECQ1 (detailed crystallographic data in SI Methods and Table S1).

The domain organization of RECQ1 closely resembles that of the bacterial RecQ (30), and the general organization of the helicase superfamily. The overall fold of RECQ1 has 3 tiers (Fig. 2A). The upper part is the common core structure of the helicase superfamily, consisting of 2 RecA-like domains: D1 (amino acids 63–281) and D2 (amino acid 282–418) (colored red and blue, respectively). These domains contain the signature motifs of helicase superfamily 2 and are thought to harbor the ATP-dependent translocation activity. The nucleotide-binding pocket is in the cleft between the 2 RecA-like domains and is surrounded by highly conserved residues (Fig. 3A).

Fig. 2.

Overview of RECQ1 structure. (A) Ribbon representation of a single RECQ1 molecule, viewed from 3 perpendicular orientations. The subdomains are identified by color: Core helicase domain D1, red; core helicase domain D2, blue; zinc motif (ZnD), yellow; helical hairpin (HH), orange; WH domain, green; and the β-hairpin in purple. ADP is shown in space-filling form. (B) E. coli RecQ, colored as in A. [PDB ID code 1OYY (30)] The molecule is viewed in the same orientation as the central view in A, using the D2 domain as a reference. (C) Side-by-side comparison of bacterial RecQ (Left) and human RECQ1 (Right). Arrows indicate the rotations of the various domains. The WH domain rotates 90° around the vertical (compare orientation of helix α1 marked by asterisk), the HH tilts up by 10°, and the D1 rotates away from D2.

Fig. 3.

Details of the ADP and Zn-binding regions. (A) The nucleotide-binding pocket. Main chain and carbons are colored according the conserved motifs: motif 0 (yellow), motif I (magenta), motif II (light blue), motif V (gray). (B) Overlay of the Zn domains of RECQ1 (orange) and E. coli RecQ (gray). The yellow/black spheres indicate the zinc ion, which nearly overlap in the 2 structures. (C) Overlay of the WH domains of RECQ1 (red), WRN (blue), and E. coli RecQ (green). The β-hairpin forming one of the wings and the hydrophobic residues at the hairpin tip are highlighted.

The next domain (yellow/orange) is characteristic of the RecQ family of helicases and includes a zinc-binding motif and a pair of antiparallel α-helices (amino acids 419–480). The C-terminal region adopts a WH, domain (amino acids 481–592; green), which is more divergent in sequence among the RecQ helicases. The structure of each domain is very similar to the corresponding domain of bacterial RecQ (Fig. 2B). The 2 proteins differ in the relative positions and orientations of the separate domains. Most prominently, the WH domain of the E. coli protein is positioned perpendicular to the D1–D2 domains (Fig. 2B), whereas in the human protein the WH domain lies directly beneath the helical hairpin (HH) portion of the Zn domain, creating a more elongated linear molecule (Fig. 2C).

A more subtle domain shift occurs between the 2 RecA-like domains. In the human enzyme, the D1 domain is tilted along the axis of the cleft between the D1/D2 so that the cleft between them is narrowed slightly (2–3 Å) at the nucleotide end and widens by up to 7 Å at the distal end compared with bacterial RecQ (Fig. S2a). We have determined multiple crystal forms of the RECQ1 and the relative orientation of the D1–D2 domains varies quite dramatically, highlighting the high degree of D1–D2 interdomain flexibility (Fig. S2b). This motion may represent the ratcheting motion along DNA, although its relation to nucleotide binding and hydrolysis remains obscure.

ADP-Binding and Signature Motifs.

The Mg–ADP-binding pocket is shown in Fig. 3A. Although initial attempts were made to model ATPγS into the experimental density, the γ-thiophosphate exhibited a local poor fit, suggesting that either the ATPγS was hydrolyzed to ADP during crystallization, or ADP may have been introduced as an impurity. Instead, the electron density in γ-phosphate position was modeled by a solvated magnesium ion (Fig. S3). We have subsequently confirmed that RECQ1 crystallized in the presence of Mg–ADP adopts an identical structure to that described here (data not shown). The ADP forms extensive contacts with domain D1, and less with domain D2. The adenine is wedged between R93 and L89, both from motif “0” (31). The R93 is conserved in other helicase structures, whereas L89 replaces an aromatic residue that usually stacks against the adenine ring (e.g., Y23 of bacterial RecQ). R93 also forms a hydrogen bond with the ribose O1′. An additional residue from motif 0, Q96, forms 2 hydrogen bonds with the adenine N⁷ and N⁶.

