1Beijing Advanced Innovation Center for Structural Biology, School of Life Sciences, Tsinghua University, Beijing 100084, China

1Beijing Advanced Innovation Center for Structural Biology, School of Life Sciences, Tsinghua University, Beijing 100084, China

1Beijing Advanced Innovation Center for Structural Biology, School of Life Sciences, Tsinghua University, Beijing 100084, China

2Institute of Biology, Westlake Institute for Advanced Study, Westlake University, Hangzhou 310064, Zhejiang Province, China

Precursor messenger RNA (pre-mRNA) splicing is executed by the spliceosome. In the past 3 years, cryoelectron microscopy (cryo-EM) structures have been elucidated for a majority of the yeast spliceosomal complexes and for a few human spliceosomes. During the splicing reaction, the dynamic spliceosome has an immobile core of about 20 protein and RNA components, which are organized around a conserved splicing active site. The divalent metal ions, coordinated by U6 small nuclear RNA (snRNA), catalyze the branching reaction and exon ligation. The spliceosome also contains a mobile but compositionally stable group of about 13 proteins and a portion of U2 snRNA, which facilitate substrate delivery into the splicing active site. The spliceosomal transitions are driven by the RNA-dependent ATPase/helicases, resulting in the recruitment and dissociation of specific splicing factors that enable the reaction. In summary, the spliceosome is a protein-directed metalloribozyme.

Precursor messenger RNA (pre-mRNA) splicing, discovered 40 years ago (Berget et al. 1977; Chow et al. 1977), is executed by a multi-megaDalton, ribonucleoprotein (RNP) complex known as the spliceosome (Brody and Abelson 1985; Frendewey and Keller 1985; Grabowski et al. 1985). The components of the spliceosome constitute five distinct classes: the uridine-rich RNPs (U1, U2, U4, U5, and U6 snRNPs), the NineTeen complex (NTC) and the NTC-related complex (NTR), the splicing factors, the RNA-dependent ATPase/helicases, and other regulatory and/or auxiliary proteins such as kinases/phosphatases and peptidyl prolyl isomerases (PPIs). The involvement of snRNPs in pre-mRNA splicing was documented by the early 1980s (Lerner and Steitz 1979; Lerner et al. 1980; Rogers and Wall 1980; Yang et al. 1981) and subsequently confirmed (Mount et al. 1983; Padgett et al. 1983; Kramer et al. 1984; Black et al. 1985; DiMaria et al. 1985; Krainer and Maniatis 1985; Berget and Robberson 1986; Grabowski and Sharp 1986; Pikielny and Rosbash 1986). Only the precatalytic spliceosome (B complex) and the pre-B complex contain all five snRNPs but lack the NTC and NTR. Despite their relatively late discovery (Tarn et al. 1994; Chen et al. 2002; Ohi and Gould 2002; Chan et al. 2003; Chan and Cheng 2005), NTC and NTR are known to be indispensable in structurally shaping the splicing active site (Hang et al. 2015; Yan et al. 2015; Shi 2017b).

The assembled spliceosome sequentially assumes eight different compositional states during each cycle of the splicing reaction: pre-B, B, activated (Bact), catalytically activated (B*), step I spliceosome (C), step II activated (C*), postcatalytic (P), and intron lariat spliceosome (ILS) (Fig. 1A). The B complex is preceded by a recently characterized pre-B complex (Boesler et al. 2016), in which the 5′-splice site (5′SS) is still recognized by the U1 snRNP. Each cycle consists of two sequential steps: branching and exon ligation, which proceed in the B* and C* complexes, resulting in the C and P complexes, respectively. Therefore, the B*-to-C and C*-to-P transitions occur unassisted. In contrast, the B-to-Bact, Bact-to-B*, C-to-C*, and P-to-ILS transitions are driven by the RNA-dependent ATPase/helicases Brr2, Prp2, Prp16, and Prp22, respectively (Fig. 1A) (Cordin et al. 2012). The disassembly of the ILS complex is mediated by the ATPase/helicase Prp43 (Arenas and Abelson 1997; Tsai et al. 2005).

The precursor messenger RNA (pre-mRNA) splicing cycle and the major structures of the spliceosome. (A) The pre-mRNA splicing cycle. Each cycle includes three phases: assembly and activation of the spliceosome, execution of the splicing reaction, and disassembly of the spliceosome. The pre-B complex represents the first fully assembled spliceosome, in which all five snRNPs are present and the 5′SS is still recognized by U1 snRNP. In addition to the pre-B complex, the assembled spliceosome exists in seven compositionally distinct states: B, Bact, B*, C, C*, P, and intron lariat spliceosome (ILS). The unidirectional conversion of each of these neighboring spliceosomal complexes to the next is driven by the conserved ATPases/helicases Brr2 (B-to-Bact), Prp2 (Bact-to-B*), Prp16 (C-to-C*), and Prp22 (P-to-ILS). The splicing factors Cwc25 and Yju2 facilitate the branching reaction, whereas Prp18 and Slu7 assist exon ligation. Disassembly of the ILS complex is executed by the ATPase/helicase Prp43. (B) The major cryoelectron microscopy (cryo-EM) structures of the spliceosome since 2015. The time axis is drawn to scale, with each white dot representing a month. The first atomic model of an intact spliceosome was elucidated for the ILS complex from Saccharomyces pombe on the basis of an EM density map at an average resolution of 3.6 Å (Hang et al. 2015; Yan et al. 2015). The 18 cryo-EM structures of the spliceosome come from four independent laboratories: nine from the Shi group (shaded light purple), four from the Nagai group (shaded yellow), four from the Luhrmann/Stark group (shaded cyan), and one from the Zhao/Zhou group (shaded light brown). (C) The cryo-EM structure of the ILS complex from S. pombe (PDB code 3JB9) (Hang et al. 2015; Yan et al. 2015). The color-coded structure is shown on the left, and the individual components resolved in the structure are tabulated on the right.

