Bulging is required for splicing, but the position of the bulge within the BS-U2 duplex is flexible
Our continued interest in the branch site-U2 snRNA duplex was spurred both by the observation of trans-
splicing to U2 in S. cerevisiae
(Smith et al., 2007
) and by the results of a screen for U2-based suppressors of branch site mutations. Because U2 is essential and sensitive to mutation in its BS-binding region, this mutagenesis screen and all subsequent experiments involving mutant U2 snRNAs were performed using a strain carrying both a wild-type (wt) copy of U2 and a second gene encoding the mutant. One U2 allele capable of suppressing the splicing defect of BS-C (5′-UACUAC
C-3′) was identified as Ψ35G (5′-GG
AGUA-3′). Other mutations at this position failed to suppress BS-C (), and other BS mutations could be suppressed by analogous mutations of position 35 to the Watson-Crick base-pairing partner of the mutated BS nucleotide (data not shown). This suggested that the splicing defect due to BS mutation was overcome via bulging and activation of the preceding adenosine – a conclusion confirmed by primer extension analysed at single nucleotide resolution (, middle panel). The ACT1-CUP1
reporter encodes a metallothionein whose expression correlates over a large dynamic range with resistance to copper in the growth medium (Lesser and Guthrie, 1993
). All primer extension data in this and subsequent experiments were confirmed by copper growth assays (, lower panel, and data not shown); this verifies that mRNA is released from the spliceosome, exported to the cytoplasm and translated.
The use of a positionally non-canonical BS nucleotide does not normally occur in S. cerevisiae
; our data suggested both that bulging of the BS was likely required for catalysis as well as assembly, and that S. cerevisiae
could be induced to display the same flexibility as higher eukaryotes with respect to branch nucleophile positioning. We introduced pairs of mutations into the BS and a second copy U2 snRNA such that the bulge was flattened by insertion or deletion of a nucleotide in the BS sequence, and restored by corresponding deletion or insertion in the second copy U2 (). As expected for an essential bulge, introns that generated flat duplexes failed to support splicing (, lanes 2&4), and restoration of the bulge via either insertion or deletion in U2 snRNA restored splicing activity, albeit with reduced efficiency relative to the wt BS-U2 pair (, lanes 3, 5&6). To distinguish between the assembly and catalytic phases, we assayed spliceosome assembly and splicing progression in vitro
using S. cerevisiae
whole cell extract and the efficiently spliced UBC4 transcript (John Abelson, personal communication) carrying a variety of mutations at and around the BS. Analysis of splicing reactions in native and denaturing polyacrylamide gels indicated that deletion of the branch nucleophile to produce a flat BS-U2 duplex allowed spliceosome assembly at a level comparable to that shown by an intron carrying a BS-C mutation: neither of these mutant introns, however, showed detectable splicing catalysis, in contrast to the efficient splicing of the wt UBC4 intron (). As expected, an intron with a severely mutated BS region failed to elicit detectable spliceosome assembly (). Further analysis of assembled spliceosomes, using low concentrations of EDTA to separate assembly intermediates (Cheng and Abelson, 1987
), indicated that the bulge-less transcript progressed at least to the stage of a U2/5/6 spliceosome – i.e. that U1 and U4 snRNAs can be lost to generate a normal assembly intermediate (Fig. S1a), and that a similar array of splicing complexes can form on wild-type and bulge-less UBC4 introns in the absence of Prp2, the ATPase whose action immediately precedes the catalytic phase of splicing (Fig. S1b&c). We can therefore conclude that the absence of a bulged branch nucleotide confers a post-assembly defect in addition to impairing spliceosome assembly, although the resolution of current assays does not allow us to conclude that the branch nucleophile is bulged at the time of catalysis per se
High-resolution primer extension of in vivo splicing indicated that the expected branch nucleophile was used for each BS-U2 pair that showed detectable splicing (, lower panel – the background band at the wt position is due to extension on endogenous actin), indicating that changing the length of the BS-U2 duplex, and correspondingly moving the bulged adenosine towards or away from its helix III-proximal end (), does not abolish the ability of this adenosine to act as the nucleophile for the first step of splicing in this context.
