Functional analysis of the zinc finger modules of the Saccharomyces cerevisiae splicing factor Luc7

Humanized Luc7 ZnF proteins support yeast viability except for LUC7L3 ZnF2

To investigate the human Luc7 homologs, we generated a yeast “shuffle” system to introduce Luc7 mutants in yeast. We deleted the chromosomal LUC7 gene while maintaining yeast viability by expression of wild type (WT) Luc7 from a low-copy, URA3/CEN6-containing plasmid. Plasmid expression of Luc7 had no effect on yeast viability in standard growth conditions (Supplemental Fig. S1A). To introduce human Luc7 homologs in yeast, we inserted yeast codon-optimized gene fragments encoding truncations of the open reading frames of LUC7L, LUC7L2, or LUC7L3 into a low-copy, TRP1/CEN6-containing plasmid containing the native LUC7 promoter and terminator sequences. These truncations coded for the predicted structured regions of the proteins but did not include the C-terminal RS domains. Each of the constructs also included a 3 × HA epitope tag at the protein C terminus.

After the transformation of yeast with these plasmids and subsequent 5-FOA selection, we found that none of the human homologs was able to support yeast growth as the sole copy of Luc7 (Supplemental Fig. S1B). We confirmed the expression of the proteins by western blotting. Although each protein was expressed, LUC7L and LUC7L2 were less abundant than either Luc7 or LUC7L3 (Supplemental Fig. S1C). In addition, we were unable to rescue viability when these proteins were expressed under the control of a strong TDH3 promoter (OE; Supplemental Fig. S1B). The inability of the human LUC7L proteins to fully compensate for yeast Luc7L has also been recently reported (Chalivendra et al. 2024). Based on these results, we instead focused on targeted mutations to humanize Luc7.

We analyzed the cryo-EM structure of the yeast spliceosome A complex to identify Luc7 amino acids located within 6 Å of the U1 snRNA/5′ss duplex (Fig. 1A,C; Plaschka et al. 2018). The amino acids predominantly come from ZnF2; however, we also decided to study ZnF1 (which is likely located further from the RNA duplex) because previous work showed genetic interactions between ZnF1 mutations and a number of splicing factors (Agarwal et al. 2016). We identified the corresponding amino acids in ZnF1 and ZnF2 of the human LUC7L paralogs based on a multiple sequence alignment generated by Clustal Omega (Fig. 1B; Sievers and Higgins 2014) and humanized yeast Luc7 accordingly by site-directed mutagenesis of the nonconserved amino acids (amino acid substitutions listed in Supplemental Table S1). We then expressed the humanized, C-terminal 3 × HA-tagged Luc7 variants in yeast on plasmids and under the control of the native Luc7 promoter. We refer to mutants of ZnF1 as ZnF1_L, L2, or L3 corresponding to ZnF1 from LUC7L, LUC7L2, or LUC7L3. The mutations needed to humanize ZnF2 for LUC7L and LUC7L2 are identical. We refer to mutants of ZnF2 as ZnF2_L/L2 or L3.

Luc7 chimeric proteins containing one or both ZnFs from LUC7L or LUC7L2 were viable (Fig. 2). Although a Luc7 chimera of ZnF1_L3 was also viable, a chimera containing ZnF2_L3 was not even though it was well-expressed in yeast (Supplemental Fig. S1D). ZnF2_L3 did not support viability even when ZnF1 was also humanized to that from LUC7L3 or any other LUC7L paralog. From these experiments, we conclude that although most ZnF modules from human Luc7 homologs can support yeast splicing, ZnF2_L3 is not functionally equivalent to either ZnF2 from Luc7 or LUC7L/L2.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

FIGURE 2.

Yeast are viable with humanized Luc7 ZnF1 and ZnF2. (A) Luc7 ZnF1 mutants are viable in yeast. The ZnF2_L/L2 mutant is also viable, whereas ZnF2_L3 is lethal. (B) Expressing Luc7 with mutations in both ZnF1 and ZnF2 does not rescue the lethality associated with ZnF2_L3. “-W” and “-W + 5-FOA” refer to synthetic media lacking tryptophan and media lacking tryptophan and containing 5-fluoroorotic acid, respectively.