The glycine-rich loop (motif 1) supports the phosphates, and the catalytic lysine K119 forms a water-mediated contact with a Mg ion. D219 and E220 of the DEVH box (motif II) form water-mediated hydrogen bonds to the Mg ion; mutating E220 to glutamine abolishes ATPase and helicase activity (23). There seems to be a single hydrogen bond between the ADP and the second RecA-like domain, involving D379 (motif V) and the ribose O3′. This contact is absent in the bacterial RecQ structure. Conversely, R404 from motif VI, termed the “arginine finger” in other DNA helicases, is distant from the projected location of the γ-phosphate.

Overall, the nucleotide-binding site is conventional for helicases, and is comparable with the ATPase-active conformation seen in other helicases (e.g., refs. 32 and 33).

RecQ-Specific Zn Domain.

The zinc domain (amino acids 419–480) is a conserved signature of the RecQ family. It includes a 4-cysteine zinc-binding motif, preceded by 2 antiparallel α-helices (the HH). The structure of the Zn domain and its position relative to the RecA-like domain D2 are virtually identical in the human RECQ1 and the bacterial RecQ protein (Fig. 3B). The main difference is the length of the HH. In the bacterial protein, both α-helices are longer than the human protein (15 and 19 aa, compared with 12 and 15, respectively). This results in the bacterial HH extending 1 helical turn farther than that of RECQ1.

C-Terminal Domain: Conserved Structure, Divergent Sequence, and Location.

The C-terminal domain (amino acid 481–592) of RECQ1 forms a WH structure, which is the DNA-binding domain of a variety of proteins [bacterial CAP, eukaryotic forkhead, HNF4 and RFX1 (34)]. Aligning the structures of the WH domains of E. coli RecQ, human WRN [PDB ID code 2AXL (35)], and RECQ1 reveals a highly conserved fold (Fig. 3C). This high structural conservation is remarkable, because there is very little sequence conservation among the WH domains of the 3 proteins.

The most conspicuous difference between the structures of bacterial RecQ and the human RECQ1 is the change in position and orientation of the WH domain, which changes the overall shape of the protein (Fig. 2C). In addition, the WH domain interacts with a different region of the core helicase and zinc domains. In RECQ1, the WH contacts exclusively the Zn domain, with the helical hairpin presenting a flat surface underneath the D2 domain of the ATPase core for interaction with the α3/wing face of the WH domain. In contrast, in E. coli RecQ, the WH domain contacts both the Zn domain and the D1 domain. Concomitantly, an exposed surface of the bacterial RecQ WH domain that includes a putative recognition helix (α3), becomes buried in human RECQ1. Thus, it would seem that the DNA binding interactions suggested for E. coli RecQ (30), in a cleft between the WH and Zn domains, are unlikely to occur in the conformation of RECQ1 seen in the crystal structure. Examination of the crystal packing shows that this domain is involved in minor lattice contacts in both structures, and so the possibility remains that the different orientations observed may have been selected by the crystallization process.

Comparison with DNA-Bound Helicase Structures.

In the absence of a DNA-RECQ1 cocrystal, we attempted to infer the DNA binding mode of RECQ1 by comparison with structures of other DNA-helicase complexes. PcrA, a Superfamily 1 DNA helicase, and HEL308, an archaeal Superfamily-2 DNA helicase, have been crystallized in complex with substrate DNA molecules. Their structures are shown in Fig. 4 A and B, respectively; the conserved recA-like domains (marked in red and blue) are similarly oriented. The DNA molecules (Fig. 4, shown in black) interact very differently in the 2 structures. However, in both structures a single-stranded DNA segment binds across the 2 recA-domains; these contacts provide the ATP-driven tracking movement along ssDNA, which is part of the mechanism of helicase action. The orientation of the double-stranded DNA segment is different in the 2 proteins. Interestingly, both proteins share a structural feature at the junction of single-stranded and double-stranded DNA: a β-hairpin structure with an aromatic residue at the tip (marked in purple and circled in Fig. 4 A and B). Multiple structures of UvrD (36) and PcrA (37–39), as well as the recent structure of HEL308 (33), suggest a role for this structure in DNA strand separation. In different reaction intermediates, the aromatic residue may stack against the last base pair or with the first unpaired base.

