U5 snRNA Interactions With Exons Ensure Splicing Precision

Abstract

Imperfect conservation of human pre-mRNA splice sites is necessary to produce alternative isoforms. This flexibility is combined with the precision of the message reading frame. Apart from intron-termini GU_AG and the branchpoint A, the most conserved are the exon-end guanine and +5G of the intron start. Association between these guanines cannot be explained solely by base-pairing with U1 snRNA in the early spliceosome complex. U6 succeeds U1 and pairs +5G in the pre-catalytic spliceosome, while U5 binds the exon end. Current U5 snRNA reconstructions by CryoEM cannot explain the conservation of the exon-end G. Conversely, human mutation analyses show that guanines of both exon termini can suppress splicing mutations. Our U5 hypothesis explains the mechanism of splicing precision and the role of these conserved guanines in the pre-catalytic spliceosome. We propose: (1) optimal binding register for human exons and U5-the exon junction positioned at U5Loop1 C₃₉|C₃₈; (2) common mechanism for base-pairing of human U5 snRNA with diverse exons and bacterial Ll.LtrB intron with new loci in retrotransposition-guided by base pair geometry; and (3) U5 plays a significant role in specific exon recognition in the pre-catalytic spliceosome. Statistical analyses showed increased U5 Watson-Crick pairs with the 5'exon in the absence of +5G at the intron start. In 5'exon positions -3 and -5, this effect is specific to U5 snRNA rather than U1 snRNA of the early spliceosome. Increased U5 Watson-Crick pairs with 3'exon position +1 coincide with substitutions of the conserved -3C at the intron 3'end. Based on mutation and X-ray evidence, we propose that -3C pairs with U2 G₃₁ juxtaposing the branchpoint and the 3'intron end. The intron-termini pair, formed in the pre-catalytic spliceosome to be ready for transition after branching, and the early involvement of the 3'intron end ensure that the 3'exon contacts U5 in the pre-catalytic complex. We suggest that splicing precision is safeguarded cooperatively by U5, U6, and U2 snRNAs that stabilize the pre-catalytic complex by Watson-Crick base pairing. In addition, our new U5 model explains the splicing effect of exon-start +1G mutations: U5 Watson-Crick pairs with exon +2C/+3G strongly promote exon inclusion. We discuss potential applications for snRNA therapeutics and gene repair by reverse splicing.

Keywords: RNA base pair geometry; U1 snRNA; U2 snRNA; U5 snRNA; U6 snRNA; group II intron retrotransposition; splice sites; splicing mutations.

