Inter-domain communication in the dimeric DEAD-box helicase Hera from T. thermophilus and implications for the mechanism of RNA unwinding

Abstract

The Thermus thermophilus DEAD-box helicase Hera consists of the conserved helicase core, followed by a dimerization domain (DD) and an RNA-binding domain (RBD). The RBD mediates high-affinity binding to an RNA hairpin; the DD mediates formation of a stable dimer. In the dimer, the active sites of the two helicase cores face each other in an ideal configuration to cooperate functionally in RNA unwinding. Here, we dissect the communication between the two RBDs and helicase cores by characterizing dimeric deletion variants with two cores, but two, one, or no RBDs, variants with both RBDs, but two, one, or no functional core, and variants with one core and one RBD, either on the same or opposite protomers. We show that RNA binds to Hera in a two-step mechanism, with an initial interaction between the RBD and a hairpin, followed by the interaction of the core with the flanking single- or double-stranded region. The duplex preferentially interacts with the core on the same protomer in the absence of ATP, but in the presence of ATP, interactions with the other core become possible. Overall, our results point to limited but significant cooperativity between the two protomers in RNA unwinding.

Graphical Abstract

Open in new tabDownload slide

Introduction

RNA helicases of the DEAD-box family unwind RNA duplexes in an ATP-dependent reaction (reviewed in [1–3]). They share a common helicase core of two flexibly linked globular domains that carry the helicase signature motifs mediating ATP binding and hydrolysis, RNA binding, and duplex unwinding. In many representatives, additional domains flanking the core region affect nucleotide binding and hydrolysis, and contribute to RNA binding and specificity, or binding of protein partners [4–13], or to duplex destabilization [14] (recently reviewed in [1]).

Hera is a DEAD-box protein from Thermus thermophilus [15] (reviewed in [16]) that consists of a helicase core, followed by a bipartite C-terminal extension that contains a dimerization domain (DD) and an RNA-binding domain (RBD) [17–20] (Fig. 1A). The DD in Hera mediates the formation of a stable dimer [19], even at picomolar concentrations [8] (Fig. 1B). The RBD, attached to the DD by a short double-β-hairpin structure [17], consists of an RNA recognition motif (RRM) with a central four-stranded β-sheet flanked by an α-helix and a disordered C-terminal tail of ten amino acids [17]. The RBD mediates binding of Hera to RNAs containing hairpins. In this binding mode, the central region of the RRM interacts with single-stranded RNA flanking the hairpin, and the C-terminal tail binds to the double-stranded stem [21].

In vivo, Hera binds to a large number of RNAs, consistent with a general role as an RNA chaperone [22]. In vitro, the RBD mediates binding of Hera to 23S rRNA fragments comprising hairpin 92 and to RNase P RNA [8]. Binding of these RNAs to Hera induces a conformational change of the helicase core to a compact, closed state [8, 22], leading to the formation of the catalytic site for ATP hydrolysis and the bipartite RNA binding site extending over both domains of the core [23]. Hera unwinds a minimal 32/9mer RNA substrate comprising hairpin 92 and an adjacent 9-bp helix derived from helix 91 of 23S rRNA in an ATP-dependent reaction in vitro [8]. The 32/9mer RNA substrate was originally identified as a minimal in vitro substrate for the DEAD-box protein DbpA (Escherichia coli) [24] and for its Bacillus subtilis homolog YxiN [11, 12, 25, 26], two helicases involved in ribosome biogenesis. This or similar hairpin-containing model substrates have also been used to characterize RNA binding and unwinding by CsdA and CshA [27, 28], helicases with the same domain architecture as Hera that also form stable dimers [27, 29]. CsdA and CshA have been associated with diverse functions, including ribosome biogenesis [30, 31], RNA decay, fatty acid homeostasis [32], quorum sensing [33], and cold adaptation [34–37].

In the Hera dimer, the active sites of the two helicase cores face toward each other, making a functional cooperation between the two cores possible. Such a cooperation may entail the interaction of both cores with the same RNA molecule or even the same duplex. However, the communication between the RBDs and the core within one protomer of the dimer, and between the RBDs and cores in different protomers is not understood. Here, we probe the role of the two RBDs by comparing the RNA-stimulated ATPase activities, RNA binding affinities and binding kinetics, RNA unwinding, and the RNA- and ATP-induced conformational change of the Hera helicase core in constructs containing both helicase cores, and two, one or no RBD(s). We then analyze the role of the two helicase cores, both in the presence and absence of the RBDs, by comparing Hera constructs containing two or no RBDs plus one or two (functional) cores. Finally, we probe the communication between RBDs and cores by comparing cis- and trans-(like) heterodimers with one (functional) helicase core and one RBD either on the same (cis, cis-like) or the other protomer (trans, trans-like). An overview about the different constructs used is shown in Fig. 1C. We show that the interaction of the RBD with the hairpin anchors Hera on the RNA, which then preferentially interacts with the helicase core on the same protomer. However, interactions with the other protomer are possible, especially in the presence of ATP. Hera with a single helicase core and an RBD on the same protomer has wildtype-like RNA-stimulated ATPase activity and RNA affinities, but its RNA helicase activity is reduced, suggesting different levels of cooperativity between the two protomers in ATP hydrolysis and RNA unwinding. The conformational change of the helicase core that is coupled to ATP hydrolysis and RNA unwinding is possible irrespective of the number of RBDs or cores, suggesting that an individual helicase core is a functional unwinding unit. Thus, domains outside the helicase core are not required for helicase activity under optimal conditions but ensure function at limiting conditions in vivo.

Materials and methods

Reagents

Reagents used in this study are detailed in Supplementary Table S1.

Biological resources

Biological resources used in this study are detailed in Supplementary Table S2.

Statistical analyses

Equations used to analyze data are detailed in the corresponding “Materials and methods” section. The number of replicate experiments (technical replicates) is indicated in the respective section as well as in the figure legends. Data from replicate ATPase assays, anisotropy titrations, and kinetic experiments on RNA binding and RNA unwinding were analyzed in concatenated fits of all data points obtained. Errors reported are the errors of these concatenated fits. Förster resonance energy transfer (FRET) efficiencies were extracted from individual FRET histograms, and mean values and the error of the mean was calculated from these individual values.

Novel programs, software, and algorithms

All software used in this work is commercially available or publicly available free of charge (see Supplementary Table S3).

Web sites/databases

This work used information from the PDB database. The web-based program Expasy [38] was used to calculate extinction coefficients of recombinant proteins (see Supplementary Table S3, S5).

Protein production and purification

All Hera constructs used in this study are summarized in Fig. 1C. Full-length Hera (Hera_1–510), Hera_1–419, and the ATPase-deficient variants Hera_1–510_K51Q and Hera_1–419_K51Q were produced with an N-terminal His₆-tag in E. coli Rosetta (DE3) cultivated in autoinducing medium [39] and purified on Ni²⁺-NTA sepharose (Ni²⁺ sepharose 6 FF, 10 ml, equilibrated in 50 mM Tris–HCl pH 7.5, 500 mM NaCl, 20 mM imidazole; elution with 500 mM imidazole in the same buffer) and by size-exclusion chromatography (HiLoad 16/60 Superdex S200, equilibrated in 50 mM Tris–HCl pH 7.5, 500 mM NaCl). If necessary, an additional purification step on heparin sepharose (HiTrap Heparin HP, 10 ml) was performed (equilibration in 50 mM Tris–HCl pH 7.5, 100 mM NaCl; elution in the same buffer with 500–750 mM NaCl). Hera_1–365 (monomeric core) and Hera_208–419 (Hera_RecA_C_DD) were produced in E. coli Rosetta (DE3) in autoinducing medium [39] and purified as previously described [8, 19]. The His₆-GST-Hera_424–510 fusion protein (Hera_RBD) was purified on glutathione sepharose (GSTPrep FF 16/10, 20 ml), followed by size-exclusion chromatography (HiLoad 16/60 Superdex 75, GE Healthcare) [17, 18].

Cysteine variants (E115C/E227C), the ATPase-deficient K51Q variants, and the K463A variant were generated by site-directed mutagenesis [see Supplementary Table S4 for primer sequences, purchased desalted and dried from Sigma (Taufkirchen, Germany)] and purified according to the protocols described above, except that 2 mM β-mercapto ethanol (β-ME) was added to all buffers for constructs containing cysteines.

For the preparation of heterodimers, His₆-Hera_1–419, His₆-Hera_1–419_K51Q, or His₆-Hera_1–510_K51Q containing a thrombin cleavage site following the His₆-tag were cleaved by thrombin after the first Ni²⁺-NTA purification step; uncleaved fusion protein was removed by a second Ni²⁺-NTA purification step. Heterodimers were generated by mixing a 5-fold molar excess of the variant without tag with the His₆-tagged variant, followed by incubation at 65°C for 1 h to accelerate subunit exchange [19]. The resulting heterodimers were purified by Ni²⁺-NTA sepharose and size-exclusion chromatography as described above.

All Hera constructs containing the DD eluted as dimers from the size-exclusion column. Protein concentrations were determined photometrically by measuring the absorption at 280 nm. Extinction coefficients were calculated for dimers from the sequence of the two protomers with Expasy ProtParam [38] (Supplementary Table S5).

RNA substrates

RNA oligonucleotides were purchased PAGE-purified from Sigma; fluorescently labeled RNAs were high-performance liquid chromatography (HPLC)-purified. The sequences were: 5′-GCAGGUCCCAAGGGUUGGGCUGUUCGCCCAUU-3′ (32mer), 5′-UUGGGACCU-3′ (9mer), and 5′-AGGUCCCAA-3′ (9mer_comp). The regions of the 32mer forming the stem of hairpin 92 are underlined. Fl-32mer and Cy5-32mer are identical in sequence to the 32mer but contain a fluorescein/Cy5 modification attached to the 5′-end. 9mer-Cy3 is identical to the 9mer but carries a Cy3 modification at the 3′-end.

