Molecular architecture of SF3B and the structural basis of splicing modulation

During splicing non-coding introns are excised from the transcribed pre-messenger RNA (pre-mRNA), and the protein-coding exons are ligated to generate the mature mRNA. In cells, the pre-mRNA splicing reaction is catalyzed by the spliceosome, a highly dynamic molecular machine composed of five small nuclear ribonucleoprotein particles (snRNPs) and additional non-snRNP factors (Wahl et al., 2009; Will and Luhrmann, 2011). At the earlier stages of spliceosome assembly, the U2 snRNP is recruited to the 3’ region of the intron for the U2 snRNA to base-pair with the branch-site (BS), in a complex and insufficiently understood process (Wahl et al., 2009). SF3B is the largest U2 subcomplex, and several of its seven subunits, including SF3B1, contact both the U2 snRNA and the intron near the BS, stabilizing the U2/BS base-pairing interaction. Recurrent somatic cancer mutations in SF3B1, and in several related splicing factors, reduce the accuracy of BS selection and, finally, lead to aberrant splicing (Dvinge et al., 2016). The compromised function of SF3B1 affects splicing of many different transcripts and thus translates to global changes in cancer cell transcriptome (Alsafadi et al., 2016; Darman et al., 2015). SF3B is also targeted by several small-molecule splicing modulators, regarded as promising chemotherapeutic agents (Bonnal et al., 2012; Effenberger et al., 2017). At the start of this project, it was unclear how SF3B is organized prior to its incorporation in the spliceosome, what structural features of human SF3B are perturbed in cancers, and how the antitumor compounds act on SF3B to modulate splicing. In the first part of this thesis work, we carried out a thorough structural analysis of the human SF3B complex (Cretu et al., 2016). Firstly, we have defined a structurally stable SF3B core complex (~254 kDa), composed of SF3B1’s C-terminal HEAT domain, SF3B3, SF3B5, and PHF5A, and determined its structure by X-ray crystallography. The crystal structure of the SF3B core complex revealed that the 20 HEAT repeats of SF3B1 adopt a distinctive superhelical conformation and share extensive contacts with the other three core subunits (Cretu et al., 2016). SF3B3 exhibits a triple β-propeller fold and accommodates the three alpha helices of the SF3B5 subunit in a deep, clam-shaped cleft (Cretu et al., 2016). Organized as a compact knot composed of three zinc finger motifs, PHF5A bridges the terminal repeats of SF3B1’s HEAT domain, contributing to the unique conformation of the superhelix. Using a set of orthogonal mass spectrometry approaches, we showed that SF3B1-PHF5A together with the more mobile SF3B6/p14 subunit form a multipartite RNA binding platform that, in spliceosomes, stabilizes the U2/BS helix and the downstream 3’ end of the intron (Fica and Nagai, 2017; Shi, 2017). Comparative analyses with recent cryo-EM structures of yeast spliceosomes (Fica and Nagai, 2017) show that the cancer-related residues of SF3B1 map to a basic groove of the HEAT superhelix where, likely, the 3’ pyrimidine-rich region of the intron binds. Altogether, our analyses suggest how changes in SF3B1 structure and interactome may lead to a compromised BS selection, thus providing insights into the molecular mechanism of SF3B1-driven cancers. In the follow-up work (Cretu et al., 2018, unpublished data), we determined co-crystal structures of SF3B core variants in complex with different compounds that modulate splicing, including some approved for clinical trials. Our work shows that splicing modulators from the pladienolide and herboxidiene families target SF3B at the same site and bind to an hourglass-shaped tunnel formed by SF3B1’s H15-H17 repeats and PHF5A. Their molecular recognition is achieved in part by shape complementarity to the tunnel, enforced by the conjugated diene group – a moiety regarded as the common pharmacophore of SF3B modulators. Importantly, while SF3B1 exhibits a “closed” conformation in fully assembled spliceosomes, we observe a more “open” state in the presence of splicing modulators. Structural comparisons indicate that the modulator binding site, available in the “open” conformation, is rearranged in the “close” state of SF3B1 to accommodate the invariant BS adenosine. Thus, our analyses suggest that splicing modulators interfere with a conformational rearrangement of SF3B1 and, in this respect, act as competitive BS antagonists. Overall, this work may serve as a conceptual framework for the structure-based design of next-generation splicing modulators