FIGURE 1

Multistage recognition of variable human splice sites by U snRNAs. (A) Only seven nucleotides in human introns are conserved above 75% (Sheth et al., 2006; Mercer et al., 2015). Apart from the terminal di-nucleotides and the branchpoint A, the most important are the two guanines of the exon end and at position +5 at the intron start (orange labels and arrows). (B) In the early spliceosome, U1 snRNA forms on average seven Watson–Crick pairs with human exon/intron boundaries (Carmel et al., 2004). (C) In the pre-spliceosome, U2 snRNA forms the BP helix with an adenosine bulge. (?) Proposed U2 G₃₁ = C_–3 pair, see DISCUSSION. (D) In the pre-catalytic spliceosome, U1 quits the complex and the start of the intron is passed on to U6 snRNA. +5G pairs with U6 42G (orange arrow). The conserved adenines +3, +4 form non-Watson–Crick pairs with U6 44G and 43A^m6 (Konarska et al., 2006; Galej et al., 2016; shown here in red according to Westhof geometric classification: 10th family, Leontis et al., 2002; role explained in Figure 14 caption). The stable U6/start of the intron helix is a checkpoint for the later spliceosome activation by Brr2 helicase (binding site on U4: blue oval). Brr2 unwinds U6/U4 duplexes and frees U6 to configure the catalytic site of the spliceosome (Nielsen and Staley, 2012). (E) The strictly conserved non-canonical G⋅⋅G (2nd Westhof geometric family, see DISCUSSION, Scadden and Smith, 1995; Costa et al., 2016). (F) At the pre-catalytic stage, U5 snRNA comes into the complex together with U6 as part of U5⋅U4/U6 tri-snRNP (Wahl et al., 2009; Wahl and Lührmann, 2015; Scheres and Nagai, 2017; Supplementary Table S1). As U1 quits the complex, the 5′exon is passed on to U5 snRNA Loop1. For the 3′exon, see DISCUSSION. Aligned together, the exons form the splice junction consensus AG|G (proto-splice site, Sverdlov et al., 2004) pictured here paired with complementary C₃₈C₃₉U₄₀ of the U5 Loop1. In this way, the most conserved exon-end G pairs with U5 39C (orange arrow). If so, in the pre-catalytic spliceosome, the intron-termini pair and U6 non-Watson–Crick pairs are stabilized by flanking U5/5′exon and U6/intron-start helices, each secured by one of the two important guanines of the human splice signals (orange arrows in D,F). Post-transcriptional base modifications of snRNAs: Ψ, pseudouridine; Superscript m, 2’O-methyl; A^m6, N6-methyladenosine; A^m6_m, 2’O-methyl,N6-methyladenosine (modification positions as in Anokhina et al., 2013).

FIGURE 2

Base pair composition for the target recognition by mobile Ll.LtrB intron helps to identify the binding register for the human exons and U5 snRNA. (A) The 11nt Id3 Loop in Domain I (DI) is the element of the Ll.LtrB intron responsible for the specific recognition of the exons. C = G pairs with guanines in Id3 positions 278 and 279 coordinate the splice junction. The Id3 loop of the excised intron can pair with genomic targets similar to the homing site and guide retrotransposition. We derived a hypothetical consensus for retrotransposition sites based on the complementarity to the Id3 Loop. The homing site (HS) in the ltrB gene differs from this consensus in the 5′ exon position −4 (thymidine). EBS1: Seven nucleotides of the Id3 loop (positions 279–285) pair with the end of the 5′ exon. δ: The remaining four nucleotides (positions 275–278) form a helix with the start of the 3′ exon. (B) Assuming that, like in Group IIA introns, Watson–Crick pairs are preferred, we derived a hypothetical consensus complementary to U5 Loop1. The actual human splice junction consensus AG|G (Figure 1F) appears incorporated into this hypothetical sequence and G = C pairs with cytosines in U5 positions 38 and 39 coordinate the splice junction (orange arrow: U5 39C pairs the conserved exon-end G). (C) We derived a grid for manual alignment of the retrotransposition sites with the Id3 loop that listed Watson–Crick and frequent mismatched pairs. In this way, we recorded base pairs involved in the recognition by the LtrB intron of 31 targets in the L. lactis genome (Ichiyanagi et al., 2002; one example is shown here). (D) Assuming that U5 snRNA Loop1 has the same base pair preferences as the Ll.LtrB Id3 Loop and that the 5′ exon helix is longer than the 3′ exon helix, we superimposed each of the 78 dystrophin gene splice junctions on the grid in the five possible binding registers (as in the example here). Alignments with most Watson–Crick pairs were chosen as most likely with 65% of dystrophin exon junctions unambiguously aligned to U5 C₃₈| C₃₉ (as in B,D) and further 30% also fit this and one or two alternative binding registers. (E) Summary for the Ll.LtrB Id3 Loop of the total n = 341 bp with 31 retrotransposition sites (Ichiyanagi et al., 2002). (F) Summary for the human U5 snRNA Loop 1 of the total n = 561 bp with only 51 dystrophin splice junctions that unambiguously aligned with U5 C₃₈|C₃₉. Base modifications as in Figure 1 caption.