ATPase activity

Steady-state ATPase activity was measured in a coupled enzymatic assay that couples the hydrolysis of ATP to NADH oxidation. Measurements were performed at 37°C with 0.15 μM Hera or Hera_1–419 (0.3 μM for all heterodimers and Hera_1–365) in 50 mM Tris–HCl, pH 7.5, 150 mM NaCl, and 5 mM MgCl₂ in the presence of 1.6 mM NADH, 0.8 mM phosphoenol pyruvate, 92 μg·ml⁻¹ pyruvate kinase, and 52 μg·ml⁻¹ lactate dehydrogenase as described [8]. Poly-U RNA was added at different concentrations from 0 to 1000 μM bases. Reactions were started by addition of 5 mM ATP. Reaction velocities v of ATP hydrolysis were obtained from the change in absorption with time:

where ΔA₃₄₀ is the change in absorption, Δt is the observed time interval, d is the pathlength (1 cm), and ϵ_{340, NADH} is the molar extinction coefficient of NADH at 340 nm [40]. The dependence of the reaction velocity on RNA concentration c_poly-U was described with a modified Michaelis–Menten equation:

v₀ is the intrinsic ATP hydrolysis rate in the absence of RNA, v_max is the maximum increase in velocity at saturating RNA concentrations, and K_1/2,RNA is the RNA concentration at which RNA-stimulated ATPase activity occurs with half-maximum velocity. The catalytic turnover number k_cat was calculated by the division of v_max by the enzyme concentration, and the catalytic efficiency is determined as the ratio of k_cat and K_1/2,RNA. Errors σ for the catalytic efficiency were calculated from the errors of k_cat and K_1/2,RNA according to the propagation of uncertainties with the following equation:

Values for k_cat and K_1/2,RNA are summarized in Table 1.

Determination of K_d values in fluorescence equilibrium titrations

K_d values for Hera/RNA complexes were determined in fluorescence anisotropy titrations of 0.05 μM 32mer or fl-32/9mer (fl-32mer with the 9mer annealed) in 50 mM Tris–HCl, pH 7.5, 150 mM NaCl, and 5 mM MgCl₂ with Hera and Hera variants as described [8] using a Jobin Yvon FluoroMax3 fluorimeter. Fluorescence was excited at 496 nm (2 nm bandwidth) and detected at 520 nm (4 nm bandwidth). Data were analyzed using the solution of the quadratic equation that describes a 1:1 complex formation:

where r₀ is the anisotropy of free RNA, Δr_max is the amplitude (r_bound – r_free), and f_bound is

with the total Hera (dimer) concentration [Hera]_tot, the total RNA concentration [RNA]_tot, and the dissociation constant K_d. K_d values for Hera/RNA complexes determined with Equation (4) are summarized in Table 2. To determine accurate K_d values from anisotropy titrations, changes in fluorescence intensity F on binding, R = F_bound/F_free, need to be taken into account. However, both analyses gave similar values for all constructs (Supplementary Table S6), and the relative K_d values for the constructs compared within each set of comparisons are unaffected by the value of R used in the analysis in all cases.

Alternatively, K_d values were determined using numerical analyses based on explicit models using Dynafit [41, 42] (see Supplementary data for Dynafit scripts). In the simplest case, binding of RNA to Hera dimers was described as independent, single-step binding to each protomer, with K_d values and the anisotropy of free (r_free) and bound RNA (r_bound) as fit parameters. These values are summarized in Table 2. For heterodimers, K_d1 and K_d2 were used to describe the interaction of the RNA with the two different protomers. As kinetic experiments on RNA binding suggested two-step binding of RNA to Hera, models describing RNA binding in two steps, with parameters K_d1, k₂ and k_-2, r_free and r_bound, were also used. The models used are provided in the respective figures alongside with the experimental data and fits.

Stopped-flow experiments

Binding of the fl-32mer or the fl-32/9mer was measured in a TgK KinetAsyst SF-61DX2 Stopped-flow system, using the fluorescence intensity of the fl-32mer as a probe for binding. The fl-32mer (0.05 μM) in 50 mM Tris–HCl, pH 7.5, 150 mM NaCl, and 5 mM MgCl₂ in the observation chamber was used as a reference signal. The intensities for the parallel and perpendicular orientation of the polarizers were adjusted by changing the voltage of the photomultiplier such that the G-factor had a value of one. Fl-32mer was excited with an LED L470A (Ocean Optics, Ostfildern, Germany), and emitted light was detected through a cut-off filter of 515 nm in photomultiplier. Stopped-flow traces were measured over 0.5 s (2048 data points) at 25°C after 1:1 mixing of the fl-32mer with Hera. The observed rate constants k_obs were determined by describing the traces with single-exponential functions. From the hyperbolic dependence of k_obs on Hera concentration, the parameters for two-step binding were extracted according to the following equation:

K_d1 is the dissociation constant of the initial Hera/RNA complex, [Hera] is the concentration of Hera, k₂ and k_-2 are the rate constants of the forward and reverse reactions of a second step following the formation of the initial complex, and their ratio k_-2/k₂ defines K_d2. The errors of k₂ and k_-2 were propagated to K_d2 according to the following equation:

The overall K_d can be calculated from K_d1, k₂, and k_-2 as:

The errors from K_d1 and K_d2 were propagated to K_d as:

In few cases, k_obs showed a linear dependence on Hera concentration, characteristic for one-step binding. In these cases, the rate constants k₁ and k_-1 for binding and dissociation were obtained from the slope and y-axis intercept according to the following equation:

Error propagation to K_d= k_-1/k₁ was done in analogy to equation (3). Rate constants and equilibrium constants determined from the concentration dependence of k_obs are summarized in Table 3.

RNA unwinding

RNA unwinding was followed as a function of time using a Cy3/Cy5-labeled 32/9mer unwinding substrate and measuring the decrease in FRET upon unwinding as a spectroscopic probe. The substrate for unwinding was generated by incubating a 2-fold molar excess of the 3′-Cy3-labeled 9mer with the 5′-Cy5-labeled 32mer in 50 mM Tris–HCl, pH 7.5, 150 mM NaCl, and 5 mM MgCl₂ at 95°C for 5 min and slowly cooled to 25°C. Unwinding reactions were performed in 50 mM Tris–HCl, pH 7.5, 150 mM NaCl, and 5 mM MgCl₂ at 25°C with 0.5 μM RNA substrate and 0.25–7.5 μM Hera in the presence of 5 μM unlabeled 9mer_comp RNA as a trap to prevent reannealing. Reactions were started by the addition of 5 mM ATP. The donor (Cy3) was excited at 554 nm (1 nm bandwidth), and unwinding was followed as a decrease in acceptor (Cy5) emission at 666 nm (2 nm bandwidth). A rate constant k_obs was obtained from describing the data with a single-exponential function. The dependence of the observed rate constant on the concentration of Hera was described by a hyperbola to determine the unwinding rate constant (turnover number), k_unw, and the corresponding equilibrium constant, K_1/2,unw:

K_1/2,unw is the concentration of Hera at which the unwinding rate is half-maximal. Values for k_unw and K_1/2,unw are summarized in Table 4.

Single-molecule FRET experiments

Single-molecule FRET experiments were performed on a Microtime 200 confocal fluorescence microscope (PicoQuant). Cuvettes were pre-incubated for 30 min with 10 μM of the ATPase-deficient Hera_K51Q variant in activity buffer. Labeling of Hera_E115C/E227C (50–100 μM) was performed in 50 mM Tris–HCl, pH 7.5, 750 mM NaCl, and 1 mM tris(2-carboxyethyl)phosphine with a 3-fold molar excess of AlexaFluor 488 (donor) and 6-fold molar excess of AlexaFluor 546 (acceptor) for 1 h at 25°C. Free dye was removed by size-exclusion chromatography (Micro BioSpin P30, Bio-Rad). Donor fluorescence was excited with the output from a pulsed LDH-PFA-488 laser diode (40 MHz), focused by a 60× water immersion objective (UPlanAPO NA 1.2, Olympus, München, Germany). Fluorescence emission was separated from excitation light by a beam splitter (500 dcr), further split into donor and acceptor emission with a dichroic mirror (Z532 rdc), passed through 535/40 (donor) and 570/LP (acceptor) filters, and detected via τ-single-photon avalanche diode (τ-SPAD) detectors (PicoQuant, Berlin, Germany). Measurements were performed with 200 pM of labeled protein (donor concentration) in 50 mM Tris–HCl, pH 7.5, 150 mM NaCl, and 5 mM MgCl₂ at 25°C for 30 min either in the absence or presence of 0.8–4.8 μM 32mer and 5 mM 5'-adenylyl-β,γ-imidotriphosphate (ADPNP) to induce conformational changes. FRET histograms were constructed from FRET efficiencies calculated from fluorescence bursts of >80 photons using SymPhoTime 64 v2.4.4874 (PicoQuant, Berlin, Germany). Cross-talk between donor and acceptor channels and differences in detection efficiencies and quantum yields were removed by applying correction parameters (α’ = 0.445, β’ = 0, and γ’ = 1.401; see [43]). FRET histograms were analyzed in OriginPro 2023 (OriginLab, Northampton, USA).

Results

Role of the RBDs: Hera containing two, one, or no RBDs

To dissect the role of the two RBDs for RNA unwinding by Hera, we generated a heterodimer with only a single RBD, containing one copy of full-length Hera and one copy of Hera_1–419 lacking the RBD (Hera/Hera_1–419; Fig. 1C). We then compared its RNA-stimulated ATPase activity, RNA binding, and unwinding properties (Fig. 2) as well as its conformational response to RNA and ADPNP (Fig. 3) with Hera (containing two RBDs) and with the dimeric core without RBDs (Hera_1–419) .

RNA-stimulated ATPase activity

The RNA-stimulated ATPase activity was measured in a steady-state ATPase assay as a function of the concentration of poly-U RNA (Fig. 2B, Supplementary Fig. S1A, and Table 1). poly-U RNA is a mixture of single-stranded RNAs of different lengths (see Supplementary Table S1): it interacts with and activates the ATPase activity of the helicase core of Hera [8]. Uridine can only be accommodated at positions 3 and 4 of the four positions forming the RNA recognition site on the RBD [21], enabling no or only weak interactions of poly-U with the RBD.

Hera, Hera/Hera_1–419, and Hera_1–419 showed a hyperbolic dependence of the reaction velocity for ATP hydrolysis on the concentration of poly-U (Fig. 2B). The k_cat value at saturating RNA concentrations was k_cat = 2.3 ± 0.3 s⁻¹ for Hera (two RBDs). For Hera/Hera_1–419 and Hera_1–419, the values were lower (k_cat = 1.0 ± 0.1 s⁻¹ for Hera/Hera_1–419, one RBD, and k_cat = 0.9 ± 0.3 s⁻¹ for Hera_1–419, no RBDs; Table 1). Note that the protein concentrations are given in terms of (hetero-)dimers; hence, these k_cat values reflect the turnover by both cores. The turnover numbers per core would be half of these values. However, as these constructs contain the same number of cores, the turnover numbers can be compared directly. The lower k_cat of Hera lacking one or both RBDs compared to the k_cat of Hera with both RBDs points to a role of the RBDs for RNA-stimulated ATP hydrolysis, even with poly-U RNA that predominantly interacts directly with the core. The K_1/2,RNA values were 240 ± 84 μM (nucleotides) for Hera, 108 ± 51 μM for Hera/Hera_1–419, and 871 ± 552 μM for Hera_1–419 (Table 1). These values reflect the K_1/2 for the RNA binding site in the helicase core. Although a conclusion on the effect of deleting the RBDs is limited by the large error of the K_1/2,RNA value for Hera_1–419, the data are suggestive of a role of the RBDs in the stimulation of the ATPase activity by poly-U RNA, both on the level of binding (K_1/2) and catalysis (k_cat).

