Molecular basis for the activation of human spliceosome

Abstract

The spliceosome executes pre-mRNA splicing through four sequential stages: assembly, activation, catalysis, and disassembly. Activation of the spliceosome, namely remodeling of the pre-catalytic spliceosome (B complex) into the activated spliceosome (B^act complex) and the catalytically activated spliceosome (B^* complex), involves major flux of protein components and structural rearrangements. Relying on a splicing inhibitor, we have captured six intermediate states between the B and B^* complexes: pre-B^act, B^act-I, B^act-II, B^act-III, B^act-IV, and post-B^act. Their cryo-EM structures, together with an improved structure of the catalytic step I spliceosome (C complex), reveal how the catalytic center matures around the internal stem loop of U6 snRNA, how the branch site approaches 5′-splice site, how the RNA helicase PRP2 rearranges to bind pre-mRNA, and how U2 snRNP undergoes remarkable movement to facilitate activation. We identify a previously unrecognized key role of PRP2 in spliceosome activation. Our study recapitulates a molecular choreography of the human spliceosome during its catalytic activation.

Here, the authors have captured the intermediate states of the human spliceosome during its activation through cryo-EM reconstruction of six distinct spliceosomal complexes.

Introduction

Pre-mRNA splicing is an essential process in eukaryotic gene expression, involving the excision of introns and ligation of exons^1,2. RNA splicing consists of two reactions: branching and exon ligation³. The spliceosome, as one of the most dynamic molecular machines in eukaryotic cells, executes splicing in four sequential stages: assembly, activation, catalysis, and disassembly^4–6. Each splicing cycle requires ten major functional states of the spliceosome: early spliceosome (E), pre-spliceosome (A), precursor to B complex (pre-B), B complex, B^act complex, B^* complex, C complex, step II catalytically activated spliceosome (C^*), post-catalytic spliceosome (P), and intron lariat spliceosome (ILS). The transition between two neighboring functional states requires major ribonucleoprotein (RNP) remodeling and is driven by a family of eight conserved RNA helicases⁷. During spliceosome assembly, the pre-B complex is converted to the B complex by the helicase PRP28^8–10.

The splicing active site is yet to be formed in the B complex. U6 small nuclear RNA (snRNA) is sequestered by U4 snRNA and cannot form the internal stem loop (ISL). Activation of the spliceosome requires remodeling of the B complex to the B^act complex, which is driven by the helicase BRR2^11–13. During this process, U4 snRNP and the B-specific proteins are released; the PRP19/CDC5L complex (also known as the NineTeen Complex, or NTC) and NTC-related complex (NTR) are recruited. The B^act complex is then remodeled by the helicase PRP2 into the B^* complex^14–16, in which the branching reaction occurs, resulting in the C complex. PRP2 remodels the B^act complex by pulling the pre-mRNA^14,15,17, which delivers the branch site (BS) into close proximity of the 5′-splice site (5′SS).

Mechanistic understanding of spliceosome activation requires detailed structural information on the activated spliceosome and related intermediate states. The B–B^act–B^* transition involves extensive shuffling of spliceosomal components and constitutes the last and arguably the most important step prior to the splicing reaction. Cryo-EM structures have been obtained for the B^18–21 and B^act22–25 complexes. The human B^act complex is thought to be highly dynamic, as evidenced by distinct conformations^22,23,26. The human B^* complex is yet to be structurally resolved. The B–B^act–B^* remodeling process, which likely involves multiple intermediate states and additional roles of the RNA helicases^27–29, remains poorly understood.

In this study, we have captured the intermediate states of the human spliceosome during its activation through cryo-EM reconstruction of six distinct spliceosomal complexes. These complexes, named pre-B^act, B^act-I, B^act-II, B^act-III, B^act-IV, and post-B^act, bridge the critical gap between the B and B^* complexes. We also re-determined the cryo-EM structure of the human C complex. Analysis of these structures reveals an ordered flux of spliceosomal components and conformational rearrangements that serve the purpose of organizing the catalytic center. We identify an unanticipated role of PRP2 in the activation of human spliceosome.

Results

Spliceosome isolation and cryo-EM analysis

To isolate the human spliceosome, we used the splicing inhibitor N-palmitoyl-l-leucine (NPL), which is known to selectively inhibit the late-stage assembly of the human spliceosome³⁰ (Supplementary Fig. 1a). The human spliceosome was assembled on a synthetic pre-mRNA supplemented with NPL in the in vitro splicing reaction, followed by affinity purification and glycerol gradient centrifugation with chemical crosslinking (Supplementary Fig. 1b). The resulting sample was examined using urea–PAGE (Supplementary Fig. 1c) and subjected to cryo-EM analysis (Supplementary Fig. 1d). A total of ~4.6 million particles, generated from 17,070 micrographs, were applied to three rounds of three-dimensional (3D) classification. We identified seven classes of the spliceosome, which yielded seven distinct reconstructions (Supplementary Figs. 2–4).

The final cryo-EM reconstructions of the human pre-B^act, B^act-I, B^act-II, B^act-III, B^act-IV, post-B^act, and C complexes display average resolutions of 3.4, 3.0, 4.2, 3.0, 3.3, 3.0, and 3.4 Å, respectively (Supplementary Fig. 3 and Supplementary Table 1). The local resolutions, which mostly reach 2.8–3.2 Å in the core regions (Supplementary Fig. 5), allow the identification of protein components and assignment of RNA elements (Supplementary Figs. 6–10). Each of the six distinct human B^act-like complexes contains a distinct set of protein components (Supplementary Table 2). In contrast, the newly determined human C complex is nearly identical to that reported previously^31,32, corroborating the conclusion of our current investigation.