FIGURE 3

U5 and U6 recognize variable human exon/intron boundaries by Watson–Crick base pairing at the pre-catalytic stage and cooperatively ensure splicing fidelity. (A,C) In the early spliceosome (complex E), U1 snRNA forms, on average, seven (minimum five) Watson–Crick pairs with the exon/intron boundary (Ketterling et al., 1999; Carmel et al., 2004). U1 can bind multiple alternative and cryptic targets (Eperon et al., 1993, 2000) and is known to initiate correct splicing when bound in the vicinity rather than at the actual exon/intron boundary (Fernandez Alanis et al., 2012; Singh and Singh, 2019), presumably leaving the fidelity check for the next stage. (B,D) In the pre-catalytic spliceosome (complex B), U1 is replaced by U5 snRNA at the exon end (pink boxes). U6 snRNA replaces U1 at the intron start (yellow boxes). The well-conserved adenines at intron positions +3 and +4 are enclosed in a red dashed box. Initially, these adenines pair with U1 pseudouridines 5 and 6, and then in the pre-catalytic complex, they form non-canonical pairs with U6 A^m6₄₃G₄₄ (Figure 1D). Intron termini pair (purple box): see Figure 1E and DISCUSSION. (C) Lack of exon complementarity to U5 is compensated by a strong intron interaction with U6 (as an example: human dystrophin intron 1). (D) Vice versa, lack of intron complementarity to U6 is compensated by U5 Watson–Crick pairs with the 5′exon: (as an example: dystrophin intron 34). Base modifications as in Figure 1.

FIGURE 4

Additional U5 Watson–Crick pairs with the 5′exon compensate for substitutions of the conserved +5G at the start of the intron. (A) Schematic of the interactions with U5 and U6 snRNAs (using human dystrophin splice junction of exons 8 and 9 and the start of intron 8 as an example). The 11 base pairs of the U5 interaction with the exons are here subdivided into four groups that correspond to sKL distributions (D–G). (B) Base pair frequencies for 445 human splice junctions related to introns lacking +5G (+5Gsub). (C) Base pair composition for 1,545 splice junctions of introns with conserved +5G. (D–G) Distributions of symmetrized Kullback–Leibler divergences (sKL) for each positional group between the +5Gsub and 10,000 random non-redundant +5G sets of n = 445 (orange histograms) and control distributions of random non-overlapping and non-redundant sets of the same size from within the +5G dataset (blue histograms).

FIGURE 5

Reciprocal effects of +5G intron substitutions on the 5′ exon base pairs with U5 and the exon-end G substitutions on the intron base pairs with U6. Distributions of 10,000 bootstrap differences for the frequency of Watson–Crick, isosteric, and non-isosteric base pairs at each position of the exon junction (U5 biding site) between junctions flanking introns that conserve +5G and those where this base is substituted (A–C). Differences for the bp frequencies of the U6 binding site at the start of the intron positions +5 to +10 between introns preceded by exons that conserve exon-end G (−1G) and those that do not (D–F). The null hypothesis probability, P(H₀), of no difference is indicated above each violin, and asterisks mark significant changes after the correction for multiple testing (see Materials and Methods for details).

FIGURE 6

The +5G effect is specific to U5 snRNA, not U1 snRNA. The bar chart on the left shows the distribution of nucleotides for the 5′exon position −3 in the presence and absence of the conserved +5G at the start of the intron. The schematics show the different nucleotide preferences for U1 and U5 paired to exon position −3 (indicated in red). Histograms for distributions of 10,000 bootstrap differences for the frequency of each nucleotide at position −3 show that significant increase for A_–3, creating a Watson–Crick pair in the U5 interaction, but not C_–3 compensates for the loss of the U6 C₄₂ = G₊₅ pair.

