Research Article

Molecular architecture of the Saccharomyces cerevisiae activated spliceosome

See allHide authors and affiliations

Science  23 Sep 2016:
Vol. 353, Issue 6306, pp. 1399-1405
DOI: 10.1126/science.aag1906


The activated spliceosome (Bact) is in a catalytically inactive state and is remodeled into a catalytically active machine by the RNA helicase Prp2, but the mechanism is unclear. Here, we describe a 3D electron cryomicroscopy structure of the Saccharomyces cerevisiae Bact complex at 5.8-angstrom resolution. Our model reveals that in Bact, the catalytic U2/U6 RNA-Prp8 ribonucleoprotein core is already established, and the 5′ splice site (ss) is oriented for step 1 catalysis but occluded by protein. The first-step nucleophile—the branchsite adenosine—is sequestered within the Hsh155 HEAT domain and is held 50 angstroms away from the 5′ss. Our structure suggests that Prp2 adenosine triphosphatase–mediated remodeling leads to conformational changes in Hsh155’s HEAT domain that liberate the first-step reactants for catalysis.

The spliceosome is a highly dynamic molecular machine that assembles de novo for each round of splicing by the ordered interaction of the U1, U2, U4/U6, and U5 small nuclear ribonucleoprotein (snRNPs) with a pre-mRNA intron (1). A key step toward a catalytically active spliceosome is the transformation of the spliceosomal B complex, which lacks an active site, into an activated Bact complex (fig. S1). This entails extensive structural rearrangements, including release of the U4 small nuclear RNA (snRNA) (catalyzed by the RNA helicase Brr2), and all U4/U6 and several U5 snRNP proteins (fig. S1). At the same time, the Prp19 complex (NTC), retention and splicing (RES) proteins, and about 10 so-called Bact proteins are recruited or stably integrated (2). Bact is transformed into the catalytically active B* complex by the Prp2 RNA helicase and its cofactor Spp2, which facilitate RNP remodeling, including destabilization of the U2 snRNP SF3a and SF3b proteins (36). B* catalyzes step 1 of the splicing reaction, namely 5′ splice site (ss) cleavage and formation of a lariat-intron-3′exon intermediate. The subsequently formed C complex then catalyzes step 2, which entails 3′ss cleavage and the ligation of the 5′ and 3′ exons.

During activation, a catalytic RNA-RNA network, very similar to the catalytic core of group II self-splicing introns (7), is established (8, 9), as documented with high-resolution cryogenic electron microscopy (cryo-EM) of endogenous, post–step 2, intron lariat spliceosomes (ILSs) from Schizosaccharomyces pombe (10). Unlike group II introns, formation of the catalytic RNA network requires spliceosomal proteins (1), foremost the U5 Prp8 protein (1012). Although the cryo-EM model of the S. pombe ILS provides molecular insight into the architecture of the spliceosome’s catalytic RNP core, it does not reveal (i) how and when the latter is generated during the spliceosome’s stepwise catalytic activation process; (ii) the architecture of the BS/U2 snRNA helix, and the U2 SF3a and SF3b proteins that stabilize the BS/U2 interaction; or (iii) the mechanisms by which the BS adenosine (BS-A, the step 1 nucleophile) is positioned for step 1 catalysis during Prp2-mediated remodeling of the Bact complex. High- to medium-resolution cryo-EM structures of the Saccharomyces cerevisiae and human U4/U6.U5 tri-snRNPs have been published (1316); however, no high-resolution structures of spliceosomes during their catalytic activation phase are available. A comparison of two-dimensional (2D) EM images of purified yeast B, Bact, and B* complexes indicates substantial structural changes during the transitions from one complex to the next (2, 17). Here, we report a 3D cryo-EM structure of the S. cerevisiae spliceosomal Bact complex at an overall resolution of 5.8 Å, which allowed us to resolve the spatial organization of most of its protein and RNA components.

