Truncating the spliceosomal ‘rope protein’ Prp45 results in Htz1 dependent phenotypes

ABSTRACT

Spliceosome assembly contributes an important but incompletely understood aspect of splicing regulation. Prp45 is a yeast splicing factor which runs as an extended fold through the spliceosome, and which may be important for bringing its components together. We performed a whole genome analysis of the genetic interaction network of the truncated allele of PRP45 (prp45(1–169)) using synthetic genetic array technology and found chromatin remodellers and modifiers as an enriched category. In agreement with related studies, H2A.Z-encoding HTZ1, and the components of SWR1, INO80, and SAGA complexes represented prominent interactors, with htz1 conferring the strongest growth defect. Because the truncation of Prp45 disproportionately affected low copy number transcripts of intron-containing genes, we prepared strains carrying intronless versions of SRB2, VPS75, or HRB1, the most affected cases with transcription-related function. Intron removal from SRB2, but not from the other genes, partly repaired some but not all the growth phenotypes identified in the genetic screen. The interaction of prp45(1–169) and htz1Δ was detectable even in cells with SRB2 intron deleted (srb2Δi). The less truncated variant, prp45(1–330), had a synthetic growth defect with htz1Δ at 16°C, which also persisted in the srb2Δi background. Moreover, htz1Δ enhanced prp45(1–330) dependent pre-mRNA hyper-accumulation of both high and low efficiency splicers, genes ECM33 and COF1, respectively. We conclude that while the expression defects of low expression intron-containing genes contribute to the genetic interactome of prp45(1–169), the genetic interactions between prp45 and htz1 alleles demonstrate the sensitivity of spliceosome assembly, delayed in prp45(1–169), to the chromatin environment.

KEYWORDS:

• Synthetic genetic array analysis
• spliceosome assembly
• co-transcriptional splicing
• chromatin modifiers
• H2A.Z

Introduction

Splicing and transcription are interdependent

The processes of transcription and transcript maturation, including capping, splicing, cleavage, and polyadenylation are executed in immediate succession [1], which results in their interdependence and allows cells to regulate them coordinately [2–4]. Co-transcriptionality of splicing was established for both yeast and Metazoa [5–8], and various mechanisms were proposed to explain the coupling between splicing and transcription [1,5,9–11]. The biological significance of splicing resides in the complexity of the decisions when and where to splice. Accumulated evidence supports the view that these „decision algorithms’’ are embedded in part in the spliceosome assembly pathways [12], which are in turn dependent on the surrounding context [3,13,14]. This context depends on the interactions of chromatin domains, RNPs, nuclear proteins and small biomolecules, and represents a continuum with the gene expression system of the cell [15]. Indeed, co-transcriptional spliceosome assembly was shown to depend on chromatin modifiers, such as the histone acetyltransferase Gcn5 [16], the H2B ubiquitylase Bre1 [17], or the histone methyltransferase Set2 [18,19]. The component of the TBP binding module of the Gcn5/SAGA acetyltransferase complex, Spt8, interacted with the spliceosomal ATPase Prp5, which promoted pre-spliceosome formation. Through this interaction, Spt8 was able to affect the proofreading of suboptimal splicing substrates, mediating the interplay between transcription initiation, elongation, and splicing [20]. The methyltransferase Set2 recruited the adaptor Eaf3 through H3K36 methylation, which in turn enhanced the association of the spliceosomal Nineteen complex and Prp45, at least on some intron containing genes [19]. Transcription elongation factors, apart from regulating PolII kinetics, can also contribute to co-transcriptional spliceosome recruitment, as was shown for Spt5, part of the DSIF complex, which enhanced U5 snRNP association with intron containing genes [21].

Splicing factors have negative genetic interactions with alleles of chromatin regulators

Splicing factors and chromatin regulators display extensive genetic interactions, but the evidence for accompanying physical interactions is lagging (see Figure 2 in [1] and references therein; [18]). To account for the functional coupling between transcription and splicing, three mutually non-exclusive modes have been considered in the literature. The first mode was direct physical binding of splicing and transcription complexes. Some splicing factors were found to be in direct contact with chromatin components [13,24], regulating the distribution of chromatin modifications in Metazoa [25]. The second mode was colocalization through liquid-liquid phase separation, which could impact the local concentrations of components available for spliceosome assembly [26–28]. Finally, kinetic coupling was postulated to account for the findings that the speed of RNA PolII matched the speed of splicing [1,29,30]. In yeast, co-transcriptional splicing was reported to affect PolII progression [31], while transcription speed was shown to have an impact on splicing efficiency [10,29]. All three modes would be receptive to regulatory inputs, such as posttranslational modifications of the C-terminal domain of RNA PolII [32,33], changes of components which affect the properties of nuclear condensates [34–36], or chromatin modifications which impact RNA PolII elongation [37–39]. In addition to chromatin modifications, deposition and remodelling of histone variants, such as H2A.Z, were shown to be important for maintaining the heterochromatin/euchromatin boundaries [40] and proper gene activation [41–47].

Prp45/SKIP

Prp45/SKIP is a spliceosomal ‘rope protein’ spanning ~ 141 Å across the structure of the yeast/human spliceosome (Figure 1A; [51,53]). The protein contains regions of predicted intrinsic disorder and assumes an extended fold, contacting several splicing factors and all snRNAs involved in catalysis across the spliceosome [54,55]. Prp45 chain assumes similar conformation in all complexes where it is traceable, i.e. from B^act to post catalytic spliceosomes [54,56,57]. The protein’s chain passes very close to the U2/U6 catalytic core, but mutations of this region in S. cerevisiae or S. pombe (SNWKN) had no apparent cellular phenotype ([58]; unpublished results). Parts of the protein were not resolved in EM models and may contact additional, not yet recognized partners.

Figure 1. prp45(1–169) impaired the inducibility of PHO5 synthetically with rad6. (A) primary structure and interacting partners of Prp45. Predicted helices are drawn as dark boxes over the primary structure. The 3D distances were taken from the spliceosomal P complex [48]. The diagram of disorder content is based on the values of the AlphaFold prediction [49]. Values below 50 were found to predict intrinsically disordered regions with high reliability [50]. The vertical lines over the primary structure indicate the extents of truncation of the variants used in this study. Interacting partners above the primary structure are drawn based on the proximity between Prp45 and the indicated components in the yeast spliceosomal structures [51]. Contacts with Prp46 (NTR) and Prp8 (tri-snRNP) are the most extensive. Contacts with Prp22 (marked with *), which are not resolved in the cryo-EM structures, are drawn according to the published two-hybrid data [52]. (B) induction of the PHO5 gene in response to phosphate depletion was delayed in cells with truncated Prp45 as compared to WT. WT and prp45(1–169) cells were incubated in non-inducing conditions, washed, and transferred to a medium without phosphate. Total RNA was isolated at the indicated time points and reverse transcribed. RNA levels were quantified by qPCR, normalized to TOM22 reference, and related to the signal obtained from WT cells before the shift. p values were calculated for the comparisons between WT and prp45(1–169) strains and a mutant strain using the t-test (see Methods). Stars indicate differences with p < 0.05. (C, D) the effect of prp45(1–169) on PHO5 induction was tested with rad6Δ (C), paf1Δ, and set1Δ (D) using the same setup as in (B). Error bars represent standard deviations calculated from independent biological replicates. p values, which were obtained for comparisons between strains using the t-test with Holm correction for multiple testing, are listed in STab. 11 (see Methods).

