Structural basis of catalytic activation in human splicing – Nature

Cloning

Recombinant proteins were produced from codon-optimized synthetic genes (GeneArt, ThermoFisher Scientific). Full-length wild-type Aquarius and its K829A mutant were fused to a C-terminal 8×His-tag and cloned into the pFL vector backbone as previously described⁶. For co-expression of IBC subunits, the master, 4-protein construct ∆ISCC, comprising SYF1 (also known as XAB2), ISY1, CCDC16 (also known as ZNF830) and PPIE (also known as CypE), was assembled by in vitro Cre–loxP recombination of the donor pSPL-CCDC16–PPIE and acceptor pFL-SYF1–ISY1 vectors⁶. The expression constructs were transformed into DH10MultiBacY Escherichia coli cells. The resulting bacmids were isolated using a High Pure Plasmid Isolation kit (Roche) and used for transfection into Sf9 (Spodoptera frugiperda) insect cells. Sf9 cells were transfected with the help of either X-tremeGENE HP (Roche) or FuGENE HD (Promega) reagents. Insect cell lines were not tested for mycoplasma contamination.

Human PRP2 (NCBI Reference Sequence identifier: NM_003587, transcript variant 1) and human PPIL4 (NCBI Sequence identifiers BC020986 and NM_139126.4) open reading frame clones were obtained from OriGene (RC202912) and Applied Biological Materials (373710120000), whereas human GPKOW and SKIP were codon-optimized for expression in insect cells and synthesized by GeneArt (ThermoFisher Scientific). All constructs were inserted by ligation-independent cloning into a modified pFastBac vector backbone in-frame with an N-terminal twin-StrepII affinity tag, which can be cleaved off with the HRV-3C protease. All recombinant constructs were verified by Sanger sequencing.

Purification of recombinant proteins

Recombinant proteins were produced in insect cells using the MultiBac system⁵⁰. The initial V₀ and V₁ baculovirus stocks were produced in Sf9 cells. Large-scale protein co-expression was conducted in High Five (Hi5) insect cells (Trichoplusia ni, BTI-TN5B1-4) through the combination of two V₁ stocks of baculovirus, coding for the C-terminal 8×His-tagged Aquarius (wild-type or mutant K829A) and the untagged ∆ISCC complex, in 1:1.5 ratio (V₁^AQR:V₁^ΔISCC). The purification of the wild-type and the K829A dominant-negative IBCs was performed according to a previous protocol established in our laboratory, but with some modifications⁶. Insect cells expressing IBC were disrupted by sonication in lysis buffer (50 mM HEPES-NaOH, pH 7.5, 400 mM NaCl, 10% (v/v) glycerol, 5 mM 2-mercaptoethanol (2-ME), 30 mM imidazole and protease inhibitors (complete EDTA-free, Roche)) and the cell debris was pelleted by centrifugation at 18,000 r.p.m. at 4 °C for 1 h (A27-8×50 rotor; Thermo Scientific). The supernatant was loaded on a 5 ml HisTrap HP Ni(II)-chelating column (GE Healthcare/Cytiva) equilibrated with the lysis buffer. The Ni(II)-bound protein species were eluted using a linear 30–400 mM imidazole gradient in Ni(II) elution buffer (50 mM HEPES-NaOH, pH 7.5, 200 mM NaCl, 10% (v/v) glycerol and 5 mM 2-ME) and analysed by SDS–PAGE. The IBC-containing fractions were loaded on an anion-exchange Q Sepharose HP column (GE Healthcare/Cytiva) equilibrated in 20 mM HEPES-KOH, pH 7.5, 200 mM KCl, 10% (v/v) glycerol and 5 mM dithiothreitol (DTT). The bound IBC was eluted off the resin with a linear 0.2–1 M KCl gradient. The peak fractions, corresponding to the stoichiometric IBC, were pooled, concentrated to 10 mg ml^–1, aliquoted, flash-frozen in liquid nitrogen and stored at −80 °C.