Structure determination and model-building

S. cerevisiae Bact spliceosomal complexes assembled in vitro were affinity purified (fig. S1), and an initial 3D model was determined by means of random-conical-tilt 3D reconstruction followed by 3D maximum-likelihood alignment and 3D classification (18). The calculated low-resolution structure was used as a starting model to determine the final 5.8 Å reconstruction from an initial data set of 1 million particle images after several steps of image-sorting and classification (fig. S2). The 3D structure of Bact encloses a total volume corresponding to a molecular mass of ~3.8 MDa (Fig. 1A). The major parts of the Bact spliceosome (~70% of the total mass) are well-defined in the high-resolution structure. However, some areas that are either more dynamic or contain components with less than stoichiometric occupancies are not visible in the final structure. The pseudo-atomic model was built for the stabler part of Bact, in which the resolution sufficed for clear identification of structured protein domains and double-strand RNA elements, allowing us to fit known x-ray structures or homology models of structured regions of Bact complex components into the EM density map (table S1). Additionally, we performed chemical protein cross-linking coupled to mass spectrometry (CX-MS) (table S2) to validate the locations of large proteins and facilitate the docking of smaller ones.

Fig. 1 3D cryo-EM structure of the yeast Bact complex and location of the U5 snRNP proteins and U5, U6, and U2 snRNAs.

(A) Different views of the Bact EM density map with the better resolved density in blue and the masked Bact regions in gray. The Bact spliceosome exhibits a mushroom-like shape, with a main body connected to a foot and a steep and a shallow slope. Details about the “front bar” domain are provided in fig. S13. (B) Position of the U5 proteins Brr2 (NC/CC, N-terminal/C-terminal helicase cassettes), Prp8 (NTD1, NTD2; N-terminal domains 1 and 2; RT, reverse-transcriptase–like; En, endonuclease-like; RH, RNase H–like; Jab1, Jab1/MPN-like), and Snu114 (domains D1 to D5 homologous to EF-G/EF-2). (C) RNA helical high-density regions (filtered to a resolution of 10 Å), representing U5, U6, and U2 RNA and the pre-mRNA intron. The position of the U5 Sm is also shown. (D) A schematic diagram of U5 and the U2/U6/pre-mRNA network and the heptameric Sm ring of U5.

The conformation of Prp8 and Brr2 differs substantially in the tri-snRNP and Bact complex

Tri-snRNP components still present in the spliceosome at the Bact stage include the guanosine triphosphatase (GTPase) Snu114, the major scaffolding protein Prp8 found at the spliceosome’s catalytic core, and the spliceosome-activating RNA helicase Brr2, in addition to the U5 Sm core and U5 and U6 snRNAs (figs. S1 and S3). Snu114 is located as a compact structure at the lower end of the main body of the Bact structure (Fig. 1B and fig. S4A). Prp8, which has an open conformation in the tri-snRNP (1316), has adopted a closed conformation in the Bact structure (Figs. 1B and 2), similar—but not identical—to its conformation in the S. pombe ILS complex (10). That is, the middle part of the RT/En domain contacts the NTD1 domain but not the tip of the En domain (Fig. 2A). The reason for this may be that the En domain has moved ~12 Å upward in Bact compared with the ILS (Fig. 2, B and C). In the S. cerevisiae and human tri-snRNPs, and in the S. pombe ILS, there is a major β-hairpin-loop (henceforth termed the switch loop) from the Prp8 linker region that runs along the long axis of the RT/En domain and touches the En domain. In the Bact complex, the EM density and protein cross-links (table S2) indicate that this switch loop (comprising Prp8 amino acids H1404 to L1436 in S. cerevisiae) has been rearranged into a loop that protrudes from the interface between Prp8’s thumb/X and linker domains and is situated close to exon 1 (Figs. 2 and 4C). Last, Prp8’s RH domain is closely associated with the En domain in Bact, in a region where in the ILS and tri-snRNPs the tip of the switch loop is instead located (Fig. 2 and fig. S4C). Thus, the Prp8 RH domain is located in different places in each of the spliceosomal cryo-EM structures analyzed so far (fig. S4, B and C).

Fig. 2 Structural differences in Prp8’s RT/En domains in the Bact complex versus S. pombe ILS.