FIGURE 7

Additional U6 Watson–Crick pairs with the start of the intron at positions +5 to +8 compensate for substitutions of the conserved exon-end G. (A) Schematic of the examined interactions with U6 and U5 snRNAs (using human dystrophin intron 64 and the splice junction of exons 64 and 65 as an example). The U6 interaction with the intron positions +5 to +10 is subdivided into dinucleotides that correspond to sKL distributions (D–F). (B) Base pair type frequencies for 392 human introns preceded by exons lacking −1G (−1Gsub). (C) Base pair composition for 1,598 introns preceded by exons with conserved −1G. (D–F) Distribution of symmetrized Kullback–Leibler divergences (sKL) for each dinucleotide between the −1Gsub dataset and 10,000 random non-redundant −1G sets of n = 392 (orange histograms) compared to a control distribution of the same size between random non-overlapping and non-redundant subsets from the −1G dataset (blue histograms).

FIGURE 8

Additional U5 Watson–Crick pairs with the 3′ exon compensate for substitutions of the conserved −3C at the end of the intron. (A) Schematic of the examined interactions with U5 and U2 snRNAs (using human dystrophin splice junction of exons 71 and 72 and intron 71 as an example). We propose (see DISCUSSION) that BP helix bridges across the PPT to the end of the intron secured by the conserved −3C paired with U2 G₃₁. This agrees with the previous mutation evidence (Brock et al., 2008; Corrionero et al., 2011), biochemical studies (Kent et al., 2003; Chen et al., 2010), and X-ray structures (Kent et al., 2003; Sickmier et al., 2006). The example here shows the U2 G₃₁--U_–3 pair instead, as dystrophin intron 71 lacks −3C. (B) Base pair-type frequencies for 792 human splice junctions related to introns lacking −3C (−3Csub). (C) Base pair-type composition for 1,211 splice junctions of introns with conserved −3C. (D–F) Distributions of 10,000 bootstrap differences for the frequency of Watson–Crick, isosteric, and non-isosteric base pairs at each position of the splice junction (U5 binding site). The null hypothesis probability, P(H₀), of no difference between the two cases is indicated above each violin; asterisks mark significant changes after correction for multiple testing (see Materials and Methods for details).

FIGURE 9

U5 Watson–Crick pairs with exon positions +2 and +3 promote inclusion of exons with +1G mutations. Fu et al. (2011) quantified the variable effect on splicing for 14 exon-start +1G mutations in human genes. (A) Boxplots show that exon inclusion (PSI, percent spliced-in) is strongly influenced by the presence of exon +2C or +3G. (B) Splice site sequences of the PKHD gene intron 24 and the flanking exons. +1G mutation in exon 25 completely blocks normal 3′ splice site and activates a cryptic 3′ss tag|ACG leading to the predominant inclusion of a longer exon and only 1% exon-skipping. (C) Base pairing scheme for the cryptic 3′ss with U5 snRNA Loop1 secured by U5 G^m₃₇ = C₊₂ and C₃₆ = G₊₃ according to our new U5 model. (D) Base pairing of the normal wt exon 25 with U5 snRNA. +1G mutation abolishes the U5 C₃₈ = G₊₁ pair, which leads to exclusive use of the upstream cryptic 3′ss. (C,D) Recognition of all splice sites is complete in the pre-catalytic spliceosome (complex B) before Brr2 promotes catalytic core formation (see DISCUSSION).

