The Prp18 protein stabilizes the interaction of both exons with the U5 snRNA during the second step of pre-mRNA splicing

Results

Screening of exon mutants

To determine directly whether Prp18 affects the interaction of loop 1 of U5 with the splicing intermediates, we studied pre-mRNAs in which the exonic bases that interact with loop 1 were mutated. We hypothesized that substrates that could base-pair more strongly with loop 1 would splice better in prp18ΔCR yeast, and we expected little effect in wild-type yeast because there is limited conservation of exon bases at the junctions (Long et al. 1997; Lopez and Séraphin 1999; Spingola et al. 1999). We used the ACT1-CUP1 system in which copper resistance reports the efficiency of splicing of the ACT1-CUP1 pre-mRNA (Lesser and Guthrie 1993). We constructed ACT1-CUP1 libraries in which either the four bases at the 3′ end of exon1 or the three bases at the 5′ end of exon2, or both, were randomized (Fig. 1A). The sequence of the wild-type ACT1 exons is TCTG|AGG, where we write the sequences of the exon ends as exon1|exon2, showing the seven bases that we randomized. The libraries were screened by replica plating in indicator yeast that had the mutant allele prp18ΔCR that encodes a Prp18 protein lacking its conserved region or in wild-type indicator yeast (see Supplementary Table 1). Examples of act1-CUP1 genes that conferred survival at the highest copper concentrations (1.5 mM for prp18ΔCR yeast and 2 mM for wild-type yeast) from the libraries were sequenced.

Selected sequences and sequence statistics for winners are shown in Figure 1. From the 5′ splice site libraries (Fig. 1B), A was the predominant base at every position except −2 in the prp18ΔCR winners, as expected if the bases pair with U₄–U₇ in loop 1 of U5. In the prp18ΔCR winners, position −1 was 100% A, whereas the wild-type winners had 25% C at −1. From the 3′ splice site libraries (Fig. 1C), there were two primary prp18ΔCR winners, |ACA and |AAA, and all sequences had an A at position +1; the wild-type winners did not have a clear consensus. The combined library winners (Fig. 1D) were more complicated and showed many of the same general trends as the individual libraries.

Wild-type winners were tested in prp18ΔCR yeast. Many conferred high copper resistance, but a significant fraction did not (Fig. 2). All of the wild-type 5′ splice site winners ending with C had the sequence NAAC| and conferred low copper resistance on prp18ΔCR yeast (Fig. 2, #5). C at position −1 does not ensure a poor 5′ exon, since exon1 ends with C in two prp18ΔCR winners, AATC|AAT (Fig. 2, #29) and AGTC|AAC. The wild-type 3′ splice site winners conferred variable copper resistance on prp18ΔCR yeast; exon sequences starting with a pyrimidine were worse (TCTG|TTG and TCTG|CAA in Fig. 2, #8 and #12) than those starting with a purine, and A was reliably better than G (#15). We also analyzed 32 act1-CUP1 plasmids from the combined library that allowed wild-type yeast to grow at 1 mM CuSO₄ (essentially unselected in the screen, with a 95% survival rate). Of these, one third, including CTTC|CAA and CCCG|ACC (Fig. 2, #23 and #24), conferred low copper resistance on prp18ΔCR yeast. Winners from all the screens were evaluated in wild-type yeast. All act1-CUP1 plasmids tested conferred very similar copper resistance on wild-type yeast (Fig. 2) despite the apparent selection of preferred sequences in the screens (Fig. 1). Small differences accounted for the selection from the libraries (see Supplementary Fig. 1).

The screen and the initial analysis showed that altering the sequences of the end of either of the ACT1-CUP1 exons could significantly change the copper resistance conferred to prp18ΔCR yeast but not wild-type yeast. Hence, the results strongly supported our initial hypothesis.

