The Human RNA Helicase DDX21 Presents a Dimerization Interface Necessary for Helicase Activity

Abstract

Members of the DEAD-box helicase family are involved in all fundamental processes of RNA metabolism, and as such, their malfunction is associated with various diseases. Currently, whether and how oligomerization impacts their biochemical and biological functions is not well understood. In this work, we show that DDX21, a human DEAD-box helicase with RNA G-quadruplex resolving activity, is dimeric and that its oligomerization state influences its helicase activity. Solution small-angle X-ray scattering (SAXS) analysis uncovers a flexible multi-domain protein with a central dimerization domain. While the Arg/Gly rich C termini, rather than dimerization, are key to maintaining high affinity for RNA substrates, in vitro helicase assays indicate that an intact dimer is essential for both DDX21 ATP-dependent double-stranded RNA unwinding and ATP-independent G-quadruplex remodeling activities. Our results suggest that oligomerization plays a key role in regulating RNA DEAD-box helicase activity.

Keywords: Biological Sciences; Molecular Biology; Molecular Modelling; Structural Biology.

Graphical abstract

Figure 1

Identification of the Dimerization Domain of DDX21 (A) Numbered linear sequence with the different domains colored in blue (HC), orange (DD), and yellow (GUCT) and the C-terminal tail with the FRGQR/PRGQR repeats in black boxes. The molecular weight of this sequence is MW_seq = 90.9 kDa (cleaved affinity tag) (see also Figures S1, Figure 1, Figure 2, Figure 3, Figure 4, Figure 5and S7). (B) Sequence alignment of the DDs from the two human paralogs DDX21 and DDX50, as well as the bacterial dimeric DEAD-box helicases Hera, CshA, and CsdA (see also Figure S3). (C) Schematic representation of the truncation constructs used in this study with their corresponding monomeric MW_seq calculated from their sequences. (D) SEC-MALS data show that all the constructs except DDX21_Core are dimers (theoretical molar masses [MW_seq] for Fl, ΔN, ΔNC, Core-DD, and Core are 90.9, 68.7, 60.6, 49, and 44.7 kDa, respectively). Experimental SEC-MALS molar masses are Fl, 198 kDa; ΔN, 143 kDa; ΔNC, 120 kDa; Core-DD, 91 kDa; and Core, 41 kDa) (see also Figures S4 and S5). (E) Homology model of the DD of DDX21 in cartoon representation, with helices α1 to α4 of each protomer colored in orange and gray, respectively. The residues in α4 mutated to create the monomeric mutants are shown in blue stick representation. (F) SEC-MALS of the Fl and ΔN monomeric mutants (in dashed lines) with molar masses of 88 and 66 kDa, compared with intact DDX21_Fl (blue) and DDX21_ΔN (orange) with molar masses of 198 and 143 kDa.

Figure 2

The DDX21 Dimer Binds Two R-loop Molecules (A) Mass photometric profiles obtained for DDX21_Fl (in blue) and DDX21_Fl monomer mutant (in gray) with the determined average molecular mass indicated above each peak. The theoretical molar mass for DDX21_Fl (with uncleaved affinity tag) is 104.9 kDa and for the DDX21_Fl monomer mutant is 90 kDa. The wild-type protein shows a monomer-dimer equilibrium, whereas the mutant is mainly monomeric. (B) Mass photometric profiles for DDX21_ΔN, DDX21_Core-DD, and DDX21_core indicate that DDX21_ΔN displays a monomer-dimer equilibrium as seen for DDX21_Fl, whereas DDX21_Core-DD and DDX21_core are mainly dimeric and monomeric, respectively. (C) Mass photometric profile for the DDX21_Fl-R-loop complex at 30 nM. (D) Fluorescence polarization binding assays measure a binding affinity of DDX21_Fl to the R-loop substrate of 9.2 ± 0.7 nM. Error bars represent the standard deviation of three independent measurements. The mass photometric experiments were performed in duplicates for the protein samples and in triplicates for the R-loop complex (see also Figure S6).

