Significance

Abstract

Results

The SPL: A Protein-Directed Ribozyme.

To characterize the SPL functional dynamics and the leading role of Spp42, the largest and the most important protein of the SPL, we have computed the cross-correlation scores (CSs) for each pair of SPL components (i.e., RNAs and proteins, explicitly considering the Spp42 domains and N-terminal motifs), which enable to qualitatively measure their interdependent dynamical coupling (27–29). The CSs histogram for ILS-1 model (Fig. 2B) is generated from the per-residue Pearson’s coefficients (CCs) cross-correlation matrix (Fig. 2A) and provides a coarse and prompt picture of the most relevant linearly coupled motions occurring across the system during the MD simulations. Among others, the following crucial connections disclose important results: (i) all of the snRNAs (U2, U5, U6) and the IL are overall positively correlated with Spp42 (especially U5) and most of the other proteins (blue squares highlighted with colored boxes in Fig. 2B), thus indicating that the SPL acts as a protein-directed ribozyme; (ii) Spp42 (Fig. 2C) is highly correlated with specific proteins, therefore governing and tuning the SPL plasticity through its unique multibranched architecture. In particular, the closest dynamical ties are established with two proteins of U5 snRNP (Cwf10 and Cwf17) and with the NTC DExD/H-box Prp5 protein. Indeed, Cwf10, a GTPase involved in SPL remodeling (6), moves in lock-step with the “lasso,” a peculiar N-terminal extended loop of Spp42, whereas Cwf17 and Prp5 are strongly coupled with two protruding “arms” of Spp42, the N-terminal α-helices binding to Cwf17 (B-Cwf17), and to reverse-transcriptase (B-RT), respectively; (iii) finally, U2 and IL positively correlate with Cwf19 and Spp42, pointing to a cooperative functional role involving these two proteins and the branch helix.

Open in a separate window

Fig. 2.

Cooperative motions underlying the functional dynamics in the spliceosomal ILS-1 model system. (A) Per-residue Pearson’s coefficients (CCs) cross-correlation matrix derived from the mass-weighted covariance matrix constructed over the last 670 ns of MD simulations of ILS-1 (replica #1) for Cα and P atoms. CCs are comprised between −1 (anticorrelation, red) and +1 (correlation, blue). The domains of Cwf10 highlighted in the matrix are (from Left to Right): N-t and G; proteins of Sm-ring are (from Left to Right): D3, B, D1, D2, E, F, and G. (B) Histogram reporting the (per-column) normalized correlation scores (CSs) calculated for each pair of spliceosomal components. Each column reports the CSs calculated between a specific component and all of the others (listed on rows). Spp42 domains and N-terminal motifs are distinctly considered. CS value ranges from −1 (anticorrelated motions) to +1 (lock-step motions). The most relevant correlations are highlighted using colored boxes. (C) Close-up of Spp42 (surface representation) in ILS-1, with the four domains shown/listed with different colors. The five N-terminal motifs are numbered and listed. Abbreviations: B-Cwf17/RT, binding to Cwf17/RT; I. Lariat, intron lariat; RT-f/p, reverse-transcriptase finger/palm; Thumb, Thumb/X.

The IL/U2 Helix Displacement.

Strikingly, behind the intriguing coupling discussed above in iii, the essential dynamics extracted from the PCA unveiled a displacement of the IL/U2 helix, whose unrolling is copromoted by Cwf19 and Spp42 (Fig. 3 and Movie S1). This original finding sheds lights on the so-far-unclear functional annotation of Cwf19 and is fully consistent with the late stage of the splicing cycle under study. In the ILS complex, in fact, the IL/U2 branch helix, formed before the catalysis to bulge out the branch point (BP) adenosine, must be displaced and unwound before IL linearization (i.e., debranching) and degradation (30–33).

Open in a separate window

Fig. 3.

Displacement (highlighted by a black arrow) of the intron lariat (IL)/U2 helix, which unrolls away from the SPL, as revealed by the PCA of MD trajectories. Arrows, colored by protein/RNA, show the motion of Cα/P atoms along the first eigenvector. Cwf19 (green) and Spp42 (cyan) are depicted as cartoon; U2 (orange) and IL (yellow) are shown as tube; U6 (blue) and the other proteins are represented as cartoons. Mg²⁺ (orange) and Zn²⁺ (yellow) are shown as spheres.