Defining better and poorer splice sites

We sought to understand the characteristics of the exon sequences that produced better and poorer prp18ΔCR splicing substrates. We started with the exons from our screens, recombining them to isolate the effects of 5′ and 3′ exon sequences, and we made new exon mutants to examine specific bases. We analyzed 170 exon mutants, and we summarize the salient features of our results. Representative mutants are shown in Figure 2 and are referred to by number in the text. All the results, together with a comprehensive description, are shown in Supplementary Table 3.

5′ splice sites

The screen showed that A at the 3′ end of exon1 was important for good splicing in prp18ΔCR yeast (Fig. 1B,D). We compared bases at the −1 position in the context AAAN|AGG (|AGG is wild-type ACT1) and found that A > G ≫ C with A conferring 10 times the copper resistance of C (#3, #5, #27, #30). In weaker overall contexts, A at position −1 was substantially better than G; thus, TCTA|TTG (#9) and CCCA|ACC conferred higher copper resistance than TCTG|TTG (#8) and CCCG|ACC. Splicing yields a stop codon in AAAT|AGG; using different 3′ exons, we found that AAAA| was somewhat better than AAAT|, which was better than AAAC| (#19, #22, #26). The analyses showed that an A at position −1 was a key determinant of efficient splicing and was sufficient to ensure good splicing of all substrates tested except those where the first base of exon2 was a C. The superiority of A and G supported the idea that base −1 pairs to U₄ in loop 1 of U5.

The 5′ exon screen suggested that T was the best base at position −2 (Fig. 1B), an idea that was bolstered by the finding that AATC| (#7, #29) conferred much higher copper resistance than AAAC| (#5). We compared bases at the −2 position in the AANC| context; A at −2 was especially poor, and the overall order was T > C > G > A (#7, #5; data not shown). The superiority of T at −2 was also shown for 5′ exons ending with G, such as TCTG|AGG (wild-type ACT1), which was fourfold better than TCCG|AGG. For position −2, pairing with U₅ does not explain the preference for T, and the ordering of base preferences seems to be opposite the base-pairing expectation.

At positions −3 and 4 the As in the consensus winner sequences (Fig. 1B,D) that could pair with Us in loop 1 were superior bases, as shown in the comparison of AATC|AGG (#7) with CTTC|AGG (#6) and the fourfold superiority of TATG|TTG to TCTG|TTG. We used AAAA|, which conferred very high copper resistance on prp18ΔCR yeast in most contexts, as a reference 5′ exon.

3′ splice site

The sequences |ACA and |AAA accounted for 10 of the 12 winners in prp18ΔCR yeast (Fig. 1C), and subsequent results confirmed that these sequences worked well (#10, #11, #31). An A at position +1 was the most important determinant in exon2. Either pyrimidine at +1 was usually poor, as shown by TCTG|TTG (#8) and TCTG|CAA (#12). G at +1 was generally better than either C or T. Pairing of base +1 to U₄ could account for the observed base preferences. |AGG, the wild-type ACT1 sequence, can pair perfectly to loop 1, but it was inferior to |ACA and |AAA (e.g., comparing #28 and #31, or #5 and #22), and |TGG was fourfold worse than |TCA. A at position +3 was selected in the prp18ΔCR screen (Fig. 1C). With a wild-type exon1 sequence, |CTA was better than |CTG (#13, #14), suggesting some role for base +3.

Combinatorial effects

The effects of mutations at both the 5′ and 3′ splice sites were generally additive, and a good site in one exon compensated for a poor site in the other. Combining the weak AAAC| site with the strong |ACA site improved copper resistance (#5 vs. #22); likewise, |TTG was improved by combination with AAAA| (#9 vs. #16). Combination of two weak sites, as in AAAC|CAA or CTTC|CAA (#24), resulted in copper resistances as low as 25 μM, and combination of two strong sites, as in AAAA|ACA (#32), gave resistances >1 mM.

Some combinatorial effects were more complicated. AAAA| markedly improved |TNN (#8 vs. #16) or |GNN exon2 sites but had little effect on weak |CNN sites (#13, #18). Curiously, one of our poorest pre-mRNAs, AAAC|CAA, was improved fourfold by the exon2 sequence |CGA (#21), although |CGA was otherwise a weak sequence.

