Structures of nonsense-mediated mRNA decay factors UPF3B and UPF3A in complex with UPF2 reveal molecular basis for competitive binding and for neurodevelopmental disorder-causing mutation

Abstract

UPF3 is a key nonsense-mediated mRNA decay (NMD) factor required for mRNA surveillance and eukaryotic gene expression regulation. UPF3 exists as two paralogs (A and B) which are differentially expressed depending on cell type and developmental stage and believed to regulate NMD activity based on cellular requirements. UPF3B mutations cause intellectual disability. The underlying molecular mechanisms remain elusive, as many of the mutations lie in the poorly characterized middle-domain of UPF3B. Here, we show that UPF3A and UPF3B share structural and functional homology to paraspeckle proteins comprising an RNA-recognition motif-like domain (RRM-L), a NONA/paraspeckle-like domain (NOPS-L), and extended α-helical domain. These domains are essential for RNA/ribosome-binding, RNA-induced oligomerization and UPF2 interaction. Structures of UPF2's third middle-domain of eukaryotic initiation factor 4G (MIF4GIII) in complex with either UPF3B or UPF3A reveal unexpectedly intimate binding interfaces. UPF3B's disease-causing mutation Y160D in the NOPS-L domain displaces Y160 from a hydrophobic cleft in UPF2 reducing the binding affinity ∼40-fold compared to wildtype. UPF3A, which is upregulated in patients with the UPF3B-Y160D mutation, binds UPF2 with ∼10-fold higher affinity than UPF3B reliant mainly on NOPS-L residues. Our characterization of RNA- and UPF2-binding by UPF3's middle-domain elucidates its essential role in NMD.

Figure 1.

UPF3B domain architecture, RNA binding and RNA-induced oligomerization. (A) Schematic representation of UPF3B domain architecture predicted by homology modelling. (B) Schematic representation of the UPF3B constructs (left) utilized for testing dsRNA binding and corresponding dissociation constants determined by fluorescence anisotropy experiments (right) mapping the nucleotide-binding interface of UPF3B to its RRM-L and N-terminal part of the middle-domain (41-262) (binding curves in Supplementary Figure S7). (C) Electrophoretic mobility shift assays of UPF3B-WT (left) and UPF3B-41–262 (right) using double-stranded 24-mer RNA (dsRNA) indicating that UPF3B-41–262 retains RNA-induced oligomerization behaviour.

Figure 2.

Crystal structure of UPF3B-41–189 + UPF2-MIF4GIII complex. (A) Cartoon representation of the crystallographic model with three representative views of UPF2-MIF4GIII (grey) complexed with UPF3B-41–189 comprising the RRM-L domain (cyan), the NOPS-L linker and α-helix (magenta). (B) Zoomed view (blue box, panel A) showing the residues involved in interactions between the NOPS-L linker with UPF2. (C) Zoomed view (red box, panel A) of residues in the NOPS-L α-helix contributing to binding of UPF2. Polar and ionic interactions are indicated (green lines). (D) WebLogo (37) alignment summarizing sequence conservation across a range of UPF3B homologs (Supplementary Figure S9) for the NOPS-L region indicating conservation of several of the residues involved in complex formation. The accumulative height of each stack (in bits) indicates the degree of conservation at that position. The height of each individual letter indicates the frequency that amino acid is found at that position.

Figure 3.