Overall structures of the intermediate-activated spliceosome

Activation of the spliceosome involves a major flux of components, including U4 snRNP, NTC, and NTR (Fig. 1a). The B^act complex already contains a splicing active site, which maintains the same configuration until spliceosome disassembly; but the BS is separated from 5′SS by a distance of ~50 Å^22,23. This distance should be reduced presumably to ~4 Å in the B^* complex to allow the branching reaction³³. The six ordered human B^act complexes share a similar overall organization but exhibit pronounced differences in composition (Fig. 1b and Supplementary Table 2). At least 21 protein components are either recruited or released during the interchange of the six B^act complexes. For clarity, only key components are discussed here. The pre-B^act complex lacks PRP2 and the NTC core (PRP19 complex and SPF27) but still contains the retention and splicing (RES) complex, which is required for efficient B-to-B^act conversion^34,35. In B^act-I, PRP2 is loaded together with the exchange of a number of splicing factors. In B^act-II, the RES complex is released, leaving a spatial vacancy to accommodate PRP2 translocation. In the latter three states, the NTC core is loaded and U2 snRNP undergoes marked translocations (Fig. 1b). In post-B^act, the BS is positioned close to 5′SS, which primes the branching reaction.

Fig. 1. Cryo-EM structures of six distinct conformational states of the human B^act complex.

a Schematic representation of the activation process of the human spliceosome. The B complex, containing U5.U4/U6 tri-snRNP, is remodeled by the RNA helicase BRR2, resulting in the release of U4 snRNP and recruitment of the NTC and NTR complexes. In the B^act-to-B^* transition catalyzed by PRP2, the BS is brought into the close proximity of 5′SS for the branching reaction. b Cryo-EM structures of six distinct intermediate states between the B and B^* complexes. These complexes, named pre-B^act, B^act-I, B^act-II, B^act-III, B^act-IV, and post-B^act, are resolved at average resolutions of 3.4, 3.0, 4.2, 3.0, 3.3, and 3.0 Å, respectively. All complexes, with components displayed in color-coded surface representation, are shown in the same orientation. c Cryo-EM structure of the human C complex at an average resolution of 3.4 Å. All structural images in this manuscript were prepared using ChimeraX⁷¹ and PyMol⁷².

In addition to compositional differences, the six ordered B^act complexes also display marked conformational variation for a number of key protein components. For example, due to drastic translocation, BRR2 is unambiguously identified in pre-B^act through B^act-III, but not in B^act-IV or post-B^act (Fig. 1b and Supplementary Table 2). This is likely due to its mobile nature during spliceosome activation. Notably, BRR2 becomes ordered again in the reconstruction of the C complex (Fig. 1c), which is almost identical to that reported previously^31,32. Based on these structures and additional evidence described below, we conclude that the six distinct conformations of the B^act complex represent intermediate states between the B and B^* complexes.

Conformational changes of the RNA elements

During spliceosome activation, 5′SS and the BS are gradually drawn together, with the formation of U6 ISL as the splicing active site^3,36. All six activated complexes contain the same set of RNA components: U2, U5, and U6 snRNAs, and pre-mRNA; but these RNAs display notable differences in conformation. In all six complexes, 5′-exon is anchored on loop I of U5 snRNA, and 5′SS forms a duplex with U6 snRNA, whereas the BS is recognized by U2 snRNA (Fig. 2a–e and Supplementary Fig. 6).

Fig. 2. Conformational changes of the RNA elements.

a Overall structure of the RNA elements in the pre-B^act complex. The color-coding scheme is preserved throughout this manuscript. In pre-B^act, U6 ISL and helices Ib and II of U2/U6 duplex are initially formed; but helix Ia is yet to form. The BS is ~60 Å away from 5′SS. The 5′-exon remains anchored to loop I of U5 snRNA. b Structure comparison of the RNA elements between the pre-B^act and B^act-I complexes. The U2/BS duplex and helix II remain unchanged. Helix Ia appears in B^act-I. The 5′SS moves toward the BS by ~10 Å, with notable movements for U5 and U6 snRNAs. All structure alignments reported in this manuscript are based on PRP8, unless stated otherwise. c Close-up views on the maturation of U6 ISL. In the pre-B^act to B^act-I transition, helix Ia is formed, and the key nucleotide U74 of U6 snRNA is bulged out. d Pairwise comparison of the RNA elements among the B^act-I, -II, -III, and -IV complexes. The distance between 5′SS and the BS remains unchanged at about 50 Å. In the transition of B^act-I to B^act-II, U2 snRNA undergoes a marked translocation and helix II is shifted slightly. In the transition of B^act-III to B^act-IV, helix II moves to the position where it is found in the C complex. e Structure comparison of the RNA elements between B^act-IV and post-B^act. U2 snRNA undergoes notable movements with the BS moving towards 5′SS by about 20 Å. f Structure comparison of the RNA elements between the post-B^act and C complexes. After the branching reaction, the BS is linked to 5′SS in the C complex, and U2 snRNA undergoes a pronounced translocation.