Mutations in the exon sequences at the splice junctions had large effects on copper resistance in prp18ΔCR yeast. Many of the effects can be explained as strengthening or weakening base-pairing interactions between the exons and loop 1 of U5, but the current, prevailing model of base-pairing between loop 1 and exon1 provided only part of the explanation for our observed sequence preferences (see Discussion).

Effect of exon mutations on splicing

Our act1-CUP1 pre-mRNAs have exonic mutations that could affect processes after splicing. To show that the changes in copper resistance were caused specifically by changes in splicing, we analyzed the splicing of the act1-CUP1 pre-mRNAs in vivo and in vitro.

In vivo splicing

We assayed splicing in vivo by primer extension. In prp18ΔCR yeast, the levels of mRNAs correlated well with the observed resistance to copper (Fig. 3A). For example, with good substrates like AAAA|ACA (Fig. 3A, lane 12) and AATC|AGG (Fig. 3A, lane 5) mRNA levels were high, whereas poor substrates like CTTC|CAA (Fig. 3A, lane 14) and CCCG|AGG (Fig. 3A, lane 7) had lower mRNA levels. Amounts of mRNA ranged from 1.3-fold more than wild-type ACT1 mRNA to fivefold less, although we are cautious in comparing RNA amounts among different yeast cultures because the copy number of the 2μ ACT1-CUP1 plasmid can fluctuate. Substrates that gave lower levels of mRNA had higher levels of splicing intermediates, consistent with a lower second-step rate; for example, AAAA|ACA had a very low level of lariat intermediate (Fig. 3A, lane 12), whereas CTTC|CAA had sevenfold more intermediate, a level similar to that of its mRNA (Fig. 3A, lane 14). The levels of lariat intermediates varied depending on the efficiency of the second step but did not accumulate to high levels (see Discussion). No significant differences in the amounts of pre-mRNA were seen among the substrates, consistent with the idea that the effects are specific to the second step.

We interpret our splicing results using the two-step kinetic scheme shown in Figure 3C and described in its legend. In this model, k₂, the rate constant for the second step, is directly proportional to the ratio of mRNA to intermediates, and this ratio is shown for each act1-CUP1 in Figure 3. The calculated second-step rates varied by 50-fold. The worst substrate, CTTC|CAA, gave a ratio of 1.7 and the best, AAAA|ACA, 90. The k₂ values correlated very well with the observed copper resistance and argue that the exon mutations specifically affect the second step (Figs. 2, 3; Table 1). The results cannot prove that k₂ is changing because the three RNA levels that can be measured in vivo are insufficient to determine the five rate constants in our simple model. Changes in k₄ and k₅ could give the results we observe; however, the consistent increase in intermediate levels when product levels decrease is most simply explained by changes in k₂.

In wild-type yeast, all the act1-CUP1 substrates spliced well and produced similar (within twofold) amounts of mRNA (Fig. 3B). Lariat intermediates, which were identified by comparison of mobility with prp18ΔCR intermediates, were present at ∼1% of the level of mRNA. In wild-type yeast the ratio of mRNA to intermediates varied about fourfold. The wild-type yeast produced about fourfold more act1-CUP1 mRNA than prp18ΔCR yeast for good substrates; some of this difference is likely related to splicing defects in prp18ΔCR yeast, but levels of intronless mRNAs are also lower in prp18ΔCR yeast (Bačíková and Horowitz 2005).

In vitro splicing

To determine definitively whether the exon mutations affected the second step of splicing, we tested the splicing of 10 representative exon mutants in vitro using extracts with no Prp18, with Prp18ΔCR, or with wild-type Prp18 protein. Time courses of splicing allow quantitative evaluation of both steps of splicing (Horowitz and Abelson 1993b) and avoid the uncertainties inherent in the in vivo measurements.

