Function of Auxiliary Domains of the DEAH/RHA Helicase DHX36 in RNA Remodeling

Abstract

The DEAH/RHA helicase DHX36 has been linked to cellular RNA and DNA quadruplex structures and to AU-rich RNA elements. In vitro, DHX36 remodels DNA and RNA quadruplex structures and unwinds DNA duplexes in an ATP-dependent manner. DHX36 contains the superfamily 2 helicase core and several auxiliary domains that are conserved in orthologs of the enzyme. The role of these auxiliary domains for the enzymatic function of DHX36 is not well understood. Here, we combine structural and biochemical studies to define the function of three auxiliary domains that contact nucleic acid. We first report the crystal structure of mouse DHX36 bound to ADP. The structure reveals an overall architecture of mouse DHX36 that is similar to previously reported architectures of fly and bovine DHX36. In addition, our structure shows conformational changes that accompany stages of the ATP-binding and hydrolysis cycle. We then examine the roles of the DHX36-Specific Motif (DSM), the OB-fold, and a conserved β-hairpin (β-HP) in mouse DHX36 in the remodeling of RNA structures. We then demonstrate and characterize RNA duplex unwinding for DHX36 and examine the remodeling of inter- and intramolecular RNA quadruplex structures. We find that the DSM functions not only as a quadruplex binding adaptor, but also promotes the remodeling of RNA duplex and quadruplex structures. The OB-fold and the β-HP contribute to RNA binding. Both domains are also essential for remodeling RNA quadruplex and duplex structures. Our data reveal roles of auxiliary domains for multiple steps of the nucleic acid remodeling reactions.

Graphical Abstract

INTRODUCTION

The DEAH/RHA helicase DHX36 (RHAU, G4R1) is conserved in metazoans and essential for embryogenesis and organogenesis in mice [1],[2]. DHX36 has been linked to AU-rich RNA elements and to DNA and RNA quadruplex structures in cells [3],[4]. DHX36 accounts for most of the quadruplex resolving activity in HeLa cell extracts [5] and associates with quadruplex structures in other cells [2, 4],[6],[7]. DHX36 remodels DNA and RNA quadruplexes in vitro [8],[9],[10]. Recent studies implicate DHX36 in quadruplex remodeling in cells [4].

Crystal structures have been reported for DHX36 from bovine and fly [11], [12]. The protein is comprised of the structural domains that are characteristic for DEAH/RHA helicases, including two RecA-like helicase domains that form the helicase core, a winged-helix (WH), a ratchet, and an OB-fold [13],[14],[15]. The domains are arranged in a pyramidal structure, as seen in other DEAH/RHA helicases [14],[15]. DHX36 contacts the quadruplex structure through an N-terminal, DHX36 specific motif (DSM) [11]. Contacts to nucleic acid outside of the quadruplex are also made by other domains, including the OB-fold, the ratchet, and the helicase core [11],[12].

DHX36 remodels quadruplex structures by ATP-driven translocation along the nucleic acid [10]. The helicase core binds to an unpaired region of the substrate, which needs to be longer than 5 nucleotides and located 3’ to the quadruplex structure [10]. DHX36 partially destabilizes the quadruplex in a non-ATP dependent fashion [8],[11],[16]. ATP-driven translocation of the helicase core on the nucleic acid then threads the unpaired nucleic acid through the nucleic acid binding tunnel, thereby resolving the structure [12]. This translocation-based mechanism explains the remodeling of intra- and intermolecular quadruplex structures, and is consistent with duplex remodeling activity by DHX36 [10],[12]. Other DEAH/RHA helicases also remodel RNA and DNA structures by an ATP-driven translocation-based mechanism [17],[18],[19]. Clear roles have been shown for the helicase core and the DSM in the quadruplex remodeling process. What roles, if any, other conserved domains play in the remodeling of nucleic acid structures is not clear.

Given the high degree of conservation of DHX36, it is important to delineate whether and how these domains impact the biochemical activity of DHX36. Here, we report the crystal structure of mouse DHX36 and, guided by the structure, interrogate the roles of several conserved domains and structural elements in remodeling of RNA duplex and quadruplex structures. Our structure of mouse DHX36 bound to ADP indicates an overall architecture similar to that previously reported for fly and bovine DHX36. Differences compared with previous structures are consistent with conformational changes that accompany the stages of ATP binding and hydrolysis and the RNA remodeling cycle.

Our structure-function analysis focuses on the roles of several nucleic acid binding domains in the remodeling of RNA structures. Although DHX36 has been functionally and physically linked to RNA in cells [4], most biochemical studies have been conducted on DNA. To our knowledge, remodeling of RNA duplexes by DHX36 has not been experimentally tested. We demonstrate ATP-dependent RNA duplex remodeling activity by DHX36. We further show that the DSM functions not only as a quadruplex binding adaptor, but also promotes the remodeling of both RNA duplex and quadruplex structures. We then demonstrate that two previously uncharacterized DHX36 regions, a 5’-β-hairpin in the helicase core, and the OB-fold domain are essential for the ability of DHX36 to bind and remodel both RNA quadruplex and duplex structures. Our data reveal that auxiliary domains of DHX36 promote not only binding to, but also the remodeling of nucleic acid structures. The results collectively suggest functional integration of auxiliary domains of DHX36 for multiple steps of the nucleic acid remodeling reaction.

RESULTS

Crystal structure of mouse DHX36

To provide a firm basis for a structure-function analysis, we determined the structure of mouse DHX36 (mDHX36). We solved crystal structures of a mDHX36 construct without the 45 N-terminal and the 26 C-terminal residues in complex with ADP (mDHX36_46–982, Fig. 1a,b). The N-terminal residues (aa: 46–153), which include the DSM, were not visible in the electron density maps, most likely due to disorder. The overall structure of mDHX36 closely resembles structures for the previously reported bovine and fly DHX36 with the six domains typical for DEAH/RHA helicases (Suppl. Fig. 1) [11],[12],[15],[14]: an N-terminal domain (NTD, aa 46–203), the two RecA-like helicase domains (aa 204–619), a winged-helix (WH) and ratchet domain, together referred to as the HA2 domain (aa 620–834), an OB-fold (aa 835–895), and a C-terminal domain (CTD, aa 896–982). Similar to the fly and bovine DHX36, mDHX36 adopts a pyramidal structure. The helicase core forms a rectangular base. The extended C-terminus with the HA2, the OB-fold, and the CTD form the apex (Fig. 1b).

Figure 1. Crystal structure of mouse DHX36 bound to ADP.

(a) Domain organization of mouse DHX36. The bottom bar indicates the construct (WT_46–982) used for crystallization; (b) Cartoon representation of the crystal structure of mouse DHX36 in complex with ADP (6UP4). Domains are colored and labeled as in panel (a). ADP is shown in green. (c) Rotation of the C-terminal domains in different nucleotide and nucleic acid bound DHX36 structures (mouse DHX36 bound to ADP: hot pink; bovine DHX36 bound to ADP.BeF3⁻(PDB: 5VHC): silver; bovine DHX36 bound to a DNA quadruplex (PDB: 5VHE): pale teal). The structures were aligned by superimposition of the RecA1 domain. Alpha-helices are shown as cylinders. The RecA1 and RecA2 domains have been omitted from the structures. (d) Movement of the RecA2 domain relative to the RecA1 domain in the different nucleotide and nucleic acid bound DHX36 structures. Structures and alignment as in panel (c). For clarity, the extended C-terminal domains are not depicted.