Twin-StrepII-tagged human PRP2(137–1022), PPIL4 and GPKOW were expressed in Sf9 or Hi5 insect cells using recombinant baculoviruses, prepared as described above, in amounts sufficient to induce cell cycle arrest in 24 h. All recombinant proteins used in functional assays were purified at 4 °C using similar purification protocols. In brief, insect cells, expressing the target proteins, were briefly sonicated and lysed in lysis buffer (50 mM HEPES-KOH, pH 7.5, 150 mM KCl, 10% (v/v) glycerol, 0.2% (v/v) Triton X-100 RNase-free, 2 mM DTT and protease inhibitors (complete EDTA-free, Roche)), with the detergent added after sonication, and the cell debris was pelleted by centrifugation at 23,000 r.p.m. for 1 h in an A27-8×50 rotor (Thermo Scientific) or 40,000 r.p.m. (approximately 164,244g) in a type 70 Ti rotor (Beckman Coulter) for 45 min at 4 °C. The filtered supernatant was applied to a 5 ml Strep-Tactin XT 4Flow pre-packed column (IBA Lifesciences) or incubated in batch for 1 h with around 1 ml or 2.5 ml Strep-Tactin beads per 1 litre of Sf9 or Hi5 culture, respectively. The bound proteins were eluted by competition with 50 mM biotin in lysis buffer or with elution buffer containing 60 mM biotin (50 mM HEPES-KOH, pH 7.5, 100 mM KCl, 5% (v/v) glycerol, 1 mM EDTA, 2 mM DTT and 60 mM biotin). The Strep-Tactin eluates were concentrated and further purified by size-exclusion chromatography (SEC) on Superdex 200 HiLoad 16/600 200 pg (GE Healthcare/Cytiva) or Superdex 200 Increase 10/300 GL (GE Healthcare/Cytiva) equilibrated in 20 mM HEPES-KOH, pH 7.5, 150 mM KCl, 10% (v/v) glycerol and 2 mM DTT. Alternatively, the PRP2(137–1022) Strep-Tactin eluates were purified on Q Sepharose HP (GE Healthcare/Cytiva) and eluted from the 5 ml anion-exchange column using a linear 0–30% gradient formed over 40 ml between buffer A (20 mM HEPES-KOH, pH 7.5, 100 mM KCl, 5% (v/v) glycerol and 0.5 mM DTT) and buffer B (20 mM HEPES-KOH, pH 7.5, 1 M KCl, 5% (v/v) glycerol and 0.5 mM DTT). With the exception of PPIL4, all proteins used in functional assays comprised the N-terminal Twin-StrepII tag. The Twin-StrepII tag of PPIL4 was removed by on-bead cleavage with the HRV-3C protease, and the untagged protein was purified by SEC on Superdex 200 Increase 10/300 GL (GE Healthcare/Cytiva). Purified proteins were concentrated by ultrafiltration to their stock concentration (PRP2(137–1022), around 19–22 µM; GPKOW, around 165 µM; untagged PPIL4, around 47 µM; tagged PPIL4, about 67 µM), frozen in liquid nitrogen and stored at −80 °C or used directly in assays.

Reconstitution of PRP2–GPKOW and PRP2–PPIL4 complexes

Using a minimal in vitro system, we first tested direct interactions and stable formation of complexes between PRP2(137–1022) and its cofactors GPKOW and PPIL4 as subjected to SEC. All SEC analyses were carried out in SEC buffer (20 mM HEPES-KOH, pH 7.5, 100 mM KCl, 1.5 mM MgCl₂ and 5% (v/v) glycerol). The PRP2 complexes were reconstituted in vitro by mixing around 30 μg recombinant helicase with a 5-fold excess of cofactor (PPIL4 or GPKOW) in a 100 μl reaction volume. The samples were then incubated for 1 h on ice and applied to a Superdex 200 Increase 10/300 GL column (GE Healthcare/Cytiva) run at 0.5 ml min^–1 using an Äkta Go system (Cytiva). SDS–PAGE analyses of the SEC fractions showed that in both cases, the peak profile of PRP2(137–1022) shifted to early fractions in the presence of a cofactor (fraction 9 in the presence of PPIL4 and fraction 7 in the presence of GPKOW) compared with the protein alone (fraction 11), which indicated the formation of a helicase–cofactor complex. As independent means of probing the direct interaction of PRP2(137–1022) with PPIL4, we co-expressed the two splicing factors in Sf9 insect cells and affinity purified their complex from baculovirus-infected cultures. The cultured cells were lysed in lysis buffer (50 mM HEPES-KOH, pH 7.5, 150 mM KCl, 5% (v/v) glycerol, 2 mM DTT, 2 mM MgCl₂, 0.1% (v/v) Triton X-100 and protease inhibitors (complete EDTA-free, Roche)) and the expressed factors were captured on Strep-Tactin XT 4FLOW affinity beads (IBA Lifesciences). The protein samples were eluted by competition with biotin in elution buffer (50 mM HEPES-KOH, pH 7.5, 150 mM KCl, 5% (v/v) glycerol, 2 mM DTT, 2 mM MgCl₂ and 60 mM biotin), concentrated by ultrafiltration to about 500 μl and subjected to SEC in SEC buffer. As for the in vitro assembled PRP2(137–1022)–PPIL4 complex, the complex prepared from insect cells peaked in fraction 9 when eluting from the Superdex 200 column. Overall, this result shows that the limited interaction interface between the NTD of PRP2 and PPIL4, observed in the B^AQR cryo-EM structure, promotes the stable recruitment of the helicase cofactor in a RNA-independent manner. The PRP2(137–1022)–PPIL4 complex used in helicase activity assays was prepared by insect cell co-expression, purified in two steps, as described above, and concentrated to about 14.5 μM. The PRP2(137–1022)–GPKOW complex used in biochemical assays was assembled in vitro, purified by SEC from GPKOW excess and concentrated to about 40.3 μM. In vitro reconstitution of the complex between PRP2(137–1022) and PPIL4 was performed two times using two independent protein preparations. The PRP2(137–1022)–GPKOW complex was assembled in vitro at least three times.

