Crystal structure of U2 snRNP SF3b components: Hsh49p in complex with Cus1p-binding domain

Abstract

Spliceosomal proteins Hsh49p and Cus1p are components of SF3b, which together with SF3a, Msl1p/Lea1p, Sm proteins, and U2 snRNA, form U2 snRNP, which plays a crucial role in pre-mRNA splicing. Hsh49p, comprising two RRMs, forms a heterodimer with Cus1p. We determined the crystal structures of Saccharomyces cerevisiae full-length Hsh49p as well as its RRM1 in complex with a minimal binding region of Cus1p (residues 290-368). The structures show that the Cus1 fragment binds to the α-helical surface of Hsh49p RRM1, opposite the four-stranded β-sheet, leaving the canonical RNA-binding surface available to bind RNA. Hsh49p binds the 5' end region of U2 snRNA via RRM1. Its affinity is increased in complex with Cus1(290-368)p, partly because an extended RNA-binding surface forms across the protein-protein interface. The Hsh49p RRM1-Cus1(290-368)p structure fits well into cryo-EM density of the B^act spliceosome, corroborating the biological relevance of our crystal structure.

Keywords: RNA binding; RRM; SF3b; U2 snRNP; splicing.

FIGURE 1.

Structure of Hsh49p–Cus1(290-368)p. (A) Domain organization of Hsh49p and Cus1p. The proline-rich region of Cus1p (290-339) contains the absolutely conserved Trp312. (B) Sequence alignment of RRM1 of Hsh49p and homologs from Drosophila melanogaster (Dm), Homo sapiens (Hs), Caenorhabditis elegans (Ce), Arabidopsis thaliana (At), and Hs RBM7. Sequence alignment of the proline-rich region of Cus1p and homologs, and the proline-rich region of Hs ZCCHC8. The sequence in gray is not present in the crystal structure of RBM7–ZCCHC8^Pro. Circles denote residues at the interface between RRM1 and Cus1(290-368)p, colored circles indicate residues that were mutated for pull-down experiments. Triangles denote residues that were mutated for RNA interaction studies. (C) Overall structure of the complex of Hsh49p (yellow) and Cus1(290-368)p (green) showing the arrangement of RRM1 and RRM2 with the C-terminal helix wedged in between. (D) Overlay of Hsh49p with RRM2 and RRM3 from Prp24p bound to U6 snRNA (4N0T). (E) Crystal packing between the C-terminal helix of Hsh49p and Cus1(290-368)p.

FIGURE 2.

Overview of RRM1–Cus1(290–368)p interaction. (A) Overview of Cus1(290–368)p (green) binding to the α-helical face of RRM1 (yellow). (B) Close up of the center region of the interaction surface. Cus1(290–368)p is lying in a hydrophobic crevice on RRM1. Key interacting residues are depicted in sticks and labeled. Hydrogen bonds are depicted as dashed lines. (C) Close up of the face-to-face arrays of Tyr65 and Tyr325 in Hsh49p and Cus1(290–368)p, respectively. (D) Close up of the face-to-face array of Tyr77.

FIGURE 3.

Overlay of RRM1–Cus1(290–350)p with RBM7 RRM–ZCCHC8(286–324) ([Falk et al. 2016]; pdb 5LXR). The RMSD of the main chains of the RRMs is 2.9 Å. (A) Overview of the interface with Hsh49p in yellow, Cus1p in green, RBM7 in pink, and ZCCHC8 in blue. Some of the variant interface residues are shown in stick representation. (B) Close-up of “patch 2” showing the extensive stacking array around Phe286 in the RBM7-ZCCHC8 heterodimer compared to the less striking interactions seen in this region of the Hsh49p–Cus1p complex.

FIGURE 4.

Coexpression of Hsh49p and GST(His)₆Cus1(290–350)p wild-type or mutant proteins and pull-down experiment. The proteins were coexpressed in E. coli and whole-cell contents are shown in the Input gel (I). Similar levels of Hsh49 are seen in all lanes but levels of tagged Cus1(290–350)p vary. Cells were lysed and after centrifugation the supernatant was mixed with Ni-NTA resin. Protein that remained bound to the resin after washing is shown in the Pulldown gel (PD). Only for complexes containing both wild-type components were stoichiometric levels of Hsh49p recovered from the pulldown; otherwise only in lane 15 [Cus1(290–350)p Y325A mutant] are significant levels observed. We attribute the lack of full-length tagged Cus1(290–350)p in many Input and Pulldown lanes to degradation of the Cus1(290–350)p, which is protease-sensitive when not bound to Hsh49p.