Both the branching and exon ligation reactions are of the bimolecular nucleophilic substitution (SN2) type transesterification. In a pre-mRNA, two neighboring exons (5′-exon and 3′-exon) are separated by an intron, which contains three conserved sequence elements: the 5′SS, the branch point sequence (BPS), and the 3′-splice site (3′SS) (Fig. 1A). The branching reaction results in a free 5′-exon and an intron lariat-3′-exon intermediate, which has a 2′-5′ phosphodiester bond linking the invariant adenosine nucleotide of the BPS to the guanine nucleotide at the 5′-end of the 5′SS (Grabowski et al. 1984; Padgett et al. 1984; Ruskin et al. 1984). Exon ligation leads to the joining of the 5′-exon and the 3′-exon and the generation of a free intron lariat (Fig. 1A).

The splicing active site of the spliceosome was postulated to contain two divalent metal ions known as M1 and M2 (Steitz and Steitz 1993), which was supported by sulfur-substituted pre-mRNA splicing studies (Sontheimer et al. 1997). In vitro splicing of Saccharomyces cerevisiae using U6 small nuclear RNA (snRNA) with sulfur-substituted nonbridging phosphoryl oxygen unambiguously identified the nucleotide U80 as a coordinating ligand for the catalytic metals (Yean et al. 2000). In the presence of thiophilic ions such as Mn2+, the spliceosome reconstituted with the sulfur-substituted U80(Sp), but not U80(Rp), U6 snRNA allowed the branching reaction to occur (Yean et al. 2000). U6 snRNA was subsequently shown to coordinate two divalent metals known as M1 and M2 (Fica et al. 2013), and structural analysis of the group II intron allowed prediction of conserved features in the active site of the spliceosome (Toor et al. 2008; Keating et al. 2010).

The entire U5 snRNP, the core of U2 snRNP, and U6 snRNA are present in all eight assembled spliceosomal complexes (Fig. 1A). Among all spliceosomal proteins, the central U5 snRNP component Prp8 is most highly conserved and plays a central role by anchoring the splicing active site. The step I splicing factors (Cwc25 and Yju2 in S. cerevisiae) only associate with the splicing active site surrounding the branching reaction, whereas the step II factors (Prp18 and Slu7 in S. cerevisiae) mainly act to promote exon ligation (Fig. 1A). In contrast, the general splicing factors Cwc21 and Cwc22 are present in the Bact through P complexes, whereas Cwc24 and Cwc27 are recruited into the Bact complex but dissociated in the Bact-to-B* transition (Ohrt et al. 2012).

Mechanistic understanding of pre-mRNA splicing and its regulation requires detailed three-dimensional information on the intact spliceosome. X-ray crystallography allowed structural determination of a number of individual spliceosomal components and subcomplexes, including the U1 snRNP core (Pomeranz Krummel et al. 2009; Weber et al. 2010; Kondo et al. 2015), the U4 snRNP core (Leung et al. 2011), subcomplexes of the U2 and U6 snRNPs (Price et al. 1998; Sickmier et al. 2006; Lin and Xu 2012; Montemayor et al. 2014; Zhou et al. 2014), the exon junction complex bound to RNA (Andersen et al. 2006), a large fragment of Prp8 bound to Aar2 (Galej et al. 2013), Brr2 bound to a carboxy-terminal fragment of Prp8 (Mozaffari-Jovin et al. 2013; Nguyen et al. 2013), the SF3a complex (Lin and Xu 2012), and the core of the SF3b complex (Cretu et al. 2016). A structure of the retention and splicing complex (RES) was determined by nuclear magnetic resonance (NMR) (Wysoczanski et al. 2014). These structures reveal detailed features of the individual components and their mutual interactions within the colossal spliceosome jigsaw puzzle. Because the intact spliceosome proved to be intractable to crystallization, its structural information comes exclusively from electron microscopy (EM).

Until 2015, EM-based studies, with the highest resolution ∼20–29 Å, had only allowed a glimpse of the overall appearance of the spliceosome at different splicing stages and tentative assignment of some snRNPs and protein components (Luhrmann and Stark 2009). These EM structures were reported for the A (Furman and Glitz 1995; Behzadnia et al. 2007), B (Boehringer et al. 2004; Deckert et al. 2006; Fabrizio et al. 2009; Wolf et al. 2009), Bact (Fabrizio et al. 2009; Bessonov et al. 2010), C (Jurica et al. 2004; Ohi et al. 2007; Fabrizio et al. 2009; Golas et al. 2010), P (Ilagan et al. 2013), and ILS (Chen et al. 2014) complexes. In 2015, the 5.9 Å cryoelectron microscopy (cryo-EM) structure of the S. cerevisiae U4/U6.U5 tri-snRNP, for the first time, allowed faithful assignment of the protein and RNA components (Nguyen et al. 2015).

Major advances have been achieved in the mechanistic understanding of the spliceosome structure and function since 2015, beginning with the public release of the first atomic model for an intact spliceosome (Fig. 1B) (Hang et al. 2015; Yan et al. 2015). Thirty-seven protein components and four snRNA molecules of the endogenous Saccharomyces pombe ILS complex were atomically modeled from the 3.6 Å cryo-EM reconstruction (Fig. 1C), in which the local resolution in the core regions of the spliceosome reached 2.9 Å (Yan et al. 2015). Most notably, structural insights derived from the S. pombe ILS complex are generally applicable to most other spliceosomes from both S. cerevisiae and human (Shi 2017a).