Systematic movement of a CAC motif through the BS-U2 duplex indicates that three bulge positions can participate in catalysis
To further investigate the apparent flexibility of nucleophile location within the BS-U2 duplex, we generated a series of BS-U2 pairs in which a 5′-CAC-3′ motif was systematically placed at all positions in the BS-U2 duplex opposite a 5′-GG-3′ motif in the second copy U2 to produce a bulge whose nature remained constant while its position within the helix was varied (). Primer extension analysis of total RNA isolated from strains carrying these BS-U2 pairs indicated that a bulged adenosine placed at any of three positions in the duplex could participate in first step splicing catalysis (). If the canonical BS position is defined as 0, these positions are −1 (immediately upstream of the BS in the pre-mRNA), 0 and +1 (immediately downstream) (, lanes 4–6). Bulged adenosines further up- or downstream showed no reactivity (, lanes 2–3 and data not shown).
Systematic variation of bulge position indicates that at least three positions within the BS-U2 duplex are competent for first step catalysis
Grossly substituted BS-U2 pairs function in splicing: development of an orthogonal BS-U2 system
The BS-binding region of U2 snRNA is universally conserved in crown group eukaryotes, although U2s from highly divergent eukaryotes (e.g. kinetoplastids) have different sequences (): minor changes both in this sequence and throughout the highly conserved 5′ end of U2 are generally lethal as a sole copy and can impair cell growth as a second copy (Parker et al., 1987
). These negative effects, together with potentially variable expression levels of mutant U2s, preclude direct, quantitative comparison of branching efficiency from different BS-U2 duplex positions because a different U2 is required for each reporter gene. We sought to generate BS-U2 pairs to allow the fair comparison of different branch positions and nucleophiles within the context of a constant U2 sequence. These second copy U2 snRNAs were designed to be orthogonal to wt U2 – that is, they would interact with the reporter transcript but not with endogenous introns, while the reporter transcript in turn would interact only with its cognate U2.
A wide variety of grossly substituted BS-U2 pairs are functional for splicing, and are orthogonal to the endogenous splicing machinery
We generated several U2 snRNAs in which the entire BS-binding region was substituted, together with corresponding reporter transcripts for each (a selection is shown schematically in ). The U2s tested included ones in which all nucleotides were mutated to C or G, with purine/pyrimidine identity preserved (, #2) or transverted (#8) at each position. Virtually all grossly substituted BS-U2 pairs could participate in both steps of splicing, although the efficiency of the first step was variable (, lanes 2–8(c), and data not shown). As expected, the substituted reporter failed to splice in the absence of its corresponding U2 (, lanes 2–8(w), and data not shown). For all BS-U2 pairs, we verified by high-resolution primer extension that the anticipated branch nucleophile was used, and confirmed splicing efficiency by copper growth ( and data not shown). Thus, despite its extreme conservation, the BS-U2 duplex is remarkably tolerant to substitution.
Using a primer extension assay in which the primer abutted the BS-binding sequence in U2, and ddTTP replaced dTTP, we were able to distinguish between, and thus assay the levels of, wt and most orthogonal U2 snRNAs (); we presume these to reflect the level of U2 snRNP, as we expect unassembled U2 snRNA to be rapidly degraded. Orthogonal U2 expression, while variable, shows no correlation with splicing efficiency (, quantitation in Fig. S2c), suggesting that U2 snRNA expression does not normally limit splicing in our system. The first step defect indicated by the accumulation of pre-mRNA () may instead reflect suboptimal packing of substituted BS-U2 duplexes with other spliceosome components. This is discussed more fully below in the context of Fig. 5.