In pre-B^act, U6 ISL, helices Ib and II of the U2/U6 duplex, but not helix Ia, are initially formed. These structural features reflect the premature nature of pre-B^act as an activated spliceosome. Consistently, the BS is positioned ~60 Å away from 5′SS (Fig. 2a), even farther apart compared to that in the published B^act complex^22,23. During the transition of pre-B^act to B^act-I, the U2/BS duplex and helix II remain unchanged; but 5′SS moves toward the BS by about 10 Å, accompanied by notable shifts of U5 and U6 snRNAs (Fig. 2b). Importantly, helix Ia is formed in B^act-I, and the nucleobase of the key nucleotide U74 of U6 ISL (U80 in Saccharomyces cerevisiae (S. cerevisiae)) is bulged out relative to its flanking nucleotides, which enables the positioning of two catalytic Mg²⁺ ions that directly participate in the splicing reaction^37–40 (Fig. 2c and Supplementary Fig. 6c).

The 50-Å distance between 5′SS and the BS remains unchanged in B^act-I through B^act-IV; in contrast, U2 snRNA undergoes pronounced changes (Fig. 2d). Compared to B^act-I, U2 snRNA in B^act-II undergoes a marked translocation, with a slight shift of helix II. Compared to B^act-II, a portion of U2 snRNA in B^act-III continues to shift position. In the transition of B^act-III to B^act-IV, helix II moves to the position where it is found in the C complex. Each of these changes in U2 snRNA is accompanied by flux of specific protein components. In the transition of B^act-IV to post-B^act, U2 snRNA undergoes a translocation of ~80 Å, accompanied by a 20-Å movement of the BS toward 5′SS (Fig. 2e). This movement shortens the distance between the two reactive moieties on 5′SS and BS for the branching reaction. In the C complex, the BS is covalently linked to 5′SS (Supplementary Fig. 6f); U2 snRNA undertakes yet another marked translocation compared to post-B^act (Fig. 2f).

In summary, analysis of the RNA conformation in these six B^act intermediates reveals remarkable changes in U2 snRNA, which allow stepwise movement of the BS toward 5′SS for the branching reaction. In contrast to U2 snRNA, U5 and U6 snRNAs remain largely unchanged in the five intermediates B^act-I through post-B^act.

A key role by PRP2 during spliceosome activation

Spliceosome activation is triggered by the helicase BRR2^11–13; but compelling evidence also implicates a key role by another RNA helicase PRP2^14,16,41. PRP2 sequentially comprises six domains: an N-terminal extension, RecA1, RecA2, WH, HB, and OB (Supplementary Fig. 11a). PRP2 is absent in pre-B^act and is recruited to the peripheral region of B^act-I where it loosely contacts the N-terminal region of SF3B1 (Fig. 3a). In B^act-I, the splicing factor SRRM1 binds to the N-terminal HEAT repeats of SF3B1. The RES complex, located between SF3B1 and PRP8, associates with the N-terminal flexible helices of SF3B1 and the C-terminal flexible sequences of SKIP (Supplementary Fig. 9b).

Fig. 3. Structure and function of PRP2 in the human B^act complex.

a PRP2 in the B^act-I complex. PRP2 is attached to the peripheral region of the spliceosome and loosely contacts the N-terminal region of SF3B1. N the N-terminus of SF3B1, C the C-terminus of SF3B1. b PRP2 in the B^act-II complex. In the B^act-I to B^act-II transition, PRP2 undergoes a pronounced translocation, moving to the cavity vacated by the RES complex. The PPT sequences of pre-mRNA are loaded into PRP2. c A close-up view on PRP2 in B^act-II. The loop between the N-terminal HEAT repeats of SF3B1 interacts with the RecA2 domain of PRP2, and the BS is anchored on the C-terminal HEAT repeats of SF3B1. The PPT sequences of pre-mRNA are bound to PRP2. d The N-terminal helices of PRP2 directly associate with the C-terminal helices of SKIP, and the linker helix and RRM domain of PPIL4. e A close-up view on the interface between the PPT of pre-mRNA and PRP2. Residues Arg503, Leu709, and Arg998 directly contact the PPT sequences. f Depletion of endogenous PRP2 from the nuclear extract. Compared to the untreated sample, the amount of PRP2 in the nuclear extract is sharply reduced as confirmed by the western blot. g Deletion of the N-terminal extension or mutation of key residues in PRP2 impairs splicing. Shown here are the reverse transcription-PCR results of in vitro splicing reaction. These results have been repeated for three times to confirm the analysis. Source data are provided as a Source Data file.

During the B^act-I to B^act-II transition, PRP2 is relocated through a distance of ~80 Å and a rotation of ~90° to the cavity vacated by dissociation of the RES complex (Fig. 3b and Supplementary Fig. 11b). The newly identified, extended helices from the N-terminal extension of PRP2 interact with PRP8, PPIL4, the C-terminus of SKIP and pre-mRNA (Fig. 3b). Meanwhile, SRRM1 is released, and the splicing factor PPIL4 is recruited. Most notably, the poly-pyrimidine-tract (PPT) sequences of pre-mRNA are loaded into PRP2 with its 3′-end reaching the RRM of PPIL4 (Supplementary Figs. 10 and 11c). The C-terminal flexible sequences of SKIP become ordered helices, contacting PRP2, PPIL4, and the 3′-end of PPT (Fig. 3c and Supplementary Fig. 11d). Furthermore, the N-terminal helices of SF3B1 become disordered in B^act-II, and the loop sequences connecting the N-terminal HEAT repeats interact with the RecA2 domain of PRP2. The BS is accommodated by the C-terminal HEAT repeats of SF3B1 (Fig. 3c).