Since 2016, at least 17 independent cryo-EM structures of the assembled spliceosome have been reported, reflecting six distinct compositional states (Fig. 1B; Table 1). Among these structures, 11 were reported for the S. cerevisiae spliceosomal complexes: B (Plaschka et al. 2017), Bact (Rauhut et al. 2016; Yan et al. 2016), C (Galej et al. 2016; Wan et al. 2016a), C* (Fica et al. 2017; Yan et al. 2017), P (Bai et al. 2017; Liu et al. 2017; Wilkinson et al. 2017), and ILS (Wan et al. 2017); six were reported for the human spliceosomes: B (Bertram et al. 2017a), Bact (Haselbach et al. 2018; Zhang et al. 2018), C (Zhan et al. 2018), and C* (Fig. 1B) (Bertram et al. 2017b; Zhang et al. 2017). In addition, three independent structures of the U4/U6.U5 tri-snRNP have been reported for S. cerevisiae (Nguyen et al. 2016; Wan et al. 2016b) and for human (Agafonov et al. 2016), and the structure of the S. cerevisiae U1 snRNP was elucidated at 3.6 Å resolution (Table 1) (Li et al. 2017).

Cryo-EM studies of the spliceosomes and the spliceosomal complexes

aThe resolution shown refers to the average overall resolution of the entire spliceosome or the spliceosomal complex. The resolution in parentheses represents the highest local resolution in the spliceosome or the spliceosomal complex.

The reported average resolution for the 18 cryo-EM structures of the entire spliceosome ranges between 3.3 Å and 9.9 Å resolution, with the highest limit reaching 2.8 Å (Table 1). Among these structures, only seven were determined at an average overall resolution of 3.6 Å or better, including the S. pombe ILS complex at 3.6 Å (Yan et al. 2015), the human Bact complex at 3.4 Å (Haselbach et al. 2018), and five S. cerevisiae complexes: Bact at 3.5 Å (Yan et al. 2016), C at 3.4 Å (Wan et al. 2016a), P at 3.3 Å (Liu et al. 2017) and 3.6 Å (Bai et al. 2017), and ILS at 3.4 Å (Fig. 1B; Table 1) (Wan et al. 2017). The average overall resolution refers to the entire spliceosome, not just the core or a specific region as evasively claimed in some studies. The local resolution for the core regions of the spliceosome in each of the six structures reaches at least 3.0 Å, which allows accurate assignment of amino acid side chains and placement of RNA nucleotides and metal ions. The ultimate test of the quality of a cryo-EM structure is not the reported resolution, but the quality of the original EM density map. Importantly, side-by-side comparison of the deposited EM density maps shows that the seven structures are indeed of the highest quality among all 18 published structures of the yeast or human spliceosome.

Two general methods were used for the isolation of the spliceosomes. One method involves direct purification of endogenous spliceosomal complexes from the nuclear extract using an affinity-tagged protein component (Ohi et al. 2007; Yan et al. 2015). This method was effectively applied to purification of the yeast, but not human, spliceosomes and allowed determination of seven of the 18 published spliceosome structures (Table 1). The pre-mRNA in these spliceosomes represents a mixture from endogenous genes; its atomic modeling was made possible by the strong sequence conservation of yeast intronic sequences (Spingola et al. 1999; Bon et al. 2003; Kupfer et al. 2004). In contrast, the other method relies on in vitro assembly of the spliceosome on MS2-tagged synthetic pre-mRNA (Jurica et al. 2002; Zhou et al. 2002). This method has been successfully applied to both human and yeast spliceosomes, although an additional step of chemical cross-linking is required for maintenance of the structural integrity for the human spliceosome (Table 1).

Except the pre-B and the B* complexes, six of the eight assembled spliceosomal complexes during each splicing cycle have been structurally characterized in recent cryo-EM studies. Among these, the pre-B and the B complex are the only spliceosomes that do not contain a well-formed splicing active site; they show a major difference compared with the other six spliceosomal complexes. The U4 snRNP remains associated and the NTC and NTR components are yet to be recruited. The other six spliceosomes share a set of strikingly similar structural features that were first observed in the cryo-EM structure of the S. pombe ILS complex (Hang et al. 2015; Yan et al. 2015).

The RNA elements in the splicing active site show a highly conserved conformation in the Bact through ILS complexes (Fig. 2A). The intramolecular stem loop (ISL) of U6 snRNA forms the center of the splicing active site and coordinates the two catalytic metal ions (M1 and M2). The ISL is stabilized by three conserved structural metal ions. Three consecutive nucleotides of the U2 snRNA from the helix Ib form a catalytic triplex with a duplex of U6 snRNA, which involves five nucleotides preceding the ISL and the key U6 nucleotide U80 in S. cerevisiae (Fig. 2B). The catalytic triplex facilitates the splicing reaction. Loop I of U5 snRNA, which recognizes the 5′-exon, is positioned in close proximity to the ISL. Beyond the active site, U5 and U6 snRNAs along their entire lengths and 30 consecutive nucleotides at the 5′-end of U2 snRNA show the same structure in the Bact through ILS complexes (Fig. 2C). Thus, among the three snRNA molecules, the only mobile region is the sequence downstream from nucleotide 30 of U2 snRNA.

Shared structural features of the spliceosome during the splicing reaction. (A) The RNA elements at the conserved splicing active site. The active site comprises the intramolecular stem loop (ISL) of U6 small nuclear RNA (snRNA) and associated metal ions, the U2/U6 catalytic triplex, helix I of the U2/U6 duplex, and loop I of U5 snRNA. Although the active site shown here is derived from the Saccharomyces cerevisiae C complex (PDB code 5GMK) (Wan et al. 2016a), the overall conformation is identical in the Bact through intron lariat spliceosome (ILS) complexes. (B) Structure of the conserved catalytic triplex. The original electron microscopy (EM) density map, shown for the S. cerevisiae C complex (Wan et al. 2016a), allows unambiguous identification of the metal ions and the nucleobases. (C) U5 and U6 snRNAs along their entire lengths and the 5′ 30 nucleotides of U2 snRNA remain largely static in the Bact through ILS complexes. The RNA elements in the S. cerevisiae C complex are color coded; whereas those from the Bact, C*, P, and ILS complexes are each displayed in a single color. The PDB codes for the Bact, C, C*, P, and ILS complexes are 5GM6, 5GMK, 5WSG, 5YLZ, and 5Y88, respectively. (D) The conserved splicing active site RNA elements are anchored in the catalytic cavity of Prp8. The cavity is formed at the interface between the N-domain and the Prp8 core and is enriched in positively charged residues. (E) Sixteen additional protein components, for their partial or entire lengths, remain largely static in the Bact through ILS complexes. These proteins, together with Prp8, maintain the rigid conformation of the snRNA elements (U5 and U6 along their entire lengths and the 5′ 30 nucleotides of U2 snRNA). These proteins and the snRNAs from the S. cerevisiae C complex are color coded here. (F) A compositionally constant group of 13 proteins in the Bact through ILS complexes. These proteins constitute two groups: nine in the U2 snRNP (the heptameric Sm complex, Lea1, and Msl1) and four in the NTC (Prp19, Snt309, Cef1, Syf1, and Clf1). Each group rigidly maintains its collective structure but is translocated during the two steps of the splicing reaction. These proteins, together with the other 20 rigid components, give rise to the characteristic appearance of the spliceosome.