(A) Open conformation of Prp8 in the yeast tri-snRNP and closed conformation in the Bact and ILS structures, respectively. The surface charge of Prp8 is shown (blue represents positively charged and red negatively charged residues). The stippled red semicircle at the lower end of Prp8 (middle) indicates the location of the lasso-like region of the NTD1 domain, whose density can only be partly discerned in the Bact structure. The stippled black circles (middle and right) indicate the differential interactions of Prp8’s En and NTD1 domains in the Bact and ILS complexes, respectively. (B) The Prp8 switch loop (comprising amino acids H1404 to L1436 in S. cerevisiae) that runs along the long axis of the RT/En domain in the S. pombe ILS is rearranged in Bact and protrudes from the interface between Prp8’s thumb/X and linker domains (indicated by the red arrow). (Single-letter abbreviations for the amino acid residues are as follows: A, Ala; C, Cys; D, Asp; E, Glu; F, Phe; G, Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro; Q, Gln; R, Arg; S, Ser; T, Thr; V, Val; W, Trp; and Y, Tyr.) The movement of the En domain in the S. pombe ILS ~12 Å upward to its position in the S. cerevisiae Bact complex is indicated by a dark red arrow. (C) The position of the switch loop in the Bact and ILS complexes is shown in blue in a space-filling model. The switch loop has the same conformation in the yeast and human tri-snRNPs as in the S. pombe ILS.

Fig. 4 A protein density element close to the 5′ss and path of the 5′ exon channel in Bact.

(A) The GU dinucleotide at the intron’s 5′ end is in close contact with a protein density element (blue). (B) A slice through the spliceosomal cryo-EM density, depicting the path of the 5′ exon leading to the catalytic center of the spliceosome. The 5′ exon is shown in red, and the proteins comprising the channel are indicated. (C) Path of the exon-binding region between Snu114’s D4 and D5 domains. The rearranged switch loop region of Prp8 (indicated with a red arrow) is positioned close to the pre-mRNA.

The RNA helicase Brr2 contains two tandemly organized helicase cassettes (fig. S3A) that we could localize in Bact in complex with the Prp8 Jab1 domain (19, 20) close to the tip of the Prp8 En domain (Fig. 1B and fig. S5A). Brr2’s general location is similar in the yeast Bact complex and yeast tri-snRNP, but in Bact, Brr2 is rotated by ~45° toward the long axis of the RT/En domain (fig. S5B) and shares an interface with Prp8’s En domain through the helix-loop-helix domain of its N-terminal helicase cassette (NC) (fig. S5A). In the human tri-snRNP, Brr2 is located in a substantially different position (compared with the yeast tri-snRNP), close to Prp8’s RT domain (fig. S5C) (16). Because the human and yeast Bact complexes are structurally similar, at least at the level of 2D EM class averages (2, 21), hBrr2 must thus be dramatically repositioned after tri-snRNP integration into the human spliceosome (16).

The locations of all high-density RNA helical elements within the Bact structure are shown in Fig. 1C. The major stem-loop (SL) of U5 fits into the long RNA helical density element, so that the 3′ terminal U5 snRNA Sm site (and thus Sm RNP core) is located at the foot of the 5.8 Å Bact EM density map, whereas U5 loop 1 is found in the upper region of the elongated main body (Fig. 1C and fig. S6). U5 snRNA has essentially the same structure in the yeast tri-snRNP (13) and Bact complex and tightly interacts with Prp8’s NTD1 domain.