The configuration of PRP2 in B^act-III and B^act-IV is nearly identical to that in B^act-II (Supplementary Fig. 12a, b). During the transition of B^act-IV to post-B^act, SF3B1 is relocated away from PRP8, helix II and U2 snRNA undergo obvious movements; but PRP2 remains unchanged. In contrast, SF3B1 undergoes dramatic conformational changes, resulting in the release of the BS from its C-terminal HEAT repeats (Supplementary Fig. 12c). U2 snRNA and U2/BS duplex also exhibit notable translocation in the transition.

One of the key structural findings is identification of the N-terminal helices of PRP2 (residues 162–341), which associate with the C-terminal helices of SKIP and the linker helix and RRM of PPIL4 (Fig. 3d). This finding suggests an important role by these PRP2 N-terminal helices and corroborates an earlier observation that the N-terminal extension of Prp2 in S. cerevisiae can be crosslinked to Prp45 (the SKIP homolog in S. cerevisiae)²⁵. Interestingly, however, deletion of the yeast Prp2 N-terminal extension was reported to have little impact on splicing⁴².

In addition, three residues in PRP2 (Arg503, Leu709, and Arg998) already closely interact with the nucleobases of PPT in B^act-II, likely preventing the backward sliding of pre-mRNA toward its 5′-end (Fig. 3e). This structural feature is reminiscent of that in the yeast B^act complex⁴³, underscoring a conserved mechanism from yeast to human¹⁶. Thr674 is located close to the binding groove for pre-mRNA (Fig. 3e); its mutation to methionine, predicted to cause a steric clash with the phosphate backbone of PPT, is thought to cause neuromuscular oculoauditory syndrome⁴⁴.

To corroborate these structural findings, we depleted PRP2 (ΔPRP2) from nuclear extract (NE) using an anti-PRP2 antibody (Fig. 3f). Consistent with previous studies⁴¹, drastic reduction of PRP2 cripples splicing (Fig. 3g, lane 4). Compared to wide-type (WT) PRP2 (Fig. 3g, lane 5), deletion of its N-terminal 361 or 305 residues abrogated splicing (lanes 6 & 7); deletion of 198 residues also reduced splicing (lane 8) (Supplementary Fig. 13). These results identify a crucial role in splicing by the N-terminal extension of human PRP2. This conclusion is supported by our structural observation and potentially explains the low sequence conservation between the N-terminal extensions of human PRP2 and S. cerevisiae Prp2 (Supplementary Fig. 14a). In addition, each of four PRP2 missense variants (R503A, T674M, L709A, and R998A), which target conversed residues from yeast to human (Supplementary Fig. 14b), exhibits a reduced level of splicing (Fig. 3g, lanes 9–12).

Compositional changes during spliceosome activation

Our structures of six intermediate B^act complexes reveal stepwise compositional changes that occur during human spliceosome activation (Fig. 4a and Supplementary Table 2). In pre-B^act, the splicing factor KIN17 places its N-terminal zinc finger (ZnF) and WH domain above the U6/5′SS duplex (Fig. 4b), consistent with its involvement in 5′SS selection⁴⁵. During the transition from pre-B^act to B^act-I, KIN17 is dissociated; the splicing factors CWC22, SRRM2, CWC27, RNF113A, and PRP2 are recruited. CWC22 is crucial for PRP2 function in promoting the release of the U2 components SF3a and SF3b⁴⁶. SRRM2, which promotes the second step conformation and modulates 3′SS selection^47,48, remains bound in the spliceosome until exon ligation and is released in the ILS complex⁴⁹. Notably, CWC27, which was reported to form a heterodimer with CWC22 using its C-terminal region⁵⁰, only transiently appears in B^act-I and is dissociated in B^act-II. RNF113A directly binds to 5′SS²². RNF113A is involved in the fidelity of 5′SS selection and may orchestrate spliceosome organization into an active configuration prior to PRP2-mediated remodeling⁵¹.

Fig. 4. Composition fluxes during activation of human spliceosome.

a The structures of six human B^act complexes, in surface representation with select components color-coded, reveal a series of compositional changes during activation of the human spliceosome. In the transition of pre-B^act to B^act-I, the splicing factor KIN17 is dissociated; CWC22, SRRM2, CWC27, RNF113A, and PRP2 are recruited. During the B^act-I to B^act-II transition, CWC27, SRRM1, SF3B6, and RES complex are released; PPIL1 and PPIL4 are recruited. In the transition of B^act-II to B^act-III, PRP17 and the NTC core are recruited. In the transition of B^act-III to B^act-IV, PPIL2 and the SF3a components SF3A1/SF3A2 are released; SYF2 is recruited. In post-B^act, RNF113A and SF3A3 are released. b A close-up view on the interface between the ZnF and WH domain of KIN17 and U6/5′SS duplex.