The relatively rigid conformation of the snRNAs in the splicing active site and beyond is maintained by at least 17 protein components, which in turn must share nearly identical structures in the Bact through ILS complexes (Shi 2017b). Among these proteins, Prp8 plays a pivotal role. A positively charged catalytic cavity, which is formed at the interface between its N-domain and the core (also known as the large domain; Galej et al. 2013), provides approximately half of the needed surface to host the conserved RNA elements at the splicing active site (Fig. 2D) (Hang et al. 2015; Yan et al. 2015). The other half is furnished by segments of the NTC components Cef1 (Cdc5 in S. pombe) and Syf2, the NTR components Cwc2 and Ecm2 (Cwf2 and Cwf5 in S. pombe), and the various splicing factors. In addition, Snu114, the heptameric Sm complex of the U5 snRNP, and six NTR components (Cwc2, Ecm2, Bud31, Cwc15, Prp45, and Prp46) display the same general conformation in the Bact through ILS complexes (Fig. 2E).

In addition to these 20 protein and RNA components that remain largely rigid in the Bact through ILS complexes, 13 additional protein components are included in each of these spliceosomes. These include nine components of the U2 snRNP (Lea1, Msl1, and the heptameric Sm complex), which interact with each other and bind the sequences around the Sm site of U2 snRNA, and four proteins of the NTC (Prp19, Snt309, Syf1, and Clf1), which together with a portion of Cef1 are intertwined to form a stable core of the NTC (Fig. 2F). The relative positioning of the components within the U2 snRNP core or the NTC core remains unchanged in the various spliceosomal complexes; but the core as a whole undergoes drastic translocation, particularly the U2 snRNP.

Together, the immobile 20 protein and RNA components, along with the compositionally stable but mobile 13 proteins, give rise to a characteristic appearance of the spliceosome that is conserved in the Bact through ILS complexes (Fig. 2F). In fact, for an experienced structural biologist, visual inspection of the two-dimensional averages of the particles not only allows identification of the spliceosomal complexes but also their specific stages in the splicing reaction. Such a structural hallmark also greatly facilitates atomic modeling of the various spliceosomal complexes. For example, the atomic coordinates of the S. pombe ILS complex (Yan et al. 2015) can be conveniently docked into the EM density map of the human C* complex (Zhang et al. 2017), followed by minor adjustment and residue replacement.

The available structural evidence proves the notion that the spliceosome is a metalloribozyme. The roles of the two catalytic metal ions M1 and M2 are thought to reciprocate in the two transesterification reactions of each splicing cycle (Steitz and Steitz 1993; Fica et al. 2013). In the branching reaction that occurs in the B* complex, M2 activates the nucleophile (2′-OH of an adenine nucleotide in the BPS) and M1 stabilizes the leaving group (3′-OH of the 5′-exon). In the exon ligation that proceeds in the C* complex, M1 activates the nucleophile (3′-OH of the 5′-exon) and M2 stabilizes the leaving group (3′-OH of the intron lariat). In the S. pombe ILS complex (Hang et al. 2015; Yan et al. 2015), both M1 and M2 are observed and their locations are consistent with published biochemical characterization (Fica et al. 2013).

The M1 metal ion is uniformly coordinated by two phosphate oxygen atoms from G78 and U80 of U6 snRNA in the spliceosomal C, C*, P, and ILS complexes (Table 2). The coordination of the M1 metal, as illustrated in the splicing active site center of the S. cerevisiae C complex (Wan et al. 2016a), involves two more ligands: a phosphate oxygen atom from the 5′-end of the 5′SS and the 3′-OH at the 3′-end nucleotide of the 5′-exon (Table 2). These four ligands are arranged in a planar fashion (Fig. 3A). The putatively octahedral coordination suggests M1 to be a divalent ion, likely magnesium (Mg2+). In addition to the two uniformly coordinating ligands from U6 snRNA, M1 is bound by the 3′-OH at the 3′-end nucleotide of the 5′-exon in both the C* (Yan et al. 2017) and P (Bai et al. 2017) complexes (Table 2). In the P complex (Bai et al. 2017), M1 is also recognized by a fourth ligand: a phosphate oxygen atom at the 5′-end of the 3′-exon. In the ILS complex (Wan et al. 2017), M1 loses all other ligands and is coordinated only by G78 and U80. Notably, M1 is yet to be loaded into the Bact complex (Yan et al. 2016).