Figure 3

dsRNA Binding and Unwinding by DDX21 (A and B) FP experiments comparing the binding affinity of the DDX21 mutants to the (A) 15-mer ssRNA and the (B) dsRNA with a 3′ 15-base overhang, respectively. Error bars represent the standard deviation of three independent measurements (see also Table S1). (C) dsRNA helicase assay control experiment (time = 5 min) showing that upon addition of ATP, DDX21 can unwind the dsRNA, such that the exposed ssRNA can be cut by RNase T1 (lane 5). Without DDX21, RNase T1 cannot cut the substrate (lane 3). As described by Hammond et al. (2018) and Valdez et al. (1997) in the absence of ATP (lane 6), DDX21 has some residual helicase activity, which disappears upon addition of 5 mM EDTA (lane 7). The nature of the RNA bands in the gel is depicted on the right, where the star represents the 5′FAM modification. (D–F) dsRNA helicase assay comparing the activity of the different DDX21 mutants at different time points (5, 10, 15, and 30 min). In native condition polyacrylamide gels, the 5′-FAM-labeled RNA was visualized by measuring the fluorescent signal at 535 nm. (G) The helicase assays shown in (D–F) were repeated in triplicate (see also Figure S8); the activity was calculated by measuring the intensity of the product (cut) RNA bands, with respect to DDX21_Fl activity; and the quantification is shown in the graph, where the points represent the mean of three independent measurements and the error bars represent the standard error.

Figure 4

DDX21 Activity on RNA G-Quadruplexes (A) FP binding curves for the different DDX21 variants and the RNA G-quadruplex Q2. Error bars represent the standard deviation of three independent measurements (see also Figure S9 and Table S2). (B) The circular dichroism (CD) spectra of the RNA G-quadruplex (in red) with a parallel structure with its characteristic maximum at 260 nm and minimum at 240 nm. In the presence of DDX21_Fl (blue), this structure is maintained, as showed by the maximum at 260 nm that is not present in the spectrum of the protein alone (black) (see also Figure S10). (C and D) RNA G-quadruplex remodeling assay comparing the activity of the different DDX21 mutants. In native condition polyacrylamide gels, the 5′-FAM-labeled RNA was visualized by measuring the fluorescent signal at 535 nm. The activity was calculated by measuring the intensity of the product (cut) RNA bands, with respect to DDX21_Fl activity. (E) Quantification of the G-quadruplex remodeling assay for the constructs with activity above background: Fl, ΔN, and the monomeric ΔN and Fl mutants. The graph on the right summarizes the results from three independent experiments (see also Figure S11). Error bars represent the standard deviation.

Figure 5

SAXS-Guided Modeling and DDX21 Conformational Flexibility (A) Experimental scattering curves generated from the SEC-SAXS traces shown in Figure S12A. (B) Pair distance distribution functions indicating the maximum dimension of the particles, D_max. (C) Dimensionless Kratky plot shows the characteristic profile for flexible multi-domain proteins (see also Figures S12 and S13). (D) The ensemble of the outcome structures from the NOLB NMA for DDX21_Core-DD displays a linear arrangement of the HC and DD domains (see also Figure S14). (E) Comparison between the calculated scattering curves from the ensemble models (black) and the DDX21_Core-DD experimental scattering curve (light blue), giving an excellent fit with χ² of 1.4. (F) Representative DDX21_ΔNC model of the ensembles created by flexible fitting, depicting the flatness of apo DDX21_ΔNC in solution. (G) Ensemble of the outcome structures from the NOLB NMA for DDX21_ΔNC. The arrows indicate the domain movements after aligning the conformers by the DD. The HC domains are shown in blue, the DD in orange, and the GUCT domains in yellow. (H) Comparison between the calculated scattering curves from the ensemble models (black) and the DDX21_ΔNC experimental scattering curve (yellow), giving an excellent fit with χ² of 1.3 (see also Figures S15–S17).