Intronic RNA of yeast RPL22 paralogs acts as an allosteric switch

Abstract

Ribosomal proteins, because of their RNA-binding capacity, may engage various cellular RNAs and fulfill non-ribosomal roles. Previously, we and others described the intergenic regulation mediated by splicing of RPL22 paralogs in Saccharomyces cerevisiae. Here, we prepared a panel of RPL22A/B intronic mutants with respect to their RNAfold-predicted features and analyzed their properties. We tested splicing efficiency and Rpl22-intron binding using an intron-containing reporter and a yeast three-hybrid system, respectively. We found that the splicing of RPL22 introns can be inhibited by stabilizing a predicted stem as part of a particular type of conformation (I structure). Stabilizing the formation of an alternate stem (P structure) permitted splicing. Binding of the Rpl22 protein to the intron led to enhanced splicing inhibition in WT and several of the mutants, which we interpret as stabilization of the I structure. Mutagenesis identified both the main and alternative 5’ss and additional stem loops as part of the regulatory mechanism. The inhibitory conformation of the intron did not prevent recognition of the 5’ss and branch point, but rather stalled splicing at a later stage, before the first catalytic step. We conclude that the structural ensemble of the RPL22 pre-mRNA behaves as an allosteric switch that responds to Rpl22 concentration.

Introduction

Ribosomal proteins (RP) constitute structural and regulatory components of ribosomes and as such remained highly conserved during evolution. Despite this constraint, some RPs acquired extra ribosomal functions by binding to RNAs other than rRNAs, mostly with an impact on gene expression [1]. As found, Rpl26 bound to the 5’untranslated region (5’UTR) of p53 mRNA, which stimulated translation and thus contributed to the regulation of the DNA damage response [2]. Rpl13a, upon IFN-γ activated phosphorylation, was released from ribosomes and translationally inhibited several groups of mRNAs as part of the GAIT system, which directed transcript selective translational control in myeloid cells [3,4]. The pre-mRNA molecules, their introns and untranslated regions in particular, can thus be viewed as evolving structures, which gained affinity for preexisting RPs [5].

Ribosomal protein coding genes (RPGs) often underwent duplication and exist as paralog pairs. In yeast, which experienced whole genome duplication followed by loss of duplicates, RPGs, in their majority, retained their paralogs. These gene pairs were implicated in splicing mediated intergenic regulation [6,7]. However, only a handful of such pairs were proven to regulate each other’s expression, for example, RPS14, RPS9, and RPL22 [8–12]. Incorporation of two different paralogs into ribosomes can give rise to ribosome heterogeneity. Proteomic and transcriptomic analyzes found evidence for the existence of ribosomal subpopulations that differed in composition, interactomes, and perhaps mRNA specificity [8,13–18]. In yeast, the localization of ASH1 mRNA translation into the emerging bud during yeast cell division, which prevents mating type switching, required specific RP paralogs [19].

Extraribosomal functions of the Rpl22 paralogs have been documented in yeast (see below), fruit fly, zebrafish, mice, and humans [20–25]. Human Rpl22 interacted with human telomerase RNA [26] and was also sequestered, apparently in competition, by the Epstein-Barr virus encoded RNAs in infected cells, possibly as part of the viral strategy to interfere with translation and growth [27,28]. More recently, Zhang and co-workers demonstrated that the zebrafish Rpl22 and Rpl22-Like1 controlled morphogenesis during gastrulation. The proteins acted antagonistically to modulate the splicing of the Smad2 pre-mRNA, presumably in cooperation with HNRNP-A1 [22]. In yeast, the pair Rpl22A/B was shown to be part of oxidative stress response [29]. Rpl22 was also required for the translation of IME1 mRNA and meiotic induction in S. cerevisiae, as rpl22Δ cells were unable to translate the IME1 transcript due to its atypical 5’UTR [30].

The yeast RPL22A and RPL22B genes form a feedback loop that modulates the ratio of paralogous transcripts in response to changing levels of Rpl22 proteins [11,12]. The regulatory loop formed by the RPL22 pair couples the fluctuations in free Rpl22 proteins ([Rpl22A] + [Rpl22B]) to the ratio of their mRNAs ([RPL22A]/[RPL22B]). The shifts in mRNA ratio were related to oxidative stress [29] or disbalances in the free RP concentration caused by exposures to MMS or Cd [11], supporting the interpretation that tuning the isoform ratio helps cells react to changing environment. The cellular processes that distinguish the difference in 19 amino acids between Rpl22A and Rpl22B remain unknown; the effects on splicing are apparently mediated by both A and B proteins [12]. Information on the interface between the intron and Rpl22, and between this complex and the spliceosome, is also limited. Gabunilas and Chanfreau postulated the intronic region between nt 153 and 225 (positions here and throughout the text are numbered relative to the first nucleotide of the RPL22B intron) as ‘regulatory element’ and found that it was responsible for protein binding (Fig 3C, D in [11]). This is similar to our conclusions based on yeast three-hybrid assays (nt 165–236; see I2 in [12]). They also showed that a predicted stem formation between nt 182–188 and 214–220 is necessary for Rpl22-mediated regulation and can rescue splicing inhibition in vitro when added in trans as part of free RNA (nt 115–255 in [11]).

Fig 3. Splicing-permissive mutations did not block RPL22Bi splicing in vivo irrespective of the presence of the Rpl22 protein.

The experiment was done and the figure was formatted as in Fig 2. (A) Mutants disrupting the inhibitory conformation of RPL22Bi were splicing permissive. See the legend of Fig 2A. (B) Summary of the reporter experiments in (A), Rpl22 binding data from the yeast three-hybrid assay, and types of stem loop arrangements predicted by RNAshapes. See the legend of Fig 2B.

RNA forms an ensemble of interchanging conformations, some of which may form with high probability and have a functional impact, as in riboswitches or snRNAs. The prevalence of particular conformations can be tuned by biological regulators or environmental conditions, such as ions or temperature [31]. The RNA structure content changes with cultivation conditions and stress, suggesting that the structural aspect is relevant for our understanding of gene expression [22,32,33]. RNA stems can mediate interactions over longer distances in transcripts, which impacts the outcome of splicing [34,35]. The splicing sequences of the 5’splice site (5’ss), branch point (BP) or 3’splice site (3’ss) were incorporated into the stems and thus blocked [36], or brought into proximity and strengthened [35,37,38]. Exons were looped out [39] or selected for alternative splicing through base-pairing interactions, such as in the case of multi-cluster mutually exclusive exons [40–43]. RNA binding proteins that modulated RNA structures thus affected the outcome of alternative splicing [44,45]. In addition to directly engaging functional sequences in cis or obliterating the access of regulatory RNA binding proteins, RNA structures can intercept the functioning of large RNP complexes operating in gene expression [46–51].