Preparation of the nuclear extracts active in splicing

HeLa S3 cells, tested for mycoplasma, were obtained from GBF (Helmholtz Centre for Infection Research). The nuclear extract active in splicing was prepared according to the standard protocol from Dignam and used as described¹³. Cells were grown in a 30 litre fermenter (Applikon Biotek) to a density of 6.5 × 10⁶ cells per ml in DMEM/F12 (1:1) medium supplemented with 5% (v/v) newborn calf serum. After collection by centrifugation for 10 min at 2,000 r.p.m. in an 8 × 2,000 ml BIOS rotor (Thermo Scientific), the cells were washed twice with cold 1× PBS buffer. The cell pellet was re-suspended in MC buffer (10 mM HEPES-KOH, pH 7.6, 10 mM potassium acetate, 0.5 mM magnesium acetate, 0.5 mM DTT and 2 tablets of protease inhibitors (complete EDTA-free, Roche) per 50 ml of the buffer). After 5 min of incubation on ice, the cells were lysed with 18 strokes of a Dounce homogenizer at 4 °C. The nuclei were pelleted for 5 min at 10,000 r.p.m. in a F14-14×50cy rotor (Thermo Scientific) and were further lysed in Roeder C buffer (20 mM HEPES-KOH, pH 7.9, 0.2 mM EDTA, pH 8.0, 25% (v/v) glycerol, 420 mM NaCl, 1 mM MgCl₂, 0.5 mM DTT and 0.5 mM PMSF) by 20 strokes of a Dounce homogenizer at 4 °C. The mixture was stirred slowly for 40 min at 4 °C and centrifuged at 12,300 r.p.m. in a F14-14×50cy rotor. The supernatant corresponding to the active nuclear extract was aliquoted and flash-frozen in liquid nitrogen and stored at −80 °C.

In vitro pre-mRNA splicing

A typical splicing reaction contained the following components: 20% (v/v) non-dialysed HeLa nuclear extract, 3 mM MgCl₂, 2 mM ATP, 20 mM creatine phosphate and 10 nM m⁷G-capped MINX-3×MS2 pre-mRNA, fluorescently body-labelled with cyanine 5-uridine-5′-triphosphate (Cy5-UTP, Enzo). To monitor splicing of the Cy5-labelled pre-mRNA substrate, the splicing reactions were incubated for 15, 30, 60, 90 and 120 min at 30 °C (Extended Data Fig. 1c). To assess the effect of recombinant IBC(K829A) on the assembly of the spliceosome, the splicing reaction was supplemented with IBC(K829A) to a final concentration of 0.45 μM and pre-incubated for 20 min at 30 °C before the addition of the pre-mRNA substrate and ATP (Extended Data Fig. 1c). In all cases, RNA was recovered by phenol–chloroform–isoamyl alcohol extraction followed by ethanol precipitation. The recovered RNA samples were analysed on denaturing urea polyacrylamide gels (14%, 0.5× TBE). The fluorescently labelled pre-mRNA substrate, splicing intermediates and the final products were detected by in-gel fluorescence using a Typhoon FLA 9500 imaging system (GE Healthcare).