During the B^act-I to B^act-II transition, SRRM1, SF3B6, and the RES complex are released (Fig. 4a and Supplementary Table 2). SRRM1, found as a coactivator of pre-mRNA splicing by promoting critical interactions among pre-mRNA-bound splicing factors^52,53, associates with pre-mRNA in B^act-I. SF3B6, also known as SF3B14 (p14), has no S. cerevisiae homolog and binds to the N-terminal sequences of SF3B1. Dissociation of the RES complex, which is required for efficient transformation of the B complex to B^act complex³⁴, allows conformational rearrangement of PRP2. This conformation is stabilized by the freshly recruited PPIL4. PPIL1 is also engaged in B^act-II. Notably, PPIL4 associates with the N-terminal extension of PRP2 and the C-terminal fragments of SKIP; this structural feature persists in all later activated states.

Different from B^act-II, PRP17 and the NTC core are engaged in B^act-III. PRP17, which is required for the second step reaction^54,55, stays in the spliceosome until its disassembly. In the transition of B^act-III to B^act-IV, PPIL2 and the SF3a components SF3A1/SF3A2 are released, and a major portion of RNF113A becomes disordered (Fig. 4 and Supplementary Table 2). SYF2 is recruited in B^act-IV. In addition, RNF113A and SF3a complex are completely absent in the post-B^act complex.

Rearrangements of U2 snRNP

Activation of the human spliceosome involves substantial changes in the position and conformation of U2 snRNP, which consists of U2 snRNA, SF3a complex, SF3b complex, and the U2 snRNP core (U2-A′, U2-B″, and U2 Sm ring) (Fig. 5). U2 snRNP undergoes a stepwise translocation during spliceosome activation, moving to the main body of the spliceosome by a distance of ~150 Å in the B-to-pre-B^act transition (Supplementary Fig. 15a). During the transitions from pre-B^act to B^act-IV, U2 snRNP undergoes only slight conformational changes but marked compositional alterations. The SF3b component SF3B6 and the SF3a components SF3A1/SF3A2 are dissociated in the B^act-I to B^act-II and B^act-III to B^act-IV transitions, respectively. The last SF3a component SF3A3 is dissociated in the B^act-IV to post-B^act transition. Notably, U2 snRNP is translocated by about 50 Å during the transition from B^act-IV to post-B^act, with U2 snRNP core moving closer to the main body of spliceosome (Supplementary Fig. 15b). In the B^*/C complexes, the SF3b complex is fully dissociated, and U2 snRNP core undergoes prominent movements.

Fig. 5. Rearrangements of U2 snRNP during human spliceosome activation.

a, b U2 snRNP, comprising U2 snRNA, SF3a, SF3b and the core, undergoes remarkable positional changes in a stepwise manner during spliceosome activation. In the B-to-pre-B^act transition, U2 snRNP undergoes a large translocation, moving toward the main body of spliceosome. U2 snRNP has a series of minor positional and conformational changes from pre-B^act to B^act-IV. Notably, the SF3b component SF3B6 and the SF3a components SF3A1/SF3A2 are dissociated in the transitions of B^act-I to B^act-II and B^act-III to B^act-IV, respectively. The last component SF3A3 of the SF3a complex is released in the B^act-IV to post-B^act transition. In the B^*/C complexes, the SF3b complex is dissociated; U2 snRNP core undergoes prominent movements.

The local conformational changes of U2 snRNP during spliceosome activation are exemplified by SF3B1, the core component of the SF3b complex, which anchors the bulged BS using its C-terminal HEAT repeats^56,57. The movement of SF3B1 generally coincides with the translocation of U2 snRNP, with the most significant changes occurring in the B to pre-B^act and B^act-IV to post-B^act transitions (Supplementary Fig. 16a). Notably, the BS is freed from the C-terminus of SF3B1 in post-B^act and moves closer to 5′SS, preparing for the branching reaction. In addition, the overall conformation of the superhelical SF3B1 becomes more extended and loosened during the activation process, getting ready for its release from the spliceosome (Supplementary Fig. 16b).

Discussion

Among all assembled spliceosomes, U4 snRNP is only present in the pre-B and B complexes; the splicing active site is only formed in the B^act complex and thereafter. These features allow unambiguous differentiation between the B-like and B^act-like complexes. Along this line, the spatial distance between the two reactive moieties (5′SS and the BS) of the branching reaction defines the B^act versus B^* complexes: about 50 Å in the former and <4 Å in the latter. These considerations allowed us to unequivocally identify six intermediate states of the B^act complex, which bridge the critical gap between the B and B^* complexes (Supplementary Fig. 2).

Ordering of these six intermediate B^act states is relatively straight forward. Pre-B^act represents a very early state, because helix Ia of the U2/U6 duplex is yet to be formed. In contrast, post-B^act represents a very late state, because the distance between 5′SS and the BS in this state is the shortest (~30 Å) among all six complexes and the SF3a complex has been released. The four intermediates between pre-B^act and post-B^act are differentiated on the basis of both prior experimental observations and rational analysis. B^act-I is ahead of B^act-II, because the RES is present in the former, but not the latter state. B^act-I and B^act-II are ahead of B^act-III, because the NTC core is only present in B^act-III. The SF3A components A1 and A2 are present only in B^act-III and prior states, but not in B^act-IV.