The HA2 and OB-fold together represent one side of the nucleic acid binding site. Both domains and the CTD are slightly rotated, compared to structures of DHX36 bound to nucleic acids [11],[12] (Fig. 1c). The rotation of these domains is even more pronounced in the bovine DHX36 structure with ADP-BeF₃⁻ [11] (Fig. 1c). We speculate that the rotation of the HA2, OB-fold, and CTD reflects conformational changes dictated by the state of the ATPase cycle and the bound nucleic acids. Consistent with this notion, several residues in the OB-fold that contact nucleic acids are reoriented in our ADP-bound structures (Suppl. Fig. 2a). We also observe movement of the RecA2 domain away from the RecA1 domain in our structure, compared to structures with nucleic acids and with the ATP-ground state analog, ADP-BeF₃⁻ [11],(Fig. 1d, Suppl. Fig. 3). Other differences between our and previous DHX36 structures include altered arrangement of residues in the helicase domains Ia, III, and IV, in the loop that connects the two helicase domains (Suppl. Fig. 2b–e), and in the conserved 3’-β-hairpin element, which is thought to act as 3’ bookend for the nucleic acid [17]. Specifically, Arg 287 is reoriented in the ADP-bound state, compared to the nucleic acid bound states (Suppl. Fig. 2f). Collectively, our observations are consistent with conformational changes that accompany different stages of the ATP binding and hydrolysis cycle and the RNA remodeling cycle. Besides establishing the arrangement of the domains in mDHX36, our structure highlights movements of the helicase core that accompany the ATP-binding and -hydrolysis cycle and RNA remodeling.

Remodeling of RNA duplexes and quadruplexes by mDHX36

Having determined the architecture of mouse DHX36, we set out to interrogate the function of conserved auxiliary domains in the remodeling of RNA structures. To provide a basis for this structure-function analysis, we first characterized the ATP-dependent remodeling activity of wild type mDHX36_46–982 (subsequently referred to as WT mDHX36). We examined the activity on three different RNA substrates: (i) a 16-bp duplex with 15 nt unpaired A 3’ to the duplex, (ii) an intermolecular G-quadruplex (GQ⁴) with five G-tetrads formed from four separate RNA oligonucleotides (3’A₁₅G₅UUA), and (iii) an intramolecular G-quadruplex (GQ^I) with five G-tetrads formed from a single RNA oligonucleotide. The RNA duplex and the intramolecular GQ^I had a single unpaired region of 15 A 3’ to the RNA structure, the intermolecular GQ⁴ had four single unpaired region of 15 A 3’ to the RNA structure (Fig. 2, Suppl. Table 2). Circular dichroism indicated that both quadruplexes formed in the parallel orientation, as expected (Suppl. Fig. 4, [20],[8]).

Figure 2. Remodeling of G-quadruplex and RNA duplex substrates by WT mDHX36.

(a) Reaction scheme for pre-steady state remodeling reactions for RNA-duplex and RNA GQ⁴. (b) Left panel: Representative PAGE for RNA duplex remodeling reaction (100 nM WT DHX36, 2 mM ATP-Mg²⁺, 0.5 nM 16-bp RNA duplex). Middle and right panels: Reaction without ATP, and without protein. Cartoons mark RNA substrate and unwound product. The asterisk shows the radiolabel. (c) Left panel: Representative PAGE for RNA GQ⁴ remodeling reaction (100 nM WT DHX36, 2 mM ATP-Mg²⁺, 0.5 nM 5G-tetrad RNA GQ⁴). Middle and right panels: Reaction without ATP, and without protein. Cartoons mark RNA substrate and unwound product. The asterisks show the radiolabel. (d) Reaction scheme for pre-steady state remodeling reactions for the intramolecular RNA- GQ^I. (e) Left panel: Representative PAGE for the RNA GQ^I remodeling reaction (100 nM WT DHX36, 2 mM ATP-Mg²⁺, 0.5 nM 5G-tetrad RNA GQ^I, 100 nM DNA trap). Middle and right panels: Reaction without ATP and without protein. Cartoons mark RNA substrate and unwound product. The asterisk shows the radiolabel. (f) Representative remodeling time courses of WT mDHX36 for RNA duplex (red diamonds), GQ⁴ (green circles) and GQ^I (blue circles). Reaction conditions as in panels B, D. Data points are averages from multiple independent reactions (N ≥ 4). Error bars mark one standard deviation. Lines show best fits to the integrated first order rate law, yielding observed rate constants (k_obs) for duplex RNA, k_obs = 0.15 ± 0.04 min⁻¹; for GQ⁴, k_obs = 3.27 ± 0.88 min⁻¹; and for GQ^I, k_obs = 3.73 ± 0.55 min⁻¹. (g) Dependence of the remodeling rate constant (2 mM ATP-Mg²⁺) on mDHX36 concentrations for the duplex substrate. Data points are averages from multiple independent reactions (N ≥ 6). Error bars mark one standard deviation. The line represents the best fit to the hill equation k_obs= k_obs^max [DHX36] ^H ·((K_{1/2, DHX36}) ^H + [DHX36] ^H)⁻¹ (k_obs: observed remodeling rate constant; k_obs^max: remodeling rate constant at DHX36 saturation; K_{1/2, DHX36}: apparent functional binding constant, H: Hill coefficient), yielding k_obs^max = 1.1 ± 0.1 min⁻¹, K_1/2 = 363 ± 54 nM, H = 1.8 ± 0.3. (h) Dependence of the remodeling rate constant (2 mM ATP-Mg²⁺) on mDHX36 concentrations for the intermolecular GQ⁴ substrate. Data points are averages from multiple independent reactions (N ≥ 3). Error bars mark one standard deviation. Curves represent the best fit to a binding isotherm k_obs= k_obs^max [DHX36]·(K_{1/2, DHX36} + [DHX36])⁻¹, yielding k_obs^max = 4.6 ± 0.3 min⁻¹, K_1/2 = 20.9 ± 4.4 nM. (i) Dependence of the remodeling rate constant (2 mM ATP-Mg²⁺, 100 nM DNA trap) on mDHX36 concentrations for the intramolecular GQ⁴ substrate. Data points are averages from multiple independent reactions (N ≥ 4). Error bars mark one standard deviation. DHX36 was saturated at the lowest experimentally accessible concentration. The average of k_obs values at all shown concentrations is k_obs^max = 4.6 ± 0.4 min⁻¹, K_1/2 < 0.06 nM.

We measured RNA remodeling under pre-steady state conditions with excess of DHX36 over RNA (Fig. 2a). These reaction conditions permit multiple cycles of enzyme substrate binding [21]. The kinetic interpretation of this reaction regime is simpler than for steady state reactions, which involve multiple substrate turnovers [21]. We detected clear, ATP-dependent duplex remodeling by mDHX36 (Fig. 2b, left panel). No remodeling was observed without ATP and without mDHX36 (Fig. 2b, middle and right panel). These data demonstrate that mDHX36 unwinds RNA duplexes.

mDHX36 also resolved the intermolecular GQ⁴ in an ATP-dependent manner (Fig. 2c, left panel). No reaction was seen without ATP and without mDHX36 (Fig. 2c, middle and right panel). Remodeling of GQ^I was measured in the presence of a DNA trap, a DNA oligonucleotide that hybridizes to the quadruplex region and thereby prevents re-formation of the quadruplex upon unfolding (Fig. 2d, [20]). The DNA trap did not measurably impact mDHX36 activity (Suppl. Fig. 5). mDHX36 resolved the intramolecular GQ^I with, but not without ATP (Fig. 2e).