Here, we aimed to characterize the Rpl22 protein – intron interaction in more detail, using constructs for in vivo measurements of splicing efficiency and yeast three-hybrid system. We were able to confirm the findings originally reported by Gabunilas and Chanfreau on the importance of a stem loop structure within the RPL22B intron [11]. Mutations that lowered the structure’s stability abolished inhibition mediated by Rpl22. On the contrary, we were unable to confirm sequence specificity with respect to Rpl22 binding or splicing inhibition within this region, which they reported. Importantly, we showed that inhibition of RPL22B intron (RPL22Bi) splicing can be caused by an RNA structure without Rpl22, provided that the structure is stabilized in cis. As the pre-mRNA can be predicted to exist in alternate conformations, we used site-directed mutagenesis to show that only one alternative base pairing arrangement mediates the inhibitory effect. Furthermore, we showed that the inhibitory mechanism requires both the 5’ss proximal and the BP proximal parts of the pre-mRNA and that it does not operate through splice site obliteration, but rather by stalling spliceosome assembly prior to the first catalytic step.

Results

The RPL22 introns are predicted to form alternative structures

Previous reports found that a stem loop between nucleotides 153 and 246 of RPL22Bi was involved in Rpl22 binding and splicing inhibition (see Introduction; [11,12]). We examined the sequence alignments of both RPL22 paralogs together with the related yeast species of the Saccharomycotina group. We found that the introns share conserved sequence elements, which could base pair in alternate arrangements (Fig 1A). Only one of the two arrangements would give rise to the stem loop between nucleotides 153 and 246 of RPL22Bi.

Fig 1. The RPL22 introns are predicted to assume alternate conformations.

(A) The RPL22 introns of Saccharomycotina yeasts contain complementary regions that suggest the formation of alternate secondary structures. The RPL22 introns of Saccharomycotina contain an absolutely conserved region of 23 nt positioned 5’ of the branch point (red), which can alternately base pair with partially complementary sequence elements located further upstream (green and blue). The homologs of RPL22A and RPL22B are grouped in the upper and lower rows of the alignment. Additional regions of complementarity are specific for the B paralogs (violet). The alignment was created using the MUSCLE algorithm https://www.ebi.ac.uk/Tools/msa/muscle/ and Jalview ver 2.11.2.6 was used for visualization. (B) Predicted structural ensembles of intronic RNAs of Saccharomyces cerevisiae RPL22A and RPL22B paralogs contain two major types of stem loop arrangements. Alternate structures of RPL22 introns were predicted by RNAfold [52]. Pairing the conserved region (red in (A)) with the 3’complementary sequence element (3’COMP; green in (A)) leads to the formation of an inhibitory structure (denoted ‘I’), while the participation of the 5’complementary sequence element (5’COMP; blue in (A)) results in permissive structures (denoted ‘P’). Both structure types are predicted to occur in many variations and regardless of the inclusion of intron-bordering regions. The sequence range used for the modeling included 56 nt in front of the intron and 41 nt from exon 2; Minimum free energy (MFE) structures were visualized by RNAfold web server [52,53] and MFE values were calculated by RNAShapes Studio [54]. The colors correspond to the boxes in (A). The additional base pair complementarity in RPL22B (violet) is denoted as “Clamp”.

We revisited the structural predictions of the introns plus their surrounding pre-mRNA sequences employing RNAfold, Mfold [52,55], RNAShapes [56] and Sfold [57]. Consistently, the introns were predicted to give rise to structures with two alternative base pairing arrangements (see Fig 1A). A conserved sequence element (CONS; labeled red) paired with (i) a proximal upstream element (3’ complementary element (3’COMP); green) or alternatively, (ii) a distal upstream element (5’ complementary element (5’COMP); blue). The conserved element was identical in the Saccharomycotina group, while its complementary elements showed more variability. In RPL22Bi, an additional short stem was predicted (Clamp; violet). We annotated the alternate structure types as inhibitory (I; Fig 1B left) and permissive (P; Fig 1B right), because we found that they differ with respect to splicing inhibition (see below). The sample conformations of these two types for both RPL22Ai and RPL22Bi are presented in Fig 1B as RNAfold predicted minimum free energy (MFE) structures. While they have similar base pair content and thus similar relative stability (see the MFE values in Fig 1B), their interconversion would require the change of well above 20 Watson-Crick base pairs (S1 Fig), suggesting that an energy barrier may kinetically limit the transitions (see the Discussion).

For RPL22A, the proportion of the two types of structure in the structural ensembles of RPL22B is shifted to I compared to RPL22A (see S6 Fig for Sfold prediction). The two-dimensional plot of base pairing probabilities in the Boltzmann weighted structural ensemble of RPL22B shows that the stem loop between nucleotides 153 and 246 forms prominently (S6A Fig), while in the plot for RPL22A, both I and P signature densities are present (S6F Fig).

RPL22i splicing can be differentially manipulated by stabilizing either I or P types of structure

The Rpl22-binding stem (I2 RNA fragment; [12]) does not involve 5’ss, BP, or 3’ss themselves. This led us to ask what the mechanism of splicing inhibition is. Using site-directed mutagenesis, we stabilized either of the predicted conformations (see above) and tested the RNA performance. We took advantage of our previous measurements of the effects of Rpl22 proteins on intron splicing of RPL22B and RPL22A in vivo [12].

In our reporter assays, we used the rpl22aΔ rpl22bΔ strain that harbors an empty or an Rpl22 producing pVTU260 plasmid. The reporter construct included the RPL22B (or other) intron surrounded by its own 5’untranslated region (5’UTR) and exon 1 and 60 nt of exon 2 fused in frame with the CUP1 reporter at the 5’ and 3’ ends of the intron, respectively (see S9A Fig). Thus, the structural and functional characteristics of the pre-mRNA molecule between nt −56 and +381 (positions are numbered relative to the first nucleotide of the RPL22B intron) were preserved, including the alternative 5’splice site (ALT5’ss; 65-GTTTGT). Therefore, we could test the splicing of RPL22B in the absence or presence of Rpl22A and directly compare the reads (the protein is not essential in S. cerevisiae). We used only the expression of Rpl22A in our assays, as we previously demonstrated that either of the paralogous proteins inhibits splicing to the same extent [12]. To complement the reporter splicing assays, we used the three-hybrid system to test the binding of Rpl22A to the I2 fragment of the RPL22Bi mutants generated in this study. The outcomes are summarized in Figs 2B–5B. The details of spot tests and RNAfold prediction of the stem structures are documented in S7 Fig.

Fig 2. Splicing-inhibitory mutations block RPL22 intron reporter splicing in vivo irrespective of the presence of the Rpl22 protein.