ZnF2 but not ZnF1 changes 5′ss usage of a yeast splicing reporter

We next used the ACT1-CUP1 reporter assay to probe how the humanization of the Luc7 ZnF domains influences 5′ss selection. In this assay, yeast growth in the presence of various concentrations of Cu²⁺ is correlated with splicing of the reporter and mRNA production (Fig. 3A; Lesser and Guthrie 1993a). In the case of ZnF1, we used ACT1-CUP1 reporters with substitutions from −3 to −1 on the exonic side of the 5′ss because this region is closest to the binding site of ZnF1 (Figs. 1 and 3A). The WT ACT1-CUP1 substrate normally only contains a single potential base pair with the U1 snRNA within the −3 to −1 region (-1G could pair with U1 snRNA C9; Fig. 1C). G-1A and G-1C substitutions in the ACT1-CUP1 should, therefore, disrupt all pairing within this region, whereas a U-2A or C-3A should strengthen the interaction by allowing pairing with either U10 or U11. To increase the sensitivity of the assay, we also incorporated mutations at the +2 (U2A) or +5 (G5A) positions because substitutions at −3 to −1 have no discernible effect on the splicing of a substrate with a consensus 5′ss.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

FIGURE 3.

Mutation of the ZnF domains of Luc7 can have specific impacts on 5′ss usage. (A) Schematic representation of the ACT1-CUP1 reporter pre-mRNA and splicing assay. Proper processing of the splicing reporter confers resistance to Cu²⁺ in the growth media. (B) ACT1-CUP1 assays using a reporter with substitutions in the 5′ss sequence U2A (GaAUGU) and the indicated changes in the exon. ZnF1 mutants have no obvious effect on splicing. Dots represent the maximum [Cu²⁺] tolerated in two independent replicates. (C) Same as in B but using a reporter with the 5′ss G5A (GUAUaU). (D) ACT1-CUP1 assays using ZnF2_L/L2 to assay intronic positions of the 5′ss. Mutation of Luc7 ZnF2 improves splicing of A3U, G5A, G5C, and G5U but does not affect other reporters. (E) ACT1-CUP1 assays using the NΔ31 truncation mutant of Luc7.

None of the ZnF1 chimeras changed yeast tolerance to Cu²⁺ for any of the ACT1-CUP1 reporters harboring substitutions at −3 to −1 (Fig. 3B,C). No changes were observed even when ZnF2 was also modified to that from LUC7L/L2 (Supplemental Fig. S2). As expected, we also did not observe changes in splicing due to ZnF1 substitutions with reporters carrying substitutions at the +1 to +6 positions, likely because these substitutions are located distal to the ZnF1 interaction site (Supplemental Fig. S3). From these results, we conclude that paralog-specific contacts of ZnF1 made by these humanized Luc7 variants are unlikely to significantly influence splicing of the ACT1-CUP1 reporter in this assay.

For studies of ZnF2, we focused on using ACT1-CUP1 reporters harboring substitutions within the intronic portion of the 5′ss (+1 to +6) because ZnF2 contacts this region (Fig. 1C). When ZnF2 was replaced by ZnF2_L/L2, we observed increased Cu²⁺ tolerance of yeast with ACT1-CUP1 reporters containing substitutions at the +3 position (A3U) and +5 position (G5A, C, U) (Fig. 3D), in agreement with contacts between Luc7 and the snRNA/5′ss duplex. Because the high levels of Cu²⁺ tolerance observed with some reporters precluded analysis (i.e., U4 and U6 substitutions), we also carried out assays in which a second mutation was incorporated at the +2 position (U2A). We were unable to find any effect of ZnF2_L/L2 with these doubly substituted reporters for changes at the +4 or +6 positions (Supplemental Fig. S4). These results suggest that a function of ZnF2 in human LUC7L and LUC7L2 could be to promote splicing at weak 5′ss containing mismatches to the U1 snRNA at the +3 and +5 positions of the 5′ss.

Deletion of a Luc7/Sm ring interaction decreases usage of nonconsensus splice sites

We wondered whether other Luc7 mutants, not confined to the ZnF domains, would also show changes in splicing at +3 and +5. The Luc7 NΔ31 mutation abolishes contact between Luc7 and the U1 snRNP Sm protein ring and was previously shown to impair splicing of some pre-mRNAs, cause synthetic lethality with a number of other mutations, and bypass the need for Prp28 (Agarwal et al. 2016; Plaschka et al. 2018). Because this mutant was previously well-characterized, we carried out ACT1-CUP1 assays with Luc7 NΔ31-expressing yeast.