Under identical conditions, mDHX36 resolved the quadruplex substrates markedly faster than the duplex (Fig. 2f). To understand whether these variations were caused by differing affinities of mDHX36 to the substrates, by differences in the remodeling rate constants, or by a combination of both, we measured apparent remodeling rate constants (k_obs) with increasing mDHX36 concentrations (Fig. 2g–i). We observed a weaker affinity of mDHX36 for the duplex, compared to the quadruplex substrates (Fig. 2g–i). The duplex remodeling rate constant at enzyme saturation (k_obs^max) was also lower than quadruplex remodeling rate constants at enzyme saturation (k_obs^max) (Fig. 2g–i). In addition, we measured a lower affinity of mDHX36 for the intermolecular GQ⁴ with four unpaired regions, compared to the intramolecular GQ¹ with only one unpaired region (Fig. 2h,i). Remodeling rate constants at enzyme saturation were roughly similar for both quadruplexes.

Functional binding isotherms for duplex remodeling were sigmoidal, with a Hill coefficient of H = 1.8 ± 0.3 (Fig. 2g). This observation suggests that multiple protomers of mDHX36 cooperate to unwind a duplex, but not the quadruplex substrates. The functional binding isotherms were hyperbolic even for the intermolecular substrate with four unpaired regions (Fig. 2i). This result suggests that remodeling of a quadruplex does not require cooperative function of multiple DHX36 protomers, regardless of the number of unpaired RNA regions. In sum, our observations showed ATP-dependent remodeling activity by mDHX36 on different RNA structures, including an RNA duplex. mDHX36 preferentially bound and remodeled substrates with a quadruplex structure, compared to those with an RNA duplex structure.

The DSM promotes remodeling of RNA quadruplexes and duplexes and binding to RNA quadruplexes

Utilizing the substrates and conditions described above (Fig. 2), we next examined the role of the DHX36 specific motif (DSM) for functional binding and remodeling of RNA substrates. The N-terminal DSM, which is conserved among DHX36 orthologues (Suppl. Fig. 6), contacts the quadruplex in a recent crystal structure ([11], Fig. 3a). The DSM promoted DHX36 binding to quadruplexes in several biochemical studies [22], [23],[16], but appeared dispensable for quadruplex binding in other work [12],[24]. Whether and how the DSM impacts remodeling of RNA duplex substrates by DHX36 is not known.

Figure 3. Impact of deletion of the DSM.

(a) Location of the DSM (green ribbon, box) in DHX36 bound to a DNA quadruplex structure (PDB: 5VHE). (b) Representative PAGE for RNA duplex remodeling reactions (100 nM mDHX36^Δ-DSM, 2 mM ATP-Mg²⁺, 0.5 nM RNA substrate) for the RNA duplex (left panel), the GQ⁴ (middle) and the GQ¹ substrates. (c) Top panel: Dependence of the remodeling rate constant (2 mM ATP-Mg²⁺) on mDHX36^Δ-DSM (red) concentrations for the duplex substrate. Reactions for WT mDHX36 (blue, Fig. 2g) are shown for comparison. Data points are averages from multiple independent reactions (N ≥ 4). Error bars mark one standard deviation. The lines represent the best fit to the Hill equation (Fig. 2g) yielding for mDHX36^Δ-DSM: k_obs^max = 0.65 ± 0.15 min⁻¹, K_1/2 = 460 ± 153 nM, H = 1.9 ± 0.6. Bottom panel: comparison of k_obs^max and K_1/2 values for WT mDHX36 (Fig. 2g) and mDHX36^Δ-DSM for the RNA duplex substrate. Error bars indicate the standard error of the datafit to the Hill equation. (d) Top panel: Dependence of the remodeling rate constant (2 mM ATP-Mg²⁺) on mDHX36^Δ-DSM (red) concentrations for the GQ⁴ substrate. Reactions for WT mDHX36 (blue, Fig. 2h) are shown for comparison. Data points are averages from multiple independent reactions (N ≥ 4). Error bars mark one standard deviation. The lines represent the best fit to the binding isotherm (Fig. 2h) yielding for mDHX36^Δ-DSM: k_obs^max = 1.97 ± 0.33 min⁻¹, K_1/2 = 120 ± 61 nM. Bottom panel: comparison of k_obs^max and K_1/2 values for WT mDHX36 (Fig. 2h) and mDHX36^Δ-DSM for the GQ⁴ substrate. Error bars indicate the standard error of the datafit to the binding isotherm. (e) Top panel: Dependence of the remodeling rate constant (2 mM ATP-Mg²⁺) on mDHX36^Δ-DSM (red) concentrations for the GQ¹ substrate. Reactions for WT mDHX36 (blue, Fig. 2i) are shown for comparison. Data points are averages from multiple independent reactions (N ≥ 4). Error bars mark one standard deviation. The lines represent the best fit to the binding isotherm (Fig. 2i) yielding for mDHX36^Δ-DSM: k_obs^max = 1.2 ± 0.1 min⁻¹, K_1/2 = 75 ± 24 nM. Bottom panel: comparison of k_obs^max and K_1/2 values for WT mDHX36 (Fig. 2i) and mDHX36^Δ-DSM for the GQ1 substrate. Error bars indicate the standard error of the datafit to the Hill equation.

To systematically characterize the contribution of the DSM for RNA quadruplex and duplex remodeling, we deleted the DSM from mDHX36_46–982 (mDHX36^Δ-DSM) and measured remodeling of the three substrates tested above (Fig. 2). mDHX36^Δ-DSM remodeled all substrates in an ATP-dependent manner (Fig. 3b). Differences between wild type (WT) mDHX36 and mDHX36^Δ-DSM were apparent (Fig. 3c–e). For the duplex substrate, the remodeling rate constant at enzyme saturation (k_obs^max) was lower for mDHX36^Δ-DSM, compared to WT mDHX36 (Fig. 3c, lower left panel). The functional affinity (K_1/2) was similar for both enzyme variants, within error (Fig. 3c, lower right). The functional binding isotherm for mDHX36^Δ-DSM was sigmoidal (H = 1.9 ± 0.6), further supporting the notion that multiple protomers of mDHX36 cooperate to unwind the duplex substrate.

For the intermolecular quadruplex substrate (GQ⁴), k_obs^max and functional affinity (K_1/2) were markedly reduced for mDHX36^Δ-DSM, compared to WT mDHX36 (Fig. 3d). For the intramolecular quadruplex substrate (GQ^I), k_obs^max and functional affinity were also reduced for mDHX36^Δ-DSM, compared to WT mDHX36 (Fig. 3e). Deletion of the DSM reduced the functional affinity for GQ^I by almost four orders of magnitude (Fig. 3e, lower right panel). Yet, the functional affinity of mDHX36^Δ-DSM for both quadruplex substrates was similar, even though WT mDHX36 binds significantly tighter to intramolecular than to the intermolecular quadruplex substrate (Fig. 3d, e).