Structures of the six human intermediate B^act complexes reveal a cascade of coordinated rearrangements during spliceosome activation. Our results highlight the interplay between protein components and RNA elements in this process. One unanticipated finding is the gradual engagement of the BS to 5′SS (Fig. 2). Analysis of the six states, together with that of the newly resolved C complex, unveils a remarkable choreography of the RNA elements, leading to the branching reaction (Supplementary Movie 1). In the pre-B^act to B^act-I transition, helix Ia is formed, the key nucleotide U74 of U6 snRNA (U80 in S. cerevisiae) is flipped inside out, and two catalytic Mg²⁺ ions are appropriately loaded, resulting in the maturation of U6 ISL. U5 snRNA undergoes a notable translocation, delivering 5′-exon-5′SS to the correct location for nucleophilic attack by the BS. Meanwhile, driven by the remarkable movement of U2 snRNA, the BS is freed from SF3B1 and moves closer to 5′SS by about 20 Å during the transition of B^act-IV to post-B^act.

Actions of the conserved RNA helicases are the ultimate driving force for spliceosomal remodeling and the splicing reaction. However, due to the transient nature of such actions, mechanistic understanding can be achieved through the recapitulation of a series of static pictures. In this study, an unanticipated finding is the remarkable rearrangement of PRP2 in the process of engaging the PPT sequences (Fig. 3). PRP2 associates with B^act-I at the periphery but, following the release of SRRM1 and RES complex, is flipped into the central region of B^act-II. In addition, PPIL4 is engaged, and the C-terminus of SKIP is ordered into helices. In B^act-II, the N-terminal extension of PRP2 plays a previously unrecognized key role in splicing through interactions with PPIL4, SKIP, and PRP8. Similar to yeast Prp2, a few residues of PRP2 may prevent backward sliding of pre-mRNA. These structure-based conjectures are supported by biochemical analyses (Fig. 3).

The transition from B to B^act is thought to be mainly driven by the RNA helicase BRR2, which pulls U4 snRNA to trigger the dissociation of U4 snRNP^11–13. However, once U4 snRNP is dissociated (as is the case in the intermediate B^act states), BRR2 can no longer exert its force to drive remodeling. Nonetheless, the conformation and composition of the spliceosome continue to change from pre-B^act to post-B^act. It is possible that the driving force for these changes is still the consequence of BRR2 action. Alternatively and perhaps more likely, PRP2 may play a key role. This conjecture is supported by the recruitment of PRP2 into B^act-I and pre-mRNA binding by PRP2 in B^act-II, both occurring at relatively early stages. The subsequent positional and conformational rearrangements of other components, as a result of continuous spliceosomal remodeling, are likely attributed to the RNA helicase activity of PRP2. Such an intricate role by PRP2 remains to be experimentally scrutinized.

The advent of the intermediate states of the human-activated spliceosome allows comparison with previously reported B^act complexes^22,26. Our pre-B^act complex is reminiscent of the previously reported pre-B^act-2 complex, which was captured using another splicing inhibitor²⁶. However, there are obvious differences in the overall architecture between our pre-B^act and the pre-B^act-2, especially in U2 and U5 snRNPs (Supplementary Fig. 17a). The B-specific proteins (PRP38, SNU23, MFAP1, and UBL5) together with CTNNBL1 are absent in our pre-B^act, suggesting that our pre-B^act may be downstream of pre-B^act-2 in the activation process. Notably, the previously reported pre-B^act-1 complex is more similar to the B complex²⁶. In addition, B^act-I in this study resembles the previously described early B^act conformation²², but SYF3 and AQR are only present in B^act-I (Supplementary Fig. 17b), suggesting the previous early B^act lies intermediary between the pre-B^act and B^act-I. Through scrutinizing the protein components, the previously described mature B^act22 resembles the B^act-III and late B^act22 appears to bridge between the B^act-IV and post-B^act. As we were preparing this manuscript, an intermediate B^AQR complex, enriched through the use of dominant negative AQR, was reported⁵⁸; B^AQR is structurally similar to our B^act-III complex (Supplementary Fig. 17c). The six distinct conformations reported in this study, together with those previously described^22,26,58, collectively define the intermediate states of the spliceosome between the B and B^* complexes.

We used the splicing inhibitor NPL, which is thought to affect the late assembly of the spliceosome³⁰. The reported effect of the inhibitor is supported by particle distribution in the EM micrographs (Supplementary Fig. 2), which gives rise to six intermediate states of the B^act complex as well as the C complex. The use of the splicing inhibitor NPL likely allows the accumulation of the spliceosomal intermediate states that are otherwise transient during normal splicing. Structural and biochemical analysis strongly suggest that these conformational states may faithfully represent those in the splicing cycle. However, there is no obvious EM density for NPL. As such, the target of NPL remains to be identified. It is possible that this inhibitor may target a protein that regulates splicing through an enzymatic activity but makes no direct interaction with the spliceosome.

Methods

In vitro assembly of human spliceosomal complexes

The human spliceosomal complexes were assembled on the MINX pre-mRNA in vitro. The pre-mRNA with three tandem MS2 binding sites at its 3′-end was in vitro synthesized using the T7 run-off transcription method. The nuclear extracts of HeLa S3 cells were prepared as described⁵⁹. The chemically synthesized small molecule NPL was incubated with the HeLa nuclear extract on ice at a concentration of ~600 μM for 1 h (Supplementary Fig. 1a). Meanwhile, the pre-mRNA was pre-incubated with the MS2-MBP fusion protein. The in vitro splicing reaction was performed at 30 °C for 1 h after mixing 15 nM pre-treated pre-mRNA with 50% (v/v) NPL bound-nuclear extract (diluted by a buffer containing 2 mM ATP, 20 mM creatine phosphate, 3 mM MgCl₂, 20 mM HEPES-KOH pH 7.9, and 50 mM KCl).