The spliceosome is a metalloribozyme. (A) Coordination of the catalytic metals M1 and M2. Shown here is metal coordination in the Saccharomyces cerevisiae C complex (Wan et al. 2016a). M1 is recognized by four ligands arranged in a planar fashion. The original electron microscopy (EM) density map is shown in panels A, B, and D. PDB code 5GMK and EMDB code EMD-9525 for the C complex; PDB code 5YLZ and EMDB code EMD-6839 for the P complex. (B) Coordination of M1 and M2 in the S. cerevisiae P complex (Bai et al. 2017). M2 is bound by five potential ligands. (C) Choreography of the catalytic metals during the splicing reaction. The coordination of M1 and M2 is shown for the spliceosomal Bact, B* (predicted), C, C*, P, and intron lariat spliceosome (ILS) complexes. Elements of the U6 small nuclear RNA (snRNA) are shown in the same orientation in the six complexes to allow meaningful comparison. Metal coordination in the B* complex is predicted based on the C complex and knowledge of the splicing reaction. (D) The structural metals are defined by an excellent EM density map (shown here for the S. cerevisiae C complex; Wan et al. 2016a) and presumably stabilize the intramolecular stem loop (ISL) fold by neutralizing the negative charges of the RNA phosphodiester backbone.

Coordination of the M1 and M2 metals in different spliceosomal complexes in the yeast of Saccharomyces cerevisiae

M2 coordination appears to be more dynamic than M1. The M2 metal ion is stably coordinated by two phosphate oxygen atoms from A59 and U80 of U6 snRNA in the spliceosomal C, C*, and P complexes (Table 2). The coordination of the M2 metal is exemplified in the active site center of the P complex (Bai et al. 2017; Liu et al. 2017; Wilkinson et al. 2017), in which M2 is bound by five ligands: the 2′-OH and 3′-OH (the leaving group in exon ligation) at the 3′-end of the 3′SS and three phosphate oxygen atoms from the nucleotides A59, G60, and U80 of U6 snRNA (Fig. 3B; Table 2). Although M2 is recognized by only two ligands from A59 and U80 in the S. cerevisiae C complex (Wan et al. 2016a), it is coordinated by one more ligand—a phosphate oxygen atom from G60—in the S. cerevisiae C* complex (Table 2) (Fica et al. 2017; Yan et al. 2017). In the S. cerevisiae ILS complex (Wan et al. 2017), however, M2 is bound only by one ligand: a phosphate oxygen atom from U80. Although M2 is loaded in the Bact complex (Yan et al. 2016), it is recognized by the phosphate oxygen atoms of A59 and G60.

This analysis reveals a choreography of the catalytic metals during each cycle of splicing (Fig. 3C). In the Bact complex (Yan et al. 2016), M2 is loaded in a preactivated fashion and M1 is absent. Both catalytic metals are presumably loaded into the B* complex, which executes the branching reaction. In the resulting C complex (Wan et al. 2016a), both M1 and M2 are properly coordinated. In the C* complex, which executes the exon ligation (Fica et al. 2017; Yan et al. 2017), M1 loses one ligand (a phosphate oxygen atom at the 5′-end of the 5′SS) owing to the relocation of the lariat junction and M2 coordination also undergoes rearrangement. In the resulting P complex (Bai et al. 2017; Liu et al. 2017; Wilkinson et al. 2017), M1 gains one more ligand and M2 has two more ligands; the coordination of M1 and M2 appears to be most complete at this stage. Finally, after exon dissociation in the ILS complex (Wan et al. 2017) M1 loses two ligands from the exon, whereas M2 has only one ligand from U80 of U6 snRNA.

In contrast to the changing patterns of the catalytic metals, three structural metal ions are steadily present in the Bact, C, C*, P, and ILS complexes as judged by the local features of the EM density map (Fig. 3D). These metals appear to stabilize the delicate fold of the U6 ISL by neutralizing the negative charges of the RNA backbone phosphates. Two of the three metal ions are also present in the S. pombe ILS complex (Hang et al. 2015; Yan et al. 2015).

The protein components in each spliceosome can be divided into four distinct groups: the structural proteins, the splicing factors, the RNA-dependent ATPase/helicases, and other regulatory proteins. The structural proteins sustain the splicing active site conformation, support the overall characteristic appearance of the spliceosome, and provide the elasticity that is needed for the splicing reaction. The splicing factors facilitate assembly and activation of the spliceosome and assist the two steps of transesterification. The ATPase/helicases remodel the spliceosomal complexes to allow flux of the splicing factors and other proteins and RNA elements. The other regulatory proteins modulate the splicing of pre-mRNA and will be briefly discussed in a later section.

Both the ribosome and the spliceosome are ribozymes. Unlike the ribosome in which the RNA components constitute the structural scaffold of the entire RNP complex (Ban et al. 2000; Nissen et al. 2000; Schluenzen et al. 2000; Wimberly et al. 2000), the snRNA elements in the spliceosome are flexible and are supported and directed by the surrounding protein components (Shi 2017b). The structural proteins include all protein components of the U2 snRNP core and the U5 snRNP, and virtually all proteins of the NTC and NTR. The structural proteins can be further classified into three distinct categories: the spring-like proteins, the rope-like proteins, and the rest.

The spring-like proteins are exemplified by the superhelical proteins Syf1 and Clf1 (Cwf3 and Cwf4 in S. pombe), which can bend and stretch, without being structurally disrupted, to allow large-degree movement of a subcomplex or a region of the spliceosome (Fig. 4A). During the C-to-C* transition, the U2 snRNP core is translocated by a distance of 60–100 Å, which is mediated by the elastic property of Syf1 (Fica et al. 2017; Yan et al. 2017). The rope-like proteins are represented by Prp45, the carboxy-terminal fragment of Cef1, and Cwc15; these proteins are flexible tethers that glue together multiple components of the spliceosome. For example, the NTR component Prp45 adopts an extended conformation, spans a distance of 150 Å, and simultaneously interacts with nine proteins and two snRNA elements in the spliceosome (Fig. 4B) (Yan et al. 2015). Other than the spring-like and rope-like proteins, each of the remaining protein components has a well-defined conformation and often contains well-folded globular domains.