(A) Mutants stabilizing the inhibitory conformation of RPL22Bi inhibited splicing irrespective of the presence of the Rpl22 protein in vivo. The splicing efficiency of RPL22B-CUP1 reporters was tested in rpl22aΔ rpl22bΔ strain harboring the plasmid pVTU260/RPL22A for overexpression of Rpl22A and in the same strain transformed with empty vector. Semi-quantitative PCR was run on cDNA prepared from RNA isolated from exponentially growing cultures using random hexamers. The loading control is included in a separate inset with the indicated gene. One of at least three independent experiments is shown. We detected alternative splicing products, which migrated above the size of the major product and which reflect the usage of alternative 5’ss. The PCR products of the corresponding length were sequenced to confirm this assumption. (B) Summary of RPL22Bi reporter experiments in (A), Rpl22 binding data from the yeast three-hybrid assay, and types of stem loop arrangements predicted by RNAshapes. Pre-mRNA accumulation was approximated on a scale of ‘++ + ’ to ‘-.’ The binding of Rpl22 to a fragment of the RPL22B intron (I2 construct; see text) was tested in a yeast three-hybrid assay. Cell growth corresponding to Rpl22 binding was summarized on a scale from ‘++ + ’ to ‘-.’ For the photographs of spot tests, see S4 Fig; n.d. - “not done”. The secondary structures were predicted using RNAshapes [54]. The structures were sorted according to the stem loop arrangement involving the conserved region and its complementary elements (see Fig 1), and the number of structure types classified as I, P, or O (other) was expressed in %. See text and S1-S3 Figs for additional information. (*) The 3’COMP/CONS base pairing in ΔCons was sensu stricto not of the I type, but the apical part of the main stem loop and the rest of the prediction remained the same as in the I structures (see also S2 Fig). (C) The perfect complementarity between the conserved element and the 5’sequence element of RPL22Ai inhibited the splicing similarly to RPL22Bi. The elements are annotated in Fig 1A. The experiment was carried out as in (A), except that the RPL22A-CUP1 reporter was used. (D) Keeping partial complementarity was sufficient to inhibit the splicing of RPL22Bi with scrambled conserved element. The mutant MutConsComp1 did not bind Rpl22 in Y3H (S7C Fig) and returned WT-like secondary structure prediction (S2 Fig). The experiment was done as in (A).

Fig 5. The splicing of the RPL22B intron was affected by the pre-mRNA context.

The experiment was carried out as described in Fig 2. (A) Both main and alternative 5’ss contributed to the RPL22Bi splicing inhibitory mechanism. The removal of alternative 5’ss increased the level of unspliced RNA in the absence of Rpl22 compared to WT (ΔALT5’ss). Mutating the 5’ss sequences to the yeast consensus decreased the ability of Rpl22 to inhibit splicing and increased the use of alternative 5’ss. The alternative splicing products migrated above the size of the major band and the PCR products of the corresponding length were sequenced to confirm this assumption. (B) Predicted secondary structure of the main stem loop of RPL22Bi. The structure was modeled by RNAfold [52], visualized using Forna [58] and annotated according to [11]. ‘Clamp’ and ‘RNA internal loop’ (RIL) are labeled red and green, respectively. In [11], ‘Clamp’ is referred to as ‘Lower distal stem.’ (C) Structure, but not the sequence of the ‘Clamp’ was important for splicing regulation. Deletion of the 3’strand of the ‘Clamp’ or weakening of its complementarity rendered intron splicing permissive regardless of Rpl22. On the contrary, maintaining complementarity while scrambling the sequence or flipping the stem arms retained the WT behavior. (D) Nucleotides within the ‘RNA internal loop’ were dispensable for the regulation of RPL22Bi splicing. In contradiction to previous findings [11], nucleotides within the ‘RNA internal loop’ (RIL) are dispensable for the regulation of RPL22Bi splicing. CCCU to AAAC mutation in the 5’arm of the loop (‘5’RIL AAAC’), the UGAA to CAUU mutation in its 3’arm (‘3’RIL CAUU’), or the deletion of the loop or its arms did not lose regulation after the overexpression of the Rpl22 protein. ΔRIL indicates the deletion of both 5’arm CCCU and 3’arm UGAA nucleotides; Δ5’RIL and Δ3’RIL indicate the per partes deletions of CCCU and UGAA, respectively.

We reasoned that if the RNAfold predictions identified features that were relevant for the splicing inhibitory mechanism, their targeted disruption or hyperstabilization would promote or inhibit splicing, respectively. Indeed, we found that most of the manipulations that stabilized the 3’COMP/CONS base pairing between nt 153–176 and 226–246 (characteristic of I) had inhibitory and Rpl22 independent effects on splicing (Figs 2 and S2, S4 Table). In the 3’CP mutant, we made the 3’COMP element perfectly complementary to CONS; in 3’CPS, we kept the complementarity of the base pairs but scrambled the base pairs (nt 153–177). We confirmed in Y3H assays that the above mutants, which had their presumed Rpl22 binding region mutated, did not bind Rpl22 (Fig 2B). 3’CP and 3’CPS were predicted to form mostly the I structures (Figs 2B and S2). In the splicing assay, we observed pre-mRNA accumulation with or without Rpl22A expressed. Part of the signal was also recovered in bands that corresponded to ALT5’ss usage (Fig 2A). The deletion of the CONS element led to a structure that kept the key characteristics of the inhibitory mechanism intact (∆Cons (nt 226–246)). Although the base pairing of the 3’COMP element was different, the apical part of the main stem loop and the rest of the prediction remained the same. When we removed the 5’COMP element or weakened its base pairing with CONS (nt 75–97), we also obtained inhibitory phenotypes, although partly sensitive to Rpl22A.

The effects of mutations were also demonstrated for the RPL22A intron (Fig 2C). Stabilization of the 3’COMP/CONS base pairing inhibited splicing irrespective of the presence of Rpl22A (3’CP in RPL22A). Finally, we created a mutant of RPL22Bi that had the CONS and 3’COMP motifs scrambled, but unlike the 3’CPS maintained partial complementarity of the bases in the main stem similar to that of WT (Fig 2D; MutConsComp1). Although the binding of Rpl22 was undetectable in Y3H (S7 Fig), splicing was still inhibited (Fig 2D).

In contrast to the manipulations shown in Fig 2, the mutations of RPL22B which destabilized the 3’COMP/CONS arrangement or favored the 5’COMP/CONS base pairing (characteristic of P) allowed intron splicing (Figs 3 and S3, S4 Table). They included the deletion of the 3’arm of the clamp stem (Δ3’Clamp, nt 214–220), the truncation of the main stem (ΔI2, nt 165–236), and the deletion of 3’COMP (Δ3’C, nt 153–176). In analogy with the 3’COMP manipulations above, the perfect complementarity between 5’COMP and CONS (5’CP, nt 75–97; predicted as P) led to a permissive outcome, although part of the inhibitory effect of Rpl22A was still retained. When the base pairs of the 5’COMP/CONS stem were scrambled, the structure was splicing permissive and Rpl22A-refractory. Most of the manipulations in Fig 3 were predicted to form P structures (Figs 3B and S3, S4 Table). In the case of ΔI2, the structural stability of the remainder of the intron was lowered (S3 Fig – upper right). We concluded that the intronic stem structure was necessary for down-regulation of the splicing of RPL22B. Favoring its formation led to splicing inhibition even without Rpl22A binding (see 3’CP and 3’CPS in Fig 2B above), while destabilization of the stem relieved inhibition. The expression of Rpl22A increased splicing inhibition in WT, Δ5’C, Δ5’CW and 5’CP, which we interpret as stabilization of the inhibitory conformation by binding of the protein rather than a ‘direct inhibition’ due to obstruction of the splice sites or BP.