Purification of human spliceosomal complexes

The assembled human spliceosomal complexes were incubated with the amylose resin (NEB) at 4 °C for 2 h, and isolated using the MS2-MBP affinity chromatography. The amylose resin was rinsed using the G150 buffer (20 mM HEPES-KOH pH 7.9, 150 mM NaCl, 1.5 mM MgCl₂, 4% glycerol). Eluted by the same buffer supplemented with 20 mM maltose, the spliceosomal complexes were further applied for a linear 10–30% (v/v) glycerol gradient with 0–0.1% EM-grade glutaraldehyde (Supplementary Fig. 1b). After centrifugation for 13 h at 4 °C at 110,000 × g, total RNA was extracted from the fractions and analyzed on a denaturing 8% urea–polyacrylamide gel (Supplementary Fig. 1c). The fractions containing spliceosomal complexes were pooled, concentrated, and dialyzed against the D150 buffer (20 mM HEPES-KOH pH 7.9, 150 mM NaCl, 1.5 mM MgCl₂).

Cryo-EM sample preparation and data collection

The purified human spliceosomal complexes were used for cryo-EM sample preparation after examination by negative staining. Holey carbon grids (Quantifoil, Cu, 300-mesh, R1.2/1.3, 2 nm C) were glow-discharged in the Plasma Cleaner (HARRICK PLASMA Company) and each aliquot (4 μL) of the sample was applied to a glow-discharged grid. After blotting for 3.5 s, the grids were quickly plunged into liquid ethane cooled by liquid nitrogen using Vitrobot Mark IV (Thermo Fisher) at 8 °C and 100% humidity.

The cryo-grids were further transferred to a 300-kV Titan Krios electron microscope (Thermo Fisher Scientific) equipped with a Gatan K3 detector and GIF Quantum energy filter (slit width 20 eV). The micrographs were collected using a normal magnification of 81,000× in the super-resolution mode (Supplementary Fig. 1d). Each movie stack of 32 frames was automatically exposed for 2.56 s with a total dose of ~50 e⁻/Ų using EPU (Thermo Fisher Scientific). The movies with the preset defocus range from −1.4 to −2.0 μm were aligned and summed using MotionCor2⁶⁰ with a binning factor of 2, resulting in a pixel size of 1.077 Å. Dose weighting was performed concurrently. The defocus value for each image was determined using Gctf⁶¹.

Cryo-EM image processing

The cryo-EM data processing is briefly described here (Supplementary Fig. 2). All steps were mainly carried out in RELION 3.0⁶² unless stated otherwise. In total, 17,070 good micrographs were manually selected from 18,277 collected images, generating ~4.6 million particles using Gautomatch (developed by Kai Zhang, https://www2.mrc-lmb.cam.ac.uk/download/gautomatch-053/). Multi-reference guided 3D classifications were applied to the total particles, and two parallel runs were simultaneously performed. Multiple conformational states of the human-activated spliceosome were clearly present. Good particles were merged, and the duplicated particles were removed. The remaining ~1.5 million particles were re-extracted using a 2× binned pixel size of 2.154 Å. To further differentiate the different conformational states, we performed another round of multi-reference 3D classification. Three different conformational states of the human-activated spliceosomes were selected. The remaining particles were re-extracted using the unbinned pixel size of 1.077 Å. Good particles for each conformation were further analyzed using 3D classification with the skip_align option in RELION⁶². Finally, six distinct conformational states of the activated spliceosomes (B^act complex) and the catalytic step I spliceosome (C complex) were reconstructed. Based on differences of protein composition and conformation, the six conformational states of the B^act complex were unambiguously ordered and named as pre-B^act, B^act-I, B^act-II, B^act-III, B^act-IV, and post-B^act. After final refinement using the Non-uniform Refinement in cryoSPARC⁶³, the reconstructions for pre-B^act, B^act-I, B^act-II, B^act-III, B^act-IV, post-B^act, and the C complexes were obtained at average resolutions of 3.4, 3.0, 4.2, 3.0, 3.3, 3.0, and 3.4 Å using 46,696, 136,665, 13,372, 111,205, 47,352, 92,596, and 18,223 particles, respectively (Supplementary Figs. 2, 3 and Supplementary Table 1).

Reported resolutions are calculated on the basis of the FSC value of 0.143⁶⁴. The angular distributions of the particles used for the final reconstructions are reasonable (Supplementary Fig. 4). Local resolution variations were estimated using cryoSPARC⁶³ (Supplementary Fig. 5). The EM density maps display clear features (Supplementary Figs. 6–10). Based on these results, we performed detailed structural and biochemical analysis on PRP2 (Supplementary Figs. 11–14), U2 snRNP (Supplementary Fig. 15), and the scaffolding component SF3B1 (Supplementary Fig. 16). We also made the structural comparison between the intermediate states reported in this manuscript and those reported previously (Supplementary Fig. 17).

Model building and refinement

The atomic models were generated by rigid docking of components with known structures and de novo modeling facilitated by AlphaFold⁶⁵. For six different conformational states of the B^act complex, identification and docking of the known components were guided by the structures of previously reported human B^act complexes^22,26. Protein components derived from these structures were fitted into the EM density maps using Chimera⁶⁶ and manually adjusted using Coot⁶⁷.