The catalytic efficiency, k_cat/K_1/2,RNA, is identical for Hera (two RBDs) and Hera/Hera_1–419 (one RBD), with values of k_cat/K_1/2,RNA = 0.009 ± 0.002 μM⁻¹ s⁻¹ (Hera) or 0.009 ± 0.003 μM⁻¹ s⁻¹ (Hera/Hera_1–419). For Hera_1–419 (no RBDs), the catalytic efficiency was reduced 9-fold, to 0.001 ± 0.0007 μM⁻¹ s⁻¹. Thus, deleting one RBD does not have much of an effect, but deleting the second RBD substantially reduces the catalytic efficiency of RNA-stimulated ATP hydrolysis.

Overall, the intrinsic ATPase activity of Hera is also stimulated in the presence of poly-U RNA in the absence of the RBDs, in agreement with direct binding of poly-U RNA to the two helicase cores present in all three constructs. While deletion of one or both RBDs leads to a decrease in k_cat, K_1/2,RNA increases only when both RBDs are deleted. This behavior points to some interaction of poly-U with the RBDs and suggests some contributions of the RBDs to RNA-stimulated ATP hydrolysis by the cores.

RNA affinity: anisotropy titrations

Next, we analyzed the effect of the number of RBDs on RNA binding in fluorescence anisotropy titrations of a fluorescein-labeled 32mer and 32/9mer with Hera, Hera/Hera_1–419, and Hera_1–419 (Fig. 2C and D, and Table 2). These RNAs contain a hairpin that binds to the RBD, and a flanking single- or double-stranded region that interacts with the helicase core [21]. The titration curves of these RNAs with Hera, Hera/Hera_1–419, and Hera_1–419 were hyperbolic, with no indication of sigmoidality, suggesting that binding of the RNA to the Hera dimer is non-cooperative. Consistent with this, all titration curves could be described by a 1:1 binding model (Equations 4 and 5). For the Hera/32mer complex, a K_d value of 0.06 ± 0.01 μM was obtained. For the 32/9mer complex, the K_d value was 33-fold higher, with K_d = 2.0 ± 0.3 μM. For Hera/Hera_1–419 (one RBD), slightly higher dissociation constants were obtained, with K_d values of 0.25 ± 0.09 μM (32mer) and 3.0 ± 0.3 μM (32/9mer; 12-fold higher than 32mer). Hera_1–419 (no RBD) showed the highest dissociation constants, with a K_d of 16 ± 7 μM (32mer) and 11 ± 5 μM (32/9mer).

Note that the K_d values determined in anisotropy titrations need to be interpreted with care because of the dimeric nature of the Hera constructs, and also because the 32/9mer can interact with different regions of Hera: for Hera_1–419, lacking RBDs, the RNA must interact with the cores, and the concentration of cores is twice the dimer concentration. In this case, the K_d values for each core of the dimer are thus twice the values determined, i.e. 32 μM (32mer) and 22 μM (32/9mer). The (isolated) helicase core of Hera thus has a slight, if any, preference for double-stranded over single-stranded RNA.

In contrast to Hera_1–419, the K_d values for the Hera/32mer and the Hera/32/9mer complexes reflect both the interaction of the RBDs with the hairpin [21] and the interaction of the flanking single- or double-stranded region with the helicase core. The apparent K_d values of K_d = 0.06 ± 0.01 μM (32mer) and K_d = 2.0 ± 0.3 μM (32/9mer) correspond to K_d = 0.12 μM per protomer (32mer, in excellent agreement with the previously reported value of 0.13 μM [21]) and K_d = 4.0 μM (32/9mer). With the assumption that the interactions of Hera with the hairpin and the flanking single- or double-stranded region are independent and represented by K_d1 (interaction with the RBDs) and K_d2,ss or K_d2,ds (interaction of the flanking single- or double-stranded RNA with the cores; see schematic in Supplementary Fig. S2A,B), the overall K_d measured must correspond to the product of K_d1 and either K_d2,ss (32mer) or K_d2,ds (32/9mer). The ratio of the overall K_d values for 32mer and 32/9mer then directly reflects the ratio K_d2,ds/K_d2,ss = 33, meaning the Hera core interacts 33-fold more strongly with single-stranded than with double-stranded RNA when the hairpin is pre-bound (note that K_d2,ss or K_d2,ds reflect the K_d values for a unimolecular interaction, not for the bimolecular interaction of the Hera core with single- or double-stranded RNA. The corresponding value for the bimolecular interaction, obtained from the comparison of the corresponding K_d values for the Hera_1–419 complexes, suggest little preference, see Table 2). Thus, the core only shows a preference for single-stranded RNA when the RNA is anchored to the RBD.

Finally, the K_d values for Hera/Hera_1–419 (two cores, one RBD) can be rationalized by binding of the hairpin to the one RBD in the Hera protomer, followed by an interaction of the flanking region with the core in this protomer (with an overall K_d of K_d1 multiplied with K_d2,ss or K_d2,ds, where K_d2 again represents the unimolecular rearrangement) and by the direct binding of the single- or double-stranded region of a second RNA molecule to the core in the Hera_1–419 protomer (K_d,core,ss or K_d,core,ds, reflecting the bimolecular association of the RNA with the core; see Supplementary Fig. S2C). The K_d values of 0.25 and 3.0 μM retrieved by analysis of the titration of Hera/Hera_1–419 with Equation (4) are between the values determined for Hera and Hera_1–419 (Table 2), suggesting that they represent apparent K_d values for binding of the RNA to the two protomers with different affinities.

Notably, the changes in anisotropy (final values) are markedly different for the different Hera variants and for the binding of single-stranded versus double-stranded RNAs (Fig. 2C and D). While RNA bound to Hera reaches an anisotropy of ∼0.3 (∼0.18 for the 32/9mer), this value is reduced to ∼0.2 (∼0.14 for 32/9mer) for Hera/Hera_1–419, and even more so, to ∼0.1 (∼0.08 for 32/9mer) for Hera_1–419. These differences are larger than expected from the only moderate differences of these variants in molecular mass (115, 104, and 96 kDa, respectively), pointing to different mobilities of the fluorophore in the different complexes, and different modes of binding. Interaction of the RNAs with the RBD (in Hera and in one protomer of Hera/Hera_1–419) is associated with a higher anisotropy increase, interaction with the core (in Hera_1–419) with a lower increase. Again, the values for the Hera/Hera_1–419 heterodimer are in-between the values for Hera and Hera_1–419, consistent with a preferential interaction of the RNA with the RBD of the Hera protomer, and an interaction with the core in the Hera_1–419 protomer. The different anisotropy values of 32mer and 32/9mer bound to Hera further suggest that these RNAs are bound differently. The higher anisotropy of the bound 32mer is consistent with its predominant interaction with the RBDs, whereas the lower anisotropy of the bound 32/9mer may suggest more contributions of the cores to the interaction.

Overall, the anisotropy titrations show that Hera interacts more strongly with hairpins flanked by single-stranded region than with the hairpin flanked by a duplex. Removal of one or both RBDs reduces the overall RNA affinity of Hera. Notably, removal of the second RBD has a larger effect (64-fold/4-fold reduction in affinity for single- and double-stranded RNA, respectively) than removal of the first RBD (4-fold/1.5-fold). The decrease in RNA affinity is more pronounced for single- than for double-stranded RNA, which reduces the discrimination between single- and double-stranded RNA from 33-fold (two RBDs) to 12-fold (one RBD) and to no preference (no RBD; see Table 2).

RNA affinity: a quantitative model for RNA binding

To derive a quantitative model for RNA binding to Hera (see Supplementary Fig. S2), we performed numerical analyses of the titration data with one-step binding models for each protomer using Dynafit (see Supplementary data for Dynafit scripts). First, we validated this approach with the Hera and Hera_1–419 homodimers. The K_d values from the numerical analyses are in excellent agreement with the values determined from analyses using the Equation (4) (see Table 2; Supplementary Fig. S3A and B, see Supplementary data for a more detailed description of Dynafit analyses).

For the Hera/Hera_1–419 heterodimer, we also obtained values in excellent agreement with the value from the analysis using Equation (4). However, this analysis does not account for the different RNA affinities of the two protomers. A simulated titration curve based on a model that assumes independent binding of the 32mer to the Hera or Hera_1–419 protomers, using the parameter values determined from the Hera and Hera_1–419 titrations either with Dynafit, or with Equation (4) (Supplementary Fig. S3C; see Dynafit Script 2), roughly reflect the overall shape of the experimental binding curve, but does not describe the amplitude of the titration data satisfactorily. When we attempted to vary the K_d values for the binding events in the two protomers during a fit, they were highly correlated and could not be determined independently from the fit, precluding further quantitative analysis.

We performed the same analyses for the titrations of the 32/9mer with Hera/Hera_1–419 (Supplementary Fig. S3C; see Dynafit Script 1). Again, we reached excellent agreement with the results from the analysis with Equation (4). Simulations of binding curves with independent binding of the 32/9mer to the Hera or Hera_1–419 protomers with different affinities (see Dynafit Script 2) gave reasonable descriptions of the experimental data (Supplementary Fig. S3C).

Collectively, our binding data show that Hera and Hera_1–419 interact differently with the 32mer and the 32/9mer. Although we cannot derive a complete quantitative model for RNA binding by Hera/Hera_1–419 based on the simple analysis of the set of anisotropy titrations performed here, RNA binding to the two protomers in Hera/Hera_1–419 is consistent with the parameters for RNA binding to the individual protomers.