FIGURE 10

RNA network of the homologous ribozymes: human major and minor spliceosomes and Group IIA intron. First catalytic step, spliceosomal complex C (successive spliceosome complexes are detailed in Supplementary Table S1). The nucleophilic attack by the BP adenosine: curved red arrow. The intron breaks off the 5′exon end and bonds with the 2’O of the branching A: double purple dashed lines indicate the scissile (purple fill) and emergent (no fill) covalent bonds. Splicing catalysis requires two Mg²⁺ ions at a fixed distance from three reactant sites (Steitz and Steitz, 1993; Fica et al., 2013). At the first catalytic step, Mg²⁺(1) activates 2’OH of the BP A in the Reactant site 1. Mg²⁺(2) stabilizes the leaving 3’OH of the last nucleotide of the 5′ exon in Reactant Site 3. Both magnesium ions form a complex with the scissile phosphate of the N₊₁ of the intron in Reactant site 2. (A) Human major spliceosome (intron 12 of the dystrophin gene as an example). The ribozyme is an assembly of three separate snRNAs with a record number of modified residues. The structure of the U6/U2 catalytic triplex is inferred from Keating et al. (2010) and Galej et al. (2016) and the U6/intron duplex as in Fica et al. (2017), see also Supplementary Table S3. Non-canonical RNA pairs are shown with Westhof geometric symbols (Leontis et al., 2002). Tertiary interactions as in Anokhina et al. (2013): blue dashed lines; Mimic Watson–Crick-like base pairing: black dashed lines; Base pairing with unknown non-Watson–Crick geometry: double dots. Base modifications as in Figure 1 and G^m2: N2-methylguanosine. (B) LtrB Group IIA intron (Lactococcus lactis). Interactions of the catalytic triplex are extrapolated from the O.i. structure (Keating et al., 2010). Core motifs of this large RNA molecule are colored as homologous RNA components of the spliceosome. Greek letters: tertiary interactions in Group II introns, shown in blue. γ–γ’ and λ–λ’ interactions do not have homologs in the spliceosome. All other core interactions and catalytic structures of the ribozyme are labeled with spliceosome homologs in italics. Domains of the Ll.LtrB ribozyme: DI-DVI; Junctions between domains II and III or V and VI: J2/3 or J5/6. Double numbering is used for the residues starting from domain V, the negative number indicating the position from the 3′ end. (C) Human minor spliceosome (Tarn and Steitz, 1996a, ; Widmark et al., 2010; Edery et al., 2011; He et al., 2011; Younis et al., 2013; reviewed in Turunen et al., 2013). U5 is the only snRNA shared with the major spliceosome. A lot fewer residues are modified in U12 and U6atac snRNAs compared to U2 and U6 paralogs. Perfect conservation of the BP helix and the U6atac snRNA AAGGAGAGA box interaction with the 5′ intron end is characteristic of the minor spliceosome. (?): an odd U12 C₄ bulge (see Supplementary Comment S1) here reproduced as in Tarn and Steitz (1996b) and Turunen et al. (2013). The minor introns are expression regulators of critical genes: the example here is intron 6 of the human Nucleolar Protein 1 (NOL1/NSUN1; Brock et al., 2008) gene encoding an RNA:5-methylcytosine-methyltransferase (known as proliferation antigen p120 overexpressed in virtually all types of cancer cells).

FIGURE 11

Watson–Crick-like geometry of G--U, A--C, C--U, and U--U pairs is supported by rare tautomerization and protonation. (A) Canonical Watson–Crick G = C and A-U pairs. (B) Predicted in the 1950s (Watson and Crick, 1953) and confirmed in 2011 by X-ray structures (Bebenek et al., 2011; Wang et al., 2011) Watson–Crick-like (isosteric to canonical) G--U pairs with either base in enol configuration and A--C pairs with imino tautomers of adenine or cytosine. Watson–Crick-like G--U is a high-frequency pair (NMR, Kimsey et al., 2015), which reflects the ease of proton repositioning provoked by the oxygen of the carbonyls. (C) Watson–Crick-like U--C and U--U pairs were reported by Rypniewski et al. (2016) in the XR structures of CCUG repeats, associated with the molecular pathology of myotonic dystrophy type 2. Possible configurations of U--C pairs are as in Rypniewski et al. (2016). The only configurations of the U--U pair that abolish the repulsion between the carbonyls and fit the reported structure (Rypniewski et al., 2016) are suggested here. Watson–Crick-like C--C pairs have not been reported; theoretical configuration requires imino tautomerization of one cytosine and protonation of the other (4-imino-C)--(2-enol-C⁺). Imino tautomerization is more difficult compared to enol, as the proton movement is between the two nitrogens.