UPF3B-UPF2 interface analysis. (A) Electrostatic surface potential of spatially separated chains of UPF2-MIF4GIII and the NOPS-L domain of UPF3B derived from our crystal structure. Blue, red and white surface indicates positive, negative, and hydrophobic surface potential, respectively. Green arrows indicate how NOPS-L binds to UPF2-MIF4GIII regions. (B) Zoom in of hydrophobic cleft (orange box, panel A) interactions between UPF3B-NOPS-L (magenta cartoon) and UPF2-MIF4GIII (electrostatic surface and grey cartoon). (C) Zoom in of UPF2-MIF4GIII basic patch residues (blue box, panel A) and their interactions with UPF3B-NOPS-L. (D) Zoom in of UPF3B-NOPS-L α-helix residues (black box, panel A) involved in hydrophobic interaction with UPF2-MIF4GIII hydrophobic cleft (including Y160). Aromatic residues selected for mutagenesis are highlighted in yellow. (E) Fluorescence anisotropy binding curves of the complex of the RRM-L domain (UPF3B-41–143) with UPF2-MIF4GIII (red curve) and the complex crystallized in this study (UPF3B-41–189 + UPF2-MIF4GIII) (black curve) to a HEX-labelled 24mer ssRNA solution, indicating an inhibition of UPF2-RNA binding in the presence of UPF3B-41–189 relative to UPF3B-41–143. Protein titrations were carried out in triplicate and error bars plotted via standard deviation before fitting a single component binding equation in GraphPad Prism to calculate K_D values. (F) Coomassie-stained native 4–20% Novex gels loaded with ssRNA + UPF3B-WT incubated with increasing amounts of UPF2L (left) and with ssRNA + UPF2L incubated with increasing amounts of UPF3B-WT (right). Green boxes highlight UPF3B:UPF2 complexes at 1:1 stoichiometry.

Figure 4.

Characterization of interaction of UPF3B variants with UPF2-MIF4GIII. (A) Table summarizing the K_D values determined in this study by SPR between immobilized UPF2-MIF4GIII and UPF3B constructs, highlighting a ∼200-fold increase in affinity of UPF3B-41–189 relative to the RRM-L alone as well as a ∼40-fold affinity decrease due to Y160D mutation (green). (B) Representative binding curves of UPF3B-41–143 (RRM-L only) and UPF3B-41–189 are shown. Sensorgrams for additional mutation analyses are shown in Supplementary Figure S11.

Figure 5.

Characterization of UPF3A interaction with UPF2-MIF4GIII. (A) Table summarizing the K_D values between immobilized UPF2-MIF4GIII and UPF3B and UPF3A constructs, as determined by SPR. The higher affinity of UPF3A-58–206 compared to UPF3B-41–189 is highlighted (green). N.B. stands for no binding. (B) Representative binding curves of UPF3A (isoform variants 1 and 2) are shown. SPR of additional constructs is shown in Supplementary Figure S12. (C) Cartoon representation of the crystallographic model UPF2-MIF4GIII (grey) complexed with UPF3A-58–206 comprising the RRM-L domain (dark blue), the NOPS-L linker and α-helix (dark red) overlaid with UPF3B-41–189 comprising the RRM-L domain (cyan) and the NOPS-L region (magenta). (D) Zoomed view (red box, panel C) of residues in the RRM-L which undergo a structural rearrangement in the case of UPF3A (left panel) relative to UPF3B (right panel). Polar and ionic interactions are indicated (green lines) and corresponding residues are boxed and highlighted in green. (E) Zoomed view (blue box, panel C) highlighting the potential interaction (grey line) of E187 of UPF3A with K767 of UPF2 (dark red).

Figure 6.

UPF3′s interactions with UPF2 and RNA. UPF2′s MIF4GIII domain (grey) can interact with dsRNA (green), with UPF3B (cyan and magenta) or UPF3A (blue and red). Formation of UPF2-UPF3 complexes interferes with RNA-binding of UPF2 (dashed red line). UPF3B Y160D mutation weakens complex formation of UPF3B with UPF2 (dashed red line). UPF3A isoform 2 does not bind UPF2 (solid red line). UPF3B and UPF3A compete (opposing solid red lines) for interaction with UPF2. UPF3B binds dsRNA via the RRM-L, NOPS-L, and CCL-1 domains (cyan, magenta and pink). RNA-induced oligomerization (bottom right) at high UPF3B concentrations is prevented by UPF2.