Affinity purification of the human B^AQR spliceosome

Spliceosomal complexes stalled with recombinant IBC(K829A) were assembled in vitro on non-labelled m⁷G-capped MINX-3×MS2 pre-mRNA, synthesized by T7 RNA polymerase run-off transcription. Before initiating the reaction, pre-mRNA substrate (used at 10 nM, final concentration) was incubated with a 20-fold molar access of MBP-MS2 fusion protein for 30 min at 4 °C. Before the addition of ATP and pre-mRNA-MBP-MS2, the large-scale splicing reactions were supplemented with 0.45 μM recombinant IBC(K829A) complex and incubated for 20 min at 30 °C. Therefore, a typical preparative splicing reaction, used from cryo-EM sample preparation, comprised (final concentrations) the following: 20 mM HEPES-KOH, pH 7.9, 2 mM ATP, 20 mM creatine phosphate, 3.2 mM MgCl_2, 84 mM NaCl, 5% (v/v) glycerol, 20% (v/v) non-dialysed HeLa nuclear extract, 10 nM MINX-3×MS2, 200 nM MBP-MS2 and 40 µM IBC(K829A).

To enable the spliceosome assembly, the splicing mixture was slowly stirred for 120 min at 30 °C in a water bath. The non-incorporated pre-mRNA was cleaved by DNA-directed endogenous RNase H digestion for 20 min at 30 °C using a 30-fold molar excess of the cmd42 (5′-TCTTACCGTTCG-3′) and cmd43 (5′-CGGGTTTCCGAT-3′) antisense DNA oligonucleotides. To prevent precipitation of assembled spliceosomes, NaCl was slowly added to the splicing reaction to a final concentration of 120 mM. The aggregates were removed by centrifugation for 10 min at 12,300 r.p.m. in a F14-14×50 rotor. The supernatant was then applied onto a pre-packed 5 ml MBPTrap HP column (GE Healthcare/Cytiva) at 1 ml min^–1, washed with 20× column volumes of G120 buffer (20 mM HEPES-KOH, pH 7.9, 1.5 mM MgCl₂ and 120 mM KCl) and the spliceosomal complexes were eluted with G120 buffer supplemented with 3 mM maltose. The peak fractions were analysed by denaturing 4–12% NuPAGE gels (Life Technologies). The gels were stained with SYBR Gold (Invitrogen) and Coomassie to detect RNA and protein species, respectively. The elution fractions containing IBC(K829A) spliceosomes (B^AQR) were loaded onto a linear 10–30% (w/v) glycerol gradient prepared in G120 buffer and centrifuged at 20,500 r.p.m. for 16 h at 4 °C in a TST 41.14 rotor (Kontron). The gradients were divided into 23 fractions (500 μl each) and manually collected from the top. Peak fractions (14, 15 and 16) from 3 different gradients were pooled and crosslinked with 0.1% (v/v) glutaraldehyde (Electron Microscopy Sciences) for 1 h on ice. The unreacted glutaraldehyde was quenched with 100 mM aspartate (final concentration). The crosslinked B^AQR was then incubated for 2 h on ice and concentrated to the sample absorbance at 280 nm of 0.6. The buffer was exchanged by ultrafiltration to the sample buffer (20 mM HEPES-KOH, pH 7.9, 120 mM NaCl, 1.5 mM magnesium acetate and 2% (w/v) glycerol) using an Amicon 50 kDa MWCO (Millipore) spin concentrator and used directly for cryo-EM grid preparation.

Preparation of cryo-EM grids

Volumes of 4 μl of concentrated sample were applied to one side of glow-discharged UltrAuFoil 200 R2/2 grids (Quantifoil) in a Vitrobot Mark IV (FEI) operated at 4 °C and 100% humidity. The grids were blotted for 2 s with a blotting force of 5 and immediately frozen by plunging into liquid ethane cooled by liquid nitrogen.

Cryo-EM data acquisition and image analysis

The electron micrographs for all datasets were acquired on an FEI Titan Krios G2 transmission electron microscope operated at 300 keV in EFTEM mode, equipped with a Quantum LS 967 energy filter (Gatan), zero loss mode, 30-eV slit width and a K2 Summit direct electron detector (Gatan) in counting mode. Automated data acquisition for dataset 1 (untilted) and dataset 2 (tilted, 25°) was performed using EPU software (Thermo Fisher) at a nominal magnification of ×130,000 (1.05 Å per pixel (Å px^–1)). Micrographs for these two datasets were collected as 40-frame movies at a dose rate of around 5 e^– Å^–2 s⁻¹ over 9 s, which resulted in a total dose of about 45 e^– Å^–2.