The catalytic U2/U6 RNA network is established in the Bact structure

In the S. pombe ILS, a group II–intron-like catalytic RNA center is found. That is, the U2 and U6 snRNAs are extensively base-paired and form a triple helix that brings the U6 ISL and the catalytic U6 AGC triad close together, allowing U6 to bind two Mg2+ ions for catalysis (7, 2225). In the precatalytic spliceosomal B complex, the U6 snRNA is based-paired with the U4 snRNA, preventing formation of the U6 ISL, U2/U6 helix I, and thus triple helix formation (fig. S1B). Biochemical and genetic evidence (7, 24), together with the closed conformation of Prp8 that we observed in the Bact complex, suggests that the U2 and U6 snRNAs are rearranged in Bact to form the catalytic U2/U6 triplex structure observed in the ILS. Indeed, there is a high-density, RNA-shaped element in the Bact model, close to U5 snRNA loop 1, into which the rearranged catalytic U2/U6 RNA structure of the S. pombe ILS can be fit (Fig. 1C and fig. S6). The well-defined density allows the placement of the U6 ISL loop (close to U5 loop 1), the kinked stem of the U6 ISL, the U2/U6 helices Ia and Ib, and the sharp U6 snRNA turn between the U6 ACAGA box helix and U2/U6 helix Ia (Fig. 3, A and B). The density of the RNA elements is also consistent with the existence of the catalytic triplex, as observed in the S. pombe ILS (25). There is, however, no density for the base pair between U6-G63 and U6-C84 at the base of the ISL (Fig. 3, A and D). Connectivity between both helices is nevertheless provided by a continuous stack between the U6 bases C61, and A62 to U65. This single-strand stack follows the kink of the U6 ISL near nucleotides (nt) U6-A62 and U6-U64 (Fig. 3D) (23, 26). As shown in Fig. 3C, the topology of the phosphate groups of U6 nucleotides U80, G78, G60, and A59, which have been shown to coordinate Mg2+ ions in the spliceosome (24), is consistent with the possibility that the two catalytic Mg2+ ions are coordinated by the phosphate groups at the 4 Å distance that is optimal for RNA catalysis (fig. S7) (24, 26). Although at our level of resolution we cannot discern whether catalytic Mg2+ ions are in place, we can conclude that no major rearrangement of the catalytic U2/U6 RNA network is subsequently required for catalysis of step 1 of splicing.

Fig. 3 Position of the 5′ss relative to RNA components of the catalytic center of the Bact complex.

(A) EM density map of the catalytic core RNA elements. U6 snRNA nucleotides of the catalytic triplex are shown in purple, with unpaired nucleotides 63 and 84 in pink. U5 snRNA (not visible) lies at the back. The orange arrow denotes the sharp U6 snRNA turn between the U6 ACAGA box and U2/U6 helix Ia. (B) Secondary-structure interactions in the RNA core of the spliceosome. The proposed tertiary interactions (7) are shown as stippled lines. Nucleotides coordinating metals (24) are labeled with a blue dot. (C) Positions of the metal-coordinating phosphate groups of U6 nucleotides U80, G78, G60, and A59 [(B) and fig. S7A] relative to each other are similar to the situation in the precatalytic group II intron (27). Exon and intron nucleotides are purple. Phosphates and their oxygens crucial for metal coordination are highlighted gray and red, respectively. (D) The helical stack of U6 bases C61 to U65. (E) Path of the pre-mRNA in the vicinity of the 5′ss. (F) Close-up of the position of the 5′ss phosphate in the EM density.

The 5′ss is shielded by protein and spatially separated from the catalytic center by 6 to 7 Å

As expected, an RNA helical element comprising the U6-ACAGA box base-paired to the 5′ end of the intron is located adjacent to the catalytic center (Fig. 3, A and E, and fig. S6). Continuing along the intron sequence (toward the 5′ss), a density element is observed that makes a U-turn close to the U6/U2 catalytic center (Fig. 3E). This density accommodates well a kinked 5′ss RNA stretch from the precatalytic group II intron structure (GUUAU/gu) (27, 28), with the scissile bond of the pre-mRNA 5′ss oriented toward the catalytic center (Fig. 3E). The scissile phosphate group is, however, located 6 to 7 Å away from the catalytic site where the Mg2+ ions are expected to be positioned (Fig. 3C and fig. S7). Moreover, the GU dinucleotides at the intron’s 5′ end are in close contact with a short protein density element (Fig. 4A). As described in fig. S8, this protein density element might represent part of the N-terminal region of the NTC protein Cef1. However, we cannot exclude that it is composed of a different protein. Irrespective of its nature, this protein element spatially separates the 5′ss from the catalytic center and at the same time hinders access of the BS adenosine—the step 1 nucleophile—to the 5′ss. Thus, during catalytic activation this protein, which contacts also the HEAT domain of the U2 Hsh155 protein (Fig. 4A), must be rearranged to liberate the 5′ss for its final docking into the catalytic center.