In RPL22A, the stabilization of the 5’COMP/CONS stem (5’CP in RPL22A) did not affect the result relative to WT (= Rpl22A still inhibited splicing), probably because the 5’COMP/CONS base pairing in WT is already relatively strong (Fig 2C). The 5’CP mutation had a mild effect in RPL22B (see above). The RNA structures thus showed the same propensity in RPL22A as in RPL22B, despite the sequence differences of the nonconserved parts of the two introns.

The main stem loop of the inhibitory structure of RPL22Bi is necessary but not sufficient for the inhibition of splicing

The inhibitory structure of RPL22Bi (see Fig 1B top left and S3 Fig) contained stem loops predicted in the vicinity of the branch point in the 5’ direction. We found that mutations in this region had strikingly distinct effects on the inhibitory mechanism. The destabilization of the BP-proximal hairpin (BPmut3, nt 280–285) left the inhibition intact, but melting of the distal hairpin (distal from BP in the 5’ direction) rendered the structure splicing permissive (BPmut4, nt 258–265; Fig 4). This phenotype was retained even in a mutant with stabilized main stem loop (3’CP + BPmut4). In contrast, complete ablation of the distal hairpin (dLoop2, nt 259–275) or disruption of the stem proximal to BP in the 3’ direction (Mutclamp2, nt 139–147) produced a WT phenotype. In a mutant that combined stabilization of the main stem loop with distal hairpin deletion (3’CP + dLoop2), the structure still inhibited splicing, although inhibition was slightly enhanced by Rpl22A expression.

Fig 4. The main stem loop of the inhibitory structure of RPL22Bi was necessary but not sufficient for splicing inhibition.

The experiment was done and the figure was formatted as in Fig 2. (A) The inhibitory mechanism required that specific structural details of the I structure remain intact. In the mutants analyzed in this set, the main stem loop is predicted to form. Melting of Loop2 of the I structure (BPmut4; see panel (C)) renders the structure splicing permissive, while ablating the loop (dLoop2) retains the Rpl22-dependent inhibitory potential. Although the stability of the main stem loop should be lowered in Mutclamp2, it is apparently still sufficient to keep the I structure Rpl22 responsive. (B) Summary of the reporter experiments in (A) and types of stem loop arrangement predicted by RNAshapes. pre-mRNA accumulation was approximated on a scale of ‘+++’ to ‘-.’ The secondary structures were predicted using RNAshapes [54]. The structures were sorted according to the stem loop arrangement involving the conserved region and its complementary elements (see Fig 1), and the number of structure types classified as I, P, or O (other) was expressed in %. See text and S1-S3 Figs for additional information. (C) Detail of the predicted secondary structure of RPL22Bi in the I conformation. The regions referred to in the text are indicated (in italics) together with the mutants (in plain text). (D) The main stem loop was not sufficient to convey splicing inhibition in an unrelated intron. The RPS17B intron without or with the main stem loop inserted was tested as in (A). M indicates the m.w. marker lane.

The effects that we observed in the mutants described in Fig 4 cannot be explained solely by the dichotomy of the alternative 5’COMP/CONS versus 3’COMP/CONS base pairing. According to RNAshapes, 3’CP + BPmut4, 3’CP + dLoop2, dLoop2 and Mutclamp2 return >80% I structures, yet the splicing results covered the whole range – from ‘permissive’ to ‘regulated’ to ‘inhibitory’. According to Sfold, BPMut4 and especially 3’CP + BPMut4 maintain the characteristics of an ensemble ‘I’ (S6 Fig), but the efficiency of splicing inhibition was diminished. In the structures of BPMut4 predicted by RNAshapes, we observed distinct arrangements between the main stem loop and the portion containing 5’ss of the secondary structure diagram (Figs 4B and S4). We hypothesize that while the main stem loop is part of the inhibitory mechanism, the efficacy with which the RNA structure interferes with the spliceosome depends on the details of its 3D contacts. In comparison, partial destabilization of the main stem loop had only a limited effect on splicing inhibition.

Gabunilas and Chanfreau inserted part of the RPL22B intron, spanning nucleotides 153–239, into the modified intron of RPS21A, but did not see an inhibitory effect on splicing after overexpression of Rpl22A [11]. According to our predictions, the intron part which the authors used did not contain the main stem loop in its whole. Therefore, we transplanted the full extent of the stem loop (nt 153–246) into the intron of RPS17B. Like RPL22B, RPS17B is a ribosomal protein gene with a suboptimal 5’ss (GTACGT) and similar intron length (314 nt). The stem loop was inserted at the same distance from the BP as it is located in RPL22B (220 nt in the intron), and the construct contained 60 nt of exon2. Stem loop formation in this new context was predicted by RNAfold (S4 Fig). The experiment did not show an inhibitory effect of this stem loop on RPS17B splicing (Fig 4D). Although this result does not exclude the possibility that the stem loop might inhibit splicing in a different setting, it supports the conclusion that additional requirements must be met for the inhibition to occur.

RPL22B intron splicing is affected by the pre-mRNA context

The 5’ss of RPL22Bi is non-canonical (GTACGT) and is accompanied by an ‘alternative’ 5’ss (ALT5’ss) 65 bp downstream. The latter site does not produce a translatable ORF and genomic data measured under standard cultivation conditions showed that it was being used only marginally [12,59–62]. We noticed the accumulation of alternatively spliced products in mutants in which the usage of the main 5’ss was inhibited (see Fig 2A). The identification of these molecules was done by sequencing the PCR product of the corresponding length.

We first asked whether the inhibitory mechanism would be capable of overcoming a stronger signal, that is, the canonical 5’ss signal. The mutant with the canonical 5’ss sequence (GTATGT; 5’ss-CONS) showed weakened but still discernible inhibition by Rpl22A (Fig 5A). Next, we mutated ALT5’ss to consensus, which lowered the Rpl22A-induced accumulation of pre-mRNA compared to WT while increasing the accumulation of alternatively spliced product(s). With increased inhibition (i.e., + Rpl22A), the band intensity of the main spliced RNA decreased, while that of the alternative product increased. The effect of consensus ALT5’ss prevailed even in the 5’CP + ALT5’ss-CONS mutant, where the 5’COMP/CONS base pairing was stabilized (and inhibition lessened; see Fig 3). We still observed alternative splice products that were correlated with the expression of Rpl22A. We reasoned that if the inhibition mechanism did not involve the 5’ part of the intron, the ratios (+Rpl22A)/(-Rpl22A) for the spliced products would be similar for both the main and the alternative 5’ss used. Their anticorrelation suggested that both parts of the intron contributed to the inhibitory mechanism. When we deleted ALT5’ss (deletion of nt 65–70), we observed enhanced inhibition in the absence of Rpl22A (Fig 5A), which further supported the role of the 5’ss recognition in the mechanism. The result showing that the deletion of ALT5’s negatively affects the splicing efficiency of RPL22Bi was already obtained by the Chanfreau group (Fig 5B in [59]; S2C Fig in [11]). We concluded that the whole intron is important for the execution of the inhibition, either because of its 3D structure or because its parts differentially affect the assembly of the spliceosome.