The two cryo-EM datasets of B^AQR, collected at 0° and 25° tilt angles, were separately preprocessed on the fly with Warp (motion correction, CTF estimation and dose weighting), and particles were picked using a retrained convolutional neural network⁵¹. Each set of Warp-extracted particles was then subjected to three parallel 2D classification runs in cryoSPARC (v.0.65) using 50 classes and applying a class uncertainty factor of 1.5. The good classes were selected from the three runs, and all particles in these good classes were merged (while removing duplicates).

Initial attempts at processing the cryo-EM images in cryoSPARC revealed that this spliceosome complex exhibited a more structurally rigid core, which continued with more dynamic and poorly resolved peripheral regions (Extended Data Fig. 2). Hence, we systematically tested different EM data processing routines with the aim of improving the more peripheral densities of the spliceosome and to computationally resolve its structural heterogeneity. In brief, the Warp-picked, combined B^AQR particle images (734,691 particles) were re-extracted in Relion 3.1 using a box size of 640/640 px (672 Å/672 Å) and then 2× binned before being subjected to 2D classification with the ‘Ignore CTFs until the first peak’ option switched on (Extended Data Fig. 2). The resulting ‘good’ particles subset (160,650 particles) was then refined in 3D using a 60 Å low-passed human B^act map as a reference volume (Electron Microscopy Databank identifier EMD-4236)¹³ and a spherical mask. The obtained consensus map of the complex (map M1) was then used as a reference in a global 3D classification round with 10 classes starting from the initial pool of B^AQR particles (Extended Data Fig. 2). The cleaned subset of B^AQR particles was then subjected to another round of global 3D classification with 8 classes and a 20 Å resolution limit. This second-round of 3D classification enabled us to resolve the two major compositional states of the complex (Extended Data Fig. 2), termed state A (29.2% particles) and state B (3.4% particles). In the state B complex, stronger density was observed for the PRP19 helical bundle and the step two splicing factor PRP17. Conversely, these map regions appeared to be poorly resolved in the state A complex, which is probably due to their flexibility and low occupancy, as observed for a previously described B^act complex¹³. In both cases, however, the peripheral building blocks of the complex (the IBC, U2 3′ core and the BRR2 helicase) were less resolved than the central part of the complex. To further improve the density of the state A map, we performed an additional global 3D classification without image alignment with 8 classes and a 20 Å resolution limit and then re-extracted and re-centred the B^AQR particle images at their original sampling rate (1.05 Å px^–1) in a 520 px box. The 3D refinement of this subset of particles (179,552 particles) with a soft mask was then followed by CTF refinement (per particle defocus, per micrograph astigmatism) and an additional round of classification in 2D. The subsequent refinement in 3D of this final subset of particles (146,157 particles) with loose or tight soft masks around B^AQR resulted in map 2 and map 3 of the complex that reached global, gold-standard Fourier shell correlation resolutions of about 3.1 Å and 2.9 Å, respectively (Extended Data Figs. 2 and 3). Using these well-resolved consensus maps, we modelled the translocating PRP2 helicase, including its extended NTD domain, as well as the remodelled SF3B complex, and two out of the four B^AQR PPIases (that is, PPIL2 and PPIL4).

Starting from the particle sets, which led to the higher resolution core maps, we also carried out local 3D classifications (Extended Data Fig. 4). Thus, by applying a mask covering the overall B^AQR volume, we subjected these particles to 3D classification without image alignment with 6 classes. Particles exhibiting more pronounced densities for the U2 3′ core module and the IBC module (27,719 particles) were subsequently re-extracted and refined in 3D with soft and spherical masks. These alternative B^AQR maps enabled the rigid body placement of the U2 3′ core and IBC modules at the periphery of B^AQR. Local classifications were also performed on the particle set assigned to state B of the complex. In this case, a soft mask was applied to the region of the spliceosome where the BRR2 helicase resides, which was poorly resolved in the consensus map. The state B particle images were then subjected to 3D classification without image alignment and 4 classes. A minority subset of the overall particle set (12,395 particles) appeared to contain a more ordered BRR2, with the helicase being generally destabilized in B^AQR cryo-EM images, which is probably because of the substantial change to the conformation of SF3B1 induced by the translocation of PRP2. These particles were then re-extracted and refined in 3D to obtain the complete map of the complex (map M4), which now enclosed the U2 3′ core, the IBC module, the PRP19 helical bundle, the BRR2 helicase and the splicing factor PRP17.