Complete assays of two substrates, AAAA|AAA and CTTC|CAA, are shown in Figure 4, and rate constants are tabulated in Table 1. Extract from a prp18-knockout strain was assayed alone or supplemented with purified Prp18ΔCR or Prp18 protein, using sufficient protein to give maximal activity. For AAAA|AAA, a superior substrate (Fig. 2, #17), second-step activities with Prp18ΔCR and wild-type Prp18 were similar; the second step with Prp18ΔCR was ∼1.7-fold slower (Fig. 4A,C). However, for CTTC|CAA, a poor substrate (Fig. 2, #24), the second step with Prp18ΔCR was 10-fold slower than with wild-type Prp18. In splicing reactions with Prp18ΔCR, intermediates accumulated and were slowly spliced to products. We carried out parallel assays of nine more substrates with exon mutations (see Supplementary Figure 2). The in vitro second-step rates in Prp18ΔCR extracts varied 14-fold (Table 1) and were well correlated with the copper resistance conferred by the corresponding act1-CUP1 plasmid in prp18ΔCR yeast. Mutations at both the 5′ and 3′ splice sites affected the rate of the second step. For closely related exon mutants, such as AAAC| and AATC|, or |ACA and |AGG, the differences in the second-step rates correlated well with the copper resistances. A comparison of AAAT| to AAAC| paired with either |ACA or |AGG (wild-type) showed that AAAT| was better in both contexts.

In the absence of Prp18, the second step of splicing was very slow, and the same trends as in Prp18ΔCR-reconstituted extracts were seen. For AAAA|AAA, spliced mRNA formed from intermediates ∼13-fold more slowly than in wild-type extract (Fig. 4). For CTTC|CAA, intermediates accumulated but only a trace of mRNA was made; our estimate of its rate was >500-fold lower than in wild-type extract (Fig. 4; Table 1). The ordering of the substrates by rate in extracts lacking Prp18 was substantially the same as in Prp18ΔCR extracts, but the rates were lower. We interpret the ΔPrp18 results as showing the loss of two distinct functions (Bačíková and Horowitz 2002): first, the loss of its conserved region, which is exon sequence dependent, and, second, the loss of its Slu7-binding activity, which is not. Exon mutations also affected the copper resistance of prp18-knockout yeast (data not shown), consistent with the in vitro results.

In extracts reconstituted with wild-type Prp18 protein, the rate of the second step varied slightly with exon sequence (Table 1). The second-step rates decreased about twofold for the substrates that are worst in prp18ΔCR yeast, suggesting that subtle effects on wild-type splicing are exacerbated in Prp18ΔCR splicing.

No significant effects on the first step of splicing were observed. For example, AAAA|AAA and CTTC|CAA had essentially the same first-step rate in all the assays (Fig. 4; see Supplemental Material). Differences in k₁ values were small and did not affect calculation of k₂. There were no significant differences in the stabilities of the pre-mRNAs, the intermediates, or the mRNAs.

The results demonstrate that the exon mutations specifically affect the second step of splicing in Prp18ΔCR-reconstituted extracts, as well as in extracts lacking Prp18, and have little effect in wild-type extracts. The comparison in Table 1 shows that the in vitro second-step rate constants and the relative in vivo rates are in excellent agreement and, further, that the rates correlate very well with the copper resistances.

Effects of exon mutations in AT-AC substrates

To determine whether exon mutations affect splicing in wild-type yeast, we used a much less efficiently spliced substrate that has an AT-AC intron (Burge et al. 1998). In yeast, AT-AC introns are spliced by the U1–U2 spliceosome. Both steps of splicing are slow (Parker and Siliciano 1993; Chanfreau et al. 1994).