A 5′ exon-binding channel is located between Prp8’s RT and NTD1 domain, close to Cwc22

The first 4 nucleotides (nt) of the 5′ exon upstream of the 5′ss are located close to U5 snRNA loop 1, and the EM density indicates that nucleotides 2 and 3 are base-paired with loop 1 nucleotides U96 and U97, respectively, which is consistent with earlier biochemical studies (29, 30). We can trace an additional 16 nt of the exon RNA, which thread through a narrow channel between Prp8’s RT and NTD1 domains and then runs next to Prp8’s rearranged switch loop and then Cwc22’s MA3 domain (Fig. 4C). Beyond the MA3 domain, the exon-binding region may extend along a path between Snu114’s D4 and D5 domains that is carpeted with positively charged amino acids (Fig. 4C and fig. S9B). Guided by protein cross-links, we can locate the MIF4G and MA3 domains of Cwc22 on both sides of the exon channel (Fig. 4, fig. S10, and table S2) (31, 32). In the human spliceosome, the evolutionarily conserved Cwc22 protein aids in the deposition of the exon junction complex (EJC) upstream of the spliced exon junction by binding the eIF4AIII helicase of the EJC through its MIF4G domain (31, 33), which is consistent with the location of Cwc22 close to the exon channel (fig. S10). Last, the N-terminal region of Cwc24 is also located close to the exon channel, as evidenced by multiple cross-links to Prp8’s NTD1, RH, and Jab1 domains (fig. S9D). However, in the EM map we cannot discern a clear density for this region of Cwc24, indicating that it is flexible in the Bact complex.

Organization of the NTC proteins

In the S. pombe ILS, several proteins of the Prp19 (NTC) complex—including Clf1 (Cwf4) (also named Syf3 in S. cerevisiae), Prp46, Prp45, and Cef1 (Cdc5)—are close to the catalytic RNA network, in addition to Prp8 (10). Except for the N-terminal regions of Cef1 (Cdc5) (fig. S8), the structural organization of these proteins is essentially the same in Bact and the ILS. Like in the S. pombe ILS (10), the tetratricopeptide repeat (TPR) proteins Syf1 (Cwf3) and Clf1 (Cwf4) form long, curved α-helical solenoids that cross one another and together form a basket-like structural element in the Bact complex, as seen in the low-resolution Bact model, which is consistent with protein cross-linking results (figs. S11 and S12, B and C). The 5′ SL of U6 and U2/U6 helix II, as well as the U6 snRNA-binding protein Cwc2 and the neighboring Ecm2 and Bud31 proteins, are also in similar positions in Bact and the ILS complexes (fig. S11). According to the density map of the unmasked low-resolution Bact model and protein cross-links, the WD40 domain of Prp17 (34), which was not mapped in the ILS, was localized above Ecm2 (fig. S12, A and B). The α-helical elements of the four copies of Prp19 (Cwf8), together with those of the NTC proteins Snt309 (Cwf7) and the C-terminal region of Cef1 (Cdc5), form a helical bundle (HB) that runs, as a self-contained arm II domain, parallel to the main body of the ILS and is only bound to it by thin structural elements (10). In the Bact complex, this entire Prp19 helical bundle density is repositioned considerably, adopting an orthogonal orientation relative to the central part of the main body (fig. S13).

The SF3b protein complex acts as a major scaffolding unit, bridging Prp8 and Brr2

In Bact, the U2 snRNP SF3a/b proteins contact the pre-mRNA intron at or near the BS (35, 36) and stabilize the U2 snRNA-BS helix, suggesting that they may help to shield the BS/U2 RNA helix and thus prevent premature step 1 catalysis. Many U2 components, including the SF3a proteins and the U2 Sm core, are either not discernable or not located in the structurally well-defined region of the yeast Bact complex, presumably because of their structural dynamics. Thus, only their general location in Bact (at the top right in the front view) can be surmised in the low-resolution Bact EM map (fig. S12D). In contrast, using the recent crystal structure of a protease-resistant human SF3b core complex (fig. S14A) (37), we could precisely localize a major portion of SF3b. This includes the C-terminal 20 HEAT repeats of Hsh155 (SF3b155 in human), as well as the Rse1 (SF3b130) three-propeller (WD40) cluster and Ysf3 (SF3b10) and Rds3 (SF3b14b), all of which interact with Hsh155’s C-terminal HEAT repeats and are organized in a similar manner in yeast Bact and the isolated human SF3b complex (Fig. 5 and fig. S14B). However, although the structures of the individual hSF3b155 and yHsh155 HEAT repeats are essentially identical, the conformation of the right-handed superhelix formed by the HEAT domain appears much more condensed in the yeast Bact complex (Fig. 5C, fig. S14, and movie S1). In the Bact structure, the HEAT domain of Hsh155 is located above the Prp8 RT/En domain (Fig. 5, A and B). The intertwined β-propellers BPA and BPC of Rse1 are located at the top of the Bact model and are aligned with the long axis of the main body (Fig. 5, A and B). The bottom part of BPB shares a major interface with the RecA domains of Brr2’s NC cassette (Fig. 5D). Last, Brr2 also contacts Hsh155 HEAT repeats H9 to H11 via its N-terminal NHD domain (Fig. 5D). Thus, Hsh155’s HEAT and Rse1’s WD40 β-propeller domains bridge Prp8 and Brr2 in the Bact complex.