Density assignment, model building and refinement

To enable model building, the B^AQR core maps (Supplementary Fig. 2) were sharpened using DeepEMhancer⁵² or locally scaled with LocScale⁵³. The quality of the cryo-EM density map at the core of B^AQR enabled careful model building and side-chain assignment, whereas modelling of the more solvent-exposed map regions was restricted to backbone tracing and rigid-body docking of known structures and computational models. Initial interpretation of the B^AQR maps was facilitated by the available models of human B^act complexes obtained in several states (PDB identifiers 6FF4, 6FF7 and 5Z57). Thus, the model building of the B^AQR complex was initiated by the docking of the human B^act models into the consensus maps (maps 2, 3 and 4) of the complex, followed by manual, residue-by-residue model adjustment in Coot⁵⁴ and refinement with phenix.real_space_refine⁵⁵. Although filtered, locally scaled maps were used for model interpretation, the real space refinement of the B^AQR model was carried out exclusively against the original, unsharpened maps. The higher local resolution of our maps enabled us to improve and correct some of the available models for the B^act core region. After initial placement and refinement, several new (or reconfigured) B^AQR density regions were observed compared with previous B^act complexes. The globular density element identified in the U2 snRNP region of the complex and surrounded by the HEAT domain of SF3B1 was assigned to the helicase domain of PRP2 (residues 388–1017). We modelled the helicase in an open state, consistent with it being trapped in a post-translocation state, and identified seven intron nucleotides accommodated by its RNA-binding tunnel. Compared with a previously published yeast B^act structure (PDB identifier 7DCO), the helicase domain of PRP2 (PRP2^core) was no longer positioned on the convex side of SF3B1^HEAT, but translocated along the intron towards the branch helix. The PRP2-bound RNA formed a continuous density stretch with the intron strand of the U2–BS duplex, showing that the helicase had translocated in a 3′-to-5′ direction from its position in the B^act complex. Supporting this model, density for the RES complex subunits RBMX2 and SNIP1, which bind intron regions (or are located close) downstream of the branch helix, could not be observed in B^AQR, whereas only a short helical region belonging to the BUD13 subunit (residues 530–557) was identified in the proximity of the MA3 domain of CWC22. Besides modelling the helicase domain of PRP2, we also built de novo a large part of the PRP2^NTD that was missing in previous B^act structures. The N-terminal most α-helical region (the pin, residues 161–193) was positioned at the periphery of the helicase domain, where it established interfaces with the RecA1, WH and HB domains of the helicase; two other long helices (the clip, residues 223–256 and residues 264–296) were engaged in tight interactions with the PPIL4 linker region and the long helix of SKIP (residues 286–340), respectively. The latter PRP2^NTD helix extended further towards PRP8 and finally reached CWC22.

The new density of PRP2 in B^AQR coincides with a significant change to the curvature of SF3B1^HEAT, with its N-terminal HEAT repeats no longer in contact with the BS–U2 helix. However, the density of BS-A was still observed in the SF3B1–PHF5A binding pocket, with the hinge region of SF3B organized as in the closed state. Consistently, the SF3A2 and SF3A3 matrin-type zinc-finger domains were organized and positioned as in the B^act complex and still engaged the 5′-end of the U2 snRNA. We modelled SF3B1 by individually docking its consecutive HEAT repeats and manually adjusting their fit to the map. The resulting structure of the reconfigured SF3B1 differed from all other known conformations of SF3B1^HEAT. Because of the PRP2-induced reconfiguration of SF3B1^HEAT, its SF3B6 and SRRM1 binding partners were destabilized from their earlier locations on the N-terminal side of the HEAT superhelix and, compared with B^act, their densities were no longer observed in B^AQR.