Several structural elements stand out from the extended fold of Prp45. It is the loop-sheet motif in its N-region which wraps around Prp46, the motif which contacts the RES complex member Slt11/Ecm2, and a long helix which interfaces Cef1, Prp8, and Pml1 (RES). These motifs can be also predicted in AlphaFold (bypassing direct use of structural templates; [50]), suggesting that they form independently of other structures and may bind prospective partners. The extended regions between these motifs may act as ‘linkers’, contributing flexibility in the initial phases of spliceosome formation [59]. Prp45 may thereby increase the cooperativity of assembly of splicing components as well as promote phase separation.

The human homolog of Prp45, SNW1/SKIP, was implicated in extra-spliceosomal roles in transcription initiation (interacting with Ski, VDR, pRb [60,61] and elongation in Metazoa [62]). This effect can be rationalized by assuming that SKIP affects transcription through its interactions with the elongating RNA PolII holocomplex or through its participation in transcription related condensates. SKIP was found to interact with the elongation factor p-TEFb [62] and splicing factor U2AF65 [63], both of which bind elongating PolII [64,65]. The SNW1 homolog of S. pombe binds U2AF35 [66].

Our earlier data hinted at the ‘rope’ character of Prp45, when we found that the stepwise truncation of the essential splicing factor did not produce any discernible growth phenotype until the rather extensive ablation of 210 amino acids in prp45(1–169). Prp45(1–169) cells showed temperature-sensitive growth phenotype, cell wall deformities, and increased sensitivity to stressors [58]. At permissive temperature, pre-mRNAs of most intron containing genes in prp45(1–169) exceeded WT levels several fold [67], similarly to mutants of early spliceosome assembly step factors [68]. Using ChIP, we demonstrated that the extended chain of Prp45 is required for co-transcriptional spliceosome assembly, affecting U2 snRNP association with pre-mRNA and subsequent assembly steps. We considered the prp45(1–169) allele to be a tool with which to further probe the function of this splicing factor.

Here, we mapped the genetic interactions of the truncated allele prp45(1–169) genome wide. Some of the phenotypes were dependent on intron presence in the SRB2/MED20 gene, which codes for the ‘head’ component of the Mediator complex [69]. Remarkably, the interaction with htz1Δ persisted in the srb2Δi background and was evident even with less truncated alleles of prp45. The results show the importance of HTZ1 for coupling transcription and splicing and the regulatory significance of co-transcriptional splicing for low expression intron containing genes.

Materials and methods

Yeast strains, media, cultivation conditions

Yeast strains used in this study are listed in STab. S1. Cells were cultivated using standard cultivation conditions in YPAD (1% yeast extract, 2% peptone, 0.01% adenine, 2% glucose, and 2% agar for solid plates) or synthetic dropout medium (SD; Formedium) supplemented with appropriate amino acids, 2% glucose, antibiotics, and other drugs as needed. Mutations in PRP45 were accomplished by homologous recombination of the integration cassette generated by PCR from the template plasmid pFA6a-3 HA-NatMX6 [70]. Intron deletion from SRB2, HRB1 and VPS75 was done using the ‘delitto perfetto’ method [71]. In the first step, the intron sequence was replaced with the URA3 gene and in the second step, the URA3 gene was removed using synthetic oligonucleotides and 5-FOA selection. Successful intron removal was confirmed by PCR and sequencing. All oligonucleotides are listed in STab. S2. Strains carrying multiple mutations were prepared by de novo introduction of a specific mutation into the desired genetic background or by crossings as indicated in STab. S1.

To measure cell growth on the Bioscreen C (Oy Growth Curves Ab Ltd), pre-cultures were grown overnight at 30°C in an orbital shaker (5 ml) or a static incubator (honeycomb plate; 350μl). Cultivation was performed in 350ul of medium with high intensity shaking for 60 sec, alternated with pausing for 60 sec. O.D. at 600 nm was measured every 20 minutes. For VarioSkan Flash (Thermo Scientific) measurements, cells were incubated in 24 well plates in 700μl of medium inoculated with 5μl of overnight pre-cultures. The instrument settings were as follows: rotation speed 240 rpm, amplitude 17 mm, and O.D. at 600 nm. Measurements were taken every 10 minutes.

In phosphate shift experiments, cells were cultivated in synthetic medium to mid-log phase, washed three times with the same volume of prewarmed water, resuspended in phosphate free medium, and cultivated for the indicated times. Cells were then collected by centrifugation (1000 g for 3 min at RT) and frozen in liquid nitrogen. Samples were subjected to Northern blot or qPCR analyses (see below).

SGA analysis

SGA analysis was performed as described in [72] in two replicates. The query strain (KAY27) was prepared by introducing the mutation prp45(1–169) in the strain Y9072. Y8835 served as a control strain expressing the nourseothricin N-acetyl transferase cassette. The query and control strains were crossed with the deletion mutant array [73]. All manipulations were performed robotically using RoToR HDA (Singer Instruments) in 1536-dot format (four replicate spots per strain). The final plates were incubated at 30° and 35°C and the growth of the double mutant progeny was documented after one, two and three days of growth using a Canon PowerShot A640 camera. Colony sizes were analysed using computer scoring software (http://sgatools.ccbr.utoronto.ca/). Hits were confirmed by random spore analysis, measurements of growth rates in liquid culture, or by tetrad dissection after crossing the strains de novo. Except for hits where the colonies were consistently much smaller at all SGA plates across all temperatures, all other SGA hits were validated. Those that remained dubious were excluded or the double mutants were prepared de novo (see STab. S4 for details). Primary SGA data, list of hits containing their computed scores and verification status, and the results of the Gene Ontology (GO) term analysis are available in STab. S3, S4, and S5, respectively.