Both the 3’COMP/CONS stem and the shorter clamp stem contribute to the formation of the inhibitory main stem loop. Removal of the 3’ arm of the clamp stem (Δ3’Clamp; deletion of nt 214–220) disrupted inhibition (see Figs 3 and 5C). To confirm the structural role of the clamp we mutated the stem to decrease its stability (Clamp’Loose’; mutation of nt 186GC to CG and 218AT to TA), scrambled the base pairs while maintaining complementarity (Clamp’Scramble’) and flipped the sequences between the 3’ and 5’ arms of the stem (Clamp’Flip’). While the Clamp’Loose’ mutant lost its ability to inhibit splicing, the latter two mutants showed WT behavior (see also [11]).

The reporters used by Gabunilas and Chanfreau [11] lacked the alternative 5’ splice site, the apical portion of the stem between nt 191 and 211 and the surrounding exons (see S8 Fig for comparison). To address whether these manipulations could have affected the outcome of the assay, we recreated some of their mutants of the ‘RNA internal loop’ (RIL; see scheme in Fig 5B) using our approach. We constructed the CCCU to AAAC mutation in the 5’arm of the loop (‘5’RIL AAAC’), the UGAA to CAUU mutation in its 3’arm (‘3’RIL CAUU’), or deletion of the loop (ΔRIL) or its arms (Δ5’RIL and Δ3’RIL), all of which were in the context of the entire intron. In contrast to the results obtained by Gabunilas and Chanfreau [11], none of the above mutations disrupted the negative regulation of splicing by Rpl22A because all reporters reacted to Rpl22A overexpression by increasing the amount of unspliced pre-mRNA to levels obtained in WT (Fig 5D). In the uninhibited state, they were also spliced with the same efficiency as the WT intron; a possible exception was the ΔRIL mutant, which showed slightly less efficient splicing. None of the RIL-targeting mutations impaired the ability of this part of the intron to interact with Rpl22 in Y3H (S7 Fig). This further supports our conclusion that mutations in ‘RNA internal loop’ nucleotides, including its DS arm GUAA (nt 178CCCU181 and 221UGAA224), do not alter Rpl22A-mediated inhibition.

Splicing inhibition occurs during cotranscriptional spliceosome assembly

Our previous analyses of RPL22 splicing identified varying degrees of pre-mRNA accumulation depending on the paralogue and the level of Rpl22, but never the lariat-exon2 intermediate, suggesting that inhibition occurs before the first splicing step [12]. Notably also, the splicing sequences were not part of the Rpl22 binding stem loop. We wanted to confirm that the inhibitory mechanism does not involve blocking splice site access but rather targets later stages, that is, the assembly of spliceosomes.

We analyzed the cotranscriptional recruitment of the components of the yeast commitment complex 1 Prp42 (U1 snRNP), Msl5 and Mud2 (BP recognition) to the RPL22B locus using ChIP-qPCR analysis as previously described [63]. Cells expressed HA-tagged versions of splicing factors and primers (P to E; Fig 6) were used to amplify DNA fragments along the RPL22B locus after chromatin anti-HA immunoprecipitation. We took advantage of the fact that the strain rpl22aΔi produces high levels of Rpl22A and compared this strain with the strain rpl22aΔ (=no Rpl22A) and the strain ‘WT’. Under the high concentration of Rpl22A (rpl22aΔi), splicing of RPL22Bi is inhibited [12]. We were able to obtain WT recruitment profiles of the tagged factors for both rpl22aΔi and rpl22aΔ strains (Fig 6). This implies that the regulatory mechanism did not block the first stage of recognition of 5’ss and BP. We were unable to analyze in detail the binding of subsequent spliceosome assembly components due to the short length of exon 2 in RPL22B.

Fig 6. Splicing inhibition did not preclude cotranscriptional recruitment of U1 snRNP and BBP to the RPL22B locus.

The recruitment of spliceosomal components to the RPL22B locus was analyzed by ChIP-qPCR analysis. We used cells expressing HA-tagged Prp42, Msl5, and Mud2 to monitor the cotranscriptionality of the formation of commitment complex 1. The signals were normalized to input values and expressed as the fold enrichment relative to the signal obtained from the telomeric region (TEL VI R). A representative graph of three independent experiments is shown; error bars represent the S.D. of three technical replicates. To compare the effect of low and high concentrations of Rpl22, we used rpl22aΔ and rpl22aΔi strains, respectively. Positions of the PCR amplicons P to E are indicated.

Rpl22 homologs complement the ability of Rpl22A to inhibit splicing in S. cerevisiae

The demonstration that Kluyveromyces lactis RPL22 (KlRPL22) can be regulated by changes in the concentration of Rpl22A/B or KlRpl22 proteins [12] led us to ask about the ability of other homologues of Rpl22 to complement the extra ribosomal function of Rpl22A in S. cerevisiae. Rpl22 binding is dependent on the presence of lysines in the RNA binding interface of the protein, because mutants in which these lysines were substituted with glutamates were not binding in Y3H and were unable to inhibit RPL22B splicing in vivo [12]. The conservation of the residues that contribute to charge attraction can be shown in the alignment of Rpl22A with other Rpl22 homologs that we used for our complementation analysis (S8 Fig). We mapped the amino acid changes among the seven selected homologs on the surface of the Rpl22 structure. As expected, the part of the protein that faces the rRNA in the ribosome [64] was the most conserved (Fig 7A).

Fig 7. Rpl22 homologs complemented the extra ribosomal function of Rpl22A in S. cerevisiae.

(A) Projection of invariant amino acids on the surface of Rpl22A of S. cerevisiae. The structure of Rpl22 and 25S rRNA was extracted from the ribosome structure (PDB: 4V88) [64] using the ChimeraX program (https://www.rbvi.ucsf.edu/chimerax) [65]. Helices 57 and 59 of 25S rRNA are highlighted in blue and orange, respectively. The pooled amino acid differences between Rpl22A and the Rpl22 homologs tested in (B) to (D) are projected onto the surface of the molecule. The colors indicate the identity to Rpl22A (red), conservative substitutions (pink; see ‘:’ and ‘. in the alignment in S8 Fig), and nonconserved positions (gray). The projection delineates parts that are likely involved in the intron-regulatory mechanism of RPL22. (B) The Rpl22 proteins from related yeast species, S. pombe and H. sapiens were tested for their ability to support WT growth of the rpl22aΔ rpl22bΔ strain. We expressed in rpl22aΔ rpl22bΔ and WT S. cerevisiae cells different RPL22 genes as intronless versions under the control of the ADH1 promoter (plasmid pVTU260) and evaluated whether the production of different Rpl22 proteins would rescue the growth phenotype. Transformation with the empty pVTU260 served as a control (-). The Rpl22 proteins tested included Rpl22A (A), the binding-dead mutant of Rpl22 (A_mut) and the homologs from Candida albicans (Ca), Candida glabrata (Cg), Debaryomyces hanseni (Dh), Yarrowia lipolytica (Yl), Schizosaccharomyces pombe (Sp), Kluyveromyces lactis (Kl) and Homo sapiens (Hs). The y-axis represents the generation time of individual strains. The plotted values are the means and standard deviations of three biological replicates. The results of the Student’s t-test with Bonferroni correction for multiple tests are indicated by (*) or (‡) for the differences between a strain and WT or the double deletant rpl22aΔ rpl22bΔ, respectively (both (*) and (‡) indicate P ≤ 0.001). (C) Endogenous RPL22B splicing can be inhibited by Rpl22 homologs in the rpl22aΔ strain. The RPL22 homologs were expressed in the rpl22aΔ strain as in (B). After the preparation of cDNA from exponentially growing cells, the level of endogenous RPL22B mRNA was measured by qPCR using primers specific for the RPL22B locus. The signal was normalized to SPT15 and to the rpl22aΔ strain transformed by an empty vector (-). Splicing inhibition was caused by the expression of S. cerevisiae Rpl22A (A) but not its binding-dead mutant (A_mut). The Rpl22 homologs were labeled as in (B). Statistical significance of the difference between the rpl22aΔ strain harboring an empty vector (-) and a Rpl22 expressor is indicated as (‡); (*) is used for the difference between (A) and other strains. Both symbols indicate P ≤ 0.005 based on a t-test with Bonferroni correction for multiple comparisons. (D) All tested Rpl22 proteins can inhibit splicing of the RPL22B reporter in vivo. The rpl22aΔ rpl22bΔ cells expressing the indicated homologues of Rpl22 (see (B)) were transformed with a plasmid carrying the reporter RPL22B-CUP1. The reporter splicing efficiency was measured and the figure was formatted as in Fig 2. TOM22 was used as a loading control. The figure represents the result of one of three independent experiments.