Strains carrying certain mutant second copy U2 alleles exhibited slow growth (Fig. S2d&e). We do not know the basis of this growth defect, as it does not correlate with reduced or excessive expression of either wt or mutant U2 snRNA (Fig. S2b&e and data not shown). Deleterious effects arising from the generation of orthogonal BS-containing introns are also unlikely: we identified candidate pre-mRNAs in which new introns may be generated by our orthogonal U2s, but were unable to detect products of such new splicing events by RT-PCR (data not shown).
Establishment of nucleophile requirements for the first step of splicing
The orthogonal BS-U2 system minimises potential interaction between the mutant BS reporter and wt U2 snRNA, allowing systematic investigation of the effect of branch nucleotide identity, context and positioning in a constant U2 background. Our previous analysis () had suggested that bulges from at least three positions within the BS-U2 duplex were competent for first step catalysis. Using reporters in which all BS flanking sequence was mutated to G/C with purine/pyrimidine identity either maintained or transverted (UACUAA
CA to CGCCGA
CG or GCGGCA
GC, respectively – the zero position branch nucleotide is highlighted), we placed a bulged adenosine at all positions from −4 to +1 () and assayed splicing by primer extension and copper growth ( and data not shown). We observed relatively efficient branching from the 0 or −1 positions in both constructs (), and additionally from the +1 position in the transverted duplex, but not from bulged nucleotides further upstream than the −1 position ( and data not shown). This confirms the positional flexibility of branching in the context of a constant U2 snRNA background – bulges at the −1 and +1 positions are expected to be separated by ~8 Å and 90° (Berglund et al., 2001
Positional flexibility and preferred identity of the branch nucleophile in orthogonal BS-U2 systems
We believe the disparity between the two constructs to be related to base pair identity flanking the bulge. Base pairing at both flanking positions strongly favours splicing (Fig. S3a), and nucleotide identity within the pair impacts splicing efficiency (Fig. S3b). Consistent with flanking base pair identity underlying the observed difference in the use of the +1 position in the pseudo wild-type CG (CGCCGCAG) and transverted GC (GCGGCGAC) duplexes (both sequences shown with a highlighted +1 adenosine), splicing is more efficient for branch nucleotides preceded by a purine and followed by a pyrimidine than the converse (Fig. S3b). Thus, one would expect the +1 position to be relatively favoured in the context of the transverted GC duplex, and relatively disfavoured in the pseudo wild-type CG context. Moving the branch site A in the context of some orthogonal BS-U2 duplexes imposes a strong block on the second step of splicing ( and data not shown): we do not know the mechanistic basis for this defect.
Splicing, both in yeast (Vijayraghavan et al., 1986
) and in higher eukaryotes (Hornig et al., 1986
), is most efficient with an adenosine branch nucleophile: guanosine is mildly suboptimal and either pyrimidine nucleotide even more so. The effects of BS mutation, however, have been studied in a small number of genes, and in an otherwise wt BS-U2 context: alternative pairings for such duplexes can be envisaged in which an ‘incorrect’ nucleotide is bulged – i.e. the observed splicing defect may arise due to bulging nucleotides suboptimal in terms of identity or of position. We produced duplexes with only one pairing register, and a BS nucleotide flanked by a pair of either guanosine or cytosine nucleotides (), to facilitate the predictable bulging of A/C/U or A/G/U, respectively. The splicing profile of these reporters mirrored that of BS mutants in a wt context, with guanosine supporting a fairly efficient first step and pyrimidines showing little to no branching (). These data support an inherent preference for adenosine at the branch position, as previously suggested by work showing recognition of multiple groups on the adenosine nucleotide during spliceosome assembly and catalysis (Query et al., 1996
A remaining question was the behaviour of branch sites with multiple consecutive bulged nucleotides. We constructed BS-U2 pairs containing one, two or three consecutive unpaired adenosines (). The presence of additional unpaired adenosines impaired splicing of the reporter (, upper panel), presumably due to the lack of base pairing downstream of the branch site adenosine (cf. Fig. S3). Splicing did, however, occur to detectable levels and high-resolution primer extension indicated that the upstream-most adenosine was invariably used as the branch nucleophile (). These results recall data regarding branch site selection in the minor spliceosome (McConnell et al., 2002
), and suggest a model whereby the most upstream unpaired adenosine in the BS-U2 or BS-U12 duplex is specified as the branch nucleophile.