Fig. 4.

Comparison of RECQ1 with DNA-bound structures of DNA helicases PcrA and HEL308. (A) PcrA (PDB ID code 3PJR). (B) HEL308 (PDB ID code 2P6R). (C) RECQ1, with DNA overlaid at the same relative positions as the experimentally determined DNA cocomplexes. The core helicase domains of the 3 proteins (depicted in red and blue) are presented in the same orientation; the DNA is depicted in black/purple, and a β-hairpin (magenta) near the point of DNA strand separation (marked by asterisk) is highlighted by green circles.

To interpret our data in this context, the structure of RECQ1 was aligned with those of 2 experimentally determined helicase-DNA complexes: HEL308 (PDB ID code 2P6R) (33), and PcrA (PBD ID code 3PJR) (39). The alignments were made as to maximize the overlap of the core β-strands of the 2 RecA-like domains (Fig. S4). Fig. 4C shows the structure of RECQ1, with hypothetical DNA molecules placed according to the each of the alignments with the structures in Fig. 4 A and B. Strikingly, a wing of the WH domain of RECQ1 is a β-hairpin whose tip is also positioned in the vicinity of the DNA strand-separation point (purple, circled). As in other helicases, RECQ1 has an aromatic residue (Y564) at the tip of the hairpin. This raises the possibility, that this hairpin of RECQ1 may be involved in DNA strand separation.

To test this hypothesis, we have introduced several mutations, including a replacement of the Y564 with an alanine, and serial deletions on both strands of the β-hairpin (Table 1 and Fig. 5A). We then tested the DNA unwinding and ATPase activities of the mutants. Remarkably, all mutants of the β-hairpin, including the single amino acid change of the aromatic residue at the tip, were nearly devoid of DNA-unwinding activity (Fig. 5A). In separate experiments, the same mutants retained DNA-dependent ATPase activity, although their k_cat values for ATP hydrolysis were slightly reduced compared with the parent protein, RECQ1^T1 Table 2). Thus, the hairpin loop and the aromatic residue are essential for coupling the core ATPase (and, presumably, the ssDNA translocation activity) to DNA unwinding. The reason for lower k_cat values measured for RECQ1^FL compared with RECQ1^T1 is probably associated to the different oligomeric forms present in a RECQ1^FL preparation (23). It is possible, that the loss of helicase activity in the hairpin mutants results from a more extensive disruption or stabilization of the WH domain. CD spectroscopy and partial proteolysis experiments (Fig. S1) indicate that the WH domain remains structured in the mutants, although not excluding more subtle changes.

Table 1.

Hairpin mutants of human RECQ1 and E.coli RecQ

Mutant	Hairpin sequence
Human RECQ1
RECQ1T1	554YLKEDYSFTAYATISYLKIG573
Y564A RECQ1T1:	554YLKEDYSFTAAATISYLKIG573
Δ2-Y564A RECQ1T1:	554YLKEDYS---AA----ISYLKIG573
Δ3-Y564A RECQ1T1:	554YLKEDY-----AA----SYLKIG573
Δ8-Y564A RECQ1T1:	554Y--------------AA------------G573
E. coli RecQ
RecQ and RecQΔC	484VTQNIAQHSALQL496
H491A RecQ and H491A RecQΔC	484VTQNIAQASALQL496
Δ3 RecQ and Δ3 RecQΔC	484VTQNIA—ALQL496

The E.coli RecQ proteins denoted ΔC encompass amino acids 1-524 and lack the C-terminal HRDC domain. The strands of the β-hairpins are underlined.

Fig. 5.

Characterization of the β-hairpin mutants of RECQ1 and E. coli RecQ. The mutant sequences are listed in Table 1. (A) Plot of the DNA fork unwinding activity as a function of RECQ1 concentration. Concentration-dependence experiments were performed by using the mutants indicated in the figure (2, 5, 10, 20, 40, 50, 80, 100, 150, 200 nM protein). All of the reactions were stopped after 15 min. Data points were the mean of 3 independent experiments with the standard deviation indicated by error bars. (B) Plot of the unwinding activity as a function of E. coli RecQ concentration. Concentration-dependence experiments were performed by using the RecQ mutants indicated in the figure (0.005, 0.01, 0.02, 0.05, 0.0625, 0.125, 0.25, 0.5, 2, 5, 10, 20 nM; only the low concentration range is shown). All of the reactions were stopped after 15 min. Data points were the mean of 3 independent experiments with the standard deviation indicated by error bars.