RNA binding kinetics

To further investigate the differences in interactions with RNA of Hera with two, one, or no RBD(s), we determined rate constants for RNA binding for all three constructs. To this end, we performed stopped-flow experiments and mixed a constant concentration of fl-32mer with increasing concentrations of Hera, Hera/Hera_1–419, or Hera_1–419. As we did not manage to obtain binding transients with sufficient signal-to-noise ratio using anisotropy as a signal, we followed complex formation as an increase in fluorescence intensity. All three variants caused a rapid increase in fluorescence within <0.1 s that could be described by a single-exponential function (Supplementary Fig. S1B). The observed rate constants k_obs determined from single-exponential fits showed a hyperbolic dependence on Hera concentration (Fig. 2E), in agreement with a two-step binding process, consisting of a first, spectroscopically silent binding step, followed by a slower, second step associated with a change in signal. Using Equation (6), K_d1 for formation of the initial complex, as well as the rate constants k₂ and k_-2 for the subsequent rearrangement of the collision complex were determined from the concentration dependence of k_obs. For Hera, we obtained values of K_d1 = 1.7 ± 0.7 μM, k₂ = 173 ± 21 s⁻¹, and k_-2 = 50 ± 9 s⁻¹, which gives K_d2 = 0.29 ± 0.06 and K_d = 0.49 ± 0.22 μM (Table 3 (a); note that K_d1 again reflects the initial interaction of RNA with the dimer. K_d1 and hence also K_d need to be doubled to calculate the value per protomer). The corresponding values for Hera/Hera_1–419 (one RBD) were K_d1 = 2.3 ± 1.5 μM, k₂ = 188 ± 44 s⁻¹ and k_-2 = 66 ± 11 s⁻¹, giving K_d2 = 0.35 ± 0.10 and K_d = 0.81 ± 0.64 μM. For Hera_1–419 (no RBD), the observed rate constants showed no clear dependence on the concentration of Hera in the concentration range tested (up to 5 μM), consistent with little binding due to the lower 32mer affinity observed in equilibrium titrations (K_d = 16 μM, see Table 2).

Two-step binding of RNA to Hera can be rationalized by an initial, rapid binding of the 32mer hairpin to the RBD, and subsequent binding of the flanking single- or double-stranded region to the helicase core (see Supplementary Fig. S2). To directly test whether the first step described by K_d1 could reflect binding of the 32mer to the RBD, we measured binding of the isolated RBD (Hera_424–510) to the fl-32mer in stopped-flow experiments. Binding of the RBD also led to a single-exponential increase in fluorescence, demonstrating that this step is associated with a change in signal. The k_obs showed a linear dependence on the concentration of the RBD in the concentration range tested. Such a linear dependence may indicate a lack of saturation, meaning we only sampled the linear part of a hyperbolic dependence. Alternatively, it could point to a one-step binding mechanism. In this case, the rate constants for binding and dissociation, k₁ and k_-1, are determined by the slope and y-intercept of the linear dependence (Equation 10), which gives k₁ = 12.5 ± 0.4 μM⁻¹ s⁻¹ and k_-1 = 34 ± 0.7 s⁻¹ (Supplementary Fig. S4), corresponding to K_d1 = 2.7 ± 0.014 μM. This value is in good agreement with the K_d1 value per protomer determined for the Hera dimer (K_d1 = 3.4 ± 1.4 μM) and the value for Hera/Hera_1–419 heterodimer (K_d1 = 2.3 ± 1.5 μM), supporting the assignment of the first step as binding of the hairpin to the RBD.

The overall K_d values for the 32mer complexes of Hera, Hera/Hera_1–419, and Hera_1–419 from kinetic data rank the proteins in the same order of 32mer affinities that were determined in equilibrium titrations (Hera < Hera/Hera_1–419 << Hera_1–419), although the absolute values differ. This difference could simply be due to the different signals measured (fluorescence anisotropy versus intensity). It could also indicate that slower processes such as slow isomerization steps not detected by stopped flow may contribute to binding. It is also conceivable that the two-step binding model, with a signal change in the first binding step and a spectroscopically silent second step, does not faithfully recapitulate all features of the underlying binding mechanism. For example, binding to the two cores and/or the RBDs could occur with different affinities and/or with different changes in signal. In fact, the observed increase in fluorescence on binding of the 32mer to the isolated RBD indicates that both steps of binding are associated with changes in intensity, and suggests that the two-step model with a change in signal exclusively in the second step is over-simplified.

We also measured binding of Hera to the fl-32/9mer in stopped-flow experiments (Fig. 2F, Table 3 (b)). Again, the stopped-flow traces showed a rapid increase in intensity that could be described by single-exponential functions. From the hyperbolic dependence of k_obs on the concentration of Hera, again consistent with two-step binding, we obtained K_d1 = 5 ± 1.5 μM, k₂ = 224 ± 33 s⁻¹, and k_-2 = 134 ± 6 s⁻¹ (K_d2 = 0.59 ± 0.09, K_d = 3.0 ± 1.0 μM). The overall K_d value reflects the lower affinity of Hera for the 32/9mer than for the 32mer detected in equilibrium titrations.

The stopped-flow data on RNA binding support all main observations from equilibrium titrations: Hera interacts more tightly with the 32mer than with the 32/9mer. Removal of the RBDs weakens binding of Hera to the 32mer. Removing one of the two RBDs has a moderate effect, removing both RBDs substantially weakens 32mer binding.

Connecting equilibrium and kinetic data

With the insight from the kinetic analyses that RNA binding to Hera follows a two-step binding model, we re-analyzed the titration data for Hera with a two-step binding model using Dynafit (Supplementary Fig. S5A, Dynafit Script 3, see Supplementary data for a more detailed description of Dynafit analyses). The equilibrium data are reasonably well-described with the parameter set obtained from the analysis of the hyperbolic dependence of k_obs on Hera concentration for the 32mer (Supplementary Fig. S5B) and for the 32/9mer (Supplementary Fig. S5C). However, the system becomes underdetermined when all parameters are varied, with large correlations between parameters, precluding an independent determination of these values from equilibrium data. Nevertheless, these comparisons show that the equilibrium and kinetic data are consistent and support a two-step binding model for the interaction of Hera with the 32mer and 32/9mer.

We next challenged this model by describing the titration data for Hera/Hera_1–419 with two-step binding to the Hera protomer and one-step binding to the Hera_1–419 protomer Supplementary Fig. S5D, Dynafit Script 4a). The simulated data with values for K_d1, k₂, and k_-2 obtained from the hyperbolic dependence of k_obs on the concentration of Hera (Table 3 (a)), and a K_d obtained from the Dynafit analysis using a one-step binding model for Hera_1–419 (as no concentration dependence of k_obs was observed to determine k₁ and k_-1) reasonably described the experimental data (Supplementary Fig. S5E). When individual or several parameters were varied, the curves matched the data more closely, but the parameters were returned with large errors, indicating that the system is underdetermined and the quantitative analysis not possible.

The simulation of the 32/9mer titration with Hera/Hera_1–419 according to the one-step-/two-step binding model (see Dynafit script 4a), with kinetic parameters for Hera [Table 3(b)] and the K_d from the Dynafit analysis (Table 2) for Hera_1–419, also matches the overall shape of the experimental curve (Supplementary Fig. S5F). Varying one or several parameters gives curves that match the experimental data, but the parameters are affected by large errors; the system is underdetermined.

Overall, titration and kinetic data are thus consistent with one-step binding of 32mer and 32/9mer to the isolated RBD (and, by inference, to the isolated helicase core), and with two-step binding to Hera, with a rapid interaction of the hairpin with the RBD in the first step, and a subsequent, slower interaction of the flanking region with the helicase core.

RNA unwinding

Our data show that the RBDs are a key element for the initial interaction of Hera with RNA and play a (minor) role in RNA-stimulated ATPase activity. Next, we asked whether the number of RBDs present also affects Hera-mediated RNA unwinding. Duplex unwinding was followed as a function of time under single-turnover conditions, using a Cy3/Cy5-labeled 32/9mer as a substrate and the decrease in FRET as a probe (Fig. 2G, Supplementary Fig. S1C, and Table 4). Reactions in the absence of ATP were performed as a negative control.

RNA unwinding by Hera, Hera/Hera_1–419, and Hera_1–419 occurred on the minute time scale, and was thus much slower than RNA binding (Supplementary Fig. S1C), indicating that steps subsequent to binding are rate-limiting for unwinding. Unwinding traces were described by single-exponential functions to extract the observed rate constant k_obs. The observed rate constants showed a hyperbolic dependence on the concentration of Hera (Fig. 2G), pointing to contributions from RNA binding. Although binding of RNA to the RBDs occurs on a much faster time scale, the concentrations of Hera used in these experiments range from sub-saturating to saturating, rationalizing the observed concentration dependence. By analyzing the concentration dependence of k_obs using Equation (11), an unwinding rate constant at saturation of k_unw= 0.062 ± 0.008 s^-1 and a concentration required for half-maximal unwinding rate of K_1/2,unw, = 1.4 ± 0.4 μM were determined for the Hera dimer (Table 4). K_1/2,unw per protomer is again twice the value determined, i.e. K_1/2,unw = 2.9 μM per protomer, which puts it into the range of the RBD affinity for RNA, whereas k_unw is concentration-independent.

Hera/Hera_1–419 (one RBD) showed a 4-fold lower rate constant of k_unw= 0.016 ± 0.003 s⁻¹ and an unwinding constant of K_1/2,unw = 2.6 ± 1.4 μM (Table 4). The value for K_1/2,unw is in good agreement with the value determined per protomer for Hera (K_1/2,unw = 2.9 μM). The rate constant of unwinding, k_unw, contains contributions from unwinding by the core of the Hera protomer carrying the RBD (k_unw = 0.031 s⁻¹) and the core of Hera_1–419 lacking an RBD. For Hera_1–419 (no RBD), unwinding was very slow, with k_unw = 0.0062 ± 0.0007 s⁻¹ (K_1/2,unw = 1.2 ± 0.4 μM), demonstrating that the unwinding by Hera/Hera_1–419 is caused predominantly, if not exclusively, by the Hera protomer.

ATP- and RNA-induced conformational changes: single-molecule FRET experiments

The Hera helicase core switches from an open state to a closed state on binding of RNA and ATP [8]. This conformational change can be observed as an increase in FRET efficiency between dyes attached to the N-terminal and C-terminal RecA domains (RecA_N, RecA_C; see Fig. 1A and B). Using single-molecule FRET experiments and Hera variants carrying a donor and acceptor fluorophore in RecA_N (Cys115) and RecA_C (Cys227), we tested whether Hera with two, one, or no RBD(s) responds to RNA and ATP binding (Fig. 3). For Hera/Hera_1–419, with two functional cores but only a single RBD, we either labeled the helicase core on the same protomer as the RBD (cis), or the helicase core on the other protomer lacking an RBD (trans; see Section “Cooperation of the RBDs with the helicase cores: heterodimers with one functional core and one RBD in cis or trans”).