RNA analysis

RNA was isolated using the MasterPure Yeast RNA Purification Kit (Epicentre Biotechnologies), including the DNaseI treatment step, essentially according to the manufacturer’s instructions. Reverse transcription was done using the RevertAid^TM First Strand cDNA Synthesis Kit (ThermoFisher) according to the manufacturer’s protocol. Two μg of RNA were used as a template and reactions were primed with random hexamers.

qPCR was performed on a LightCycler 480 II (Roche). Each reaction consisted of 2 μl of HOT FIREPol® EvaGreen® qPCR Supermix (Solis Biodyne), 0.3 mM primers, and 1.5 μl of cDNA (total volume of 10 μl). Each sample was run in triplicates. Results were calculated using the ΔΔC_t method [74]. For statistical analysis, Student’s t-test was employed using ΔCt data and the correction for multiple testing where appropriate (see STab. S11 for p-values). The calculations were performed with the ‘R’ statistical package version 3.2.3 (www.r-project.org/) using the t.test() function with parameters ‘paired = FALSE, alternative = “two.sided’’’. The function p.adjust() with Holm correction was used; p < 0.05 was considered significant.

Data analyses

RNA-seq data of the prp45(1–169) strain were published in [67] and are available from the Array Express Database (https://www.ebi.ac.uk/arrayexpress/) under accession number E-MTAB-5149. Where appropriate, RNA-seq data were normalized to obtain TPM (transcripts per million) values to account for transcript length and sequencing depth. TPM is described in http://www.arrayserver.com/wiki/index.php?title=TPM. Splicing efficiency was calculated as described in [75]. Briefly, splicing efficiency for each intron was determined using the formula: Efficiency = transread count/intron end base coverage (the 5’splice site was used in calculations). While the amount of transreads reflects mature mRNA abundance, intron coverage reflects pre-mRNA levels.

The yeast genetic interaction similarity network was generated by the spring-embedded network layout [4]. SAFE (Spatial Analysis of Functional Enrichment; [76]) was used to map the GO terms over the network of genetic interaction similarities. The select lists of genes were overlaid over the regions of genetic interaction similarity network using the web interface of TheCellMap database (https://thecellmap.org/).

Results

prp45(1–169) impaired the inducibility of the PHO5 and GAL1 genes

The prp45(1–169) truncation produced a mild phenotype, yet it frustrated the function of Prp45, so it could be used as a tool. In preliminary experiments, we found the PHO genes to be among the downregulated genes in the prp45(1–169) mutant. Because these genes were extensively studied with respect to gene induction and chromatin remodelling at promoters [77,78], we decided to examine their expression in more detail. We assessed the induction kinetics of several genes of the PHO regulon upon phosphate removal, which is known to induce the expression of factors regulating phosphate homoeostasis [77,78]. The tested genes exhibited an induction delay in the prp45(1–169) mutant. The shift, which was most pronounced in PHO89, PHO12, and PHO5, was also temperature dependent (Figure 1B and SFig. S1A). Because of the capacity of SNW1/SKIP to affect both transcription initiation and elongation [60,62], we examined genetic interactions of prp45(1–169) with a panel of alleles of chromatin regulators which were known to interact with pre-mRNA splicing and processing. We found negative genetic interactions with a few alleles, including bur2Δ, paf1Δ, lge1Δ, and rad6Δ (SFig. S1B).

We next probed the inducibility of PHO5 in some of the double mutants identified in our preliminary screen. The double mutant rad6Δ prp45(1–169) showed a synthetic effect (Figure 1C), while set1Δ, which did not synthetically impair the growth of prp45(1–169), did not impact PHO5 inducibility (Figure 1D, right panel). paf1Δ, which interacted strongly with prp45(1–169) in the growth test, abrogated the induction to a great extent (Figure 1D, left panel; see also [79]). The inducibility defect of the PHO5 gene was not limited to the PHO regulon. We found that the induction of the GAL1 and GAL10 genes upon the glucose to galactose nutrient shift was likewise hampered (SFig. S1C). We reasoned that the above induction defects were likely part of a general deficiency in gene inducibility, either in the general mechanism of transcription induction or in the permissibility of chromatin to the inducing effects of regulators.

We also noticed that prp45(1–169) had only limited impact on splicing of a reporter minigene derived from ACT1 ([58]; SFig. S1D). The intron with WT splice sites showed no defect in the formation of spliced product, both in prp45(1–169) cells and in double mutants with paf1 or rad6. Only the recombinant introns 3’gAG or 5’A3C, which were limiting for the first or second splicing step [80,81], respectively, showed splicing defects in the prp45(1–169) cells. The rad6Δ and paf1Δ mutants spliced the limiting substrates well and did not enhance the defect of prp45(1–169).

prp45(1–169) genetically interacted with alleles encoding chromatin regulators

We decided to take advantage of the thermosensitive allele prp45(1–169) (see Figure 1A) and examine its genetic interactions on the whole genome level. To this end, we employed the genome-wide SGA screen with a collection of 4291 viable deletion strains using prp45(1–169) as a query [72]. The deletion strains bearing prp45(1–169) were considered as hits when the average difference in colony size as compared to the control strain was at least 10% in at least one cultivation temperature. We found that prp45(1–169) negatively genetically interacted with 175 alleles of the deletion collection; a complete curated list of our SGA findings is available in STab. S4.

The genetic interactome was characterized by a high proportion of genes participating in transcription and chromatin regulation, and mRNA processing, which account for 33 and 14% of NGIs, respectively, as well as in vesicle trafficking, inositol and lipid metabolism, signalling, and DNA replication and repair (Figure 2A). A full list of NGIs according to GO terms is available in STab. S5. Several of the GO term categories showed high enrichment in NGIs, such as histone modifications and exchange, chromatin remodelling, regulation of RNA PolI transcription and nucleotide excision repair (SFig. S2A). As expected, the genes participating in spliceosome assembly, splicing and mRNA processing also showed high prevalence of NGIs. The NGIs included components of complexes regulating the chromatin environment – SWR1, INO80, SAGA, Paf1, Set3, Rpd3L/Rpd3S, and COMPASS. The genetic interaction of prp45(1–169) with htz1Δ was the strongest NGI that we found in our screen. The double mutant spores had problems germinating, but we eventually succeeded in isolating the strain, which was severely growth impaired (see Figure 2B upper left panel and Figure 4C). In addition to htz1Δ, SWR1C components (Figure 2B) and other chromatin regulators also showed strong double mutant phenotypes.