Table 2.

Comparison of the k_cat and K_m values for ATP hydrolysis of RECQ1^FL, RECQ1^T1, and RECQ1^T1 mutants

Protein	kcat (min−1)	Km, μM
RECQ1FL	90 ± 2	115 ± 8
RECQ1T1	675 ± 20	135 ± 13
Y564A RECQ1T1	445 ± 24	100 ± 18
Δ3Y564A RECQ1T1	569 ± 11	96 ± 6
Δ8Y564A RECQ1T1	515 ± 12	92 ± 7

The β-hairpin structure in human RECQ1 is substantially longer than the corresponding structure in the E. coli enzyme (Fig. 3C). Alignment of the predicted β-hairpin motifs of several RecQ-family helicases (Fig. 6) reveals significant differences; bacterial enzymes have short hairpins, with histidine often present in place of an aromatic residue; and some BLM proteins have no aromatic residue at the tip. As an initial test of the functional significance of these differences, we have constructed 2 mutants of the E. coli RecQ: a replacement of the histidine with alanine (H491A) and a deletion of 3 amino acids including H491 (Table 1). As previously observed, the E. coli enzyme is much more active than human RECQ1, so the amount of enzyme used is 100-fold lower (Fig. 5B). In contrast with the hairpin mutants of RECQ1, both mutants of the bacterial RecQ are active in DNA fork unwinding (Fig. 5B). Similar results were obtained when we deleted the C-terminal HRDC domain of E. coli RecQ: The truncated proteins bearing the native and the 2 mutated versions of the hairpin had very similar helicase activity to the wild type.

Fig. 6.

Alignment of the putative β-hairpins of RecQ family helicases. The sequences are arranged in 4 clusters according to the overall similarity in the entire RecQ catalytic region (PFAM domains DEAD, HelicC, RQC, and HRDC); residues are colored according to similarity within a subgroup or across the family. An aromatic residue, thought to be near the tip of the hairpin, is marked in purple; note that some of the BLM proteins lack such a residue. Sequences 12–14 are bacterial RecQ enzymes. The β-hairpin is significantly shorter that that in the eukaryotic enzymes. The vast majority of known bacterial RecQ sequences have a histidine (red) in place of an aromatic residue, as in E. coli. Two bacterial sequences shown have a phenylalanine in that position, indicating that the histidine may have a similar role to that of the aromatic residue in other RecQ helicases.

Discussion

We present here the structure of the human RECQ1 helicase, a member of a highly conserved family of DNA helicases involved in maintaining chromosome stability. A combination of structural, mutational, and biochemical analyses reveal a number of interesting features. The protein used in these experiments (designated RECQ1^T1) is truncated at both the N and C termini; the truncated protein resembles the catalytic core of the bacterial RecQ protein that was previously crystallized (30). The truncated protein differs from full-length RECQ1 (denoted RECQ1^FL) in several respects. First, the deletion of the first 48 aa results in loss of high-oligomeric forms seen in the full-length protein. In parallel, the truncated protein lacks HJ unwinding activity, as well as a DNA-annealing activity. This confirms the previous observations that the higher-oligomeric forms of RECQ1 are involved in DNA strand annealing (22, 23) and suggest that they might be also be required for the disruption of HJ structures (22). In contrast, the truncated RECQ1^T1 protein is as active as the full-length protein in unwinding forked duplexes, confirming the earlier observation that a monomer or dimer of RECQ1 can perform this activity (22, 23).

As expected, the protein shares the general fold of the bacterial RecQ protein. The core helicase domain, as well as the RecQ family-specific zinc domain, have very similar structures in both proteins. However, the more divergent C-terminal region, which folds as a WH domain, is positioned very differently in RecQ and RECQ1^T1. Interpretation of the differences between the bacterial and human proteins is not straightforward. It is possible, that the 2 proteins have diverged in structure and mechanism of activity. Alternatively, the different domain orientation (and, possibly, the different quaternary structures) may reflect alternative conformations that are available to each protein. Domain movements are seen in other DNA helicases, and are associated with nucleic acid binding; a structural rearrangement of the WH domain upon binding the DNA substrate may be part of the enzymatic mechanism. Interestingly, previous studies showed that the WH domain of WRN is involved in heterologous protein recognition (40), suggesting that different domain orientations might also mediate protein binding, Finally, both E. coli RecQ and RECQ1 were crystallized as truncated proteins, which may differ in conformation from the full-length proteins. Further work will be required to address these possibilities.