Donor–acceptor-labeled Hera_E115C_E227C showed a FRET efficiency of E_FRET ≈ 0.6 in the absence of ligands that increased to E_FRET ≈ 0.85 on addition of 32mer and the nonhydrolyzable ATP analog ADPNP (Fig. 3A). This increase in FRET efficiency is indicative of closing of the helicase core [8]. Hera/Hera_1–419, labeled in cis or in trans relative to the single RBD present, showed a narrower distribution of FRET efficiencies, with E_FRET ≈ 0.5 in the apo-state. The more defined histogram is the result of the presence of only two cysteines in Hera/Hera_1–419, and hence a single distance between donor and acceptor fluorophores. In contrast, the Hera dimer contains four cysteines, and labeling with donor and acceptor dyes leads to a heterogeneous mixture of species carrying the two dyes in different positions [115–227 (same protomer), 115–227 (different protomers), 115–115, and 227–227], resulting in different donor/acceptor distances and different FRET distributions that are superimposed in the overall histogram. Addition of RNA and ADPNP to Hera/Hera_1–419 resulted in the appearance of a high-FRET state at E_FRET≈0.65–0.7, irrespective of which core was labeled. Hence, both cores in Hera/Hera_1–419 can close on binding of RNA and ADPNP, independent of the relative position of the RBD. Note that the RNA is in large excess and present in high concentrations in these experiments. It is conceivable that, at these concentrations, it may directly bind to the core of the Hera_1–419 protomer, causing it to close. The FRET histogram of Hera_1–419 (no RBD) showed a broad peak at E_FRET≈0.6 that increased slightly to E_FRET≈0.65 on addition of ADPNP and RNA (Supplementary Fig. S6). Again, the broad histogram is the result of the four cysteines present, and the contributions of different species with different donor/acceptor distances formed on statistical labeling. As this broad distribution may preclude the detection of a change in FRET efficiency, we generated a Hera_1–419 heterodimer from a protomer with two cysteines and a cysteine-free protomer. This heterodimer showed a well-defined FRET histogram with E_FRET ≈ 0.5 in the apo-state and exhibited a clear shift to E_FRET ≈ 0.65–0.7 on addition of ADPNP and 32mer (Fig. 3A), confirming that the helicase core closes on direct binding of RNA and ADPNP. This is consistent with closing of the (monomeric) Hera helicase core on binding of poly-U RNA and ADPNP as reported previously [8].

The FRET efficiencies as a function of 32mer concentration (Fig. 3B) reflect the reduced affinities of Hera/Hera_1–419, and Hera_1–419 compared to Hera for the 32mer, as observed in anisotropy titrations (see Fig. 2C).

Altogether, these experiments demonstrate that the helicase core of Hera closes on binding of RNA and ADPNP even in the absence of the RBDs, indicating that the RBDs are not necessary for closing. However, their presence favors formation of the closed state at lower concentrations of 32mer, consistent with their contribution to RNA binding.

Overall, these experiments are consistent with Hera binding the 32mer by initial interactions of the RBD with the hairpin, followed by interactions of the core with the adjacent duplex. In the presence of ATP, the interaction of RNA with the core leads to core closing, independent of the interactions of the RNA with the RBD, and to unwinding of the duplex by the core. The reduction in the rate constant of unwinding under saturating conditions observed for Hera/Hera_1–419 (Fig. 2G and Table 4) suggests an appreciable effect of deleting one RBD on cooperativity between the protomers in RNA unwinding. A similar effect is seen for RNA-stimulated ATPase activity in the presence of poly-U RNA (Fig. 2B and Table 1), but not for closing of the helicase core.

Role of the cores: Hera containing two or one functional helicase core(s)

To dissect the role of the two helicase cores for RNA unwinding by Hera, we compared the RNA-stimulated ATPase activity, RNA binding and unwinding, and conformational changes of Hera containing one or two functional helicase cores. To probe the role of the two cores in the presence of the RBDs, we compared Hera (two functional cores, two RBDs) with a heterodimer generated from one copy of Hera and one copy of Hera_K51Q, carrying a mutation in the Walker A motif (motif I) that renders it ATPase- and unwinding-deficient (one functional core, two RBDs; Figs 1A and C, and 4A). The Hera_K51Q dimer served as a negative control (no functional core, two RBDs). To address the role of the two helicase cores in the absence of the RBDs, we compared the properties of the dimeric core (Hera_1–419) with the monomeric core (Hera_1–365; Figs 1C and 5A; [8]).

RNA-stimulated ATPase activity

Again, we first assessed the effect of the cores on the RNA-dependent ATPase activity (Fig. 4B, Supplementary Fig. S7A, and Table 1). Hera_K51Q (no functional core, two RBDs) did not show any ATPase activity (Fig. 4B). Hera/Hera_K51Q (one functional core, two RBDs) had a turnover number k_cat of 1.19 ± 0.05 s⁻¹ and K_1/2,RNA of 109 ± 18 μM. The turnover number is about half of the corresponding values determined for Hera (k_cat=2.3 ± 0.3 s⁻¹), as expected from inactivation of one helicase core in the dimer (in the absence of cooperativity). The K_1/2,RNA of 109 ± 18 μM is also half of the value for Hera (240 ± 84 μM; see Table 1). As a result, the catalytic efficiency k_cat/K_1/2,RNA of 0.009 ± 0.002 μM⁻¹ s⁻¹ of Hera/Hera_K51Q is identical to that of Hera (k_cat/K_1/2,RNA = 0.009 ± 0.003 μM⁻¹ s⁻¹).

In the absence of the RBDs, the effect of the number of cores can be gleaned from the comparison of Hera_1–365 and Hera_1–419 (monomeric and dimeric cores; Figs 1C and 5A). We showed before that Hera_1–419 (two cores, no RBD) has a reduced turnover number k_cat of 0.9 ± 0.3 s⁻¹ compared to Hera (2.3 ± 0.3 s⁻¹), but a higher K_1/2,RNA of 871 ± 552 μM (240 ± 84 μM for Hera, see Table 1), in agreement with a lower RNA affinity of the core. The catalytic efficiency of k_cat/K_1/2,RNA of 0.001 ± 0.0008 μM⁻¹ s⁻¹ is ∼9-fold reduced compared to Hera (0.009 ± 0.003 μM⁻¹ s⁻¹). For Hera_1–365 (one core, no RBD), no RNA-dependent ATPase activity was detected (Fig. 5B and Supplementary Fig. S7A). These data suggest that dimerization is important for RNA-stimulated ATPase activity of the Hera core in the absence of the RBDs.

Together, these data show that the inactivation of one of the cores in Hera does not significantly affect the RNA-stimulated ATPase activity of the other core. In the absence of the RBDs, both cores seem to cooperate in RNA-dependent ATP hydrolysis.

RNA affinity

We next tested the effect of the number of (active) cores on RNA binding in fluorescence equilibrium titrations and stopped-flow experiments (Figs 4C and D, and 5C; Supplementary Fig. S7B; and Table 2). By analyzing the titration curves with Equation (4), we obtained K_d values for Hera/Hera_K51Q (one functional core, two RBDs) of K_d = 0.08 ± 0.02 μM (32mer; Fig. 4C and Table 2) and K_d = 2.5 ± 0.4 μM (32/9mer; Fig. 4D). For Hera_K51Q, the values were K_d = 0.17 ± 0.02 μM (32mer; Fig. 4C) and 2.1 ± 0.4 μM (32/9mer; Fig. 4D). Thus, the affinities of Hera_K51Q for the 32mer and 32/9mer are comparable to Hera [K_d = 0.06 ± 0.01 μM (32mer) and 2.3 ± 0.3 μM (32/9mer); see Table 2].

Numerical analyses with Dynafit, based on a one-step binding model and independent, non-cooperative binding to the two protomers (see Dynafit Script 1; Supplementary Fig. S8), were again virtually identical to the ones determined per protomer by analysis with Equation (4) (see Table 2). When we tested analysis of the Hera_K51Q titrations of 32mer and 32/9mer with a cooperative binding model (see Dynafit Script 2), K_d1 and K_d2 for the two binding events were again highly correlated, precluding their determination.

When we used the parameters for the Hera and Hera_K51Q homodimers determined from the numerical analyses to describe the experimental data for the titration of the 32mer and 32/9mer with the Hera/Hera_K51Q heterodimer (see Dynafit Script 2b), the resulting binding curves only roughly described the experimental data (Supplementary Fig. S8A). When K_d1 and K_d2 were varied, their high correlation again precluded their independent determination from the fits, both for the 32mer and for the 32/9mer titrations (see Supplementary data for a more detailed description of the Dynafit analyses).

Collectively, the titration data demonstrate that the K51Q mutation has essentially no effect on binding of Hera to the 32/9mer and little effect on binding of the 32mer (Table 2). Similar to the effect of the RBDs, the effect of the number of cores on RNA affinity was more pronounced for the 32mer, while the affinity for the 32/9mer was virtually independent of the number of functional cores. In contrast to the RBDs, the effect of inactivating one core was similar to the effect of inactivating the second core. These observations are consistent with RNA binding by the RBDs, with little contribution from the core in the absence of ATP.

The requirement of the RBDs for RNA binding was also evident when we compared the effect of the number of cores on RNA affinities in the absence of the RBDs. Hera_1–419 (two cores, no RBD) had increased K_d values of 16 ± 7 μM (32mer) and 11 ± 5 μM (32/9mer; see Table 2; K_d values per core are again twice these values). For Hera_1–365 (one core, no RBD), the affinity for the 32mer was decreased much further, to 81 ± 29 μM (Fig. 5C), which would rationalize the lack of RNA-stimulated ATPase activity (see Fig. 5B).

Overall, the RNA affinities of Hera and Hera_K51Q are similar, the absence of the RBDs is associated with markedly reduced RNA affinities, and the apparent RNA affinity is significantly higher with two cores, at least when the RBDs are absent.