Interestingly, electrostatic calculations performed on the cryo-EM structure and on representative MD snapshots reveal the presence of a positively charged cavity, rich in arginines and lysines, exposed by the RT domain of Spp42 and by Cwf19 (SI Appendix, Fig. S5). This pocket acts as an electrostatic trap for the negatively charged nucleotides of U2 and IL. The gradual IL/U2 displacement (Fig. 4) appears to be triggered by two key events: (i) the electrostatic recruitment of IL/U2 in the vicinity of the positive pocket; (ii) the electrostatic lock of the BP region, that is, the BP adenosine (A501), and the flanking G100 and C502 nucleotides. This lock is modulated by K364, K366, and R388 of Cwf19, which form a polar tweezers stably anchoring the BP region (Fig. 5). The 3′-terminus of the IL is firmly embedded in the positive pocket, thus favoring the 5′-downstream unrolling of the branch helix. The positive clamp surrounding the BP region constitutes the pivot of the IL/U2 unrolling (Fig. 5), which, assisted by the concomitant action of Cwf19 and Spp42, promotes the IL/U2 displacement. Remarkably, the IL/U2 helix is truncated in the cryo-EM structure (SI Appendix, Fig. S1) (5, 6). This, consistently with our results, may be related to a significant mobility of IL/U2, which most likely hampered its complete reconstruction.

Open in a separate window

Fig. 4.

The electrostatic origin of the branch helix displacement. Snapshots from MD simulations showing the gradual displacement of the intron lariat (IL)/U2 helix. (Left) Proteins are shown with electrostatic surface (blue/red colors for positive/negative charges, respectively). (Right) Cwf19 (green) and Spp42 (cyan) are represented with surface; U2 (orange), IL (yellow), and U6 (blue) are shown as cartoon. (A) The positively charged pocket, formed by Cwf19 and Spp42 recruits IL/U2. (B) IL/U2 interacts with this pocket and starts unrolling. (C) IL/U2 displacement due to the action of Cwf19 and Spp42.

Open in a separate window

Fig. 5.

The polar tweezers formed by K364, K366, and R388 of Cwf19. Cwf19 and Spp42 are shown with the electrostatic surface (blue/red colors for positive/negative charges, respectively). U2 (orange), U6 (blue), and IL (yellow) are depicted as cartoons. IL A501 (spheres) and 3′-termini (licorice) are colored by atoms.

The Enlarged (ILS-2) Model.

Here, results from the MD simulation of the enlarged ILS-2 model (Fig. 1B) are compared with those of ILS-1. The PCA scatterplot shows that the conformational space explored along the first two PCs is similar in the two systems (SI Appendix, Fig. S6). A narrower distribution of points is observed for ILS-2, with the contribution of the first three PCs to the overall SPL motion being roughly equivalent to that of PC1-only in ILS-1 (SI Appendix, Fig. S7). The contraction of the conformational space explored by ILS-2 can be ascribed to the proteins (Prp45 and Prp17) and domains (of Spp42 and Cwf10) additionally included in this model. Prp45 in particular appears to stabilize the overall SPL structure, damping and softening the functional motions of ILS-2 (at least within the timescale investigated). Despite this, the coarse CSs histogram (Fig. 6B) obtained from the per-residue cross-correlation matrix (Fig. 6A) confirms the central role of Spp42 (Fig. 6C) in directing the motions of distinct SPL components as observed in ILS-1. Interestingly, in ILS-2, the IL/U2 displacement is shown by PC2 and PC3 (Fig. 7), which display an increased relative contribution to the total variance with respect to that observed in the ILS-1 model (SI Appendix, Fig. S7). Finally, the MD simulation of ILS-2 further assesses the critical role exerted by the polar tweezers, which forms the hinge of the IL/U2 unrolling by tightening the BP region (SI Appendix, Fig. S8). Here, in addition to K364, K366, and R388, also the R385 residue exposed by Cwf19 contributes to lock the BP region.

Open in a separate window

Fig. 6.

Cooperative motions underlying the functional dynamics in the spliceosomal ILS-2 model system. (A) Per-residue Pearson’s coefficients (CCs) cross-correlation matrix derived from the mass-weighted covariance matrix constructed over the last 920 ns of MD simulations of ILS-2 (replica #1) for Cα and P atoms. CCs are comprised between −1 (anticorrelation, red) and +1 (correlation, blue). The domains of Cwf10 highlighted in the matrix are (from Left to Right): N-t, G, D-II, D-III, D-IV, D-V, and C-t; proteins of Sm-ring are (from Left to Right): D3, B, D1, D2, E, F, and G. (B) Histogram reporting the (per-column) normalized correlation scores (CSs) calculated for each pair of spliceosomal components. Each column reports the CSs calculated between a specific component and all of the others (listed on rows). Spp42 domains and N-terminal motifs are distinctly considered. CS value ranges from −1 (anti-correlated motions) to +1 (lock-step motions). The most relevant correlations are highlighted using colored boxes. (C) Close-up of Spp42 (surface representation) in ILS-2, with the six domains shown/listed with different colors. The five N-terminal motifs are numbered and listed. Abbreviations: B-Cwf17/RT, binding to Cwf17/RT; Endo., endonuclease; I. Lariat, intron lariat; RNase, RNase H-like; RT-f/p, reverse-transcriptase finger/palm; Thumb, Thumb/X.