We used the rpl22aΔ rpl22bΔ strain of S. cerevisiae and first tested the ability of the Rpl22 homologues to complement the ribosomal function of endogenous Rpl22A/B. The deletion of both RPL22 genes was shown to result in 1.7 times slower growth compared to WT S. cerevisiae [66]. Upon expression of Rpl22A, the doubling time was reduced to WT level (Fig 7B). The expression of Rpl22 homologs from Candida albicans, Candida glabrata, Debaryomyces hanseni, Kluyveromyces lactis, or Schizosaccharomyces pombe had a similar effect. The exceptions were the proteins of Yarrowia lipolytica and Homo sapiens. While these proteins were expressed at detectable levels in the rpl22aΔ rpl22bΔ strain (S10 Fig), they were unable to reduce the doubling times. The expression of human Rpl22 increased the doubling time of WT, similarly to the effect of the binding-dead mutant of Rpl22A (A_mut). We assume that the toxicity resulted from interference with normal ribosome function.

To see the effects of the homologs on RPL22 splicing, we measured the levels of endogenous RPL22B mRNA in a rpl22aΔ strain by qPCR. All tested proteins, except for Rpl22mut, decreased the levels of RPL22B mRNA (Fig 7C). The effects of the homologs of C. glabrata, D. hanseni, K. lactis, and S. pombe inhibited splicing comparably to Rpl22A of S. cerevisiae. The effects of Y. lipolytica and human homologs (< 40% of the empty vector strain) could have been due to lower protein levels (S10 Fig) and/or toxicity. In addition, we also analyzed inhibition of RPL22B splicing using our RPL22B-CUP1 reporter system. The experimental setup was described in Figs 2–4. The reporter showed no trace of unspliced product when the binding-dead mutant was expressed, in contrast to the effects of S. cerevisiae Rpl22A/B, which led to pre-mRNA accumulation. Among the homologs, Y. lipolytica and human products showed the weakest effects, but all the proteins tested inhibited splicing to some extent (Fig 7C).

The map of the amino acid differences among the Rpl22 homologs shows that while the rRNA interacting helix contains only conservative substitutions, the opposite side of the protein harbors non-conservative changes (Fig 7A). Because both the ribosomal and extra ribosomal functions of Rpl22A were complemented by the various homologs, it is likely that these functions use the same interface on the protein.

Discussion

The inhibitory mechanism requires the intron as a whole

We mutagenized the RPL22 introns to test whether their splicing efficiency depends on the formation of a particular RNA structure. We employed RNAfold and other algorithms which predict secondary structures by findingthe base-pairing pattern with the lowest folding free energy. Although these algorithms are established as predictive tools for in vitro conditions [52–57], they do not account for the effects of proteins or crowding inside cells. Therefore, we were pleased to find independent support for the existence of the predicted structure within the intron of RPL22B in vivo [67]. Rangan and coworkes undertook a comprehensive analysis of Saccharomyces cerevisiae intron structures in vivo using transcriptome-wide dimethyl sulfate probing and were able to obtain sufficient coverage for the intron of RPL22B (not for RPL22A, which is much better spliced). The authors found four short stems (> 70% confidence estimates), three of which exactly match the stem regions predicted by RNAfold. The DMS reactivity data thus support the prediction of the middle of the intron structure, the “CLAMP” stem (nt 182–188/ 214–220) and the apical stem in the main stem loop of the “I” conformation of RPL22Bi (see S12 Fig). The fourth region (nt 133–138/ 314–319) of low DMS reactivity is predicted as a stem with a bulge. We believe that these independent in vivo structure data lends the necessary support to the RNAfold predictions used in this manuscript.

Our experimental setup allowed us to differentiate mutations that were splicing inhibitory regardless of the presence or absence of the Rpl22 protein (see Fig 2) from those that were permissive in its presence, rendering the intron unresponsive to Rpl22-mediated inhibition (see Fig 3). Furthermore, some mutations showed graded effects and some affected the relative use of ALT5’ss. We grouped mutants into three categories according to the predicted structures they contained, those with (i) stabilized inhibitory conformations (I), (ii) stabilized permissive conformations (P) and (iii) ‘disabled’ (= splicing permissive) I conformations. In contrast to the previous study [11], we retained ALT5’ss and included intron proximal sequences in our constructs. This was because we found that the deletion of RPL22Bi ALT5’ss impaired the splicing efficiency at the major site (Fig 5A). The in vivo splicing of RPL22B uses alternative 5’ and 3’ splice sites, albeit at low frequencies [12,59–62]. Although ALT5’ss does not lead to a functional product, it may be part of the spliceosome assembly process, for example, as an adjunct assembly landing platform [37,68].

Among the three categories of mutations, the ‘disabled’ I conformations were the most interesting, as they apparently targeted the inhibitory mechanism outside of the main stem loop, which was nevertheless still predicted to form (BPmut4 and 3’CP + BPmut4; see Fig 4). The ensemble of secondary structures of BPmut4 was predicted to be similar to WT (I > P), but the splicing in vivo was not blocked. We concluded that the predicted formation of the stem loop itself was not sufficient for splicing inhibition. We cannot exclude that in this mutant the binding of Rpl22 in vivo was not affected even though the motif that bound the protein in Y3H was intact. There may be a part of the RNA structure other than the main stem loop that contacts Rpl22. However, the destabilization of Rpl22 binding in other mutants (= nonbinding Rpl22 in Y3H) did not automatically result in splicing permittivity. This is why we assumed that the defect in BPmut4 was not related to the absence of Rpl22 binding. We speculate that the defect in inhibition of splicing in BPMut4 was due to (i) changes in 3D contacts within the intron structure or (ii) changes in the energy barrier that separates I and P conformations. We noticed that the predicted conformations of BPmut4 differed from those of WT in details that included the 5’COMP base pairing and the nucleotides between 5’COMP and 3’COMP. Structure predictions postulate two stem loops 5’ proximal to BP (see Fig 4C). The melting of the more proximal of the two (in BPmut3) did not affect the inhibition outcome, while melting of the more distal one did (in BPmut4). Moreover, ablating this latter loop kept the inhibitory effect intact. If the I/P ratio was under kinetic control, BPmut4 can work by lowering the energy barrier that separates the I and P conformations and thus allowing more P structures to form cotranscriptionally (and splice). We concluded that blocking BP access is not the way the mechanism operates and that it relies on features throughout the intron, including the 5’ splice sites.