The data thus show that the DSM promotes binding to RNA quadruplexes, but not to duplexes. The differences in the functional affinity of WT mDHX36 for the intra- and the intermolecular quadruplex structures suggest that mDHX36 either prefers one or more specific orientations of unpaired RNA to the quadruplex structures, that multiple options of mDHX36 binding to a single quadruplex is disadvantageous for functional association, or a combination of both scenarios. The DSM appears to confer this characteristic to mDHX36, because deletion of the DSM largely eliminates the differences in the functional affinity for both quadruplex substrates. However, the functional affinity of mDHX36^Δ-DSM for the duplex substrate is still lower than that for the quadruplex substrates, indicating that removal of the DSM does not completely abolish the preferential binding of mDHX36 to quadruplexes.

Our data further reveal a previously unappreciated role of the DSM in the remodeling step of quadruplex and duplex substrates, reflected in the reduced k_obs^max values for mDHX36^Δ-DSM, compared to WT mDHX36 (Fig.3c–e). Although the size of this effect varies for the different substrates, the data suggest a role for the DSM in the conformational re-arrangement of the DHX36 domains that accompany the RNA remodeling step, translocation events, or both [11]. In sum, our data show that (i) the DSM is a binding adaptor for quadruplex, but not for duplex structures (ii) that the DSM plays a role in the remodeling step of both quadruplex and duplex substrates.

The OB-fold promotes binding and remodeling of quadruplex and duplex structures

We next probed the role of the OB-fold (Fig. 4a). This domain is located C-terminal relative to the helicase core and conserved across the DEAH/RHA family [15],[14],[19]. The OB-fold in DHX36 contacts the nucleic acid ([11], [12] Fig. 4a). Based on these contacts, a role of the domain in quadruplex recognition was proposed [11], [12]. However, mutations of residues that contact the nucleic acid did not impact remodeling [11].

Figure 4. Impact of deletion of the OB-fold.

(a) Location of the OB-fold in DHX36 bound to a DNA quadruplex structure (PDB: 5VHE). (b) Representative PAGEs for RNA remodeling reactions (100 nM mDHX36^ΔOB-fold, 2 mM ATP-Mg²⁺, 0.5 nM RNA substrate) for the RNA duplex (left), the GQ⁴ (middle) and the GQ¹ (right) substrates. (c) Left panel: Reaction scheme for steady-state ATPase of RNA mDHX36 with RNA substrates. Right panel: Initial reaction velocity (V₀) of WT mDHX36 (blue) and mDHX36^ΔOB-fold (orange), without or with RNA substrates as indicated (266 nM mDHX36, trace amounts of [γ−³²P] ATP, 0.5 mM ATP-Mg²⁺, 2 μM RNA substrate). Values for V₀ were obtained by multiplying the initial reaction rates with the total ATP concentration for the linear part of the progress curve (< 20% product formation). Data represent the average of 3 independent measurements. The error bars mark one standard deviation. V₀ values were for WT mDHX36: no RNA, V₀ = 0.9 ± 0.1 μM min⁻¹; no RNA and RNase, V₀ = 1.0 ± 0.3 μM min⁻¹; with ss RNA, V₀ = 10.2 ± 1.3 μM min⁻¹; with the RNA duplex substrate, V₀ = 7.8 ± 2.8 μM min⁻¹; with the GQ⁴ substrate, V₀ = 6.6 ± 2.1 μM min⁻¹; and with the GQ¹ substrate, V₀ = 10.6 ± 0.8 μM min⁻¹. For mDHX36^ΔOB-fold: no RNA, V₀ = 1.1 ± 0.5 μM min⁻¹; no RNA and RNase, V₀ = 0.9 ± 0.1 μM min⁻¹; with ss RNA, V₀ = 1.0 ± 0.2 μM min⁻¹; with the RNA duplex substrate, V₀ = 1.3 ± 0.3 μM min⁻¹; with the GQ⁴ substrate, V₀ = 1.3 ± 0.5 μM min⁻¹; and with the GQ¹ substrate, V₀ = 0.8 ± 0.2 μM min⁻¹. (d) Representative PAGEs of equilibrium binding reactions with increasing concentrations of WT mDHX36 (left panels) and mDHX36^ΔOB-fold (middle panels) and RNAs as in panel (c) Cartoons mark free and protein-bound RNA, asterisks show the radiolabel. Right panels: plots of fraction of bound RNA as a function of enzyme concentration. Data points represent an average of three independent measurements. Error bars mark one standard deviation. Lines represent trends.

To clarify the role of the OB-fold for DHX36 activity, we deleted the domain from mDHX36_46–982 (mDHX36^ΔOB-fold). We examined overall folding of the construct by circular dichroism (CD, Suppl. Fig. 9). At 30 °C, mDHX36 ^ΔOB-fold and WT mDHX36 had highly similar spectra, indicating significant α-helical content and thus overall folding of mDHX36^ΔOB-fold (Suppl. Fig. 9). We examined the remodeling activity of mDHX36^ΔOB-fold on the RNA substrates tested above. We did not detect significant remodeling for any of the substrates tested (Fig. 4b). To nevertheless obtain insight into functional characteristics of mDHX36^ΔOB-fold, we measured the RNA-stimulated ATPase activity (Fig. 4c). Without RNA, mDHX36^ΔOB-fold showed basal ATPase activity comparable to WT mDHX36 (Fig. 4c, Suppl. Fig. 7). In contrast to WT mDHX36, no RNA stimulation of the ATPase activity was seen with mDHX36^ΔOB-fold (Fig. 4c). The ATPase activity remained at the level of the unstimulated activity (Fig. 4c, Suppl. Fig. 7). These data suggested that mDHX36^ΔOB-fold retained the capacity to hydrolyse ATP, but lost the ability to bind RNA, couple RNA binding to ATP turnover, or both.

To examine the role of the OB-fold for RNA binding, we measured equilibrium binding of mDHX36^ΔOB-fold to the RNA substrates tested above and to a 31 nt single stranded RNA (Fig. 4d). We observed reduced binding of mDHX36^ΔOB-fold to all substrates, compared to WT mDHX36 (Fig. 4d). WT mDHX36 showed the lowest affinity to ssRNA, followed by the duplex substrate and the quadruplex substrates (Fig. 4d). Affinities of mDHX36^ΔOB-fold followed the same trend, but were generally lower than for WT mDHX36 (Fig. 4d). For the intermolecular GQ⁴ substrate, which contains four unpaired RNA regions, we noticed multiple bound species (Fig. 4d), consistent with binding of several mDHX36 protomers to the RNA. Apparent equilibrium constants of K_1/2 ~ 50 nM for mDHX36^ΔOB-fold binding to the quadruplex substrates indicated a clear ability to bind the quadruplex. Comparison of the apparent affinities of WT and mDHX36^ΔOB-fold for the three substrates suggested that deletion of the OB-fold results in a loss of binding energy that is similar for all substrates (Suppl. Fig. 8). This observation implies that the OB-fold contributes a free binding energy to nucleic acid binding roughly similar for all substrates and largely additive to the free binding energy provided by other nucleic acid binding sites in mDHX36.