RNA binding kinetics

While the RNA affinity is not affected much by the functionality of the helicase core, we also tested whether the kinetics of RNA binding are different for Hera with two, one, or no functional helicase core(s) in stopped-flow experiments. First, we compared Hera (two functional cores, two RBDs), Hera/Hera_K51Q (one functional core, two RBDs), and Hera_K51Q (no functional core, two RBDs; Fig. 4E, Supplementary Fig. S7B, and Table 3). We observed a hyperbolic dependence of k_obs for RNA binding on Hera/Hera_K51Q and Hera_K51Q concentrations, suggesting similar two-step binding of RNA as observed for Hera. The K_d1 is 0.5 ± 0.2 μM for Hera/Hera_K51Q, and increases to 2.8 ± 1.2 μM for Hera_K51Q (Fig. 4E and Table 3). The rate constants for the second binding step were determined as k₂ = 179 ± 12 s⁻¹ and k_-2 = 46 ± 14 s⁻¹ for Hera/Hera_K51Q (K_d2 = 0.26 ± 0.08, K_d = 0.13 ± 0.07 μM), and k₂ = 194 ± 32 s⁻¹ and k_-2 = 107 ± 6 s⁻¹ for Hera_K51Q (K_d2 = 0.55 ± 0.09; K_d = 1.5 ± 0.7μM); the values determined for Hera were K_d1 = 1.7 ± 0.7 μM, k₂ = 173 ± 21 s⁻¹, and k_-2 = 50 ± 9 s⁻¹ (K_d2 = 0.29 ± 0.06, K_d = 0.49 ± .022 μM; see Fig. 2E and Supplementary Fig. S1B). The overall K_d values from kinetic data reflect the moderate reduction in RNA affinity observed for Hera_K51Q in equilibrium titrations (Fig. 4C and D). Additional measurements with 32/9mer were therefore not performed.

Again, we used the knowledge on two-step binding of the 32mer to Hera, Hera_K51Q, and Hera/Hera_K51Q to analyze the equilibrium titration data with two-step binding models using Dynafit (Supplementary Fig. S9A and C, see Dynafit Scripts 3a–d). Using different sets of K_d1, k₂, and k_-2 for the Hera protomer and for the Hera_K51Q protomer, determined in the kinetic analysis of the Hera and Hera_K51Q homodimers, gave a simulated curve that describes the data reasonably well (Supplementary Fig. S9D), demonstrating internal consistency of the data. As we lacked kinetic data for 32/9mer binding to Hera/Hera_K51Q and Hera_K51Q, we simulated binding curves for the 32/9mer to Hera_K51Q and Hera/Hera_K51Q with a two-step binding model to test whether we can extract the underlying parameters (see Dynafit Script 3a, Supplementary Fig. S9B and D). The agreement between simulation and data further supports that Hera and Hera_K51Q interact similarly both with the 32mer and with the 32/9mer.

Comparing the kinetics of RNA binding to Hera with two or one functional core(s) in the absence of RBDs was not possible. Hera_1–419 (two cores, no RBD) showed no clear dependence of the observed rate constant on the concentration of Hera (see Fig. 2E). The low RNA affinity of Hera_1–365 (one core, no RBD) precluded stopped flow measurements at reasonable concentrations.

Overall, the inactivation of one or both helicase cores by the K51Q mutation does not affect RNA affinity or binding kinetics much when the RBDs as the major RNA-binding platform are present. In the absence of the RBDs, the helicase core has a much lower RNA affinity, making the difference in affinity between monomeric and dimeric cores difficult to assess. Equilibrium data indicate that two cores lead to an increase in RNA affinity compared to a single core (Fig. 5C and Table 2).

RNA unwinding

We next tested the role of the two cores for RNA unwinding, again both in the presence and absence of the RBDs (Figs 4F and 5D; Supplementary Figs S1C and S7C; Table 4). Hera/Hera_K51Q (one functional core, two RBDs) showed RNA unwinding with k_unw= 0.019 ± 0.003 s⁻¹, ∼3-fold more slowly than Hera (k_unw= 0.062 ± 0.008 s^-1), suggesting that inactivation of one core reduces the unwinding activity of the remaining functional core (Fig. 4F and Table 4). K_1/2,unw for the functional core in Hera/Hera_K51Q was similar to Hera, with K_1/2,unw = 2.0 ± 0.8 μM (1.4 ± 0.4 μM for Hera), in agreement with the negligible effect of the K51Q mutation on RNA affinities. Hera_K51Q (no functional core, two RBDs) did not show unwinding activity.

In the absence of the RBDs, the two cores of Hera_1–419 (two cores, no RBDs) supported RNA unwinding with a rate constant k_unw = 0.0062 ± 0.0007 s⁻¹ (Fig. 5D, see also Fig. 2G), ∼10-fold lower than the rate constants for RNA unwinding by Hera. Hera_1–365 (one core, no RBDs) unwinds RNA with an even lower rate constant of k_unw = 0.0016 ± 0.0001 s⁻¹ (Fig. 5D), confirming that the monomeric core does bind 32/9mer RNA in the presence of ATP, despite the low affinity for this RNA in the absence of ATP. Deletion of one core in the absence of the RBDs thus is associated with a 4-fold reduction of the unwinding rate constant, similar to the effect of inactivating one of the cores in the presence of the RBDs. The concentration dependence of the unwinding rate constants was similar, with K_1/2,unw = 2.4 ± 0.8 μM for Hera_1–419 (dimeric core, per core), and K_1/2,unw = 2.3 ± 0.6 μM (monomeric core). Thus, the interaction with RNA in the presence of ATP during unwinding is not affected by dimerization.

ATP- and RNA-induced conformational changes: single-molecule FRET experiments

Finally, we also tested whether the number of cores present affected the ATP- and RNA induced activation and closing of the (functional) helicase core in the absence of the RBDs (Hera_1–419, Hera_1–365; Fig. 5E and F). Similar to Hera_1–419_E115C_E227C (two cores, no RBDs; see Fig. 3A and Supplementary Fig. S6), the histogram of Hera_1–365_E115C_E227C (one core, no RBD) showed a distinct peak at E_FRET ≈ 0.5 and a clear shift on addition of RNA and ADPNP to E_FRET≈ 0.65–0.7 (Fig. 5E). Collectively, these data demonstrate that neither the presence of the RBDs nor dimerization are prerequisites for closing of the Hera helicase core. The dependence of the FRET efficiency of the high-FRET, closed state on the concentration of the 32mer reflects the reduced 32mer affinities of Hera_1–419 and Hera_1–365 compared to Hera (Fig. 5F).

In summary, we do not observe strong evidence for cooperation of the two helicase cores of the Hera dimer in RNA-stimulated ATP hydrolysis or RNA binding. The RNA- and ATP-induced closing of the helicase core also occurs irrespective of the number of cores present. In contrast, inactivation or removal of one core reduces the RNA unwinding activity of the remaining core.

Cooperation of the RBDs with the helicase cores: heterodimers with one functional core and one RBD in cis or trans

So far, our data show that a single helicase core of Hera can interact with RNA and nucleotide and undergo a conformational change into a closed state in the presence of these ligands. RNA substrates are bound in a two-step process; they initially interact with the RBD, which facilitates the interaction of flanking regions with the core, leading to stimulation of the ATPase activity and RNA unwinding. Although we observe beneficial effects of two RBDs and two cores for RNA binding and unwinding, our data suggest that Hera with a single RBD and a single core should be an efficient ATP-dependent RNA helicase. However, it remains unclear whether RNA anchored to the RBD of one protomer can interact only with the core on the same protomer, or also with the core of the second protomer.

To further dissect the cooperation of the two RBDs with the two helicase cores of Hera, we generated heterodimers of Hera that contain one (functional) helicase core and a single RBD that is either part of the same protomer that provides the helicase core (cis- or cis-like heterodimer) or part of the second protomer that does not contain a (functional) helicase core (trans- or trans-like heterodimer). To this end, we used two different strategies. First, we generated a cis-heterodimer formed by Hera and Hera_208–419, comprising the RecA_C and DD domains, and a trans-heterodimer formed by Hera_1–419, comprising the core and the DD, and Hera_370–510, comprising the DD and the RBD (Figs 1C and 6A). The presence of RecA_C in the cis-heterodimer was necessary to stabilize the dimerization domain, which we could not produce separately. The RecA_C domain does not contribute to RNA binding. In the second approach, we used Hera/Hera_1–419, a heterodimer with two cores and one RBD, as a scaffold, and inactivated either the core of the Hera_1–419 protomer lacking the RBD (cis-like heterodimer) or the core of the Hera protomer containing the RBD (trans-like heterodimer, Fig. 6A) by the K51Q mutation. We then compared the ATPase activities, RNA binding and unwinding, and conformational changes of the helicase core of these heterodimers with Hera (Fig. 6).

RNA-stimulated ATPase activity

The cis-, cis-like, trans-, and trans-like heterodimers showed a hyperbolic dependence of the reaction velocity on the concentration of poly-U RNA (Fig. 6B and Supplementary Fig. S10A). The k_cat values at saturating RNA concentrations were 1.2 ± 0.1 s⁻¹ and 0.3 ± 0.09 s⁻¹ for the cis- and cis-like heterodimers, and 1.9 ± 0.4 s⁻¹ and 0.7 ± 0.2 s⁻¹ for the trans- or trans-like heterodimers (Table 1), compared to k_cat = 2.3 ± 0.2 s⁻¹ for Hera (two RBDs), and k_cat = 1.0 ± 0.1 s⁻¹ for Hera/Hera_1–419 (one RBD). The K_1/2,RNA value for the cis-heterodimer (K_1/2,RNA = 200 ± 52 μM) was similar to Hera (240 ± 84 μM). For the cis-like (K_1/2,RNA = 519 ± 357 μM), trans- (K_1/2,RNA = 697 ± 227 μM), and trans-like heterodimers (K_1/2,RNA = 1113 ± 486 μM; Table 1), the values were not well-determined, but consistent with an increased K_1/2,RNA. The resulting catalytic efficiencies of all variants were reduced compared to Hera (k_cat/K_1/2,RNA = 0.009 ± 0.003 μM⁻¹ s⁻¹) but to different extents. For the cis-heterodimer, the catalytic efficiency decreased 1.5-fold (k_cat/K_1/2,RNA = 0.006 ± 0.002 μM⁻¹ s⁻¹), for the trans-heterodimer 3-fold (k_cat/K_1/2,RNA = 0.003 ± 0.001 μM⁻¹ s⁻¹). The catalytic efficiencies of the cis- and trans-like heterodimers, with two cores present, but only a single functional core, showed 15-fold reduced catalytic efficiencies, with k_cat/K_1/2,RNA = 0.0006 ± 0.0004 μM⁻¹ s⁻¹ (cis-like) and k_cat/K_1/2,RNA = 0.0006 ± 0.003 μM⁻¹·s⁻¹ (trans-like). Note that the poly-U RNA used in these experiments interacts directly with the core. Hence, no conclusions can be made regarding the cooperation of the single RBD with the core(s).