Fig. 5 Localization of the SF3b complex proteins in the Bact structure.

(A) Front view of the Bact complex and fit of the SF3b proteins Rse1, Ysf3, Rds3, and Hsh155 at the top of the Bact model. (B) Slice of the upper region of the Bact complex highlighting the arrangement of SF3b proteins, position of U2/BS, and neighboring Brr2 and Prp8 proteins. Red “A” marks the BS adenosine. (C) Location of the U2/BS helix between the N- and C-terminal HEAT repeats of Hsh155 near Rds3. All 20 HEAT repeats of Hsh155 form one turn of a solenoid, located above the Prp8 RT/En domain. (D) Interface of Rse1’s β-propeller BPB with both RecA domains of Brr2’s NC cassette [see also (A) and (B)]. The Prp8 RH domain also contacts Hsh155’s HEAT domain.

The BS/U2 RNA helix is sequestered by Hsh155’s HEAT repeats and the BS-A is ~50 Å away from the 5′ss

The U2/BS helix is located between the terminal HEAT repeats of Hsh155 (Figs. 5C and 6A). The BS RNA faces the C-terminal HEAT repeats, and the BS-A is located in a protein pocket built by the β-helices of HEAT repeats H15 to H17 and capped by Rds3 (Fig. 5C). On the other side of the BS/U2 helix, the β helices of HEAT repeats 1 and 2 pack against the U2 snRNA. The orientation of the BS/U2 helix places the 5′-terminal nucleotide of U2 ~27 Å above the 3′-terminal U2 nucleotide of U2/U6 helix Ia (Fig. 1C and fig. S15), and both helices are connected by the 4-nt-long U2 linker (fig. S6). The 2′ hydroxyl of the bulged BS adenosine is spatially separated from the scissile bond of the 5′ss by 50 Å (fig. S15). Given the very high conservation of most SF3b proteins, it is very likely that the BS/U2 helix will be recognized in the human Bact complex in a similar way. However, S. cerevisiae lacks the SF3b protein p14, which could be ultraviolet (UV) cross-linked in human spliceosomes to the bulged BS-A (38, 39); it will thus be interesting to see where p14 is located in the human Bact cryo-EM map. Our results reveal that the BS/U2 RNA helix not only is held at a remote distance from the catalytic RNA center, but also that the BS-A is occluded in an SF3b protein pocket, ensuring that the first catalytic step of splicing does not occur prematurely. This in turn explains why catalytic activation proceeds through two distinct stages (Bact and subsequently B*). Prp2 facilitates the final catalytic activation, including release of the BS adenosine and 5′ss, only after the extensive rearrangements accompanying Bact formation have occurred.

Fig. 6 Path of the intron’s 3′ region across the HEAT domain of Hsh155 and location of the RNA helicase Prp2 and RES proteins.