Protein components and catalytic motifs of the spliceosome. (A) The spring-like proteins, exemplified by Syf1 and Clf1 in Saccharomyces cerevisiae, allow large conformational rearrangements of the spliceosome while maintaining their structural integrity. The S. cerevisiae C complex (Wan et al. 2016a) is used for illustration in panels A–C. Syf1 and Clf1 are colored yellow and cyan, respectively. (B) The rope-like proteins, represented by Prp45 and Cwc15, stabilize spliceosomal organization through direct interactions with multiple proteins and RNA elements. Prp45, colored red and displayed in surface representation, simultaneously interacts with at least nine proteins and two small nuclear RNA (snRNA) elements (U2 and U6). (C) The step I factors Cwc25 and Yju2 constitute an integral part of the splicing active site during the branching reaction. The amino termini of both proteins reach into the center of the active site RNA elements and stabilize their conformations. (D) The step II factors Prp17 and Prp18 stabilize the active site conformation. Shown here is the S. cerevisiae C* complex (Yan et al. 2017). (E) The splicing factor Cwc24 protects the guanine nucleotide at the 5′-end of the 5′SS in the S. cerevisiae Bact complex (Yan et al. 2016). Two aromatic residues Tyr155 and Phe161 sandwich the guanine nucleobase. (F) The splicing factor Cwc21 stabilizes the binding of the 5′-exon to loop I of U5 snRNA. Cwc21 simultaneously interacts with nucleotides of the 5′-exon and the switch loop of Prp8 in the Bact through P complexes. The interactions, shown here for the S. cerevisiae Bact complex (Yan et al. 2016), are identically preserved in the C, C*, and P complexes. (G) The 1585-loop of Prp8 in the S. cerevisiae Bact (Yan et al. 2016) (left panel) and P (Bai et al. 2017) (right panel) complexes. (H) The β-finger (from the RNaseH-like domain of Prp8) in the S. cerevisiae C (Wan et al. 2016a) (left panel) and P (Bai et al. 2017) (right panel) complexes.

The splicing factors constitute three classes: those that are specific for the branching reaction (step I factors) or the exon ligation (step II factors), and those that are generally required for splicing. The third class is further divided into two categories: those that are required for the early phase of active site formation (Cwc24 and Cwc27) and those that are present throughout both steps of the transesterification (Cwc21, Cwc22).

The step I factors Cwc25 and Yju2 play an indispensable role in stabilizing the active site RNA conformation (Chiu et al. 2009; Warkocki et al. 2009; Schneider et al. 2015), as observed in the cryo-EM structure of the S. cerevisiae C complex (Fig. 4C) (Galej et al. 2016; Wan et al. 2016a). Notably, for both Cwc25 and Yju2, the amino-terminal fragment is positioned close to the center of the active site, with conserved residues directly contacting the BPS/U2 duplex. Cwc25 also binds to the helix I of the U2/U6 duplex and the ISL. Yju2 interacts with the 5′-exon, Isy1, and Prp8. These elaborate interactions render Cwc25 and Yju2 integral parts of the splicing active site composite.

The step II factors Prp17, Prp18, and Slu7 play an important role in exon ligation (Jones et al. 1995; Aronova et al. 2007), although they appear to be less critically involved in the splicing active site compared with the step I factors. Indeed, the absence of Prp18 or Slu7 still allowed splicing of some pre-mRNA to proceed when the RNA sequence between the BPS and the 3′SS is shortened (Brys and Schwer 1996; Zhang and Schwer 1997; James et al. 2002). Nonetheless, these factors are located close to the splicing active site and may facilitate exon ligation through interactions with the protein and RNA elements. In the S. cerevisiae C* (Fica et al. 2017; Yan et al. 2017) and P (Bai et al. 2017; Liu et al. 2017; Wilkinson et al. 2017) complexes, the WD40 domain of Prp17 is located between the BPS/U2 duplex and the 5′SS/U6 duplex (Fig. 4D). Prp18 makes no physical interactions with the RNA elements but directly contacts the RNaseH-like domain of Prp8. The RNaseH-like domain, in turn, interacts with the BPS/U2 duplex in a nearly identical fashion in the spliceosomal C, C*, and P complexes (Galej et al. 2016; Wan et al. 2016a; Bai et al. 2017; Fica et al. 2017; Liu et al. 2017; Wilkinson et al. 2017; Yan et al. 2017). Both the RNaseH-like domain of Prp8 and the WD40 domain of Prp17 undergo marked translocation in the C-to-C* transition. Only a small portion of Slu7 was tentatively identified in the S. cerevisiae C* complex (Fica et al. 2017), although the bulk of this protein has been assigned in the human C* complex (Zhang et al. 2017).

The splicing factor Cwc24, which is recruited into the spliceosome in the B-to-Bact transition, plays an important role in the Bact complex by insulating the 5′-end guanine nucleobase of the 5′SS (Yan et al. 2016). The aromatic side chains of Tyr155 and Phe161 of Cwc24 sandwich the guanine nucleobase (Fig. 4E), whereas Lys160 mediate hydrogen bonds (H-bonds) to the nucleobase and the ribose. Through the action of Prp2, Cwc24 is dissociated in the Bact-to-B* transition. The splicing factors Cwc21 and Cwc22 are recruited into the Bact complex and dissociated in the P-to-ILS transition. The amino-terminal fragment of Cwc21 directly interacts with the 5′-exon to stabilize its binding to loop I of U5 snRNA (Fig. 4F). Cwc21, in turn, is stabilized by the switch loop of Prp8. The 5′-exon–Cwc21–switch loop interactions remain unchanged in the spliceosomal Bact through P complexes. Cwc22 is located close to Cwc21 and interacts with Prp8, whereas the splicing factor Cwc27 binds both Prp8 and Brr2 and is present only in the Bact complex.