RNA affinity

Next, we analyzed the effect of the relative position of the RBD toward the helicase core on RNA binding in fluorescence equilibrium titrations of 32mer and 32/9mer as described (Fig. 6C and D). From equilibrium titrations of the 32mer with the cis- and trans-heterodimers, we obtained K_d values of K_d = 0.126 ± 0.009 μM (cis, 32mer) and K_d = 1.4 ± 0.5 μM (cis, 32/9mer), or K_d = 3.1 ± 0.5 μM (trans, 32mer) and K_d = 16 ± 6 μM (trans, 32/9mer; Fig. 6C and D). For comparison, we also determined the RNA affinities of the isolated RBD [K_d = 1.0 ± 0.2 μM (32mer) and K_d = 17 ± 3 μM (32/9mer; Table 2)]. The cis-heterodimer thus binds RNA with comparable affinities to Hera. In contrast, the affinities of the trans-heterodimer are reduced 52- and 8-fold compared to Hera, resulting in RNA affinities similar to the isolated RBD. Thus, the helicase core and the RBD can functionally cooperate in RNA binding when they are located on the same protomer. When they are on different protomers, their interaction is strongly disfavored.

RNA binding kinetics

We also tested whether the kinetics of RNA binding are dependent on the relative orientation of the RBD toward the helicase core (Fig. 6E and F, Supplementary Fig. S10B and Table 3). The cis-heterodimer showed a hyperbolic concentration dependence of k_obs, giving K_d1 = 0.17 ± 0.08 μM, k₂ = 178 ± 25 s⁻¹, and k_-2 = 18 ± 29 s⁻¹ (K_d2 = 0.1 ± 0.2, K_d = 0.02 ± 0.03 μM), compared to K_d1 = 1.7 ± 0.7 μM, k₂ = 173 ± 21 s⁻¹, and k_-2 = 50 ± 9 s⁻¹ (K_d2 = 0.29 ± 0.06 and K_d = 0.49 ± 0.22 μM) for Hera. For the trans-heterodimer, the concentration dependence of k_obs was linear in the concentration range tested. The linear dependence may indicate a lack of saturation, meaning we only sampled the linear part of a hyperbolic dependence. Alternatively, it could point to a one-step binding mechanism. In this case, we obtain from slope and y-axis intercept the values k₁ = 10 ± 1 μM⁻¹ s⁻¹ and k_-1 = 75 ± 4 s⁻¹ (K_d = 7.3 ± 1.0 μM). These values rank the affinities in the order cis > Hera >> trans, which reasonably agrees with the ranking from equilibrium titrations (cis ≈ Hera >> trans). As seen in the equilibrium data, the RNA affinity of the trans-heterodimer from kinetic data is similar to the affinity of the RBD, whereas the affinity of the cis-heterodimer is similar to Hera.

Due to the generally lower affinities for the 32/9mer, and the reduced RNA affinity of the trans-heterodimer, we determined rate constants for 32/9mer binding only for the cis-heterodimer (Fig. 6F, Supplementary Fig. S10B). The observed rate constants k_obs showed a hyperbolic dependence on the concentration of the cis-heterodimer, giving K_d1 = 4 ± 1.9 μM, k₂ = 262 ± 28 s⁻¹, and k_-2 = 103 ± 17 s⁻¹ (K_d2= 0.40 ± 0.08; K_d = 1.8 ± 0.8 μM). These values are comparable to the ones determined for Hera [K_d1 = 5 ± 1.5 μM, k₂ = 224 ± 33 s⁻¹, and k_-2 = 134 ± 6 s⁻¹ (K_d2 = 0.59 ± 0.09, K_d = 2.8 ± 1.0 μM), see Fig. 2F], further supporting similar RNA binding of Hera and the cis-heterodimer.

The titration data of the 32mer and the 32/9mer with the cis- and trans-heterodimers were reasonably well described by a simulated curve with K_d1, k₂, and k_-2 (cis) or k₁ and k_-1 (trans) from the analysis of the concentration dependence of k_obs (Supplementary Fig. S11, see Supplementary data for a more detailed description of the Dynafit analyses), demonstrating that equilibrium and kinetic data are internally consistent.

Collectively, these data suggest that the communication between the helicase core and the RBD on the same protomer is a key factor for two-step binding of RNA to Hera, at least in the absence of ATP.

RNA unwinding

As the relative position of the RBD to the helicase core determines the cooperation in RNA binding, we also investigated RNA unwinding by the cis-, cis-like, trans-, and trans-like heterodimers (Fig. 6G, Supplementary Fig. S10C, and Table 4). All constructs showed a hyperbolic dependence of the observed rate constants on the protein concentration. Unwinding rate constants were k_unw= 0.015 ± 0.002 s⁻¹ (cis) and k_unw= 0.022 ± 0.008 s⁻¹ (trans). The cis-like heterodimer showed a rate constant of RNA unwinding comparable to the cis-heterodimer (k_unw = 0.012 ± 0.003 s^-1). The trans-like heterodimer had an unwinding rate constant of k_unw = 0.009 ± 0.002 s^-1. These values are 3–7-fold lower than the rate constant of unwinding for Hera (k_unw = 0.062 ± 0.008 s^-1), consistent with the reduced unwinding activity of a single helicase core.

The concentrations for half-maximal unwinding rates were K_1/2,unw = 2.5 ± 0.8 μM (cis) or K_1/2,unw = 8.7 ± 3.9 μM (cis-like), and K_1/2,unw = 4 ± 3.0 μM (trans) or K_1/2,unw = 2 ± 1.1 μM (trans-like). Although these values are affected by large errors, they indicate that, again, the cis-heterodimer (one core) behaved similar to wild-type (two cores). The trans-heterodimer is more similar to the dimeric core, reflecting the importance of the position of the RBD relative to the core for RNA binding. It is unclear why the cis-like and trans-like heterodimers are so different from the cis- and trans-dimers with respect to K_1/2,unw. It seems that the effect of the non-functional core present in cis- and trans-like heterodimers has opposite effects: less efficient interaction with RNA in the cis-like heterodimer compared to the cis-heterodimer, and more efficient interaction with RNA in the trans-like heterodimer compared to the trans-heterodimer.

ATP- and RNA-induced conformational changes: single-molecule FRET experiments

Finally, we also tested whether the relative position of the RBD toward the helicase core affects the ATP- and RNA-induced activation and closing of the (functional) helicase core (Fig. 7A). The cis-, cis-like, trans-, and trans-like heterodimers showed unimodal FRET histograms with E_FRET ≈ 0.5. In the presence of RNA and ADPNP, all variants showed an increase in FRET efficiency, but to different extents. The mean FRET efficiencies of the high-FRET, closed states were similar, with E_FRET = 0.65–0.75. In all cases, the FRET efficiency of the high-FRET state increased with the concentration of the 32mer RNA, in agreement with the reduced RNA affinity of these heterodimers, indicating that saturation has not been reached yet (Fig. 7B). These data show that the helicase core can close independently of the relative position of (functional) core and RBD when ADPNP and RNA are present.

Collectively, we have shown that the RBD interacts predominantly with the helicase core of the same protomer in RNA binding, although RNA binding to the RBD of one protomer can facilitate interaction of the core on the opposite protomer with this RNA and induce core closing.

Discussion

Role of the RBDs and cores for RNA binding and unwinding

Here, we probed the contributions of the RBDs and the cores of the dimeric DEAD-box helicase Hera to RNA-stimulated ATPase activity, RNA binding, ATP-dependent RNA unwinding, and to closing of the helicase core in the presence of RNA and ATP.

RBDs

We show that RNA binds to Hera in a two-step mechanism, with an initial interaction between the C-terminal RBD and a hairpin, followed by the interaction of the core with the flanking single- or double-stranded region. While the Hera core shows little preference for single-stranded versus double-stranded RNA, it strongly prefers single-stranded RNA once the RNA is anchored to the RBD. The isolated RBD interacts more strongly with a hairpin flanked by single-stranded RNA, demonstrating that part of the preferential binding of Hera to a flanking single-stranded region stems from the RBD itself. The RBDs seem to play a role beyond RNA binding, however, and are also involved in the stimulation of ATP-hydrolysis by RNA and RNA unwinding. Interestingly, removing one of the two RBDs has only a moderate effect on RNA binding, but is sufficient to reduce the RNA-stimulated ATPase activity of Hera (per core) and the rate constant of RNA unwinding. Removing the second RBD strongly reduces the RNA affinity and has an additional deleterious effect on the unwinding activity but does not further compromise the ATPase activity of Hera. The conformational change in the helicase core of Hera on RNA and ATP binding that is necessary for RNA-stimulated ATP hydrolysis and ATP-dependent RNA unwinding can occur in the absence of the RBDs. These differential effects point to different levels of crosstalk and cooperativity within and between the protomers of the Hera dimer.

Cores

The effect of the Hera cores on the RNA-stimulated ATPase activity depends on the context: inactivating one of the cores in the Hera dimer has no effect on the ATPase activity or RNA affinity of the remaining functional core (Hera versus Hera/Hera_K51Q). Hera with only one core and one RBD also shows wildtype-like ATPase activity and RNA affinity (Hera versus the cis-heterodimer). Both observations suggest little cooperativity between the cores in ATP hydrolysis and RNA binding. However, in the absence of the RBDs (dimeric versus monomeric core), the number of cores does matter, both for ATP hydrolysis and for RNA binding, pointing to some cooperativity that may be masked in the presence of the RBD(s). In contrast, RNA unwinding shows (some) cooperativity between the two protomers both in the presence and absence of the RBDs: inactivating the core in one protomer (Hera versus Hera/Hera_K51Q) or removing one of the cores (dimeric versus monomeric core) both reduces the helicase activity of the remaining protomer. Consistent with these observations the unwinding activity of the cis-heterodimer is reduced compared to Hera. In contrast, RNA- and ATP-induced closing of the core is not only independent of the RBDs but also independent of the presence of the core in the second protomer.