Taken together, the RNA binding data and the lack of ATPase stimulation by any RNA substrate by mDHX36^ΔOB-fold indicate a role of the OB-fold in enabling the coupling of the ATPase cycle to nucleic acid binding, and thus a function beyond RNA binding. A mere impact of the OB-fold on RNA binding during the ATPase cycle would result in a measurable ATPase stimulation with the quadruplex substrates. However, no stimulation was observed (Fig. 4c). Our data thus collectively indicate that the OB-fold, beyond a role in promoting binding to nucleic acids, is essential for coupling of the ATPase cycle to nucleic acid binding and thus for the ATP-driven remodeling of quadruplex and duplex structures. The impact of the OB-fold deletion on the binding to quadruplex structures further suggests that the DSM alone is not sufficient to confer the high affinity for quadruplexes to mDHX36 (Fig. 4a). Instead, high affinity binding of mDHX36 to quadruplex structures appears to involve multiple mDHX36 domains.

The β-hairpin promotes binding and remodeling of quadruplex and duplex structures

The OB-fold interacts with another conserved structural feature of the DEAH/RHA family: a β-hairpin (β-HP), which is inserted into the RecA2 domain and contacts the OB-fold across the nucleic acid (Fig. 5a, [15],[14],[19]). The β-HP is also present in Ski2-like helicases, where it has been proposed to promote translocation and strand separation [25]. However, the function of the β-HP has not yet been experimentally validated in any helicase.

Figure 5. Impact of deletion and truncation of the 5’-β-HP.

(a) Top panel: Residues comprising the 5’-β-HP. Residues deleted to generate the mDHX36^Δβ-HP construct are marked in red. Residues replaced by a glycine to generate the mDHX36^{short β-HP} construct are marked in brown. Bottom Panel: Location of the 5’-β-HP in DHX36 bound to a DNA quadruplex structure (PDB: 5VHE). (b) Representative PAGEs for RNA remodeling reactions. Left panels: Reactions with 100 nM mDHX36^Δβ-HP, 2 mM ATP-Mg²⁺, 0.5 nM RNA substrates as indicated by the cartoons on the left. Right panels: Reactions with 100 nM mDHX36^{short β-HP}, 2 mM ATP-Mg²⁺, 0.5 nM RNA substrates as indicated. (c) Initial ATPase velocities (V₀) for WT mDHX36 (blue, for comparison, Fig. 4c), mDHX36^{short β-HP} (purple) and mDHX36^Δβ-HP (green) with and without RNA substrates as indicated (266 nM mDHX36, trace amounts of [γ−³²P] ATP, 0.5 mM ATP-Mg²⁺, 2 μM RNA substrate). Data represent the average of three independent measurements. Error bars mark one standard deviation. Values for V₀ were for mDHX36^Δβ-HP: without RNA, V₀ = 2.9 ± 0.3 μM min⁻¹; without RNA and with RNase, V₀ = 2.3 ± 0.1 μM min⁻¹; with ss RNA, V₀ = 3.3 ± 0.6 μM min⁻¹; with RNA duplex substrate, V₀ = 3.1 ± 1.2 μM min⁻¹; with GQ⁴ substrate, V₀ = 5.0 ± 1.1 μM min⁻¹; and with GQ¹ substrate, V₀ = 3.3 ± 0.7 μM min⁻¹. For mDHX36^{short β-HP}: without RNA, V₀ = 0.6 ± 0.2 μM min⁻¹; without RNA and with RNase, V₀ = 0.4 ± 0.2 μM min⁻¹; with ss RNA, V₀ = 0.6 ± 0.1 μM min⁻¹; with RNA duplex substrate, V₀ = 0.6 ± 0.1 μM min⁻¹; with GQ⁴ substrate, V₀ = 0.8 ± 0.3 μM min⁻¹; and with GQ¹ substrate, V₀ = 0.6 ± 0.1 μM min⁻¹. (d) Representative PAGEs of equilibrium binding reactions with increasing concentrations of mDHX36^Δβ-HP (left panels) and mDHX36^{short β-HP} (middle panels) for RNA structures as indicated. Cartoons mark free and protein-bound RNA, asterisks show the radiolabel. Right panels: plots of fraction of bound RNA as a function of enzyme concentration. Data points represent an average of three independent measurements. Error bars mark one standard deviation. Lines represent smoothed trends.

To determine the role of the β-HP for DHX36 activity, we generated a mDHX36 variant where we deleted this domain from the mDHX36_46–982 construct (mDHX36^Δβ-HP, Fig. 5a). In addition, we generated a variant where we replaced the top of the hairpin, which contacts the OB-fold, with a flexible glycine linker, leaving a shortened β-HP (mDHX36^{short β-HP}, Fig. 5a). Overall folding of the constructs was examined by circular dichroism (CD, Suppl. Fig. 9). At 30 °C, both, mDHX36 ^Δβ-HP and mDHX36^{short β-HP} showed spectra similar to WT mDHX36, indicating significant α-helical content and thus overall folding of the constructs (Suppl. Fig. 9).

We next examined the remodeling activity of both mDHX36^Δβ-HP and mDHX36^{short β-HP} on the RNA substrates tested above. No significant remodeling for any substrate was detected (Fig. 5b). We then measured RNA-stimulated ATPase activity mDHX36^Δβ-HP and mDHX36^{short β-HP} (Fig. 5c). In the absence of RNA, ATPase activity increased with mDHX36^Δβ-HP, compared to WT mDHX36 (Fig. 5c). Addition of RNA did not significantly stimulate this ATPase activity of mDHX36^Δβ-HP, although it did for WT mDHX36 (Fig. 5c). The increase in basal ATPase activity in mDHX36^Δβ-HP, compared to WT mDHX36, raises the possibility that complete removal of the hairpin relaxes the coupling of RNA binding and ATP binding or turnover. mDHX36^{short β-HP} showed reduced ATPase activity at all conditions, with no stimulation by RNA (Fig. 5c). Together, these data for both mDHX36^Δβ-HP and mDHX36^{short β-HP} suggested that mDHX36^Δβ-HP and mDHX36^{short β-HP} could still hydrolyse ATP, but were defective either in RNA binding, in the ability to couple RNA binding to ATP turnover, or in both.

We next measured equilibrium binding of mDHX36^Δβ-HP and mDHX36^{short β-HP} to the RNA substrates tested above (Fig. 5d). For the ssRNA, we detected reduced binding of mDHX36^Δβ-HP, compared to WT mDHX36 (Fig. 5d, left and right panel in upper row). Virtually no binding of mDHX36^{short β-HP} to the ssRNA was seen (Fig. 5d, middle and right panel in upper row). For all other substrates we observed similarly reduced binding of mDHX36^Δβ-HP and mDHX36^{short β-HP}, compared to WT mDHX36 (Fig. 5d). Of note, removal or shortening of the β-HP (mDHX36^Δβ-HP and mDHX36^{short β-HP}) lowered affinity for structured RNAs less than removal of the OB-fold (Fig. 4d), even though deletion of either domain abrogated remodeling activity.

Comparison of the apparent affinities of WT and mDHX36^Δβ-HP and mDHX36^{short β-HP} for the three substrates suggested that both removal and shortening of the β-HP resulted in a loss of binding energy that is similar for all substrates (Suppl. Fig. 8). As seen for the OB-fold, this observation implies that the β-HP contributes free binding energy to nucleic acid binding that is roughly similar for all substrates and largely additive to the free binding energy provided by other nucleic acid binding sites in mDHX36.

Collectively, our data indicate that the β-HP, much like the OB fold, is essential for remodeling of quadruplex and duplex structures and promotes binding to nucleic acids. Together with the data for mDHX36^ΔOB, these observations suggest that the conserved domains contacting the nucleic acid are not only critical for promoting binding of mDHX36 to the RNA, but also for organizing the helicase domains to allow for coupling of RNA binding to the ATPase cycle thus for the ATP-driven remodeling of duplex and quadruplex structures.