Open in a separate window

Fig. 7.

Displacement (highlighted with a black arrow) of the intron lariat (IL)/U2 branch helix, which unrolls away from the SPL, as revealed by the PCA of MD trajectory of ILS-2 model. Arrows, colored according to the protein/RNA filament to which they belong, show the motion of Cα/P atoms along the second eigenvector (PC2, thicker arrows) and third eigenvector (PC3, thinner arrows). Cwf19 (green) and Spp42 (cyan) are depicted as cartoon; U2 (orange) and IL (yellow) are shown as tube; U6 (blue) and the other proteins are represented as cartoons. Mg²⁺ (orange) and Zn²⁺ (yellow) are shown as spheres.

Discussion

The SPL is an extremely dynamic machinery, undergoing continuous conformational and compositional changes. Although recent cryo-EM experiments have provided significant insights on this sophisticated macromolecule, the SPL structural/functional plasticity still calls for an in-depth understanding, aimed at better dissecting the specific role of the different proteins recruited along the splicing cycle. In the present study, we disentangle the most important cooperative motions underlying the functional dynamics of the ILS complex from an unprecedented atomic-resolution perspective. As major results, our work elucidates the leading role of the Spp42 protein and supplies key hints for a functional annotation of the NTR Cwf19 protein.

As highlighted by correlation and PCA analyses, our MD simulations corroborate that the SPL acts as a protein-directed ribozyme. Indeed, the three snRNAs of the ILS complex, all included in our models, are extremely tied with the proteins (Spp42 in particular) in their dynamical evolution, suggesting that their motions are directed by the protein scaffold (Figs. 2B and ​and6B).6B). Significantly, this unique trait of the SPL, so far highlighted by structural considerations based on the cryo-EM models solved in the last years (12, 25), is now sustained also by our simulations. In this scenario, the large Spp42 (Prp8 in S. cerevisiae) protein (Figs. 2C and ​and6C),6C), which exhibits a peculiar structural organization made of seven distinct domains (i.e., N-terminal, RT finger/palm, thumb/X, linker, endonuclease-like, RNase H-like, Jab1/MPN), is the central hub of the SPL (1–4, 25). Despite the dramatic compositional and conformational remodeling observed along the splicing cycle, the cryo-EM models of the SPL (mostly from S. cerevisiae) have pictured that Prp8 (Spp42) constitutes a rigid scaffold from the B^act to the ILS complex, with the exception of its highly mobile RNase H-like domain and some peculiar structural motifs and loops (25). Apparently, small conformational changes involving these local structural elements and RNase H-like domain of Prp8 (Spp42) may contribute to finely control the progression along the SPL cycle (25). Remarkably, these experimental observations are fully consistent with the functional dynamics of Spp42 captured from our simulations, showing that Spp42 heads the gene maturation in the ILS by modulating the motions of specific proteins through its sprawling multibranched architecture. The leading role of Spp42 is particularly emphasized by the plasticity of its N-terminal domain, which bears peculiar structural motifs such as the lasso loop or α-helices acting as protruding arms. Indeed, the CSs histograms calculated for ILS-1 (Fig. 2B) and ILS-2 model (Fig. 6B) reveal that, through these N-terminal motifs/loops, Spp42 establishes a close dynamical connection with specific proteins, among which are Cwf10, Cwf17, and Prp5. These couplings are remarkably evident and conserved in both ILS-1 and ILS-2 models. Since it has been hypothesized that the stage-specific splicing factors and enzymes are recruited onto a generally similar SPL scaffold (12, 25), it is tempting to suggest that the functional dynamics of Spp42 (Prp8) elucidated here may be shared also by other SPL complexes assembled along the cycle.