Facing the PRP2 density element and following the RNA density exiting the helicase cassette, a new small globular domain appeared to be recruited to B^AQR. We assigned it to the C-terminal RRM region of PPIL4 on the basis of its continuity, through an α-helical density, to the predicted cyclophilin-type PPIase domain that interacts with the long helix of SKIP(286–340) and proteomics analysis. A similar SKIP-bound PPIase density has previously been observed in the later C complex (PDB identifier 5yzg), at an almost identical position; in the published C complex model (PDB 5YZG), the authors assigned it to the PPIase domain of the PPIase PPIG³⁰. The latter lacks a RRM domain and is more abundant in C complexes, but only in trace amounts in B^AQR. The other PPIase identified in the B^AQR proteome, which comprises a PPIase and an RRM domain, is PPIE, an IBC module component located at the periphery of the spliceosome.

In addition to PPIL4 and PPIE, we modelled two other PPIases: PPIL1, interacting with and, probably, stabilizing the PRP19 helical bundle onto the B^AQR core and PPIL2. The latter PPIase adopted an extended conformation with its N-terminal tandem U-box motifs, interacting with SNU114 and separated from its C-terminal PPIase domain by about 90 Å. An ordered linker region (residues 234–266) connected the two PPIL2 moieties, as previously observed in a B^act cryo-EM structure⁵. We did not observe density for the PPIase domain of CWC27, which is consistent with it being destabilized by the propagated changes induced by the remodelling of SF3B1. However, its interacting partner CWC24 was still present in B^AQR, with densities observed for both its zinc finger motif (residues 190–238), sequestering the 5′SS and its C-terminal moiety (residues 262–309) bound to the BPB domain of SF3B3. The final model of the complex (Supplementary Data 2), including its more dynamic peripheral modules, consisted of four RNA molecules (the U2, U5 and U6 snRNA, and the MINX pre-mRNA substrate) and 45 individual polypeptide chains, among which 3 are splicing helicases (PRP2, BRR2 and Aquarius) and 4 are PPIases (PPIL1, PPIL2, PPIL4 and PPIE).

Morphing and generation of movies

The PRP2 helicase translocation trajectory was generated using ChimeraX (v.1.3), with the morph functionality starting from the B^act state of the SF3B module (that is, SF3B1 in a closed conformation with PRP2 positioned on the convex side of SF3B1^HEAT). The NTD domains of SF3B1 (SF3B1^NTD) and PRP2 (PRP2^NTD) were not considered in the morphing analysis, and the U2 sequence was limited to the U2 stem-loop IIa structure and the BS-interaction region. The initial (B^act) and final (B^AQR) trajectory snapshots were aligned using the PHF5A subunit of SF3B as a reference before morphing. The RNA-bound state of PRP2 in B^act was modelled based on the yeast B^act structure⁵. Supplementary Videos 1–4 were generated using ChimeraX (v.1.3), and Supplementary Video 5 was generated using PyMol.

Helicase assays

The helicase assays were performed as previously described for other splicing helicases^{6,56,57,58,59}. To assess the ability of human PRP2 to unwind the U2–BS helix in the presence of its different cofactors, we performed in vitro helicase assays on fluorescently labelled model substrates comprising a perfect double-stranded RNA duplex followed by a single-stranded 3′ overhang. The unwinding activity of PRP2 activity was monitored using either a gel-based readout⁶, in which the single-stranded Cy5-labelled product is separated from the helicase substrate on a native polyacrylamide gel, or by recording the time course of decrease in the fluorescence of the substrate⁵⁸ as the duplex is unwound. In this case, the dual-labelled RNA strand, displaced because of the helicase activity, forms an intramolecular hairpin that brings in proximity the terminal Cy5 probe and its spectrally overlapping dark quencher (BHQ-2), thereby leading to fluorophore emission quenching and fluorescence decay.

The helicase substrate for the gel-based assay was prepared by mixing 30 μM Cy5-labelled strand (5′-CACCAGCUCCGUAGGCGC-Cy5-3′) with 45 μM unlabelled RNA oligonucleotide (5′-GCGCCUACGGAGCUGGUGGCGUAGGCGCAAAAAAAAAAAAAAAAAAAA-3′, the complementary region is shown in bold) in 20 mM HEPES-KOH, pH 7.5. The RNA substrate used in the fluorescent-based helicase assay⁵⁸ was prepared by mixing in a similar molar ratio the dual-labelled RNA oligonucleotide (5′-Cy5-GCGCCUACGCCACCAGCUCCGUAGGCGC-BHQ-2-3′) with the unlabelled strand (5′-GCGCCUACGGAGCUGGUGGCGUAGGCGCAAAAAAAAAAAAAAAAAAAA-3′, the complementary region is shown in bold) in RNA annealing buffer (6 mM HEPES-KOH, pH 7.5, 50 mM KCl and 0.2 mM MgCl₂). In both cases, the single-stranded RNA oligonucleotides were annealed by sequential incubation at 95 °C for 2 min, then at 80 °C for 10 min and then slowly cooled down to room temperature and stored on ice. The RNA and DNA oligonucleotides used in the helicase assays were obtained from IDT (Integrated DNA Technologies) or Microsynth.