Figure 2. Synthetic genetic array analysis revealed strong negative genetic interactions of prp45(1–169) with genes involved in transcription and chromatin regulations. (A) pie-chart of manually curated categories illustrating the proportions of negative genetic interactions (NGIs) of prp45(1–169). The categories ‘transcription and chromatin’ and ‘RNA processing’ represent 33% and 14% of the NGIs found. Genes were grouped into categories using their GO terms as listed in STab. 6. (B) the prp45(1–169) allele had strong negative interactions with deletions of HTZ1 and SWR1 complex members. prp45(1–169) mutant cells were crossed with htz1Δ, swr1Δ, vps71Δ, vps72Δ, swc3Δ, and swc5Δ strains from the yeast deletion collection. Haploids with indicated combinations of mutations obtained from tetrad dissections of diploid strains harbouring an URA3 plasmid with full length PRP45 (p416ADH-His6-PRP45; [22]) were cultivated to the mid-log phase, serially 5× diluted and spotted on SD plates and SD plates with 5-FOA to get rid of the complementing plasmid. (C) prp45(1–169) shares a high proportion of its genetic interactions with other components of transcription and chromatin regulatory complexes. We overlaid the NGIs of prp45(1–169) onto the network of 165 chromatin regulators constructed in a previously published perturbation analysis [23]. The parts of the network which were enriched for the NGIs of prp45(1–169) (see STab. 7) were re-clustered, using the relative proportions of shared NGIs/all NGIs as a measure (increasing Jaccard indexes were used to cluster the heat-map). As a data source of NGIs genomewide, BioGRID (S. cerevisiae; thebiogrid.Org; downloaded 210127) categories ‘negative genetic’, ‘synthetic growth defect’, ‘synthetic lethality’ and ‘phenotypic enhancement’ were used. Multiple gene list comparator tool (https://www.molbiotools.com/listcompare.Php) was employed for the comparisons between genetically interacting genes and the calculation of Jaccard indexes. Interaction profiles of individual genes were hierarchically clustered and visualized as a heatmap.

We compared our list of NGIs with previously published networks of genes involved in chromatin regulation and RNA metabolism. First, we analysed the distribution of prp45(1–169) NGIs among the clusters of chromatin regulators constructed in the perturbation analysis of Lenstra and co-authors [23]. Clusters containing members of the Paf1, Rad6/Bre1, Swr (including Htz1), and Set3 complexes showed the highest overlap with our NGI list (see Figure 1 in [23] and STab. S7). We constructed a matrix of all pairwise intersections of NGIs among the alleles from the enriched subclusters (37 alleles; STab. S7) and prp45(1–169) using the Multiple gene list comparator (STab. S8). The analysis showed that prp45(1–169) shared most genetic interactions with the Swr1 complex components vps72Δ, swc3Δ, vps71Δ, arp6Δ, and swr1Δ (sorted from the highest Jaccard index; [82]). The matrix is presented in Figure 2C as a heatmap, using the relative proportions of (identical NGIs)/(all NGIs) as a measure. The map illustrates that prp45(1–169) is an integral part of this highly interconnected group. The density of genetic interactions among the chromatin regulators and their genetic interaction partners in our map recapitulated the clusters based on transcriptomics profiles of deletion mutants (see the cladogram in Figure 1 in [23]). The connectivity among the SWR1 complex members plus htz1Δ (subclade 6D in STab. S7) and prp45(1–169) is further documented in a node diagram constructed in Cytoscape (SFig. 2B). Importantly, the impact of the gene deletions on PRP45 expression was mostly minimal, with the lowest and highest mut/WT fold changes being 0.7 and 1.25, respectively (see Sheet2 in STab. S7), which included htz1Δ (0.8–0.86). In the htz1Δ dataset of Gu and co-authors [83], the mut/WT ratio for PRP45 was 0.68.

Second, we examined the distribution of our NGIs among the splicing- versus retention-promoting gene deletions analysed by Sorenson and Stevens [84]. Using a unique single-cell yeast reporter, the authors tested 4,967 gene deletions for their effect on intron-bearing reporter cassettes. The gene deletions promoting intron retention included 17 NGIs, among which were the components of the Swr1 complex (swc3Δ, arp6Δ, vps72Δ, vps71Δ), htz1Δ, spt8Δ, and set2Δ (see Figure 7 in [84]; STab. S9). Mediator subunits Med9 and Med31 (cse2Δ and soh1Δ) were also part of this cluster. Only four null alleles promoted splicing, while 88% of prp45(1–169) NGIs were among the alleles which had only a mild effect on intron retention and remained outside of the clusters mentioned above.

Last, we compared the overlap of our NGI list with mRNA surveillance genes, which were functionally analysed by Sun and co-workers [85]. These authors used metabolic RNA labelling to cluster 46 surveillance factors according to their effects on mRNA degradation rates. Most NGIs of prp45(1–169) overlapped with a cluster containing components of the LSM1–7 complex (lsm1Δ, lsm6Δ, lsm7Δ), which targets RNA substrates for decapping, pat1Δ, an enhancer of decapping, and pop2Δ, a deadenylation factor (see Figure 5 in [85] and the complete list of overlaps in STab. S10). In contrast, we did not find GIs with components of cytosolic NMD pathway xrn1Δ or upf3Δ (SFig. S3A) and these NMD alleles did not produce any synthetic effect with respect to pre-mRNA accumulation when tested as double mutants with prp45(1–169) on ECM33 [88].

prp45(1–169) preferentially impacted mRNA levels of low expression intronic genes

The truncation prp45(1–169) had only a modest effect on the transcriptome as a whole as judged from mean total RNA counts compared between mutant and WT [67]. We found 285 differentially expressed genes, out of which 17 were DOWN > 2 times and 9 were UP > 2 times. The lowest and highest log2(fold change) values were −2.7 and + 2.9. Among the differentially expressed genes, 33 contained an intron and 12 of these were ribosomal protein genes. Using our workflow for the calculation of splicing efficiency [75,89], we re-examined the prp45(1–169)-affected transcripts in comparison with our SGA hits and other features of intron containing genes in the literature (Figure 3). The pipeline was used to identify potential splice sites genome wide and calculate splicing efficiency as ratios of transread count/intron first base coverage. To characterize the effect of the truncation on mRNA levels, differentially scored transread ratios mut/WT were evaluated (see Methods). For most intron containing genes, these ratios were buffered around WT levels (1.1–0.8). We noticed, however, that the transread ratios decreased more substantially in intron containing genes with low levels of expression. The downregulated differentially scored transreads are referred to as ‘subset_45’ in the text (Figure 3A). Unlike the transread ratios, the relative splicing efficiencies in the mutant were uncorrelated with gene expression levels (Figure 3B). Considering intron length, subset_45 genes clustered with short intron-containing genes (Figure 3C), and the graph recapitulated the pattern from Figure 3A. This is probably because most intron containing RPGs, which are highly expressed, have longer introns (>200 nt), whereas non-RPGs cluster at short intron lengths.