FIGURE 12

Our new U5 Loop1 interactions model compared to the current Cryo-EM model. The exon junction logo of the +5Gsub group (frequencies as in Figure 4B) reflects that substitutions of the conserved +5G in human introns are associated with the significant increase in −5A, −3A, −2A, and −1G in the 5′exon (Exon 1) and no changes in the sequence of the 3′exon (Exon 2)—compare to the exon junction logo for all human introns in Figure 1F. Here, we fit this +5Gsub exon junction logo alternatively to our new U5 model and the current CryoEM model and argue that the new model is a better match. (A) Our U5 Loop1 model is based on the initial alignment of human splice junctions with U5 Loop1 in parallel with the alignment of bacterial retrotransposition sites with the homologous Ll.LtrB Id3 loop. According to our model, substitutions of the conserved +5G in human introns are compensated by the additional Watson–Crick pairs with U5 Loop1 in the 5′ exon positions −1, −2, −3, and −5. In addition, our model explains the effect of mutations of exon-start G (Fu et al., 2011) by Watson–Crick base pairing of exon positions +2 and +3 with U5 Loop1 C₃₆G^m₃₇ (Figure 9). The intron termini pair is shown in the configuration of the second Westhof geometric family in agreement with the previous mutation analyses (Scadden and Smith, 1995). This pair must be formed in the pre-catalytic spliceosome (complex B) to play a central role at the transition stage (complex C*). The intron termini pair brings the 3′ Exon 2 in contact with U5 Loop1 in the pre-catalytic spliceosome (see DISCUSSION). (B) The CryoEM model for U5 that currently prevails features a 7nt Loop1 and places the 5′ exon paired with U5 U₄₀U^m₄₁U₄₂ in the pre-catalytic complex B (Zhang et al., 2017, 2018, 2019). This eliminates the energy benefit of the G = C pair for the 82% conserved intron-end G. Accordingly, +5G substitutions are only supported by the increase in A-U pairs in exon positions +2 and +3. The intron termini pair was captured only in the post-catalytic spliceosome (complex P, Zhang et al., 2017), although the authors suggest that it must be present at the transition stage (complex C*). The configuration of this pair corresponds to the third Westhof geometric family, which is not consistent with the previous mutation analysis, as the covariant A⋅⋅C pair or compensatory A⋅⋅A and I⋅⋅I pairs are impossible in this configuration (Scadden and Smith, 1995, see DISCUSSION). Base pairing for the 3′ exon is still not resolved (Zhang et al., 2019). We placed 3′ exon aligned with only two possible unpaired positions of the U5 Loop1 (base pairing with question marks). However, exon +2C or +3G cannot form Watson–Crick pairs with U5 Loop1 in this binding register, so the fact that exon +2C/+3G promotes inclusion of exons with +1G mutations cannot be explained by the CryoEM model. The 7nt U5 Loop1 is too small to accommodate specific interactions with both exons. (C) Our de novo structural model of the U5 Loop1 duplex with the splice junction of exons. We used hypothetical exons complementary to U5 Loop1, while our comparison with the Ll.LtrB Id3 loop suggests that the real exon junctions form Watson–Crick-like pairs to fit diverse sequences and preserve the shape of the helix. Remarkably, a turn of the A-helix contains 11 bp, so the 11nt loop as shown here can well accommodate specific interactions with both exons. The U5 Loop1 helix appears hollow along the axis, which is typical of the A-helix (Heinemann and Roske, 2020).