In a typical gel-based helicase assay, 50–100 nM fluorescently labelled RNA substrate was mixed in 20 μl on ice with increasing concentrations of PRP2 constructs or PRP2 complexes (1–10 μM) in helicase assay buffer (20 mM HEPES-KOH, pH 7.5, 50 mM KCl, 2 mM MgCl₂, 0.1 mg ml^–1 BSA (NEB), 0.08 U μl^–1 RNasin (Promega), 5% (v/v) glycerol and 0.5 μM competitor DNA (5′-GCGCCTACGGAGCTGGTG-3′), final concentrations). The GPKOW cofactor was added in a 5-fold molar excess over PRP2(137–1022) or PRP2(137–1022)–PPIL4. After pre-incubation for 10 min at 37 °C (or 20 min at 30 °C), the unwinding reaction was initiated by the addition of 2 mM ATP, and the samples were incubated for 1 h at 37 °C (or 1 h at 30 °C). The reactions were stopped by the addition of 5 μl quenching buffer (0.83 mM Tris-HCl, pH 7.6, 5 mM EDTA, 5% (v/v) glycerol, 0.0025% bromophenol blue, 0.0025% xylene cyanol FF and 0.04 U μl^–1 proteinase K (NEB), final concentrations) and incubated for 20 min at 37 °C. RNA duplex unwinding was assessed on a 14% polyacrylamide native gel prepared in 1× TBE or 1× Tris-glycine buffer and was run in 0.5× TBE (or 1× Tris-glycine) at 100 V for approximately 90 min at room temperature. The yeast Prp22p helicase, used as a positive control, was prepared as previously described²⁷. The RNA gels were scanned at the Cy5 excitation peak using an iBright 1500 imaging system (Invitrogen). The gel-based helicase assays were repeated at least two times.

The molecular beacon helicase assays were performed using a Fluorolog spectrofluorometer (Horiba). In brief, 200 nM dual-labelled substrate was mixed with 1–5 μM purified helicase in reaction buffer (20 mM HEPES-KOH, pH 7.5, 50–150 mM KCl and 2 mM MgCl₂) in the absence of ATP and preincubated for at least 10 min at room temperature. As for the gel-based unwinding assays, GPKOW was added in a 5-fold molar excess over PRP2(137–1022), whereas PPIL4 was added in a 3-fold molar excess. The reactions were subsequently transferred to a 1.5 mm ultra-microcuvette (105.252-QS, Hellma) and equilibrated at the measurement temperature (25 °C or 30 °C) in the spectrometer. The unwinding reaction was initiated by the addition of 2 mM ATP (final concentration), and the decay in Cy5 fluorescence was immediately recorded at an interval time of 1 s and an integration time of 0.1 s. The Cy5 fluorescent probe was excited at 645 nm (3 nm slit width) and its emission was measured at 667 nm (3 nm slit width) and corrected for detector noise. Data were analysed and plotted using OriginPro 2020 (9.7.0.188). The fluorescence-based helicase assay was repeated at least three times. Three independent preparations of PRP2(137–1022) were tested in these assays.

Electrophoretic mobility shift assay

To investigate the RNA binding activity of the different PRP2 complexes tested in the unwinding assays, 50 nM fluorescent substrate was mixed with increasing concentrations of helicase or helicase cofactor samples (0.1–8 μM). The 20 μl binding reactions were prepared in RNA binding buffer (20 mM HEPES-KOH, pH 7.5, 50 mM KCl, 2 mM MgCl₂, 0.1 mg ml^–1 BSA (NEB) and 5% (v/v) glycerol, final concentrations) and incubated for 30 min at room temperature before being loaded on a 5% polyacrylamide native gel prepared in 1× Tris-glycine buffer. The native RNA gel was pre-run at 40 V for 30 min at room temperature and run then at 60 V in 1× Tris-glycine. The gels were scanned at the Cy5 excitation peak. The RNA binding assays were repeated at least two times, and similar results were obtained.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.