Base pairing of conserved elements across introns was previously described as a means of enhancing splicing in RPS17B and several other RPGs [37,69–71]. Rogic and co-workers tried to demonstrate that a ‘structural distance’ between 5ss and BP, defined as the shortest walk on a 2D map of secondary structure prediction, was correlated with splicing efficiency. Considering the ‘structural distances’ between 5ss and BP in the conformations of RPL22Bi, we found that they cannot explain the inhibition observed in RPL22. In the I conformation, ca 100–180 nucleotides are ‘removed’ by stem loops (depending on the stability of the stem), whereby the 5’ss BP distance is thus ca 40–50 nucleotides (measuring a shortest path on a graph of a secondary structure). Although the P conformation orients the 5’ss and BP in closer proximity, neither of the two conformations should pose a challenge of distance. Spliceosome assembly proceeds co-transcriptionally, with the U1 snRNP complex being anchored by the cap binding complex and the PolII holocomplex [72–75]. The BP recognition complex is also in contact with Prp40, Prp39 and Nam8 of U1 snRNP and genetically interacts with the TGS1 cap binding complex enzyme [76–79]. Without some obstacle effect that prevents alignment of the 5’ss and BP binding complexes, the assembly of the spliceosome would not be blocked.

Inhibition of RPL22Bi does not block recognition of 5’ss and BP

The cotranscriptional recruitment profiles of the U1 snRNP component Prp42 (see Fig 6) and the heterodimer BBP/Mud2 were not affected by splicing inhibition, suggesting that the steps further downstream in the spliceosome assembly pathway were responsible for the blockage. We presume that the critical steps lie between the formation of the prespliceosome A and the maturation to the spliceosomes pre-B and B [77,80,81]. Gabunilas and Chanfreau, who also obtained ChIP data for Prp42 recruitment at the RPL22B locus, concluded that they found partial inhibition of Prp42 binding, which may not fully explain the observed inhibition of splicing. Indeed, their ChIP results (Fig 4G in [11]) showed only moderate signal reduction. An interesting comparison can be made with inhibition of RPL30 splicing by the Rpl30 protein [82,83]. Rpl30 binds a kink-turn structure that encompasses the 5’ss of its own pre-mRNA and thus inhibits splicing. Inhibition did not block U1 snRNP and BBP/Mud2 recruitment, but affected U2 snRNP binding, likely also blocking subsequent rearrangement steps.

Although the inhibitory mechanism did not block the first step of recognition of 5’ss, the result of inhibition was dependent on the presence of ALT5’ss and the strength of both the main and alternative splice sites. A canonical main 5’ss made the inhibition less efficient (Fig 5A). The non-canonical sequence 5’GUACGU(U) contains U at the 7th position, which should weaken the interaction with U6 snRNA (through the 5’caauACa motif of U6). In yeast splice sites, C at the 4th position was found to be correlated with A at the 7th position [84]. In ALT5’ss GUUUGU(C), the base pairing with U6 would be canonical, but the interaction with U1 would be weaker. Perhaps this is part of the reason why the main and alternative sites behave distinctly with respect to Rpl22-induced inhibition.

Rpl22 homologs complement the ability of Rpl22A to inhibit splicing in S. cerevisiae

The Rpl22 proteins of other yeasts, as well as human RPL22, complemented the inhibitory function of Rpl22A (see Fig 7). The Rpl22 homologs were constrained in evolution to bind to highly conserved rRNA and function in the ribosome context [85]. As we did not find any example of an Rpl22 homolog that did not inhibit the splicing of RPL22B at least partially, it is highly probable that the extra ribosomal function of Rpl22 uses the protein’s rRNA binding site. At the same time, given that only 26 out of 121 amino acids of Rpl22A are identical among the seven homologs, the interaction seems to be flexible. The imperfect stem of the main stem loop of RPL22 intron contains a motif that is identical in both paralogs as well as all RPL22 genes of the species of the Saccharomycotina group (Fig 1A). The proteins accommodate both the 25S rRNA loops [64] and the intronic RNA without any apparent similarity between the two. In higher eukaryotes, but not in yeast, some of the RNAs whose expression is regulated by Rpl22 contain a consensus hairpin that fits the Rpl22-interacting part of 25S rRNA [22].

The RPL22 intron behaves as an RNA switch

According to cross-linking experiments and structural information, intron – spliceosome contacts are not limited to the splice sites themselves but include additional sequences around them [86,87]. A striking physical association was documented between an intronic stem placed between 5’ss and BP and a spliceosome in the ACT1 intron [78,87]. The stem loop was placed across positively charged residues of Prp39, Prp42, and the C-terminal domain of U1C as part of the E complex. Mutations destabilizing the stem loop led to pre-mRNA accumulation of the reporter gene, suggesting that interactions via this stem loop aided splicing efficiency [78].

The spliceosomal RNP complex goes through a series of conformational rearrangements before it can catalyze the first transesterification step. Intronic structures could affect these processes if they form anywhere within the footprint of the assembling spliceosome [88]. Kumar and coauthors performed a structure-guided prediction analysis of the effects of mutations on alternative splicing of the exon 10 – intron 10 junction of the MAPT gene. They obtained the best prediction accuracy when considering a footprint of 43 nucleotides for the B^act complex. The authors argued that the RNA structure “functions as a rheostat” in this range of nucleotides, which must be unfolded to participate in the spliceosome – pre-mRNA interface.

According to the prediction of Boltzmann weighted structural ensembles for RPL22i, the tendency to form I conformations is higher in RPL22Bi than in RPL22Ai (S6 Fig). In fact, endogenous RPL22B was more stringently regulated than RPL22A - after Rpl22 overexpression, RPL22B mRNA production dropped to ~10% [12,89]. We hypothesize that a nascent pre-mRNA structure of RPL22 interacts with the assembling spliceosome and prevents its own splicing. Structures form as the RNA exits the PolII complex [50], suggesting a kinetic control. As splice sites emerge during transcription, they are recognized by U1 snRNP and BBP/Mud2. By the time the 3’ss proximal part is synthesized, the 3’COMP/CONS (= inhibitory) or the 5’COMP/CONS (= permissive) stem loops prevail, depending on the level of free Rpl22. Rpl22 may act here as a factor tuning the I/P ratio through controlling the energy barrier that separates the two conformations (see Figs 1 and S1). The I/P ratio, in turn, dictates the degree to which splicing is inhibited (and the pre-mRNA degraded). The results of our splicing assays suggest that the inhibition requires the presence of both the main and alternative 5’ss and that it depends on the noncanonical character of the 5’ss. Hypothetically, the presence of the inhibitory RNA structure (I) could lead to aberrant pre-mRNA interactions with snRNAs and / or recruit disassembly factors to spliceosome on an otherwise slow substrate. The spliceosome disassembly process was recently shown to target aberrant B^act conformations and to be part of a proofreading mechanism of incorrectly interacting pre-mRNAs [90,91].