Grossly substituted BS-U2 duplexes predominantly impair spliceosome assembly: inappropriately bulged nucleotides limit catalysis
The BS-U2 duplex, as a highly conserved element containing a first-step substrate, likely forms several tertiary interactions in the spliceosome core; our constructs, although stably base-paired, may be defective in some such interactions. Although virtually all grossly substituted BS-U2 duplexes supported splicing, first-step efficiency was invariably diminished relative to wt BS-U2 (). Spliceosome assembly precedes branching catalysis, and pre-mRNA could accumulate due to a defect in either or both of these steps.
To clarify the nature of the defect shown by our constructs, we investigated their splicing in the context of mutant alleles of the spliceosomal ATPases Prp5 and Prp16. Prp5 has been proposed to monitor the stability of BS-U2 interaction during spliceosome assembly via kinetic competition between its ATPase activity and BS-U2 pairing – failure to establish stable interaction prior to ATP hydrolysis by Prp5 leads to pre-mRNA discard (Xu and Query, 2007
). Mutant prp5
alleles, by relaxing the stability requirement for the nascent complex, thus facilitate spliceosome assembly on introns with suboptimal BS-U2 pairing and presumably also those in which a stable duplex is imperfectly packed. Prp16 enhances splicing fidelity by acting in competition with splicing catalysis – failure to complete first step catalysis prior to ATP hydrolysis by Prp16 leads to pre-mRNA discard (Burgess and Guthrie, 1993
). A suboptimally packed BS-U2 duplex could inhibit first step catalysis by globally destabilising the first step conformation of the spliceosome.
For each ATPase we assayed the splicing efficiency, in a wt and a mutant strain, of substrates with a CGCCGA
C branch site with an adenosine at either the −1, 0 (highlighted here), or +1 positions (). The prp5 N399D
allele (Xu and Query, 2007
) stimulated splicing from the 0 position ~1.5-fold, yet had little effect on the splicing of a reporter with a −1 position adenosine (). By contrast, the prp16-101
allele (Burgess and Guthrie, 1993
) stimulated the splicing of the −1 position reporter 1.8-fold but had little effect on the 0 position reporter (). Taken together, these observations suggest distinct limiting steps in the splicing pathway for the −1 and 0 position reporters. In this context, spliceosome assembly is impaired by duplex substitution: the mutant prp5
allele suppresses this assembly defect. In a substrate with an optimal 0 position bulge, enhanced assembly leads to enhanced splicing. With the nucleophile bulged from the −1 position, however, increased assembly does little to stimulate splicing because the overall efficiency of the reaction is limited at a later stage: in this case, the slowed exit from the first step conformation afforded by prp16
mutation strongly stimulates branching, suggesting that the position of the bulged nucleophile within the BS-U2 duplex is important for first step catalysis by the spliceosome.