FIGURE 13

Interactions of the intron termini are base pairs with parallel strand orientation. (A) The configuration of mammalian intron termini pairs was defined by mutation analyses (Scadden and Smith, 1995) as G⋅⋅G: N1, carbonyl symmetric, A⋅⋅C: reverse wobble, both corresponding to the second Westhof geometric family. CryoEM configuration corresponds to the third geometric family, which is impossible for A⋅⋅C pairs (for further explanation, see text). (B) Base pair configuration for Group II intron first and sub-ultimate nucleotides captured in the recent crystal structure (after Costa et al., 2016, confirmed by personal communication with Professor Eric Westhof) and shown here with additional hydrogen bonds formed by 2’O of the riboses (after Leontis et al., 2002). Parallel strand orientation is characteristic of these pairs. Ribose is located on the perpendicular plane and is shown as a schematic blue pentagon with + and o indicating the opposite directions of the sugar-phosphate backbone. For further explanation, see Supplementary Figure S3. The IsoDiscrepancy index is a numerical measure of geometric similarity (isostericity) of base pairs (online RNA base pair catalog, http://ndbserver.rutgers.edu/ndbmodule/services/BPCatalog/bpCatalog.html).

FIGURE 14

The central role of the intron termini pair in the transition between the two catalytic steps of splicing. The bond between the intron and the 5′exon is broken at the first catalytic step, and the formation of the covalent bond between the BP adenosine and the 5′ end of the intron (branching) triggers a rotation of the BP helix on its axis (Somarowthu et al., 2014; Bertram et al., 2017a). The correct repositioning of the branched intermediate absolutely requires U6 non-Watson–Crick pair(s) at the start of the intron at position +3 and/or +4 (Konarska et al., 2006). The momentum of the revolving BP helix is transmitted by the intron termini pair and drives the translocation of the 3′ exon. This movement is enabled by the relaxation of the U5 Loop 1 due to the disruption of the covalent bond between the intron and the 5′ exon. The overall configuration of the U6/U2 metal binding site stays unchanged, and the reacting residues are transitioned to the fixed Reactant sites (Steitz and Steitz, 1993; Fica et al., 2013; Semlow et al., 2016— Supplementary Comment S2). (A) RNA network of the first step spliceosome—as in Figure 10. (B) The RNA–RNA interactions at the moment of transition between the two steps of splicing: The biggest purple arrow indicates the repositioning (rotation) of the BP adenosine after branching, and the purple triple dashed line shows the transmission of the motion via the intron termini pair to the 3′ exon. Dashed purple arrows trace the movement of the residues out of the Reactant Sites 1 and 2, and the incoming nucleotides follow the path of continuous purple arrows. The position of the 5′ exon is unchanged in the Reactant Site 3. For the second reaction, the metal ions reverse their actions: the Mg²⁺ (2) activates the 3’OH group of the 5′ exon in Reactant Site 3, an attack is launched at the 5’PO₄ of the 3′ exon in Reactant Site 2, while Mg²⁺ (1) stabilizes the leaving 3’OH of the last nucleotide of the intron at Reactant Site 1. The fact that the two metal ions play the activation role in turn enables a single catalytic core to accommodate both steps of splicing and also implies the ease of the reverse process (Fica et al., 2013). Although the second reaction (curved red arrow 2) happens in the C* complex following the transition, here the emergent covalent bond is shown in the successive P complex. (C) RNA network after the second catalytic step. Double purple dashed lines indicate the emergent (purple fill) and previous (no fill) covalent bonds. The exons are ligated and are still paired with U5 Loop1. The intron lariat stays paired with U6 and U2 snRNAs (Zhang et al., 2019). (D) The joined exons are disassociated from U5 Loop1 by Prp22 (Wan et al., 2017; Zhang et al., 2019); without RNA partners, the loop changes to “closed” (7nt) conformation. The intron lariat stays paired to U2/U6 and associated with U5 snRNP. The ILS complex is homologous to Group II intron RNP ready for reverse splicing (RNA network dynamics of successive spliceosomal complexes is summarized in Supplementary Table S1).