We constructed AT-AC ACT1-CUP1 reporters (Kivens and Siliciano 1996) with wild-type and mutant exon sequences, and we assayed splicing by copper resistance and primer extension. The wild-type exon act1, TCTG|AT—AC|AGG (exon1|intron|exon2), conferred very low copper resistance (Fig. 5), about twice that of a blank plasmid, as reported (Kivens and Siliciano 1996). Altering the act1 exons to AAAA|AT—AC|AAA—better sequences from the prp18ΔCR results—improved the copper resistance about fourfold (Fig. 5). Most of the improvement resulted from changing the 5′ exon to AAAA| with the 3′ exon change to |AAA contributing only marginally. The increase in copper resistance was abolished if either terminal exon base was changed to C (Fig. 5; data not shown), paralleling the behavior of GT-AG introns in prp18ΔCR splicing.

The changes in copper resistance were accompanied by commensurate changes in the level of correctly spliced mRNA, determined by primer extension (Fig. 6A) and sequencing of RT–PCR products. The levels of mRNA correlated well with the measured copper resistances; TCTG|AT—AC|AGG yielded ∼3% as much mRNA as wild-type ACT1 (Fig. 6A, lane 3), whereas AAAA|AT—AC|AAA yielded ∼10% as much (Fig. 6A, lane 4). Intermediates accumulated with all the AT-AC substrates, and the levels of intermediates and mRNA were inversely correlated for all the mutants. The ratio of products to intermediates provides an estimate of the rate of the second step (Fig. 3C). The second step with AAAA|AT—AC|AAA was about sixfold faster than with TCTG|AT—AC|AGG, and second-step rates correlated well with copper resistance. Primer extension results for two other substrates (Fig. 6A, lanes 5,6) paralleled their copper resistances. Levels of pre-mRNAs did not change appreciably among the mutant substrates, suggesting that first step rates were not changing significantly among the mutants, although the possibility that large k₃ values obscured changes in k₁ could not be excluded.

Exon mutations improved the second step of splicing of AT-AC substrates in vitro: AAAA|AT—AC|AAA was spliced better than the “wild-type” TCTG|AT—AC|AGG (Fig. 6B). The AT-AC substrates were spliced very inefficiently (Fig. 6B, lanes 7–12). Both AT-AC substrates shown in Figure 6B were spliced slowly through the first step, but only the AAAA|AT—AC|AAA intermediates were spliced significantly through the second step to make product mRNA. Comparing amounts of intermediates and products strongly suggested that k₂ for the AAAA|AT—AC|AAA was higher than for TCTG|AT—AC|AGG, consistent with the in vivo results, but the slow splicing and difficulty of quantitating the RNAs precisely precluded application of our kinetic model.

Kivens and Siliciano (1996) found that mutation of bases upstream of the 3′ splice site improved the second step of splicing of ACT1 AT-AC introns. We used the two-base deletion Δ−9,−10 as an example of this type of mutant. The TCTG|AT—Δ−9,−10—AC|AGG and AAAA|AT—AC|AAA ACT1-CUP1 plasmids conferred similar copper resistance (Fig. 5) and gave similar levels of spliced mRNA (Fig. 6A, lanes 4,7). We combined the two mutations to make AAAA|AT—Δ−9,−10—AC|AAA. The effects of the two mutations were additive, yielding an AT-AC ACT1-CUP1 substrate that allowed growth up to 500 μM copper (Fig. 5) and produced ∼25% as much mRNA as wild-type GT-AG ACT1-CUP1 (Fig. 6A, lane 8). The rate of the second step in vivo was correspondingly increased, by 50-fold compared with TCTG|AT—AC|AGG.

Better exon sequences only increased copper resistance if the introns had both the AT and AC changes. AT-AG and GT-AC introns provided essentially no copper resistance above background (Fig. 5; data not shown). For GT-AC intron substrates, whose first step proceeds at normal rates (Rymond and Rosbash 1985), practically no intermediates accumulated and mRNA was not detectable (Fig. 6A, lanes 10,11). AT-AC intermediates are likely stabilized by interaction of the terminal bases of the intron (Scadden and Smith 1995; Tarn 1996). Without this interaction the GT-AC intermediates may be degraded early in the second step before the exon mutations affect splicing, or the GT-AC introns may have unfavorable interactions that cannot be overcome by the exon mutations.