DISCUSSION

In this structure-function analysis of mouse DHX36 we report the crystal structure of the protein bound to ADP, and define the biochemical function of three auxiliary domains that are conserved among DHX36 orthologs. Our crystal structures of mDHX36 bound to ADP show a domain architecture resembling fly and bovine DHX36 and structures of other DEAH/RHA helicases [11],[12],[19],[26]. Differences to previous DHX36 structures include a rotation of the WH, ratchet, and OB-fold, compared to structures with bound nucleic acids and/or nucleotide analogs, and a closed conformation of the two helicase core domains (Fig.1). These structural differences are consistent with conformational changes that accompany stages of ATP binding and hydrolysis and the RNA remodeling cycle. While these structural differences are predominantly in highly conserved regions, it is possible that differences, especially in less conserved regions, could also reflect variations between the bovine and the mouse proteins.

Our structure-function analysis focused on roles of the DSM, the OB-fold and the β-HP, three conserved domains that contact nucleic acids. We examined the roles of the domains in the remodeling of RNA structures. Although DHX36 is functionally and physically linked to RNA in cells [4],[27], the majority of biochemical studies had been conducted on DNA substrates. Remodeling of an intermolecular RNA GQ⁴ substrate by DHX36 had been demonstrated previously [10], but differences in experimental conditions between this and our work preclude a direct comparison of obtained values. Yet, the biochemical characterization of RNA quadruplex and duplex remodeling activities is, to our knowledge, the first comparative biochemical analysis of remodeling of different RNA structures and the first demonstration of RNA helicase activity by DHX36.

The functional affinity of mDHX36 for quadruplex substrates is higher than for duplex substrates. This finding is consistent with preferential association of DHX36 to quadruplex regions in cells [4], and mirrors observations for DNA substrates [22],[9]. The functional affinity for the intramolecular quadruplex substrate was higher than for the intermolecular quadruplex substrate with identical number of G-tetrads, suggesting that mDHX36 prefers a substrate with a single unpaired region over substrates with multiple unpaired tails. At saturating concentrations of mDHX36 both quadruplex substrates are unwound at virtually identical rates. This observation implies one or more similar remodeling steps for quadruplex substrates with different architectures.

The remodeling stage of the duplex substrate is traversed slower than that for the quadruplex substrates, even though the thermodynamic stability of the quadruplex structures is higher than that of the duplex substrate [20],[28]. We speculate that this difference is caused by the partial ATP-independent destabilization of the quadruplex, which leaves fewer G-tetrads for the ATP-dependent, translocation-based remodeling step [11],[12],[16]. To remodel the duplex, DHX36 presumably has to translocate more nucleotides, which could conceivably result in a slower remodeling rate constant. However, rigorous mechanistic interpretations of the differences in remodeling rate constants between quadruplex and duplex substrates have to await a model for the mechanism of duplex remodeling by DHX36. In this context, it is likely important to investigate the cooperativity of DHX36 during duplex remodeling. Our data suggest that multiple mDHX36 protomers cooperate to unwind the duplex (Fig. 2). By contrast, quadruplex remodeling requires only a single protomer [16].

Our structure-function analysis of the DSM, the OB-fold and the β-HP reveal functions of these domains in the RNA remodeling cycle beyond RNA binding. This finding is particularly remarkable for the DSM, a small segment that contacts the quadruplex [11]. Consistent with these contacts, deletion of the DSM confers binding defects for quadruplex, but not for duplex substrates (Fig. 3), indicating that the DSM is a binding adaptor for quadruplex, but not for duplex structures. Yet, deletion of the DSM also decreases the rate constant for the remodeling step(s), notably also for duplex remodeling (Fig.3). These findings reveal a role for the DSM in the remodeling step of both quadruplex and duplex substrates. We speculate that the impact on the remodeling step(s) arises from the linker that connects the DSM to the helicase domains and thereby presumably contacts several DHX36 domains, akin to a brace. This linker might not be positioned properly across DHX36 without the DSM, thereby impairing the remodeling step.

The OB-fold, which is conserved among DEAH/RHA helicases, is an integral part of the DHX36 architecture and contacts unpaired nucleotides of the substrate [11],[12]. Accordingly, we find that the OB-fold promotes RNA binding for all substrates, contributing approximately the same free binding energy for all substrates. However, we find that the OB-fold is also essential for coupling of the ATPase cycle to nucleic acid binding and thus for the ATP-driven remodeling of quadruplex and duplex structures (Fig.4). Our findings mirror observations for the DEAH helicase Prp43p, where the OB-fold is essential for activity in vivo [29],[30], and deletion of this domain impairs RNA binding and ATPase stimulation by RNA [29]. In the RHA helicase MLE, mutations in the OB-fold residues diminish RNA-stimulated ATPase activity of the helicase [18]. While the defects in nucleic acid binding in DHX36 and other DEAH/RHA helicases are readily explained by the loss of the RNA contacts, delineation of the exact function of the OB-fold in the coupling of RNA/DNA binding to the ATPase cycle requires more experimentation. We speculate that the OB-fold, together with other domains, is essential for the conformational changes needed to couple RNA binding to the ATPase cycle. This notion would be consistent with the conformational changes in our structure of mDHX36 bound to ADP, relative to structures at different stages of the ATP hydrolysis cycle [11].

We also examined the role of the 5’ β-hairpin (β-HP), which protrudes from the RecA2 domain between motifs V and VI and interacts with the OB-fold (Fig.1). The β-HP is conserved across the DEAH/RHA family and also present in Ski-2-like and NS3/NPH-II helicases [25],[31]. Our data show that the β-HP, like the OB-fold, is essential for remodeling of quadruplex and duplex structures and also promotes binding to nucleic acids. The β-HP contributes approximately the same free binding energy to RNA binding, regardless of which substrate is bound. This observation is consistent with contacts of β-HP to the unpaired nucleic acid region [11]. In addition, both deletion and shortening of the β-HP abrogate ATP-driven RNA remodeling and coupling of the ATPase cycle to RNA binding (Fig.5). These observations suggest that the β-HP also organizes conformational changes of perhaps multiple mDHX36 domains in response to RNA and ATP binding and ATP hydrolysis. This function appears sensible, because the β-HP bridges several enzyme domains (RecA2, HA2, OB-fold) and also contacts the NA [11]. Deletion of the β-HP, which eliminates contacts between the RecA2 and the extended CTD domain, increases the basal, non-RNA stimulated ATPase activity of mDHX36, compared to WT and other mDHX36 variants (Fig.5). This observation suggests a role of the β-HP in organizing mDHX36 domains even without RNA bound. A critical role of the β-HP for enzyme activity, while not previously demonstrated for DHX36, is consistent with findings for other helicases with a β-HP. In Prp43p, point mutations in the β-HP confer cold-sensitive and slow growth phenotypes [30],[32]. In DHX29, point mutations in the β-HP lead to defects in 48S formation in vivo, and a reduction in RNA-stimulated ATPase activity in vitro [33].