(A) Density fit of Prp2 and the RES core complex on the convex side of Hsh155’s HEAT domain and placement of the U2/BS helix and the branch site adenosine. The intron’s 3′ end region is shown as a solid or dotted red line. The Hsh155 HEAT domain also contacts the top of the Prp8 RT domain. Additional close-up views of the location of Prp2 and RES proteins are available in fig. S16. (B) Cancer-related mutations in the human SF3b155 (the human Hsh155 homolog) mapped on the yeast Hsh155 HEAT domain in the Bact complex (viewed from the bottom). The mutated amino acids are shown as pink space-filling models. The amino acids shown in red correspond to hot spot mutations that map close to the exit site of the BS-3′ss RNA from the HEAT domain and to the binding sites of RES and Prp2. The amino acid shown in blue indicates the position of R1074 in hSF3b’s HEAT domain, close to the BS adenosine, whose mutation to histidine confers resistance to the tumor drug pladienolide B (51).

Prp2 and RES bind to the convex side of Hsh155’s HEAT domain opposite the BS/U2 helix

In the Bact 3D model, Prp2 is located on the convex side of the HEAT domain opposite the BS/U2 helix, where it is attached via its C-terminal OB fold domain to the upper part of HEAT repeats H7 to H9 of Hsh155 (Fig. 6A and fig. S16, A and B). This is consistent with this domain playing an essential role in mediating Prp2’s interaction with the spliceosome (40). We were able to fit the core structure of the RES complex, including the RRM protein Snu17 (table S1), close to Prp2 and below Hsh155’s HEAT repeats H6 to H8 (Fig. 6A). Additional parts of RES proteins, including Pml1’s FHA domain, were mapped adjacent to Prp8’s RT domain (fig. S16, C and D), indicating that the RES complex helps to bridge the Hsh155 HEAT and Prp8 RT domains. Our protein cross-linking results indicate an extensive protein-protein interaction network between Prp2, Spp2, and the RES proteins (fig. S17A), which is consistent with the latter proteins potentially playing a role in the efficiency and/or accuracy of the Prp2-facilitated remodeling of Bact (41). This protein network comprises the U2 SF3b proteins, the C-terminal region of Prp45, and the extended MA3 domain of Cwc22 that plays an important role in coupling the adenosine triphosphatase (ATPase) activity of Prp2 to catalytic activation of the spliceosome (32). Last, in the low-resolution model of the Bact complex there is additional density between Prp2 and the RES core complex, where Spp2 and unstructured regions of the RES proteins likely will be located (fig. S17B).

The path of the intron’s 3′ region across the HEAT domain

The cryo-EM map of the Bact complex contains density for most of the first 20 nt of the intron downstream of the BS, allowing us to trace the path of this RNA region. It passes along the concave side of the N-terminal HEAT repeats, exiting the HEAT domain at the bottom of repeats H4 to H6, where it passes Snu17 and is eventually bound by Prp2, which is consistent with positively charged protein patches along the proposed RNA path (Fig. 6A and fig. S18). The distance from the 3′ end of the BS sequence to Prp2’s RNA-binding channel would require 25 to 30 nt for a single-strand RNA. This path of the intron’s 3′ end (BS-3′ss RNA) agrees with UV cross-linking studies that revealed cross-links of Snu17 and Prp2 to intron nucleotides 15 to 20 and 25 to 35 downstream of the BS, respectively (3, 35, 42), whereas Hsh155 cross-linked throughout the intron’s 3′ end (35, 36).

Cancer-related mutations in SF3b155’s HEAT domain are close to RES, Prp2, and the 3′ end of the intron

Cancer-related mutations occur frequently in SF3b155’s HEAT domain close to or within the intrarepeat loops of HEAT repeats H4 to H7 (4345), and they have been proposed to change the curvature of the HEAT solenoid (43). These mutations lead, via an unknown mechanism, to alternative BS selection and ultimately to aberrantly spliced mRNAs (43, 44). Assuming that the spatial organization of components near the SF3b155 HEAT domain is similar in human and yeast Bact, cancer-related hot spot mutations would be located close to the intron’s 3′ end and also the binding sites for the RES complex and Prp2 in the human Bact complex (Fig. 6B). Thus, structural changes in the HEAT domain resulting from cancer-related mutations could potentially affect not only its interaction with the BS/U2 helix but also with the 3′ end of the intron and indirectly influence BS choice. Alternatively, or in addition, mutation-dependent changes in the curvature of Hsh155’s HEAT domain could affect its interactions with RES or other proteins and thereby alter the kinetics of alternative splicing and thus the levels of certain mRNAs.