Unlike the ZnF2_L/L2 substitution, Luc7 NΔ31 caused a loss of yeast tolerance to Cu²⁺ using reporters with multiple substitutions at the +2 to +6 positions (Fig. 3E; Supplemental Fig. S4). The only position that was unaffected was +1. This suggests that pairing at this position and enforcing selection of the nearly invariant +1G at the 5′ss may be due to other snRNP components—likely Yhc1/U1-C, which makes extensive contacts with the duplex at this position (Fig. 1C; Kondo et al. 2015; Hansen et al. 2022). Surprisingly, Luc7 NΔ31 even decreased Cu²⁺ tolerance in yeast with U4A and U4G substitutions, which should result in increased pairing to Ψ5 in the snRNA. This would seem to indicate that WT Luc7 has evolved to function preferentially on substrates containing mismatches (i.e., a U) at the +4 position of the 5′ss. Interestingly, the +4 nucleotide is ultimately juxtaposed with A49 in the U6 snRNA during splicing catalysis, and +4U could form a canonical base pair with A49 (Kandels-Lewis and Séraphin 1993; Lesser and Guthrie 1993b). Luc7 may help the snRNP to tolerate a 5′ss/U1 snRNA mismatch that ultimately increases 5′ss base-pairing to U6.

Luc7 variants perturb splicing of endogenous yeast pre-mRNAs

5′ss sequences found in endogenous yeast introns are less diverse than the reporters we used in the ACT1-CUP1 assays. For example, we do not believe there are any naturally used 5′ss with adenosine located at the +5 position even though we observed increased Cu²⁺ tolerance with the G5A reporter and ZnF2_L/L2. We first tested two endogenous transcripts, SUS1 and RPL22B, for changes in splicing due to Luc7 ZnF2_L/L2 or Luc7 NΔ31. We chose these transcripts because changes in SUS1 splicing had previously been reported for Luc7 NΔ31 (Agarwal et al. 2016) and RPL22B has a cryptic, intronic 5′ss (GUUUGU) with an A3U substitution relative to the consensus (Kawashima et al. 2014). Because the splicing of ACT1-CUP1 reporters with A3U is stimulated by Luc7 ZnF2_L/L2 (Fig. 3D), we wondered if the humanized protein would also stimulate the use of the cryptic site in RPL22B.

We analyzed the splicing of SUS1 and RPL22B by RT-PCR using yeast strains expressing Luc7-ZnF2_L/L2 or Luc7 NΔ31. UPF1 was also deleted in these strains to block nonsense-mediated decay (NMD) and facilitate detection of RPL22B isoforms (Kawashima et al. 2014). Both Luc7 variants somewhat inhibited splicing of both SUS1 and RPL22B with Luc7 NΔ31 showing a greater splicing defect (Supplemental Fig. S5). We did not observe a substantial increase in the use of the cryptic 5′ss in RPL22B due to Luc7 ZnF2_L/L2. However, Luc7 NΔ31 accumulated unspliced RPL22B and had comparatively less usage of the cryptic 5′ss than either WT Luc7 or Luc7 ZnF2_L/L2. This is consistent with the ACT1-CUP1 assays and with Luc7 NΔ31 inhibiting splicing of weak, nonconsensus 5′ss, including A3U (Fig. 3E).

To further probe the effects of the Luc7 variants on the yeast transcriptome, we used total RNA-seq to analyze changes in splicing efficiency genome-wide. RNA-seq was performed in an upf1Δ genetic background deficient for NMD. The upf1Δ genetic background not only helps to stabilize unspliced mRNAs harboring premature termination codons but also facilitates a more direct analysis of splicing efficiency without confounding effects of differential isoforms stability (Sayani et al. 2008; Kawashima et al. 2014).

mRNA expression was globally unchanged when Luc7 ZnF2_L/L2 was expressed compared to WT, as no genes showed significant differential expression (P < 1 × 10⁻⁵). The Luc7 NΔ31 strain had nine differentially expressed genes (Supplemental Fig. S6A): YBL059W*, ATP5, IMD2, DGR2, YLR366W, RPS22B*, YML131W, HRB1*, and LYS9 (* indicates an intron-containing gene). HRB1 is interesting because of its reported role in the selective export of spliced mRNAs (Hackmann et al. 2014). Its modest upregulation (∼1.6-fold) is likely due to the accumulation of unspliced transcripts. It is worth noting that analysis of HRB1 RNAs also showed a significant fraction of unspliced reads in the control upf1Δ strain expressing WT Luc7 (∼1:1), indicative of suboptimal splicing even in the presence of the WT protein.