Three prominent structural motifs, all from Prp8, appear to facilitate the splicing reaction through close interactions with the active site elements. These motifs include the switch loop and the 1585-loop (also known as α-finger; Nguyen et al. 2016) of the linker domain and the β-finger (Yang et al. 2008) of the RNaseH-like domain. The switch loop, which comprises an extended β-hairpin-containing structure, stabilizes 5′-exon binding. In all spliceosomes (Bact, C, C*, and P) in which the 5′-exon is loaded onto loop I of U5 snRNP, the switch loop is identically arranged (Fig. 4F). Compared with these spliceosomes, the switch loop is flipped by 180° in all structurally characterized complexes in which the interaction of the 5′-exon with U5 loop I is absent, including in the U4/U6.U5 tri-snRNP (Nguyen et al. 2016; Wan et al. 2016b) and in the B (Plaschka et al. 2017) and ILS (Hang et al. 2015; Yan et al. 2015; Wan et al. 2017) complexes.

In the Bact complex (Yan et al. 2016), the 1585-loop interacts with the U2/U6 duplex and binds the structural elements of Cwc24 and Prp11 that together protect the 5′-end guanine nucleotide of the 5′SS (Fig. 4G, left panel). In the C* (Fica et al. 2017; Yan et al. 2017) and P (Bai et al. 2017; Liu et al. 2017; Wilkinson et al. 2017) complexes, the tip of the 1585-loop is located close to the 5′-exon and the catalytic metals, contacting the ISL and the lariat junction (Fig. 4G, right panel). In addition, the 1585-loop directly interacts with the pre-mRNA nucleotides of the 3′SS and two nucleobases of the 3′-exon (Bai et al. 2017; Liu et al. 2017; Wilkinson et al. 2017). At the end of the splicing cycle, the 1585-loop is no longer needed and becomes disordered in the ILS complex (Hang et al. 2015; Yan et al. 2015; Wan et al. 2017).

In the Bact (Rauhut et al. 2016; Yan et al. 2016) and ILS (Hang et al. 2015; Yan et al. 2015; Bai et al. 2017) complexes, the β-finger remains unengaged, not interacting with any protein or RNA element. In the C (Galej et al. 2016; Wan et al. 2016a) and presumably B* complexes, the β-finger interacts with the BPS/U2 duplex (Fig. 4H, left panel). In the C* (Fica et al. 2017; Yan et al. 2017) and P (Bai et al. 2017; Liu et al. 2017; Wilkinson et al. 2017) complexes, the β-finger interacts with the lariat junction on the opposite side of the 1585-loop, located between the U2/BPS and U6/5′SS duplexes (Fig. 4H, right panel), and contacts the step II splicing factor Prp17.

The cryo-EM structures of the various spliceosomes reveal mechanistic insights into the B-to-Bact, Bact-to-B*, C-to-C*, and P-to-ILS transitions. Except for the U5 snRNP, nearly all other protein and RNA components of the B complex undergo marked changes—either dissociation or translocation and conformational change—in the transition to the Bact complex. This transition is mediated by Brr2, which is thought to bind the single-stranded U4 snRNA sequences downstream from the U4/U6 duplex (Fig. 5A). Propelled by ATP binding and hydrolysis, Brr2 presumably pulls on the single-stranded RNA sequence (Plaschka et al. 2017), leading to the dissociation of the associated protein factors and unwinding of the U4/U6 duplex. The B-to-Bact transition is speculated to occur in several distinct steps (Plaschka et al. 2017; Sidarovich et al. 2017; Zhang et al. 2018).

Spliceosomal transitions are driven by the ATPase/helicases. (A) The B-to-Bact transition is propelled by Brr2. Shown here is a surface view of the Saccharomyces cerevisiae B complex (Plaschka et al. 2017). Brr2 is thought to pull on the single-stranded U4 small nuclear RNA (snRNA) sequences, triggering the unwinding of the U4/U6 duplex and subsequent changes. (B) The Bact-to-B* transition is mediated by Prp2. Shown here is a surface view of the S. cerevisiae Bact complex (Yan et al. 2016). Prp2 likely pulls on the 3′-end sequences of the pre-mRNA, leading to dissociation of the retention and splicing complex (RES) complex, the SF3b and SF3a complexes, and the splicing factor Cwc24. (C) The C-to-C* transition is executed by Prp16. Shown here is a surface view of the S. cerevisiae C complex (Wan et al. 2016a). Prp16 presumably pulls on the 3′-end sequences of the intron lariat-3′-exon intermediate. (D) The P-to-intron lariat spliceosome (ILS) transition is driven by Prp22. Shown here is a surface view of the S. cerevisiae P complex (Bai et al. 2017). Dissociation of the ligated exon is mediated by Prp22 pulling on its 3′-end sequences. (E) The disassembly of the ILS complex is performed by Prp43. Shown here is a surface view of the S. cerevisiae ILS complex (Wan et al. 2017). Prp43 may function by binding to and pulling on either the 3′ end of U6 snRNA or the intron lariat. The PDB codes for the B, Bact, C, P, and ILS complexes are 5NRL, 5GM6, 5GMK/5LJ5, 5YLZ, and 5Y88, respectively.

At the periphery of the Bact complex (Yan et al. 2016), Prp2 binds the SF3b protein Hsh155 and is located close to the RES complex (Fig. 5B). The RNA sequences downstream from the BPS are bound by the RES complex, with the disordered 3′-end sequences presumably recognized by Prp2 (Liu and Cheng 2012; Schneider et al. 2015). This structural arrangement is fully consistent with the notion that pulling on the 3′-end intron sequences may result in the dissociation of the RES complex and the SF3b complex during the Bact-to-B* transition. In the C complex (Galej et al. 2016), the RNA-binding site of Prp16 is about 70 Å away from the structurally identified nucleotide at the 3′ end of the intron (Fig. 5C). In the C-to-C* transition, Prp16 would presumably pull on the 3′-end sequences of the intron, leading to the dissociation of the step I factors Cwc25 and the amino-terminal fragment of Yju2 (Tseng et al. 2011) and subsequent recruitment of the step II factors Prp18 and Slu7 (Ansari and Schwer 1995; James et al. 2002). In the P complex (Fig. 5D), Prp22 directly interacts with the Linker and the RT Finger/Palm domains of Prp8; but the RNA-binding site of Prp22 is ∼80 Å away from the structurally identified 3′-end nucleotide of the 3′-exon. This structural feature explains why the release of the ligated exon from the P complex is inhibited by shortening the 3′-exon sequence (Schwer 2008; Ilagan et al. 2013). In the P-to-ILS transition, Prp22 is thought to pull on the 3′-end RNA sequences to dissociate the ligated exon (Company et al. 1991; Schwer and Gross 1998; Semlow et al. 2016), triggering the dissociation of Cwc21, Cwc22, Prp18, Slu7, and the WD40 domain of Prp17 and subsequent recruitment of Cwc23 and the disassembly Ntr complex (Prp43, Ntr1, and Ntr2).