Figure 3. Negative genetic interactions of prp45(1–169) do not overlap with genes impacted on mRNA levels. To compare the genetic interactions of prp45(1–169) with the effect of the prp45(1–169) mutant allele on transcription, we plotted mRNA levels of intron containing genes as transread ratios mutant/WT, or splicing efficiency, using previously published RNA-seq data and our workflow [67,75]. (A) scatter plot shows the effects of prp45(1–169) on mRNA levels of intron containing genes, expressed as transread ratios mutant/WT, plotted against mean gene expression levels (total RNA count across all samples). Differentially scored transreads are shown in red. (B) for comparison, the effects of prp45(1–169) on splicing efficiency (S.E.) is plotted as in (A). Splicing efficiency was calculated by dividing transread counts by 5’ intron end first base coverage and expressed as mutant/WT ratios. Black data points are genes with sufficient coverage in both WT and mutant strain, i.e. ≥5 transreads and ≥5 reads covering intron end base. Gray data points represent genes which did not fully meet our criteria of sequence read data coverage (see S Fig.3A for details). (C) scatter plot showing relative transread ratios of intron containing genes (same data set as in (A)) as a function of intron length. (D) Genetic interaction enrichment landscapes generated using SAFE annotated genetic interaction similarity network of S. cerevisiae [4,76]. The maps were obtained from the web interface of TheCellMap database (https://thecellmap.org/). Downregulated differentially scored transreads of prp45(1–169) (Subset_45_genes) (STab. 12) (1), complete set of intron containing genes of S. cerevisiae (STab. 13) (2), and all NGIs we found (SGA hit list including manually curated alleles; STab. 14) (3) were visualized in green overlay using the interface. NGIs of prp45(1–169) but not the subset_45 genes showed overlap with the subnetworks of transcription and chromatin regulators. (E)–(G) relative mRNA levels of intron containing genes (same data set as in (A)) were plotted as a function of splicing and chromatin related parameters measured previously. Relative co-transcriptional efficiency (taken from [5]) (E), relative Htz1 density at transcription start sites (TSS) of genes [83] (F), and relative nucleosome turnover at transcription start sites [87] (G) were plotted on the x axis. Only the genes present in both datasets used for the comparison are shown. In (F), the graph does not show the gene YJR145C (coordinates 9.06112; 0.017245623). The parameter ‘observed delta’ increases with increasing co-transcriptional splicing efficiency (‘1’ = 100% co-transcriptional splicing). The parameter ‘Z-score’ increases with decreasing nucleosome turnover. Highly expressed genes, which are mostly ribosomal protein coding genes, cluster at longer intron length (in C), high co-transcriptional splicing efficiency (in E), low levels of Htz1 at transcription start sites (in F), and low relative nucleosome turnover at transcription start sites (in G). See text for discussion.

To compare the subset_45 genes with our SGA hits, we used the NGI profile similarity map of Saccharomyces cerevisiae (Figure 3D [4,90–92]. Genetic interactions between categories of chromatin regulators and components of gene expression/surveillance systems were analysed using Cytoscape [93]. We compared the projections of our SGA hits, the subset_45 genes, and the complete set of intron containing genes. Clearly, only the SGA hit map overlapped the clusters of genes which are associated with mRNA processing, chromatin, and transcription (Figure 3D).

To identify features which would characterize the subset_45 genes, we looked into published data on co-transcriptional splicing efficiency [5], Htz1 levels measured by ChIP [83], or relative nucleosome turnover rates [87]. The graphs in Figure 3E, F show the distribution of the differentially scored transreads (see above; genes in red) in the datasets. Relative mRNA versus co-transcriptional splicing efficiency plot (= parameter ‘observed delta’ calculated in [5]; Figure 3E) showed that the genes with the highest parameter values (= highly efficiently spliced co-transcriptionally; observed delta between 0.7 and 1) are the least affected in the prp45(1–169) mutant. These genes are enriched in RPGs and thus highly expressed (out of 54 genes with ‘observed delta’ value > 0.8, 38 are RPGs).

Next, we compared our RNA-seq data with the data on Htz1 levels measured by Gu and co-workers ([83]; Figure 3F). In the prp45(1–169) mutant, genes containing very low Htz1 signal at transcription start site clustered at higher average relative mRNA than the rest of intron containing genes. This clustering again reflects the properties of RPGs – most of the genes with Htz1 ChIP signal below ‘8’ are RPGs. When the intron containing genes were sorted according to expression level, the highest expressors showed the lowest Htz1 ChIP-Seq signal around transcription start site and the lowest antisense transcription (see SFig. 3B).

Finally, the scatter plot of relative mRNA versus relative nucleosome turnover at transcription start sites (Z-scores; [87]) showed that our differentially scored transreads tend to be enriched among high turnover rate genes (genes with low Z-scores; Figure 3G). As Dion and co-workers demonstrated, high nucleosome turnover correlated with increased Htz1 content. We found no dependence of relative splicing efficiency on Htz1 content, co-transcriptional splicing efficiency, or the length of intron or ORF (SFig. 3C).

Intron deletion of SRB2 repaired part of the phenotypes of prp45(1–169) and double allele mutants

prp45(1–169) reduced mRNA levels of a group of low expression intron containing genes, which could have contributed to the synthetic interactions that we observed. We tested this hypothesis by deleting introns of three such genes and measuring the phenotypes of the intronless mutants. We prepared intronless knockins (Δi) of genes coding for Hrb1, poly(A+) RNA-binding protein involved in spliced mRNA export; Vps75, histone binding HAT activator participating in DSB repair; and Srb2, subunit 20 of the Mediator complex [69], all of which have functions related to RNA metabolism. Intron deletions of either of these genes increased their mRNA levels in the prp45(1–169) background, with SRB2 reaching ~ 2-fold excess over WT mRNA (SFig. S4A-I). Neither of the knockins influenced the growth of prp45(1–169) mutants (SFig. S4A-II and Figure 4B).

Figure 4. Intron deletion of SRB2 repaired part of the growth defect of prp45(1–169). (A) intron deletion in SRB2 partially repaired the growth defects of prp45(1–169) double mutants carrying deletions of RAD6, PAF1, and GCN5. Cells were cultivated to mid-log phase, serially diluted (ratio 1:4), spotted onto YPAD plates, and incubated at 30°C or 37°C for 6 days. (B) intron deletion in SRB2, but not in HRB1 or VPS75, partially repaired the growth phenotype of prp45(1–169) rad6Δ double mutant. The growth of liquid culture in the YPD medium at 37°C was measured in VarioSkan. (C) SRB2 intron removal had only negligible capacity to rescue the strong negative genetic interaction between htz1Δ and prp45(1–169). Cells were cultivated to mid-log phase, serially diluted (ratio 1:4), spotted onto SD plates containing 5-FOA to dispose of complementing plasmid (p416ADH-His6-PRP45; [22]), and incubated at 30°C, 37°C, or 16°C for 7 days.