To calculate splicing efficiency, we quantified reads that could be unambiguously assigned as either spliced junctions or unspliced reads (i.e., spanning both exonic and intronic regions). For each intron annotation, we first calculated the ratio of unspliced reads to spliced reads. We then compared these ratios for each mutant relative to the WT control. Both Luc7-ZnF2_L/L2 and Luc7 NΔ31 had a negative impact on global splicing efficiency compared to WT (Fig. 4A). However, the effect was not universal as both mutants showed increased splicing efficiency of a select few transcripts. We grouped each intron by its change in splicing efficiency in each mutant compared to WT and analyzed splicing sequences for each group by generating sequence logos of the 5′ss, branch points, and 3′ss (Fig. 4A; Supplemental Fig. S6B).

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

FIGURE 4.

Analysis of RNA-seq data from cells expressing WT, NΔ31, or ZnF2_L/L2 Luc7 Proteins. (A) Plots showing changes in splicing efficiencies for each intron in Luc7 NΔ31 (top) and Luc7 ZnF2_L/L2 (bottom). p-values are the results of unequal variances t-tests using geometric means between three replicates. Introns with statistically significant changes in splicing efficiency were categorized into increased or decreased intronic reads in mutants over WT, p-value < 0.05 (dashed line). Sequence logos for 5′ss (middle) and BP (right) are shown for each category. (B) Changes in splicing efficiencies in Luc7 NΔ31 (left) and Luc7 ZnF2_L/L2 (right) versus WT for ribosomal protein genes (RPGs). Cyan points indicate RPGs. Histogram p-values are the results of unequal variances t-tests using geometric means. (C) Plots showing transcriptional start site to 5′ss distances versus changes in splicing efficiency for Luc7 NΔ31 (left) and Luc7 ZnF2_L/L2 (right) versus WT Luc7. Color indicates p-values. For plots in A–C, note that the x-axis in A and the y-axis in B and C represent fold changes. The black vertical line indicates 85 nt. (MD) Difference of geometric means.

We found that introns with increased splicing efficiency with Luc7 NΔ31 had near-perfect consensus 5′ss (GUAUGU). In line with the ACT1-CUP1 data, introns with non-“GUAUGU” 5′ss had an average of 45% more unspliced reads with Luc7 NΔ31 compared to WT (Supplemental Fig. S6C). The 5′ss sequence had surprisingly little effect with the Luc7 ZnF2_L/L2. Only one of the common, nonconsensus 5′ss showed a statistically significant change in splicing efficiency: Introns with a GUACGU 5′ss had 18% fewer unspliced reads on average (Supplemental Fig. S6C). This is consistent with the position of the ZnF2 region in contacting the U1 snRNA/5′ss duplex near the U4 position, but we did not observe a similar change at U4 substitutions in our ACT1-CUP1 assays (Fig. 3D; Supplemental Fig. S4).

The introns with increased splicing efficiency in both the Luc7 ZnF2_L/L2 and NΔ31-expressing strains showed a strong enrichment for the consensus U at position one of the branch point consensus (UACUAAC, underlined) as well as a more modest enrichment for the consensus C at position 7 (UACUAAC). Interestingly, the introns with increased splicing efficiency using the Luc7 ZnF2_L/L2 variant showed an enrichment for an A immediately downstream from the branch point (UACUAACA). An adenosine at this position would facilitate the formation of an additional base pair with the U2 snRNA at U2 nucleotide U33. Base-pairing between the U2 snRNA and intron is seen at this position in cryo-EM structures of the U2 snRNP (Plaschka et al. 2018; Zhang et al. 2024). This suggests that increased base-pairing between U2 and the branch point may compensate for the deleterious effects of the ZnF2 L/L2 variant.

We also found that the strain expressing Luc7 ZnF2_L/L2 was able to splice pre-mRNAs of RPGs protein pre-mRNAs with much higher efficiency than pre-mRNAs of non-RPGs, whereas the strain expressing Luc7 NΔ31 showed little difference between the two (Fig. 4B). In an attempt to explain this observation, we checked for correlations between intron length and splicing efficiency as well as between the transcription start site (TSS) to 5′ss distance and splicing efficiency, because introns found in RPGs are typically longer than those of non-RPGs (Spingola et al. 1999). Although both were statistically significant with the Luc7 ZnF2_L/L2 mutant, neither was a better explanatory variable for splicing efficiency than ribosomal versus non-RPG (Fig. 4C; Supplemental Fig. S6D). We do note, however, that the expression level of the RNAs could play a role in how 5′ss compete for U1 snRNP binding because RPGs are more highly expressed than most other intron-containing genes (Munding et al. 2013).