FIGURE 5.

RNA-binding studies of Hsh49p–Cus1(290–368)p with different 5′ end U2 snRNA oligos. (A) 5′ end of S. cerevisiae U2 snRNA. The pppG at the 5′ terminus is a result of the in vitro transcription, and the nucleotides denoted in red are post-transcriptionally modified to pseudouridine in vivo. (B) Summary of 5′ U2 snRNA oligos used for bandshift assays. (C) Bandshift of different 5′ U2 snRNA oligos. (D) Fluorescence anisotropy measurements of different RNA oligos with Hsh49p–Cus1(290–368)p or Hsh49p. In the absence of Cus1(290–368)p, Hsh49p binds significantly weaker. The error bars represent the SD of each data point calculated from three independent fluorescence anisotropy measurements. (E) Fluorescence anisotropy measurements of different Hsh49p–Cus1p constructs with 5′SLI oligo. As in D, it is immediately obvious that the presence of Cus1p proline-rich domain enhances the affinity. The data of RRM1 had to be fitted with a Hill coefficient of 1.8; in addition the higher maximum anisotropy shows that multiple RRMs could be binding to the RNA. (F) Fluorescence anisotropy measurements of mutants of Hsh49p or Cus1(290–350)p with 5′SLI. Mutation of the canonical RNA-binding residues of RRM1 significantly impair RNA binding. In addition, the Cus1p Arg290Ala mutant also decreases the RNA-binding affinity.

FIGURE 6.

Electrostatic surface of Hsh49p RRM1–Cus1(290–350)p. Crystals contained Cus1(290–368)p but residues 351–368 were disordered. (A) The electrostatic surface of the interface displays an extended positively charged region (blue). The surface potential was calculated using the program PDB2PQ (Dolinsky et al. 2007) and APBS (Baker et al. 2001). A pink dotted line represents where RNA may bind this basic region as well as the β-sheet of RRM1 on the right-hand edge, which is the canonical RNA-binding surface, to partially explain Cus1p's contribution to Hsh49p RNA binding. (B) Cartoon representation of the surface shown in A, upon which the residues mutated in the RNA-binding studies are indicated. (C) Rotated view of A showing the canonical RNA-binding surface. A possible RNA-binding site is indicated by the pink dotted line. (D) Cartoon representation of the surface shown in C.

FIGURE 7.

Comparison of Hsh49 modeled into EM density of yeast activated spliceosome (B^act) ([Yan et al. 2016], EMD-9524) with the crystal structure presented here. (A) Originally, RRM1 (blue) and RRM2 (pink) of Hsh49 were modeled side-by-side into the map ([Yan et al. 2016], pdb 5GM6). (B) The Hsh49p–Cus1(290–368)p structure was fitted into the EM density by superposition on RRM1 of the model of Yan and colleagues (RMSD 1.23 Å). Cus1(290–368)p (green) fits well into the density ascribed to RRM2, but the RRM2 from the crystal structure (yellow) would locate in an empty region of the map. (C) RRM2 may fit in unassigned density seen in the bottom left of this panel. Some of the pre-mRNA has been modeled (Yan et al. 2016) and the nucleotides seen to the left of RRM1 partially occupy unassigned density across the face of its β-sheet. There is no clear density for the nucleotides that lie across the edge of RRM1 and Cus1(290–368)p.

FIGURE 8.

Comparison of Hsh49p RRM1–Cus1(290–350)p interface with the interaction seen in the UHM–ULM complex of U2AF35–U2AF65 (pdb 1JMT). The structures are superposed on the RRMs with an RMSD of 1.91 Å. In both cases a tryptophan is interacting with the α-helical side of the RRM but the Cus1 Trp312 side chain is not inserted in a cleft between the α-helices like the ULM Trp92 of U2AF65. Apart from the general proximity of the tryptophans, there is no resemblance between Cus1(290–350)p and U2AF65 ULM.