Intriguingly, the fine modulation of ILS cooperative motions occurs while preserving a stable catalytic site placed at the core of the SPL and composed of four Mg²⁺ ions coordinated by U6 snRNA phosphate ligands (1–4). This intricate RNA-based active site has been extraordinarily conserved across evolution from bacteria to humans for catalyzing the pre-mRNA splicing reactions (1–4). However, at variance with G2IRs, where the active site is stabilized by a network of RNA/RNA interactions, in the SPL the catalytic center is embedded in a protein cavity formed by Spp42 (Prp8) (25). In line with these experimental observations, our simulations confirm the stability and the integrity of the catalytic site even at the late stage of splicing studied here (ILS), further assessing the reliability of the adopted simulation protocol.

Of utmost importance is that the cryo-EM structure investigated here provides structural information of Cwf19 (5, 6), whose function has remained so far obscure (32). The S. pombe Cwf19 (CWF19L2 in humans) (19) is the paralog of the S. cerevisiae Drn1, as they show only a conserved C-t domain (32). In S. cerevisiae, Drn1 enhances the activity of its homolog Dbr1 (32), the enzyme that linearizes the lariats by cleaving the 2′,5′-phosphodiester linkage (33). Garrey et al. (32) hypothesized that Cwf19 (as well as CWF19L2 in humans), lacking a complete evolutionary homology with Drn1, may have developed supplemental functions that go beyond the regulation of Dbr1 debranching activity. In this respect, our results appear to foster this theory, suggesting that Cwf19 is effectively involved in the displacement of the branch helix (Figs. 3, ​,4,4, and ​and7),7), a critical event preceding (and necessary for) IL debranching. As a final remark, the electrostatics, which plays a fundamental role in the SPL for the recruitment and the placement of the snRNAs, as well as for the formation of the catalytic cavity (5, 6, 25), is here suggested to be also at the origin of this functional motion affecting the ILS disassembly (Fig. 5 and SI Appendix, Figs. S5 and S8).

In summary, multi-microsecond-long all-atom MD simulations of the heart of the ILS complex provide unprecedented insights on the SPL functional plasticity. This work repaints the key role of Spp42 as that of an orchestra conductor of a gene maturation symphony by strikingly revealing its fine modulation of the functional motions of many distinct SPL components. Among those, the essential dynamics analysis has unveiled an electrostatically driven unrolling and displacement of the branch helix copromoted by Cwf19 and Spp42. This finding provides evidences for a functional involvement of Cwf19 in the ILS disassembly. A detailed comprehension of the eukaryotic splicing has potential breakthrough implications for revolutionary gene modulation therapies to fight complex diseases. In this scenario, our work represents an unprecedented attempt to characterize the SPL functional dynamics with MD simulations and dispenses a noteworthy piece of knowledge for a thorough mechanistic understanding of this fundamental step of gene expression.

Materials and Methods

MD simulations were based on the S. pombe SPL reconstructed with cryo-EM at the average resolution of 3.6 Å (PDB entry 3JB9) (5, 6). The ILS-1 and ILS-2 models were built on the core region of this structure, representing the most important and conserved part of the SPL throughout the splicing cycle. De novo model building, as implemented in Modeler 9, version 16 (34), was used to reconstruct short missing loops where necessary. MD simulations were performed with Gromacs 5 (35) software package. The AMBER-ff12SB FF was adopted for proteins (36), while the ff99+bsc0+χOL3 FF was used for RNAs (37, 38), since these are the most validated and recommended FFs for protein/RNA systems (26). Mg²⁺ ions were described with the nonbonded fixed point charge FF due to Åqvist (39) as it was shown to properly describe binuclear sites (9). Na⁺ ions parameters were taken from Joung and Cheatham (40), while Zn²⁺ ions were modeled with the cationic dummy atoms approach developed by Pang (41). A slow and careful equilibration protocol was adopted before the productive phase as suggested by Šponer et al. (26). The trajectories were inspected and visualized with VMD software (42). All of the analyses, including rmsd, R_g, PCA, and the calculation of the cross-correlation matrices, were done with cpptraj module of Ambertools 16 (43) and with Gromacs 5 (35) suite. The correlation scores (CSs) were derived from the cross-correlation matrices (Figs. 2A and ​and6A6A and SI Appendix, Figs. S9 and S10), following the approach adopted in previous studies (27–29). Electrostatic calculations were performed on the proteins included in ILS-1 and ILS-2 with the adaptive Poisson-Boltzmann solver (APBS 1.4) software (44). A detailed description of the structural models (ILS-1 and ILS-2), systems setup, MD simulations, PCA, correlation analysis, and electrostatic calculations are provided in SI Appendix.

Supplementary Material

Acknowledgments

Footnotes

References