Collectively, our structure-function analysis indicates that mDHX36 domains which contact nucleic acids have roles beyond creating the DNA or RNA binding site. Consistent with the compact architecture of mDHX36, the DSM, the OB-fold and the β-HP domains contribute to the coupling of RNA binding to the ATPase cycle and thus to RNA remodeling. Although the architecture of mDHX36 consists of clear modules, all of these appear to work closely together during several reaction steps. The degree of conservation of the domain architecture in DEAH/RHA helicases raises the possibility that the domains in other enzymes of this family play similar, tightly integrated roles. Despite this integration, each helicase appears adapted to specific functions, at least in cells. For DHX36 this is association with quadruplex substrates. Yet, the enzyme readily unwinds RNA duplexes.

MATERIALS AND METHODS

Construction of expression plasmids

The cDNA sequence of mouse DHX36 (GenBank reference BC138061, AA138062, residues 46–982) was PCR amplified using Phusion® Hot Start Flex DNA polymerase (New England Biolabs) and inserted into pSMT3 vector (provided by Dr. Christopher Lima, Sloan-Kettering Institute) between SalI/ NotI restriction sites. The mutant constructs were also generated by the standard PCR cloning strategy. All constructs were verified by DNA sequencing. The primers used in this study as follows, with restriction sites underlined:

1. WT, sense: 5’ GTTGTTGTCGACAAATCCCGCACACCTTAAGGGTCGCG 3’ and antisense: 5’ GTTGTTGCGGCCGCTCAAGTTTTGATCAAGTCTAGAATAGCTG 3’

2. Δ β-HP, sense: 5’ GGAGGAGTTAGTAAAGCTAATGCCAAACAG 3’ and antisense: 5’ CTTTACTAACTCCTCCATCTATTACATAAACC 3’

3. short β-HP, sense: 5’ AAAGAAGGATCTGCTGAGTGGGTTAGTAAAGC 3’ and antisense: 5’ CAGCAGATCCTTCTTTAATTTTTCCTCCATCT 3’

4. Δ DSM, sense: 5’ CGTCGACAAGCCAAGAAGCAGACGCAGAAGAAC 3’ and antisense: 5’ CTTGGCTTGTCGACGGAGCTCGAATTCGGATC 3’

5. Δ OB-fold, sense: 5’ TCCCAAAACAGAAGTGTCACCATACTGCCTCC 3’ and antisense: 5’ CACTTCTGTTTTGGGATATAAACCAGCACAGAT 3’

DHX36 expression and purification

Mouse dhx36-pSMT3 (46–982) and truncation mutants were expressed as N-terminal His₆-SUMO fusion protein in Escherichia coli BL21-Codonplus (DE3)-RIPL cells (Agilent Technologies, Santa Clara, CA). The culture was grown at 37°C in LB medium containing 50 μg/ml kanamycin until the absorbance at 600 reaches 0.8–1.0. The protein expression was induced by 0.2% lactose (MilliporeSigma, Burlington, MA) or 0.2 mM IPTG at 18°C for 48 h. The cells were harvested and resuspended in lysis buffer containing 50 mM Tris-HCl pH 8.0, 13% glycerol, 12 mM imidazole, 300 mM NaCl and stored at −80°C until ready for use.

Frozen cells were thawed and disrupted by sonication in lysis buffer in addition of 1 mM DTT, 1 mM PMSF and Protease inhibitor cocktail (complete, EDTA-free, Roche). The soluble protein was collected by centrifugation at 35,000 rpm, 30 min, and 4°C. The crude soluble protein was treated with 0.1% polyethyleneimine (Sigma) to eliminate endogenous nucleic acid contaminants. The nucleic acid pellet was then removed by centrifugation at 14,000 rpm, 20 min, 4°C. The nucleic acid free protein solution wa s passed through Ni-column (HisPrep FF 16/10; GE Healthcare) equilibrated in 50 mM Tris-HCl, pH 8.0, 300 mM/1 M NaCl, 20% glycerol, 20 mM imidazole, 1 mM DTT. Protein fractions containing His₆-SUMO fusion protein was incubated with Ulp1 protease (plasmid provided by Sloan-Kettering Institute) and dialyzed against buffer containing 50 mM Tris-HCl, pH 8.0, 350 mM NaCl, 30% glycerol, 1 mM DTT for overnight, at 4°C. The proteolysis reaction was loa ded onto Ni-column to remove free His₆-SUMO tag, non-cleavage protein, and Ulp1 protease. The unbound fractions containing mDHX36 was dialyzed against a buffer containing 25 mM HEPES, pH 7.0, 20% glycerol, 1 mM DTT and then loaded on to Source 15S cation exchange column (GE Healthcare Biosciences, Pittsburgh, PA). mDHX36 protein was eluted by a step gradient of 0–1 M NaCl. Purified protein was precipitated by addition of 87% ammonium sulfate and resuspended in storage buffer containing 25 mM Tris-HCl, pH 7.5, 150 mM NaCl, 5% glycerol, 1 mM DTT and loaded onto S200 gel filtration (Superdex200 HR 10/300 GL; GE Healthcare Biosciences, Pittsburgh, PA) pre-equilibrated with the same buffer. Purified mDHX36 fractions were pooled and concentrated by centrifugation at 2800 g using an Amicon Ultra (MW cutoff 30K) concentrator. The concentrated samples (9–15 mg/ml) were stored in 25 mM Tris-HCl, pH 7.5, 150 mM NaCl, 50% glycerol, and 1 mM DTT at −80°C. The mutant protein expression and purification were the same as native protein.

Crystallization, data collection, and structure determination

Crystals were grown by hanging drop vapor diffusion at 18°C by mixing equal volume of protein and crystallization reagent. The complex of mDHX36 in complex with 2 mM ADP was crystallized with 10% PEG 5000 MME, 0.1 M HEPES pH 7.0, 5% v/v Tacsimate. Crystals were cryoprotected by 25% (v/v) glycerol in crystallization condition. X-ray diffraction data were collected at the SER-CAT 23IDD and IDB beamline at Advanced Photon Source (APS, Argonne, IL, U.S.A.). All data were processed and scaled by HKL2000 [34] and XDS [35]. The mDHX36 structure was determined using a Multi-wavelength anomalous diffraction (MAD) dataset collected from crystals obtained with Se-Met protein samples. The structure was built with the program COOT [36] and refined with PHENIX [37]. The crystal structures were validated by the MolProbity server [38].

Preparation of oligonucleotides and radiolabeled RNA substrates

RNA and DNA oligonucleotides used in this study were purchased from Sigma and Dharmacon. Sequences of oligonucleotides used in this study are as follows (G-tetrads and duplex region underlined). GQ⁴: 5’ UUAGGGGGAAAAAAAAAAAAAAA 3’. GQ^I: 5’ UUAGGGGGAGGGGGAUGGGGGAGGGGGAAAAAAAAAAAAAAA 3’. R16 (top strand): 5’ AGCACCGUAAAGACGC 3’ and R31 (bottom strand with 3’ end overhang): 5’ GCGUCUUUACGGUGCUAAAAAAAAAAAAAAA 3’. The sequence of DNA TRAP used for GQ^I remodeling assays: 5’ TCCCCCATCCCCCTCCCCC 3’. The 5’-end of the RNA (GQ⁴, GQ^I and R16) was radiolabeled using T4 polynucleotide kinase (NEB) followed by purification on denaturing PAGE. The duplex substrate was generated by annealing the radiolabeled top strand (R16) to its corresponding bottom strand (R31) in 10mM MOPS, pH 6.5, 50mM KCl and 1mM EDTA, followed by purification on non-denaturing PAGE. The intermolecular and intramolecular G4 substrates were formed by adding the respective radiolabeled oligos to a buffer containing 10mM MOPS, pH 6.5, 50mM KCl and 1mM EDTA. The solution was heated to 98 °C for 10 min, slow cooled to 0 °C overnight. Following this, the structured substrates were purified on non-denaturing PAGE. The annealed radiolabeled G4 and duplex substrates were gel eluted and stored in a buffer containing 10mM MOPS, pH 7, 50mM KCl and 0.1 mM MgCl_2.