We then examined the effects of intron deletions on double and triple mutants, re-testing some of the alleles that we found previously in the SGA screen. Intron deletions of either HRB1 or VPS75 showed negligible impact on the growth phenotypes of rad6Δ, paf1Δ, or gcn5Δ as single or prp45(1–169) double mutants. By contrast, srb2Δi partially rescued the growth defect in prp45(1–169)rad6Δ and prp45(1–169) gcn5Δ at 37°C, while it almost fully repaired the growth defect of prp45(1–169) paf1Δ (Figure 4A, B). Intriguingly, srb2Δ negatively genetically interacts with gcn5Δ but not with either rad6Δ or paf1Δ (BioGRID Version 4.4.223). The genetic interaction between prp45(1–169) and htz1Δ was strong even in the srb2Δi background, despite partial growth improvement of prp45(1–169) htz1Δ srb2Δi at 30°C (Figure 4C). We did not observe any growth of the triple mutant at 37°C or 16°C.

We also tested the effect of intron deletions on the induction delay of PHO5 (Figure 1) and found that intron deletion of SRB2, but not HRB1 or VPS75, repaired the kinetic defect caused by PRP45 truncation (SFig. 4B). Decreased Srb2 was thus most likely the cause of the delayed induction seen with PHO and GAL genes (see Figure 1B and SFig. 1B), as they depend on Mediator recruitment to achieve full activation [94–96].

Less extensive truncation of Prp45 still exhibited a synthetic phenotype with htz1

Because of the strong synthetic growth defect of prp45(1–169) htz1 mutant (Figure 4C), we constructed less extensive PRP45 truncations (see truncation extents in Figures 1A&5A) and compared their phenotypes with the prp45(1–169) version. Single alleles prp45(1–247) and prp45(1–330) had no apparent growth defects at 16/30/37°C (Figure 5B), but they still accumulated excess levels of pre-mRNAs. The extent of pre-mRNA accumulation depended on Prp45 truncation but was not proportional to its extent, i.e. it was more severe in prp45(1–330) than in prp45(1–247) (Figure 5D). This would suggest that the truncation between aa 247 and 330 impaired the recruitment of some factors which inhibited retention. Alternatively, the Prp45 chain between aa 247 and 330 could interfere with the spliceosome assembly process due to the lack of the protein’s C-terminus.

Figure 5. Less extensive truncation of Prp45 is sufficient to cause htz1-dependent pre-mRNA accumulation. (A) extended chain of Prp45 connects distinct parts of the spliceosomal architecture. In yeast B^act spliceosome (structure 5gm6; [54]), Prp45 contacts extensively Prp8 (grey) and more than 10 other proteins. We highlighted Prp46 (dark blue), Cef1 (purple), Bud13 (green), Snu17 (blue grey), Pml1 (light blue), and Hsh155 (ocre). The Prp45 fragments are coloured in dark red (1–169), orange (170–247), light orange (248–330), and yellow (331–350). The complex was visualized using ChimeraX [86]. Prp45(1–169) lacks parts of the chain that lines the interface between Prp8 (RT-like domain) and Cef1 (left). Prp45(1–247) is devoid of the contacts to the RES complex components (Pml1, Snu17 and Bud13) and the RNAse H-like domain of Prp8 (right). Truncation of the C-terminal 49 amino acids in Prp45(1–330) should impinge on the interactions with Prp8, Ist3, and Hsh155. Amino acids 1–30 and 350–379 (C-term) were not resolved in the structure and may contain additional interacting partners. (B) growth rate comparison of prp45(1–169), prp45(1–247), and prp45(1–330) cells. Cultivations in YPAD at 30°C were monitored using VarioSkan. In contrast to prp45(1–169) cells, the growth of prp45(1–247) and prp45(1–330) mutants was indistinguishable from the WT strain. (C) Intron deletion of SRB2 partially repaired the cold sensitive growth phenotype of htz1Δ mutant but not of the double mutant htz1Δ prp45(1–330). Cells were cultivated to mid-log phase, serially diluted (ratio 1:4), spotted onto YPAD plates, and incubated at 16°C or 30°C for the indicated number of days. (D) prp45(1–247) and prp45(1–330) mutants accumulated increased levels of pre-mRNA. The mRNA (left) and pre-mRNA levels (right) of ECM33 and ACT1 genes were measured by qPCR in WT, prp45(1–169), prp45(1–247), and prp45(1–330) cells. While the mRNA levels were approximately the same in all four strains, the pre-mRNAs were accumulated to the highest extent in prp45(1–169), followed by prp45(1–330). qPCR values were normalized to TOM22 mRNA and expressed relative to WT strain. Error bars represent the standard deviation of four biological replicates for WT and prp45(1–169) cells and six biological replicates for prp45(1–247) and prp45(1–330) cells. (E) prp45(1–330) and htz1Δ negatively interacted on the level of pre-mRNA accumulation. The mRNA (left) and pre-mRNA levels (right) of ECM33 and COF1 genes were measured by qPCR in WT, htz1Δ, prp45(1–330) and htz1Δ prp45(1–330) cells. The pre-mRNAs showed highest accumulation in the double mutant. qPCR values were normalized to TOM22 mRNA and expressed relative to WT strain. Error bars represent the standard deviation of 8 biological replicates. Statistical significance of the differences between strains in (D) and (E) is indicated as (*) for p ≤ 0.05, (**) for p ≤ 0.01, and (***) for p ≤ 0.001 based on the t-test with Holm correction for multiple testing (see methods and STab. 11).

Allele prp45(1–330) showed genetic interaction with htz1Δ on the level of growth at 16°C. This phenotype was not repaired by SRB2 intron deletion (Figure 5C). Importantly, prp45(1–330) also showed synthetic effect with htz1Δ on the level of pre-mRNA accumulation. We measured pre-mRNA and mRNA levels of ECM33, which is an efficiently spliced gene, and COF1, which is spliced less efficiently, and found the effects in both genes. Even the least extended truncation of PRP45 thus resulted in genetic interaction with htz1Δ on the level of pre-mRNA accumulation (Figure 5E).

The hyperaccumulation of pre-mRNA, brought about by truncated prp45 alleles, depends on the degree of truncation. Intriguingly, the defect can be repaired in part by lowering the demand for spliceosomes. We showed this by incubating cells with the TOR1C inhibitor rapamycin, which inhibits the transcription of RPGs in yeast via a feedback mechanism [97,98] (SFig. 5). Stalling of RPG transcription and splicing then lowers the demand for spliceosomes, as splicing of RPGs represents most of the splicing in S. cerevisiae. Rapamycin treatment lowered the high pre-mRNA levels of ECM33 and ACT1 in the mutants, but the pre-mRNAs remained at > 2 times WT levels, even after 30 minutes of incubation. This suggests that there are at least two components of the defect, one of which is rescued by increasing the availability of the spliceosomal components.