Although splicing efficiency in the Luc7 NΔ31–expressing strain was not significantly correlated with ribosomal versus non-RPG or intron length, it did show a significant correlation with the distance from TSS to the 5′ss (Fig. 4C) with longer distances from the TSS corresponding to generally lower splicing efficiency. Because our data (Figs. 3E and 4) and that by Agarwal et al. (2016) suggest that Luc7 NΔ31 is less able to stabilize the U1 snRNA/5′ss interaction, it would seem that this becomes even worse the further from the TSS and (possibly) the 5′ end of the nascent transcript. We do not know exactly what factors are responsible for this behavior, but Luc7 NΔ31 is synthetically lethal with deletion of the nuclear cap-binding complex (CBC) (Agarwal et al. 2016). It is possible that the ability of the CBC to stabilize U1 on the pre-mRNA (Görnemann et al. 2005; Larson and Hoskins 2017) and compensate for Luc7 NΔ31 is less effective at greater TSS to 5′ss distances. Similar RNA cap to 5′ss distance dependence for the function of Luc7 has been remarked upon previously with the presence of Luc7 favoring the selection of cap-proximal 5′ss (Puig et al. 2007).

Luc7 ZnF2_L/L2 alters the requirement for Prp28 during splicing

We next assayed the Luc7 mutants for an altered requirement of the DEAD-box ATPase Prp28 for U1/U6 base-pairing exchange at the 5′ss during spliceosome activation (Fig. 5A; Staley and Guthrie 1999). It has previously been shown that Luc7 NΔ31 can completely bypass the need for Prp28 activity and that the PRP28 gene is no longer essential when Luc7 NΔ31 is present (Agarwal et al. 2016). Unlike Luc7 NΔ31, Luc7 ZnF2_L/L2 does not bypass the requirement for the presence of Prp28 during splicing (Fig. 5B). We then tested whether Luc7 ZnF2_L/L2 could alter the demand for Prp28 activity using a Prp28 ATPase mutant (Prp28 R499A, a mutation in DEAD-box motif Va) that is proposed to impair ATP hydrolysis and leads to a cs growth phenotype (Fig. 5A). We predicted that if Luc7 ZnF2_L/L2 places increased demand on Prp28, then this Luc7 mutant should exacerbate the growth defects present with Prp28 R499A. Alternatively, if Luc7 ZnF2_L/L2 facilitates U1/U6 exchange by Prp28 R499A, it should result in suppression of the cs phenotype. Luc7 ZnF2_L/L2 not only suppresses the cs phenotype, it surpasses Luc7 NΔ31 in its ability to do so (Fig. 5C). Together, these data demonstrate that Luc7 ZnF2_L/L2 alters the requirement for ATP hydrolysis by Prp28 during splicing in a way that still requires a Prp28 presence and is distinctly different from Luc7 NΔ31. Although it is often assumed that Prp28 mutant suppression is due to weakening of U1/5′ss interactions, our results suggest that this can still occur even when a subset of U1/5′ss interactions are potentially strengthened as demonstrated by Luc7 ZnF2_L/L2 in the ACT1-CUP1 reporter assay (Fig. 3) and with endogenous 5′ss with U4C substitutions relative to the consensus (Fig. 4).

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

FIGURE 5.

Luc7 ZnF2_L/L2 suppresses the cs phenotype of Prp28 R499A. (A) Schematic showing the impact of Prp28 alleles that modulate the U1 to U6 exchange during splicing. Truncation of Luc7 bypasses the requirement for Prp28 in splicing and allows splicing to proceed whereas Prp28 ATPase mutants inhibit this step and stall splicing. (B) Luc7 ZnF2_L/L2 does not bypass the requirement for Prp28 in splicing unlike Luc7 NΔ31. “-HW” and “-HW + FOA” refer to synthetic media lacking histidine and tryptophan in the absence or presence of 5-fluoroorotic acid, respectively. (C) Luc7 ZnF2_L/L2 suppresses cold sensitivity associated with Prp28 R499A.