For the splicing factors KIN17 in pre-B^act, SRRM1 in pre-B^act and B^act-I, and PPIL2 in pre-B^act to B^act-III, individual domains from the predicated models were first docked into the respective EM density regions; the flexible linker sequences were de novo built. The N-terminal extension of PRP2 was assigned on the basis of its predicted helical feature and the positions of the bulky residues. Initially identified by mass spectrometry, the splicing factor PPIL4 was located through examination of the EM map and modeled on the basis of its AlphaFold-predicted structure. Based on our previously reported structures³¹ and the current 3.4-Å EM map, the atomic model of the human C complex in this study was improved.

The final atomic models of the human-activated spliceosomes and C complex were refined according to the cryo-EM maps using phenix.real_space_refine program in PHENIX with secondary structure restraints⁶⁸. The structures were further validated through examination of the MolProbity scores and statistics of the Ramachandran plots. MolProbity scores were calculated as described⁶⁹.

Expression and purification of PRP2

The cDNA sequences of human full-length PRP2 (DHX16, Uniprot ID: O60231), three truncated constructs (Δ1-198, Δ1-305, Δ1-361), and four missense variants (R503A, T674M, L709A, R998A) were generated using PCR-based strategy, and individually subcloned into a pCAG vector with an N-terminal 3xFlag tag⁷⁰. The constructs were confirmed using DNA sequencing. Each plasmid was prepared with the GoldHi EndoFree Plasmid Maxi Kit (CWBio Inc., Cat# CW2104M). The purified plasmids were transiently transfected into Expi293F cells using 40K polyethyleneimines (PEIs, Polysciences) when the cell density reached about 2 × 10⁶. The transfected cells were further cultured in SMM 293T-II medium at 37 °C with 5% CO₂ for ~60 h.

The harvested cells were resuspended in the lysis buffer (25 mM HEPES-KOH pH 7.9, 150 mM NaCl) supplemented with 1 mM phenylmethylsulfonyl fluoride, 2.6 μg/mL aprotinin, 1.4 μg/mL pepstatin, and 5 μg/mL leupeptin, and subsequently lysed through sonication. Cell debris was removed through centrifugation, and the supernatant was applied to the anti-Flag M2 affinity gel (Sigma). The washed resin was further rinsed with the buffer containing 25 mM HEPES-KOH pH 7.9, 50 mM NaCl, 0.2 mg/mL Flag peptide. The protein eluent was applied to a heparin column (HiTrap Heparin HP, Cytiva) to remove endogenous nucleic acids. The desired fractions were concentrated, and subjected to a size-exclusion chromatography (Superdex 200 10/300 GL, Cytiva) using the lysis buffer. The peak fractions were examined using SDS–PAGE (Supplementary Fig. 13), and stored at −80 °C for further reconstitution assays.

Immuno-depletion of PRP2 and reconstitution assay

To deplete endogenous PRP2, 20 μL PRP2 antibody (Proteintech, Cat# 11021-1-AP) was incubated with 10 μL Protein A/G Magnetic beads (Pierce) at room temperature for 2 h in a buffer containing 10 mM Tris-HCl pH 8.0, 150 mM NaCl, 0.01% NP-40. Four hundred microliters HeLa nuclear extract was successively incubated with three aliquots of 10-μL anti-PRP2-conjugated magnetic beads at 4 °C, each for 2–3 h. The PRP2-depleted (ΔPRP2) nuclear extract was collected. Depletion efficiency was examined through western blots using an anti-PRP2 antibody. For mock depletion, 20 μL of 0.35 mg/mL BSA were incubated with 10 μL of Protein A/G Magnetic beads.

The small-scale in vitro splicing assays were performed using 40% (v/v) WT, mock and PRP2-depleted (ΔPRP2) nuclear extracts supplemented with 2 mM ATP, 20 mM creatine phosphate, 3 mM MgCl₂, 20 mM HEPES-KOH pH 7.9. The splicing reaction was carried out at 30 °C for 1 h. The total RNA was isolated through phenol extraction and ethanol precipitation after proteinase-K digestion. The splicing efficiency was examined using reverse transcription-PCR.

For PRP2 reconstitution assay, each of the purified recombinant proteins of the PRP2 variants was pre-incubated with the ΔPRP2 nuclear extract on ice for 30 min at a concentration of 400 nM. Then the protein-supplemented ΔPRP2 nuclear extract was applied to the small-scale splicing assay, and the final concentration of reconstituted recombinant proteins was about 200 nM.

Reporting summary

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

Author contributions

Y. Shi conceived and supervised the project. X. Zhan and Y. Lu designed and performed the experiments, and processed the EM data. X. Zhan built the atomic models. X. Zhan, Y. Lu, and Y. Shi wrote the manuscript.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

The atomic coordinates for human pre-B^act, B^act-I, B^act-II, B^act-III, B^act-IV, post-B^act, and C complexes have been deposited in the Protein Data Bank (PDB) under the accession codes 8I0P, 8I0R, 8I0S, 8I0T, 8I0U, 8I0V, and 8I0W, respectively. The EM maps of human pre-B^act, B^act-I, B^act-II, B^act-III, B^act-IV, post-B^act, and C complexes have been deposited in the EMDB with accession codes EMD-35105, EMD-35107, EMD-35108, EMD-35109, EMD-35110, EMD-35111, and EMD-35113, respectively. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-024-50785-0.