Discussion

Genetic interactions of prp45(1–169) overlap with the networks of chromatin regulators

Because of its flexibility and propensity to interact with multiple partners, Prp45 could bring together components during spliceosome assembly as well as couple splicing to other processes [19,51,53,99]. The truncation prp45(1–169), which drastically lowered the valency of this linker protein (see Figure 1A), was nevertheless viable in S. cerevisiae and produced only a mild phenotype (see Introduction; [58,67]). Notably, C-terminal truncations of Prp45 became lethal when the N-terminus was also ablated [58]. In this study, we performed an SGA screen to identify negative genetic interactions of a truncated version of PRP45 (prp45(1–169)). Our data extend the networks of genetic interactions between splicing and transcription related factors, chromatin remodellers and modifiers observed previously [1,46,47].

The prp45(1–169) genetic interaction map that we obtained revealed htz1Δ and SWR1 complex components as the strongest interacting alleles (Figure 2). The histone variant Htz1 of S. cerevisiae (H2A.F/Z family of H2A variants) is deposited mostly in -1 and +1 nucleosomes at RNA PolII promoters, and its controlled turnover is important for proper gene activation [41,42,100,101]. Chromatin remodelling complexes SWR1C and INO80C are responsible for its deposition and targeted eviction, respectively. Alleles of SWR1C components and htz1Δ were reported to interact genetically with other splicing factors in S. pombe and S. cerevisiae [4,46,47]. A genetic screen with swr1 in S. pombe identified splicing complexes across the splicing cycle, as well as htz1 (pht1 in S. pombe), Mediator complex and pTEFb, as enriched for NGIs [46]. In S. cerevisiae, targeted genetic screen with htz1 found NGIs with alleles from all major splicing subcomplexes. Deletions of U2 snRNP related MSL1, LEA1, or SNU17 conferred synthetic lethality [47]. The effect of htz1 deletion on splicing efficiency was measured [47], but it was much less severe than the phenotype of prp45(1–169). We did not find any correspondence between the two data sets.

There is a notable discrepancy between the ample genetic interactions connecting the alleles in splicing and chromatin GO_TERM categories and the relatively fewer instances of documented physical coupling between proteins (see Figure 2 in [1] and references therein). Nevertheless, overlaps in genetic networks are interpreted in the literature to reflect functional similarities [90]. For example, general transcription factors show gene specific defects upon deletion, which can be used to cluster them to reveal functional interconnections [102,103]. We noticed that the alleles of chromatin regulators that negatively genetically interacted with prp45(1–169) form a highly connected group ([23]; see Figure 2D). Remarkably, the extent of connectivity of prp45(1–169) to nodes of the mentioned subnetwork is similar to the number of edges among the chromatin components themselves, suggesting functional involvement of Prp45 (see both the subcluster 6D diagram in SFigure 2C as well as our reclustered heatmap in Figure 2C).

Besides the SWR1 and INO80 complexes, several other heterooligomers of chromatin regulators were enriched in NGIs with prp45(1–169), such as the complexes of Paf1 [104,105], Rad6/Bre1 [106], Set1/COMPASS [107,108], Set3/Rpd3L [109], and Set2-Rpd3S [18,110]. All of these complexes are known to couple transcription with transcript processing through the changes of chromatin environment [111,112] and a number of their components genetically interact with htz1Δ. rad6Δ or bre1Δ is lethal in the absence of HTZ1 [113], which hints at partial redundancy of H2B ubiquitylation and H2A.Z deposition. set3Δ also becomes growth limiting in htz1 knockout, with loss-of-function suppressors being the SWR1 complex, the H2B deubiquitination module of SAGA, and Set1, among others [109].

Expression of low copy-number genes is particularly sensitive to prp45(1–169) truncation

Prp45 truncation resulted in the decrease of splicing efficiency for most intron containing genes, while the span of the splicing efficiency levels remained like WT. While splicing efficiency seems to have been affected globally, more pronounced decreases of mRNA levels were mostly confined to low expression genes (subset_45; Figure 3A). Two nonexclusive possibilities can be considered. The mutation could have a gene specific effect, similar to what was found for general splicing factors in S. cerevisiae [114]. However, we did not find any correlations which would hint at any common intronic feature of the affected genes in the subset_45. Alternatively or in part, Prp45 truncation could impair splicing (spliceosome assembly) globally, leading to generalized splicing efficiency decrease. The low expression genes would then be disproportionately affected because of their lower capacity to adjust turnover rates and thus buffer mRNA levels (see Fig 7 in [115]).

Splicing of the Mediator component may explain part of the prp45(1–169) phenotype

Intron deletion of the Mediator component SRB2 repaired part of the growth phenotype of prp45(1–169)-interacting alleles (Figure 4). It should be noted that srb2∆i elevated SRB2 mRNA levels to ~ 2 times WT levels, so some of the effects might be caused by the overexpression (SFig. S4A-I). While Srb2 decrease (=defect in Mediator function) may explain part of the negative synthetic growth defects of the SGA-positives, the srb2∆i does not repair the phenotype fully and is therefore unlikely to explain the whole NGI set. Also, the NGIs of the two alleles do not overlap. Notably, SRB2 intron removal partly repaired the cold sensitivity of htz1Δ, but not of the htz1Δ prp45(1–330) double mutant (Figure 5C). SRB2 mRNA levels in htz1Δ were previously found to be similar to WT [47], but because SRB2 intron is spliced inefficiently, its deletion may lead to SRB2 mRNA increase in htz1Δ. Similarly to prp45(1–169) htz1Δ srb2Δi, the prp45(1–330) htz1Δ srb2Δi mutant showed synthetic growth defect at 16°C, which was independent of SRB2 splicing. The genetic interaction between htz1Δ and prp45 can thus be demonstrated under conditions where just the C-terminal 49aa of Prp45 are ablated and it persists in the srb2Δi background.

SRB2 splicing was reported to be highly dependent on the functioning of the RES complex [116]. This is intriguing as Prp45 contacts the RES components in the spliceosome (Figure 5A; [54,117,118]) and its ablation could result in impaired RES recruitment. We compared the effect of prp45(1–169) on relative pre-mRNA levels with the effect of RES complex alleles bud13Δ and ist3Δ/snu17Δ [119,120], but we found no correlation. RES component Ist3/Snu17 is required for Mer1 dependent induction of AMA1 and other Mer1-regulated meiotic genes [121]. Because prp45(1–169) cells undergo meiosis, the mutant must allow for the induction of AMA1, which implies that the recruitment of RES is achieved even in the absence of Prp45 chain (Figure 5A; [67]. We conclude that prp45(1–169) does not phenocopy RES complex alleles.

The Mediator complex is functionally linked to remodelling of chromatin in yeast promoters [95,96,122,123]. This can be illustrated by NGIs between alleles of Mediator components and complexes of SAGA, Paf1, and Ccr4-Not in the network of chromatin regulators constructed by Lenstra and co-workers (see Figure 4 in [23]). Two of the Mediator complex alleles, srb2Δ and soh1Δ, are included in the subnetwork enriched in prp45(1–169) NGIs (Figure 2C). The co-clustering of srb2Δ and prp45(1–169) may reflect the inefficient splicing of SRB2.