Remodeling reactions

Reactions were performed at 30°C in a temperature-controlled heat block in a buffer with 40mM Tris-HCl (pH 8.0), 100 mM KCl, 0.5 mM MgCl₂, 6% glycerol (vol/vol), 0.01% IGEPAL, 2mM DTT, 0.3 U/ul RNase Inhibitor (Roche). Prior to the reaction, radiolabeled RNA substrate (0.5 nM final concentration) was incubated with the indicated concentration of mDHX36 (WT or mutants). Reactions were initiated by addition of equimolar ATP and MgCl₂ (2 mM final concentration, unless otherwise stated). For remodeling reactions with GQ^I, 200-fold excess of DNA trap (100 nM) was added at reaction start. At the times indicated, aliquots were removed and the reaction was stopped by addition of equal volume of a buffer containing 1 %SDS, 50 mM EDTA, 10% glycerol, 0.05% (w/v) xylene cyanol and 0.05% (w/v) bromophenol blue. Samples were then applied to a 15% non-denaturing PAGE and the structured and single stranded RNAs were separated by electrophoresis at 15V/cm. Gels were dried. Bands were visualized on a Typhoon Phosphorimager (GE health care) and quantified using ImageQuant 5.2 software (Molecular Dynamics). The fractions of single-stranded and structured RNA were determined from the relative amounts of radioactivity in the respective bands. Observed rate constants (k_obs) for the remodeling reactions were determined by fitting time courses of the fraction unwound RNA (Frac P) to an integrated first order rate law according to:

$$\text{Frac}\mathrm{P}=\mathrm{A}\cdot \left(1-{\mathrm{e}}^{-(kobs\cdot t)}\right),$$ (Eq. 1)

A is the reaction amplitude, t is time. When both, unwinding and annealing reactions take place, the observed first-order rate constant k_obs is the sum of rate constants for unwinding (k_unw) and annealing (k_ann) reactions, according to :

$$k_{obs}\text{=}{k}_{unw}\text{+}{k}_{ann}.$$ (Eq. 2)

The reaction amplitude also contains contributions from unwinding (k_unw) and annealing (k_ann) rate constants, according to

$$\mathrm{A}=k_{unw}\cdot {\left(k_{unw}+k_{ann}\right)}^{-1},$$ (Eq. 3)

A detailed derivation of the equations and corresponding discussion has been published in reference [39].

RNA-Protein equilibrium binding reactions

Binding reactions were performed at 30°C in a temperature-controlled heat block in a buffer with 40mM Tris-HCl (pH 8.0), 100 mM KCl, 0.5 mM MgCl₂, 6% glycerol (vol/vol), 0.01% IGEPAL, 2mM DTT, 0.3 U/ul RNase Inhibitor (Roche), 0.5 nM radiolabeled RNA substrate, indicated concentration of mDHX36 (WT or mutants) and incubated for 30 min [40]. Samples were then applied to a 4% non-denaturing PAGE and the mDHX36-RNA complexes and structured RNAs were separated by electrophoresis at 4°C. Gels were dried and bands were visualized on a Typhoon Phosphorimager (GE health care).

ATPase reactions

ATPase measurements were performed at 30°C in a buffer containing 40mM Tris-HCl (pH 8.0), 100 mM KCl, 0.5 mM MgCl₂, 6% glycerol (vol/vol), 0.01% IGEPAL, 2mM DTT, 0.3 U/ul RNase Inhibitor (Roche) and indicated concentration of mDHX36 (WT or mutants). mDHX36 was preincubated with 2 μM RNA substrate for 5 min before reaction start. Reactions were initiated by addition of a mixture of trace amounts of [γ−³²P] ATP and 0.5 mM ATP. All reactions contained equimolar ATP and MgCl₂ with 0.5 mM MgCl₂ excess. At various time points, 1 μL aliquots were removed and applied to a PEI-cellulose thin-layer chromatography plate (20 cm X 20 cm; Selecto Scientific). Hydrolysis of [γ−³²P] ATP was monitored by TLC, as described [40]. The PEI plate was developed with 0.5 M LiCl and 1.5 M formic acid and subsequently dried. Radioactivity was quantified with a Phosphorimager (GE) and the ImageQuant software (Molecular Dynamics). Initial rates of ATP hydrolysis were determined by linear least-squares fit to the initial phase of the reaction.

Circular Dichroism Spectroscopy

RNA G4Qs were prepared in buffers containing 10 mM MOPS (pH 6.5) in the presence of 50 mM KCl at a concentration of 4 μM (GQ⁴) and 1 μM (GQ¹), annealed by heating to 95°C and then cooling slowly to room temperature. CD of RNA oligonucleotides was determined at 30°C by an Applied Photophysics PiStar 180 spectropolarimeter equipped with a temperature controller. An average of three CD spectra ranging from 220 to 340 nm was recorded in a 10 mm path length cuvette at a scan rate of 1nm/sec with a 1-sec response time, 1 nm bandwidth, and continuous scan mode.

CD spectra for wildtype and mutant mDHX36 variants were recorded at 30 °C in a Jasco J-815 spectrophotometer in a quartz cuvette with a path length of 1 mm. Protein stocks were diluted into a 10-fold excess of 25 mM NaH₂PO₄, pH 7.5, 150 mM NaF for a final volume of 200 μl and a final protein concentration of 0.05 mg/ml. For each construct, three independent CD spectra (195 – 260 nm) were recorded in a 1 mm path length cuvette (scan rate: 50 nm/min, response time: 2s, bandwidth: 1 nm, continuous scan mode). The average of the spectra was then computed. The molar ellipticity was calculated from the observed ellipticity values and protein concentrations, and plotted against the standard wavelengths from 195 to 260 nm.

Accession Codes

Coordinates and structure factors of mouseDHX36-ADP complex (PDB ID: 6UP4) have been deposited in the Protein Data Bank.

Highlights.

• Crystal structure of mouse DHX36 bound to ADP.

• Demonstration of ATP-dependent RNA duplex unwinding by DHX36.

• Comparison of quadruplex and duplex remodeling activities by mDHX36.

• The auxiliary DSM, OB-fold, and β-HP domains are involved in multiple steps of the RNA remodeling reaction.

Abbreviations

RNA

Ribonucleic acid

ADP

Adenosine di-phosphate

ATP

Adenosine tri-phosphate

DSM

DHX36-specific motif

OB-fold

Oligonucleotide/oligosaccharide-binding fold

β-HP

β-hairpin

WH

Winged helix

mDHX36

mouse DHX36

NTD

N-terminal domain

CTD

C-terminal domain

HA2

Helicase associated domain 2

GQ⁴

intermolecular G-quadruplex

GQ¹

intramolecular G-quadruplex

k_obs

apparent remodeling rate constants

kobsmax

remodeling rate constant at enzyme saturation

K_1/2

functional affinity

H

Hill coefficient

WT

wild type