The disassembly of the spliceosome is mediated by the Ntr complex (Arenas and Abelson 1997; Tsai et al. 2005, 2007; Fourmann et al. 2016). The structure of the ILS complex suggests two potential mechanisms by which the spliceosome is disassembled by the ATPase/helicase Prp43 (Fig. 5E) (Wan et al. 2017). Prp43 binds the middle portion of the superhelical protein Syf1, whereas Ntr1 and Ntr2 are located away from Prp43. The 3′-end nucleotide of U6 snRNA that has been structurally identified is located only ∼50 Å away from the RNA-binding site of Prp43. In contrast, the intron lariat is located 85–100 Å away. Therefore, it appears more likely that Prp43 grabs onto and pulls on the 3′-end of U6 snRNA as opposed to the intron lariat (Bohnsack et al. 2009).

These transitions are, by definition, dynamic and can be inferred only from comparative analysis of the static cryo-EM structures. The steps and mechanisms separating two neighboring spliceosomal complexes could be considerably more dynamic than the qualitative comparison. For example, the human Bact complex has at least three compositional states—early, mature, and late (Zhang et al. 2018)—and the mature state comprises at least eight distinct conformational states (Haselbach et al. 2018). Importantly, however, the essential features of the various transitions have been structurally observed. These features to a large extent confirm published biochemical and genetic data that have accumulated in the past three decades. Despite the tantalizing interpretation of the structural images, how the ATPase/helicases function mechanically remain to be elucidated, perhaps through single-molecule biophysical investigations. The need for in-depth biophysical studies is reinforced by the fact that these ATPase/helicases are located mostly in the periphery of the spliceosomes and are characterized by medium-to-low resolution, which does not allow accurate assignment of atomic features that are required for faithful mechanistic understanding.

The conserved sequence motifs of the pre-mRNA are mainly recognized by conserved RNA elements from both snRNA and pre-mRNA. Three nucleotides at the 3′-end of the 5′-exon are anchored on the loop I of U5 snRNA in the Bact through P complexes. The BPS and surrounding sequences of the intron form an elongated duplex with complementary sequences of the U2 snRNA in the A through ILS complexes. The dinucleotides GU at the 5′ end of the 5′SS are recognized by the Linker domain of Prp8 in the S. cerevisiae U4/U6.U5 tri-snRNP (Wan et al. 2016b) and the human B complex (Bertram et al. 2017a) and by Cwc24 in the Bact complex (Yan et al. 2016); the six consecutive nucleotides downstream from GU of the 5′SS are recognized by U6 snRNA in the Bact through ILS complexes. In the P (Bai et al. 2017; Liu et al. 2017; Wilkinson et al. 2017) and possibly C* complexes, the dinucleotides AG of the 3′SS are recognized by the lariat junction that is formed after the branching reaction (Fig. 6A). Notably, AG of the 3′SS pair up with the nucleophile-containing A of the BPS and the invariant G at the 5′-end of the 5′SS, respectively, through noncanonical Watson–Crick H-bonds.

Choreography of the protein and RNA components during the splicing reaction. (A) Recognition of the AG dinucleotide at the 3′SS by the lariat junction in the P complex (Bai et al. 2017; Liu et al. 2017; Wilkinson et al. 2017). Notably, the AG nucleobases at the 3′SS form noncanonical Watson–Crick base-pairing interactions with two consecutive AG nucleobases of the lariat junction. (B) Recognition of the nucleophile-containing adenine nucleotide of the branch point sequence (BPS) by Hsh155. The adenine nucleobase is already flipped out of the registry of the intron/U2 duplex and is in close contact with a few residues of Hsh155 and Rds3. (C) A cartoon diagram depicting the activation of the Bact complex, the branching and exon ligation, and the disassembly of the intron lariat spliceosome (ILS) complex.

The spliceosome is a protein-directed metalloribozyme. The RNA elements, both the immobile portion (U5, U6, the 5′-end portion of U2, and the 3′-end portion of the 5′-exon) and the mobile segments (U2 downstream from nucleotide 30, the bulk of the intron, and the 3′-exon), are specifically recognized by protein components. This subject has been thoroughly reviewed recently (Shi 2017a) and is thus only briefly discussed here. The immobile portion is mainly recognized by the 17 static protein components. For example, the active site RNA elements are cradled in the positively charged catalytic cavity that is formed between the N-domain and the core of the Prp8 (Fig. 2D). The mobile RNA segments are recognized and directed by conserved protein elements. For example, in the Bact complex (Rauhut et al. 2016; Yan et al. 2016), the BPS/U2 duplex is surrounded by Hsh155 and the nucleophile-containing adenine nucleobase of the BPS is flipped out of the U2/BPS duplex registry and recognized by specific residues of Hsh155 (Fig. 6B). The 5′-end guanine nucleotide of the 5′SS is recognized by the splicing factor Cwc24 (Fig. 4E) and the SF3a protein Prp11.

Three snRNA molecules, U5, U6, and 30 nucleotides at the 5′-end of U2, remain largely static in the spliceosomal Bact through ILS complexes, with only minor shift (mostly