The vulnerability of splicing may be a useful thing to have

The outcome of the prp45(1–169) truncation can be summarized as an Htz1-sensitive failure to couple splicing to ongoing transcription. The interaction with HTZ1 was bordering synthetic lethality and persisted even in the srb2Δi background. Less truncated variants of Prp45 maintained synthetic effect with htz1Δ on the level of pre-mRNA accumulation. We assume that the truncation of the ‘rope’ of Prp45 hampered the processes of snRNP recruitment during co-transcriptional spliceosome assembly [67]. For the same reason, the recycling of snRNPs may have been affected as well, breaking the ‘supply chain’ of splicing snRNPs. Intriguingly, lowering the demand for spliceosomes (see rapamycin experiment in SFig. S5A) partly rescued the splicing efficiency defect. The ‘supply chain’ problem could be related to deficient Prp22 recruitment [58], presumably because the truncation leaves Prp45(1–169) without the Prp22 interacting region (see Figures 1A and 5A). It was suggested that Prp22 and Prp16 depletion causes stalling of the first assembly steps via precursor depletion [124].

In addition to delayed recruitment of trans-acting splicing factors, the truncation of Prp45 could affect the subsequent steps of mRNA maturation, including export. Because of the phenotype of its truncated versions, Prp45 could be regarded as a factor contributing to early steps of intron recognition. Defects in early acting splicing factors were found to be associated with intron retention [125]. It was assumed that the defective intron recognition [125,126] made the pre-mRNAs prone to escape because they would no longer interact properly with the surveillance and export systems [127–129]. As these latter systems also depend on nuclear structure [130] and chromatin state, including Htz1 metabolism [131], they also must be taken into account when considering the crosstalk between the splicing defect and htz1 deletion [132]. In our screen, the alleles of the perinuclear quality control factors Mlp1 and Mlp2 [130] showed synthetic growth defects with prp45(1–169).

The htz1-sensitive phenotype of the prp45(1–169) mutant can be interpreted using any of the three models which describe the coupling between transcription and splicing in the literature (see Introduction). In the direct physical coupling scenario, Prp45 truncation could result in severing some of the specific physical contacts which are necessary to recruit the splicing apparatus to the emerging RNA. As part of an assembled spliceosome, the C terminal part of Prp45 is in contact with several Nineteen complex proteins, such as Cef1 and Syf2 (Figure 1A [133]). Nineteen complex is in turn implicated in connecting the transcription, splicing, and export complexes together with chromatin regulators [19,112,134,135]. The recruitment of the Nineteen complex, together with Prp45 itself, to some intron containing genes was assisted by the chromodomain protein Eaf3 [19]. In our screen, however, eaf3 did not qualify as a hit. The physical connections between the complexes participating in gene expression may be redundant [13,136], some of them serving a particular type of promoter, chromosomal position of a gene or export pathway [137,138].

In the kinetic coupling scenario, the absence of splicing complexes from RNA PolII during delayed assembly in prp45(1–169) cells could affect RNA PolII parameters. Interference with splicing in budding yeast was found to affect RNA PolII distribution along intron containing genes, which led to the proposal of splicing-coupled transcription checkpoints [31,139]. Acting in the opposite direction, RNA PolII mutants were shown to modulate chromatin modifications along the body of genes and the degree of unidirectionality (the proportion of antisense transcription) at bidirectional promoters [44]. Slow PolII led to 3’ to 5’ shift of H3K36me modification, deposited by Set2, probably because PolII CTD Ser2 phosphorylation, responsible for Set2 recruitment, shifted as well [44].

In the indirect colocalization scenario, the concentration of properly matured snRNPs and splicing factors would not be sufficient for assembly because of defective or misplaced phase separation [140]. Both the lowered valency of the truncated Prp45 chain and htz1 deletion could play contributing roles. Diffusion constants of nuclear RNPs are low, suggesting that the complexes engage in multiple transient interactions [141]. Liquid-liquid phase separation may result in their differential distribution, concentrating them to the vicinity of the processes for which they are needed. Misplaced (i.e. posttranscriptional) splicing may then face rate-limiting concentration of components needed for spliceosome assembly [142]. Notably, Prp45 has extensive regions of predicted intrinsic disorder [143]. It remains to be seen whether its conformational flexibility is only needed for spliceosome assembly [51,144], or whether it contributes to condensate formation [59,145]. The phase separation phenomena may also explain the disconnect between location and effect, which was repeatedly observed with regard to Htz1 [42,146–148].

Our data on Prp45 variants thus hint at two interesting aspects of splicing. First, the interactions of splicing factors preceding the moment of their final assembly (into spliceosome) decide on the speed of assembly and eventually determine the splicing efficiency. The problem of synergistic recruitment of proteins to RNA reflects the limits imposed by diffusion rates of snRNPs, and the speed and efficiency of recycling. A similar problem was addressed in the ribosome field, where the late associating ribosomal proteins were found to affect the early stages of ribosome assembly as chaperones through transient interactions with rRNA [149,150]. Second, Htz1 absence can aggravate the problem of hampered spliceosome assembly, suggesting the role of the chromatin environment in the splicing cycle. The cycle is prone to such vulnerability, as the spliceosome is assembled and disassembled for each cycle of catalysis and in many distinct locations. This vulnerability of splicing makes it a regulatory interface which the cells may use to modulate gene expression [11,142,151–153]. It should be interesting to learn more about the ‘chaperoning phase’ of splicing, which delivers the properly matured snRNPs to the sites of transcription at the right time. It is perhaps here that much of the chromatin-splicing interplay takes place.

Abbreviations

GO	=	gene ontology
NGI	=	negative genetic interaction
RPG	=	ribosomal protein gene
Subset_45	=	down regulated differentially scored transreads

Acknowledgments

We are grateful to the Boone Lab for the SGA data analyses. We thank members of the Sunnerhagen Lab for stimulating discussions. We thank Eva Krellerová for expert technical assistance and Dr. Marian Novotný for the molecular visualizations in Figure 1.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The authors confirm that the data supporting the findings of this study are available within the article and its supplementary materials.

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/15476286.2024.2348896

Additional information

Funding

The work of K.A. was funded by The Education for Competitiveness Operational Program (ECOP) and co-financed by the European Social Fund and the state budget of the Czech Republic (CZ.1.07/2.3.00/30.0022). We acknowledge the support of K.A. through the Marie Curie Host fellowship (QLK-CT2000-60036) and the EMBO Short Term Fellowship (ASTF 226-2005) during her stays in P. Sunnerhagen’s Laboratory. The research was further supported by Charles University grants SVV260083, GAUK119710, GAUK441711, and GAUK8214, and the Swedish Cancer Fund (22-2014).