A model for the mechanism of spliceosome catalytic activation

During catalytic activation, the BS/U2 snRNA helix must be released from its Hsh155- and Rds3-binding pocket, so that the BS adenosine can carry out a nucleophilic attack at the 5′ss. This remodeling is achieved through Prp2-mediated adenosine 5′-triphosphate (ATP) hydrolysis, during which Prp2 is thought to move in a 3′-to-5′ direction along the BS-3′ss RNA, leading to displacement of U2 proteins from the BS region (42). However, Prp2 is bound to the convex side of the Hsh155 HEAT domain, ~60 Å away from the BS/U2 helix, and thus does not have direct access to the region of the BS-3′ss RNA close to the BS/U2 helix, which is consistent with the observation that pre-mRNAs with a short piece of RNA downstream of the BS allow Bact complex formation but not catalytic activation by Prp2 (2, 42, 46). Thus, Prp2/Spp2–mediated remodeling of the BS has to occur from a distance. We thus propose instead that ATP hydrolysis by Prp2 mediates a change in the curvature of Hsh155’s HEAT domain, destabilizing its interaction with the BS/U2 helix. Indeed, HEAT domains can generally adopt diverse conformations (47). In addition, the conformations of the HEAT domains of importins and exportins, for example, are regulated by the Ran GTPase (48, 49). A change in the HEAT domain could be caused by a Prp2-mediated alteration in the interaction between the BS-3′ss RNA and the HEAT domain and/or be induced via changes in protein-protein interactions between Prp2/Spp2 and the HEAT domain. Whether this involves initial translocation of Prp2 along the BS-3′ss RNA remains an open question. Changes in the curvature of Hsh155’s HEAT domain could also destabilize the interaction of Prp2 and the RES proteins, which is consistent with these proteins leaving after B* complex formation (4, 17, 42). Proteins that occlude the 5′ss GU dinucleotide would also have to be rearranged to “liberate” the 5′ss (Fig. 4A) and to allow its repositioning closer to the catalytic center, paving the way for the nucleophilic attack by the BS-A. In Bact, the BS-A is oriented so that a rotation of the BS/U2 helix about the 4-nt-long U2 snRNA linker could potentially juxtapose the two step 1 reactants (Fig. 7). Recently, the cryo-EM structure of an endogeneous S. cerevisiae spliceosome whose structure closely resembles our in vitro assembled Bact complex was reported (50). It will therefore be interesting to perform a detailed comparison of the structure of both complexes to ascertain whether they are stalled at an identical stage during activation. Future cryo-EM studies of the catalytically activated B* complex will be needed to reveal the structural changes that accompany the transformation of the Bact to B* complex.

Fig. 7 Model of Prp2-mediated catalytic activation of the spliceosome.

Schematic of the catalytic RNA network, showing the proposed rotation of the BS/U2 helix about the 4-nt-long U2 snRNA linker that would juxtapose the OH group (red sphere) of the BS adenosine with the 5′ splice site (gray sphere), after liberation of the catalytic step 1 reactant during Prp2-mediated B* formation. Green spheres indicate catalytic magnesium ions. Movement of the BS/U2 RNA helix into the catalytic center is likely supported and coordinated by proteins such as the nearby Prp8 α-finger, which was UV–cross-linked to the BS+2 nucleotide within spliceosomes (13), or Yju2, which can be cross-linked to the U2 snRNA close to the BS/U2 helix (52) and whose interaction is stabilized during B* formation (4).


Materials and Methods

Figs. S1 to S19

Tables S1 to S3

References (5374)

Movies S1 and S2


  1. Acknowledgments: We thank T. Conrad for yeast production in a bioreactor and M. Raabe and U. Pleßmann for excellent technical assistance. We are very grateful to N. Rigo and C. L. Will for advice and many helpful discussions and to J. Schmitzovà for the gift of the dominant-negative Prp2 mutant. The EM map has been deposited in the Electron Microscopy Data Bank with the accession code EMD-4099. The atomic coordinates have been deposited in the Protein Data Bank with the accession code 5LQW. This work was supported by grants from the Deutsche Forschungsgemeinschaft (SFB 860) to R.L., H.S., and H.U.
View Abstract

Stay Connected to Science

Navigate This Article