Another significant difference between the bacterial and human RecQ structures is the β-hairpin structure that forms part of the WH fold. This hairpin is significantly longer in RECQ1 than in RecQ of E. coli and other bacteria. By analogy with structures of other DNA-bound helicases, we suspected that the β-hairpin in RECQ1 may be important in DNA strand separation. Indeed, mutating the aromatic residue to alanine (Y564A), or deletion of portions of the hairpin, severely reduce helicase activity, without affecting DNA-dependent ATPase activity. A possible interpretation of these observations is that the hairpin in RECQ1 functions in a similar manner suggested for HEL308, UvrD, and PcrA, whereby an aromatic residue at the tip of the hairpin interacts with the DNA base pairs.

An alignment of putative β-hairpin sequences from several RecQ-family proteins is shown in Fig. 6. The sequences of the hairpins cluster according to the overall sequence similarity of the proteins; an aromatic residue appears in most proteins, with the notable exception of BLM and related proteins. Bacterial proteins have a much shorter hairpin, which, in most cases, has a histidine in place of an aromatic residue at the tip (the 2 sequences shown, #12 and #14, were selected because they are exceptional in having phenylalanine at this position). There are several reports of differences in substrate specificity between RecQ family members. For example, differences between full-length RECQ1 and BLM suggest that there must be some key structural features that distinguish the catalytic domains of these 2 enzymes (22). The substrate specificity of RECQ1 is also distinct from that of E. coli RecQ because, for example, the bacterial enzyme can unwind blunt-ended duplexes and G-quadruplex DNA, which cannot be resolved by RECQ1 (41, 42). Gel mobility shift experiments showed that the RQC domain of E. coli RecQ and BLM is required for G-quadruplex DNA binding, suggesting that regions of this domain play an important role of DNA substrate recognition (43). We propose that the β-hairpin may also represent an important determinant for the distinct substrate specificity of RECQ1. Our analysis of mutations in the β-hairpin of E. coli RecQ lend further support to a mechanistic difference between RecQ family members; it would be interesting to explore how these differences relate to biological specificity or catalytic efficiency of RecQ enzymes.

Methods

Full details of the methods used are provided in SI Methods.

Protein Expression and Purification.

Human RECQ1^T1 and all mutant derivatives were expressed from constructs encompassing amino acids 49–616. These are fused to a C-terminal extension, AENLYFQ*SHHHHHHDYKDDDDK, containing a TEV-protease cleavage site (*), a hexahistidine and a Flag tag. E. coli RecQ and its mutant derivatives, were expressed from constructs fused to an N-terminal extension, MHHHHHHSSGVDLGTENLYFQ*SM. The proteins were expressed in E. coli and purified by using nickel-affinity purification, followed by preparative gel filtration. Full-length RECQ1 was expressed in insect cells as described previously (29).

Protein Crystallization and Structure Determination.

Purified RECQ1^T1 protein was crystallized in the presence of ATP-γ-S and oligonucleotides by vapor diffusion from sitting drops equilibrated against 0.2 M sodium bromide, 20% PEG 3350, 10% ethylene glycol, 0.1 M bis-Tris propane (pH 7.5). Data were collected from frozen crystals on beamline X10SA at the Swiss Light Source (Paul Scherrer Institut, Switzerland), to a resolution of 2.05 Å. Poor initial phases obtained by using molecular replacement with the coordinates of E. coli RecQ (PDB ID code 1OYY) were improved substantially by multicrystal averaging with a different crystal form grown from protein derived from an N-terminally tagged version of RECQ1 (termed RECQ1^T2). The quality of the resultant maps was excellent and allowed the entire chain to be traced. The final model, comprising residues 63–592, was refined with REFMAC5 to an R cryst/R free of 22.7/27.4, respectively. The data collection and refinement statistics are summarized in Table S1.

DNA helicase, ATPase assays, and analytical size exclusion chromatography were performed as described previously (23, 29). The sequences of the oligonucleotides used for the helicase assays are reported in Table S2.