The features of the RPL22 intron invite comparison with the riboswitch mechanism [92]. Binding of small molecules induces a conformational change of their aptamers, which is then transduced as an output via the regulatory platform (Fig 8). An intriguing example of a riboswitch that recognizes a large molecule (tRNA) and represents an ensemble of elongated structures encompassing ~300 nucleotides, is the T-box riboswitch [94,95]. An example of an RNA switch regulated by protein ligands is the ‘stress responsive RNA switch’ in the 3’untranslated region of mRNA for human vascular endothelial factor A [96]. The switch responds to changing levels of the hnRNP L and GAIT complexes to regulate vascular endothelial factor A expression in response to hypoxia signaling [97,98].

Fig 8. Model of toggling between permissive (P) and inhibitory (I) conformations in RPL22 pre-mRNAs.

5’COMP/CONS stem (in P) is mutually exclusive with 3’COMP/CONS stem (in I); drawn in analogy to U2 stem II toggling model [93]. The I conformation inhibits splicing. The effects of mutants and Rpl22p binding are indicated.

Alternate RNA structures may serve as an interface between the inputs, e.g., transcription speed or activity of RNA binding proteins, and the outputs, e.g., alternative splicing or translation. Transcription and transcript processing complexes can act as chaperones, selecting alternate structures from ensembles in response to environmental stimuli [48]. As Saldi pointed out, cotranscriptional RNA folding may be one of the means by which to regulate posttranscriptional splicing [47]. The RPL22i - Rpl22 feedback loop represents an intriguing example of an intronic structure, stabilized by the Rpl22 protein, which hinders its own splicing. If confirmed by structural biology (see also [67]), the RPL22 toolkit of S. cerevisiae should help our understanding of how splicing responds to regulatory cues.

Materials and methods

Yeast strains and cultivation methods

Details of the yeast strains used here are provided in Table S1. Yeasts were transformed using the lithium acetate method [99]. The primers and plasmids utilized in this work are listed in S2 and S3 Tables. For the splicing assays, yeast cells were inoculated from overnight grown precultures and let to grow for two generations in synthetic medium without uracil and histidine. For the measurement of the doubling times, cells were cultured in medium lacking uracil and tryptophan at 30° C. Optical densities of cell suspensions were recorded spectrophotometrically using Varioskan LUX (λ = 600 nm). The analysis of the yeast three-hybrid system was performed essentially as previously described [12].

Splicing analysis

RNA isolation and reverse transcription were performed as previously described [12]. Semiquantitative PCR was performed in 25 μl-reactions with 5 μl of 50 times diluted cDNA as template and with primers listed in Table 2 for 25–28 cycles. The whole sample was loaded onto 2.5% agarose gel and DNA was stained with ethidium bromide. The photos were taken with a Panasonic DMC-FZ7 camera, processed in the GNU Image Manipulation Program 2.10.6 (https://www.gimp.org) and assembled in Inkscape 0.92.3 (https://www.inkscape.org).

Plasmid preparation

The intron mutants of RPL22B/A were synthesized by GeneArt (Thermo Fisher Scientific) and swapped using the restriction sites BamHI and EcoRI with the WT version of RPL22B in p423GPD-RPL22B-CUP1. For the three hybrid tests, the mutated versions of RPL22B were amplified from GeneArt DNA Strings with primers specific for each mutant (S2 Table) and cloned at the SphI site in the vector p3HR2. Plasmids pVTU260 expressing RPL22 genes from different organisms were prepared as described in [12] using cDNAs as templates for PCR amplification. The cloning outcome was verified by restriction analysis and sequencing.

ChIP

For ChIP analysis, rpl22a∆ and rpl22a∆i cells containing a HA-tagged version of a particular spliceosomal subunit (Prp42, Msl5, or Mud2) were cultivated in YPD medium. The precultures were grown overnight and then reinoculated in 50 ml of fresh YPD medium and cultivated for 2–3 generations to reach the mid-log phase. After harvesting, all the steps were performed exactly as described [63]. The DNA fragments were then amplified with primers covering various parts of the RPL22B locus using HOT FIREPol® Evagreen® qPCR Supermix on LightCycler 480 II (Roche).

RNA structure modeling

For RNAfold [52,53] and RNAshapes [56], we used default parameters except temperature, which was set to 30° C. No folding constraints were applied.

Supporting information

S1 Fig. I ⇋ P interconversion of the intronic RNA would require an extensive change of base pairs.

(GIF)

S2 Fig. RNAfold predicted structures of introns and their mutants tested in Fig 2.

(PDF)

S3 Fig. RNAfold predicted structures of introns and their mutants tested in Fig 3.

(PDF)

S4 Fig. RNAfold predicted structures of introns and their mutants tested in Fig 4.

(PDF)

S5 Fig. RNAfold predicted structures of introns and their mutants tested in Fig 5.

(PDF)

S6 Fig. Base pairing probabilities of Sfold predicted RNA ensembles of RPL22 introns.

(PDF)

S7 Fig. Yeast three-hybrid analyses of Rpl22A binding to the I2 fragment of the intronic RPL22B RNA.

(PDF)

S8 Fig. Sequence alignment of RPL22 homologs tested in this study.

(PDF)

S9 Fig. The reporter constructs used to analyze RPL22B splicing in vivo.

(PDF)

S10 Fig. Levels of heterologous Rpl22 proteins in the experiments of Fig 7.

(PDF)

S11 Fig. Raw images of gels and blots.

(PDF)

S12 Fig. Secondary structure prediction of the RPL22B intron by RNAfold is supported by DMS reactivity data published previously.

(PDF)

S1 Table. Strains used in this study.

(XLSX)

S2 Table. Primers used in this study.

(XLSX)

S3 Table. Plasmids used in this study.

(XLSX)

S4 Table. List of intronic manipulations tested.

(XLSX)

S5 Table. Original data for graphs in Figs 6 and 7.

(XLSX)

Abbreviations

3’COMP

3’complementary element

3’ss

3’splice site

5’COMP

5’complementary element

5’ss

5’splice site

5’UTR

5’untranslated region

ALT5’ss

alternative 5’splice site

BP

branch point

CONS

conserved sequence element

I2

stem loop region of RPL22B intron used in Y3H (nt 165 to 236; Abrhámová et al. 2018)

MFE

minimum free energy

RP

ribosomal protein

RPG

ribosomal protein coding gene

RPL22Bi

RPL22B intron

SD

standard deviation

Y3H

yeast three-hybrid system

Data Availability

All relevant data are within the manuscript and its Supporting Information files.

Funding Statement

This work was supported by the Charles University programs PROGRES Q43 and UNCE 204013. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.