Grossly substituted BS-U2 duplexes predominantly impair spliceosome assembly: inappropriately bulged nucleotides predominantly impair splicing catalysis
The preferred nucleophile position depends on distance from the U2 portion of U2/U6 helix Ia, and does not determine 5′SS selection
The analysis described above demonstrates the flexibility of branch site activation within a constantly positioned BS-U2 duplex, without addressing how movement of the duplex itself may affect branch site selection. U2 snRNA is involved in several intra- and intermolecular structures within the catalytic spliceosome (Burge et al., 1999
). We therefore sought, again using a constant BS sequence from which an adenosine could be bulged at multiple positions, to investigate the impact on preferred bulge position of changing the distance between BS-U2 and these spliceosomal structures. We were relatively unconstrained in our choice of mutations, and made a large number of insertion and deletion mutants throughout the 5′ end of U2, including into either the stem or loop of the 5′ stem loop, up- and downstream of helix I, and upstream of helix III (Fig. S4). Several of these mutant U2s failed to show stable expression, presumably due to an inability to assemble into snRNPs (data not shown); all expressed U2s with insertions/deletions anywhere upstream of helix I showed at worst a mild splicing defect with no impact on branch nucleophile selection (data not shown). Of most interest were insertions predicted to alter the distance between helices Ia & III and the BS-U2 duplex, which lies between the U2 components of each (), as this network of interactions may help to juxtapose key catalytic components and splicing substrates.
An all-CG branch site with the (0 position branch) sequence CGCCGAC shows branching from the −1 and 0 positions in the presence of its cognate U2 snRNA (). Insertion of two nucleotides into U2 upstream of helix III, with accompanying deletion of two nucleotides downstream of helix Ia, predicted to shift the BS-U2 duplex towards helix Ia and away from -III, altered this splicing profile such that the −2 position was now highly active for branching and the 0 and −1 positions less favoured than in the original context (). An analogous 1-nt insertion/deletion pair also shifted the observed branching preference towards upstream bulges (data not shown). Conversely, deleting one nucleotide upstream of helix III and inserting one downstream of helix Ia, predicted to shift BS-U2 away from helix Ia and towards -III, disfavoured the use of the −1 position while maintaining robust use of the 0 position BS (). Collectively, these data suggest that branch position relative to either or both of these helices impacts the bulge positions within the BS-U2 duplex that are competent for first step catalysis.
The preferred position of the BS nucleophile within the BS-U2 duplex is partially determined by distance from the U2 component of U2/U6 helix Ia
To investigate the relative importance of BS distance from helices Ia and III, we produced constructs with single-site deletions either downstream of helix Ia or upstream of helix III. Deletion of two nucleotides downstream of helix Ia again activated branching from the −2 position (), recapitulating the result of such a deletion in the context of an accompanying helix III-proximal insertion (cf. ). By contrast, a 1-nt deletion upstream of helix III produced a splicing profile similar to that of the original orthogonal U2 (cf. ). These data show that the distance from the branch nucleophile to the U2 component of helix Ia is an important determinant of nucleophile specification for catalysis. We note that this distance requirement may be relative either to helix Ia or to a rigid higher-order structure containing this helix. Despite the activation of upstream bulge positions for catalysis by helix Ia-proximal deletion in U2, branching remains most efficient from the zero position (). We hypothesise that this position within the duplex remains the optimal one, and that the distance between the bulge and helix Ia is an (at least partially) independent factor in nucleophile determination.
To confirm the above analysis, we recapitulated the results in the context of the transverted GC duplex. Again, reducing the distance between the BS-U2 duplex and helix Ia led to relative activation of upstream bulge positions for first step splicing catalysis (Fig. S5 a&b). We note that the distribution of upstream branch use differs between the two constructs; as previously discussed, we believe that flanking base pair identity underlies this observed difference.
Although either or both of the branch nucleophile and the entire duplex from which it was bulged were moved in the above analysis (&), and high-resolution primer extension confirmed the potential to use various different branch sites, the 5′SS remained unchanged and no activation of cryptic sites was observed in any of this work (, panel iii, and panel iv of ). If 5′SS positioning were dependent on BS positioning, or if the structures that contained the 5′SS and BS, rather than the reactive groups themselves, were positioned relative to one another, movement of the branch site nucleophile and/or BS-U2 duplex would be expected to result in an analogous movement of the 5′SS. Therefore, the observation that moving the BS or the duplex containing it fails to impact 5′SS selection suggests that the position of the BS nucleophile does not determine the position of the 5′SS electrophile for the first step of pre-mRNA splicing.