Orientation of RBD and core

Hera with a single functional core and a single RBD in wildtype-like configuration (cis-heterodimer) can catalyze RNA-dependent ATP hydrolysis and binds single- and double-stranded RNA with wildtype-like properties. Despite this, the unwinding activity is reduced, consistent with some cooperativity between the protomers in the Hera dimer in unwinding. The orientation of the RBD with respect to the helicase core is critical for RNA binding: high-affinity RNA binding is achieved only if the RBD and the core are on the same protomer. If the RBD is located on the other protomer, RNA affinities resemble those of the isolated RBD. This picture changes in the presence of ATP, though: the RNA-stimulated ATPase activities of cis- and trans-heterodimers are similar, the conformational change of the helicase core is possible in both configurations, and the rate constants of RNA unwinding are similar. These observations show that the RBD on one protomer can functionally interact with the core on the other protomer, and suggest that ATP binding overrides the preference of the RBD to preferentially interact with the core in cis.

Comparison of Hera to other dimeric DEAD-box helicases

Hera is the founding member of a small sub-family of DEAD-box helicases that form stable dimers. Similar dimeric structures, formed by interactions between dimerization domains homologous to the one in Hera, have also been reported for E. coli CsdA/DeaD [29] and CshA from Geobacillus stearothermophilus [27]. Thus, dimeric DEAD-box helicases are not exclusive to thermophilic organisms, suggesting that dimerization is not predominantly a strategy to achieve thermostability, but may have functional implications. For both CsdA and CshA, some of the deletion variants corresponding to the ones we analyzed here for Hera have been characterized with respect to ATPase activity as well as RNA binding and unwinding [27, 29]. Strikingly, the affinities of CsdA, the dimeric core, and the isolated RBD for the 32mer (0.07, 28, and ∼1 μM, respectively) are very similar to Hera. The RNA affinities of full-length CshA, the dimeric and the monomeric core, and the isolated RBD (0.31, 18, 6, and 0.30 μM for a hairpin RNA) are very similar to those of CsdA and of Hera, demonstrating that the RBD is the major binding platform and the anchor for CsdA, CshA, and Hera on RNA. The full-length protein shows the highest RNA affinity in all three helicases, demonstrating that RBD and core cooperate in RNA binding. In contrast to our observations with Hera, the monomeric core of CshA shows a higher affinity for RNA (6 μM) than the dimeric core (18 μM). CshA binds to a hairpin substrate in a 2:1 stoichiometry, with one dimer interacting with a single hairpin [27]. While the authors interpret this as a cooperation of both helicase cores in RNA unwinding, the underlying mechanism remains unclear.

The residues on the CsdA RBD involved in RNA binding differ from the ones that mediate RNA binding to Hera [21], suggesting different modes of binding even within this sub-family. Despite the different RNA binding modes, the CsdA RBD shows a preference for G-rich single-stranded RNAs similar to the GGGPur motif recognized by the Hera RBD [21].

Similar to Hera, the dimeric core of CsdA shows RNA-stimulated ATPase and ATP-dependent RNA unwinding activities, while no activity is detected for the monomeric core [29]. A CsdA deletion construct comprising the core, the DD, and the RRM of the RBD, but lacking the C-terminal tail, shows increased ATPase and RNA unwinding activities, suggesting that the RBDs of CsdA also contribute to these activities as we observe for Hera. For CshA, RNA-stimulated ATP hydrolysis was ∼50-fold faster for the full-length protein compared to the dimeric and monomeric cores, suggesting a role of the RBD in RNA-stimulated ATP hydrolysis [27].

Cooperativity in other DEAD-box proteins

Beside the dimeric helicases CsdA and CshA, with an architecture similar to Hera, other DEAD-box helicases have been captured as dimers or functional multimers when bound to RNA. One example is the E. coli DEAD-box protein DbpA that has been crystallized in complex with an RNA substrate similar to the 32/9mer [44]. DpbA forms a dimer in the crystal, with the monomers cross-linked by the bound RNA. The relative arrangement of the two helicase cores in this dimer differs substantially from the relative orientation of the two cores in Hera (see Fig. 9B), and most likely does not have any functional relevance: previous work has shown that DbpA is a monomer over a wide range of conditions in solution [45]. In contrast, the Saccharomyces cerevisiae DEAD-box protein Ded1p has been described as a functional trimer, in which two “loading protomers” bind to single-stranded regions and recruit a third protomer that catalyzes unwinding of an adjacent duplex [46]. A similar multimerization, induced by the RNA substrate, has been suggested for eIF4A [47]. Structural studies of the human ortholog of Ded1p, DDX3X, bound to double-stranded RNA, revealed two helicase molecules, each interacting with one of the two strands of the same duplex [48] (see Fig. 9A). This arrangement has been interpreted as cooperative action of the two helicase cores on duplex destabilization. Again, the relative geometry of the two monomers is very different from the juxtaposition of the two helicase cores in Hera, suggesting different modes of action.

A structural model for the interaction of Hera with RNA?

How can the cooperation between RBD and core in RNA binding be rationalized on a structural level? The crystal structure of the RBD of B. subtilis YxiN [49] with an RNA containing hairpin 92 of the 23S RNA has revealed the molecular details of the interaction of this RBD with hairpin 92 in the 32mer and 32/9mer (Fig. 8A), with interactions between conserved residues of the RBD and bases in the apical loop of hairpin 92. A very similar binding mode was observed for E. coli DbpA [44] (Fig. 8A). The RBD of Hera is structurally distinct from the RBDs of YxiN and DbpA [17], and RNA binding to the RBD of Hera is markedly different [21]. Binding of RNA to Hera is mediated by a different interface and set of RBD residues, leading to a binding site with different specificity [21]. The position of single-stranded RNA bound to the Hera RBD is different from the position of the loop bound to YxiN and DbpA (Fig. 8A). In fact, the Hera RBD does not recognize the loop, but instead interacts with a flanking single-stranded region through the RRM and with the stem of the hairpin though its C-terminal tail [21], which is reflected in the higher affinity of the RBD for the 32mer compared to the 32/9mer, as we report here.

In the structures of the RNA complexes of YxiN and DbpA, the duplex region that is unwound by the helicase core points away from the core (Fig. 8B and C), implying a substantial conformational change before unwinding can occur. Although single-molecule FRET experiments revealed a conformational change of YxiN on RNA binding to the RBD, involving a large translational and rotational motion of the RBD relative to the YxiN core that leads to its allosteric activation [50], there is no supporting evidence for such a motion of the RBD on RNA binding in DpbA [51].

The possible position of a duplex bound to the Hera helicase core can be gleaned from structural information on the helicase cores of DbpA [44], the DEAD-box helicase DDX3 [48], and Mss116 [52] bound to RNA duplexes. Superposition of these cores with one of the helicase cores of Hera (Fig. 9A–C) places the RNA duplex between the two helicase cores, with the backbone of one of the strands superimposing closely with the single-stranded RNA as bound to the helicase core of Vasa [23]. The RNA binding site on the second core of Hera is barely 10 Å away (Fig. 9D), demonstrating that only a small conformational change of the dimer would enable cooperative action of both cores on the same duplex. Such a binding mode would suggest a stoichiometry of one RNA bound per Hera dimer, which has been observed for binding of an RNA hairpin to the dimeric helicase CshA [27].

In the structure of the DbpA dimer, each RNA binds to the RBD of one DbpA molecule via hairpin 92, and to the helicase core of the other DbpA molecule through the duplex region. The distance between the hairpin of one RNA bound to the RBD and of the duplex of the second RNA molecule bound to the core of the same DbpA molecule can be bridged by the three nucleotides connecting hairpin and duplex in the RNA substrate, affording a molecular model of 32/9mer binding to DbpA [51] (Supplementary Fig. S12). The RBD in Hera has a very different position relative to the helicase core in Hera compared to DbpA and YxiN, and interacts with the stem of hairpin 92 and the 5′-sequence flanking the hairpin 92 instead of the apical loop. Constructing a model for the 32/9mer bound to Hera is not as straight-forward as for DbpA (Supplementary Fig. S12). The two ends of RNA that are covalently linked in the 32mer, the 5′-end of the RNA bound at the RBD and the 3′-end of the duplex strand superimposing with the single-stranded RNA as bound to the helicase core of Vasa, are 58 Å apart. Although the distance between the 5′-end of the RNA bound at the RBD in the opposite protomer is much smaller, 36 Å, it is still much too large for a covalent linkage without any conformational change. The molecular details of the interactions of the Hera domains with RNA will only be unraveled by structural data for a Hera–RNA complex.

Conclusions

From our data, a picture emerges in which Hera binds RNAs through its RBD that serves as an anchor point and enables the helicase core to unwind adjacent duplexes. If the duplex to be unwound is placed at the active site of one helicase core in such a way that the second helicase core would have to move just a few Ångstrom to act on the same duplex, both cores could act cooperatively in unwinding of this duplex. Notably, the juxtaposition of the two helicase cores in Hera is such that they would act on the same strand of the duplex, not on opposite strands as in the DDX3X–RNA complex. Such a mode of action would actually be more efficient as it enables the two protomers to destabilize a longer duplex region. It would also explain why deletion of one RBD in Hera is not as deleterious as deleting both RBDs: one RBD would be sufficient to facilitate binding of the helicase core to an adjacent duplex, and the second core is always within close reach. It would further rationalize the activities of the cis- and trans heterodimers: in both cases, the single RBD initiates RNA binding, facilitating the interaction of the helicase core with the duplex. The intrinsic plasticity of the DD enables changes in the relative orientation of the RBD and the two helicase cores. Unwinding with a single core is less efficient, which could be linked to the destabilization of a shorter region of the duplex by the action of a single core. The presence of two RBDs in the Hera dimer might constitute an additional benefit and make the capture of RNAs for unwinding more efficient. Altogether, the Hera dimer thus might be optimized for efficient unwinding of RNA substrates by the concerted action of two cores.

Acknowledgements

We thank Jochen Reinstein for helpful discussions, Christoph Flüchter for preliminary experiments with Hera containing two helicase cores and a single RBD, and Jessica Guddorf and Daniela Schlingmeier for excellent technical assistance.

Author contributions: D.K. conceptualized research and acquired funding. P.D., C.K., L.S., B.S., and A.Z.Z. performed experiments. P.D., B.S., A.Z.Z., and D.K. designed experiments. P.D., C.K., B.S., A.Z.Z., and D.K. analyzed data. P.D. and D.K. wrote and revised the manuscript.

Supplementary data

Supplementary data is available at NAR online.

Conflict of interest

None declared.

Funding

This work was supported by the Deutsche Forschungsgemeinschaft (DFG) (KL1153/7-1, 7-2 to D.K.). Funding to pay the Open Access publication charges for this article was provided by the University of Muenster.

Data availability

All data described are presented in the manuscript figures and/or tables, or in the Supplementary data. Original data are available from the corresponding author on request.

References