Spp382p Interacts with Multiple Yeast Splicing Factors, Including Possible Regulators of Prp43 DExD/H-Box Protein Function

Abstract

Prp43p catalyzes essential steps in pre-mRNA splicing and rRNA biogenesis. In splicing, Spp382p stimulates the Prp43p helicase to dissociate the postcatalytic spliceosome and, in some way, to maintain the integrity of the spliceosome assembly. Here we present a dosage interference assay to identify Spp382p-interacting factors by screening for genes that when overexpressed specifically inhibit the growth of a conditional lethal prp38-1 spliceosome assembly mutant in the spp382-1 suppressor background. Identified, among others, are genes encoding the established splicing factors Prp8p, Prp9p, Prp11p, Prp39p, and Yhc1p and two poorly characterized proteins with possible links to splicing, Sqs1p and Cwc23p. Sqs1p copurifies with Prp43p and is shown to bind Prp43p and Spp382p in the two-hybrid assay. Overexpression of Sqs1p blocks pre-mRNA splicing and inhibits Prp43p-dependent steps in rRNA processing. Increased Prp43p levels buffer Sqs1p cytotoxicity, providing strong evidence that the Prp43p DExD/H-box protein is a target of Sqs1p. Cwc23p is the only known yeast splicing factor with a DnaJ motif characteristic of Hsp40-like chaperones. We show that similar to SPP382, CWC23 activity is critical for efficient pre-mRNA splicing and intron metabolism yet, surprisingly, this activity does not require the canonical DnaJ/Hsp40 motif. These and related data establish the value of this dosage interference assay for finding genes that alter cellular splicing and define Sqs1p and Cwc23p as prospective modulators of Spp382p-stimuated Prp43p function.

EIGHT phylogenetically conserved DExD/H-box proteins act at discrete steps to regulate the assembly, activation, and dissociation of the splicing apparatus (reviewed in Brow 2002; Konarska and Query 2005; Linder 2006). How these RNA-dependent ATPases are temporally and functionally regulated remains poorly understood. One putative regulator, the 83-kDa Spp382/Ntr1 protein (henceforth referred to by the Saccharomyces Genome Database standard name, Spp382p), was discovered in a screen for mutants capable of suppressing defects in yeast spliceosome assembly (Pandit et al. 2006). While spp382 null alleles are lethal, partial loss of function suppresses mutations in several other splicing factors, including the genes encoding the essential spliceosomal proteins Prp8p and Prp38p. Spp382p is a spliceosomal protein that binds the Prp43p DExD/H-box protein to promote efficient dissociation of spliceosomal factors after completion of splicing in vitro (Tsai et al. 2005). Some but not all spp382 mutants also accumulate the excised intron product of splicing in vivo, ostensibly due to protection of the intron within a hyperstabilized spliceosome (Pandit et al. 2006; Tanaka et al. 2007). The suppression of spliceosome assembly defects by spp382 mutation is proposed to occur by impairing Spp382p-stimulated dissociation of kinetically impaired or otherwise inefficient spliceosomes by Prp43p (Pandit et al. 2006). Consistent with this hypothesis, prp43 mutations also suppress spliceosome assembly defects in a manner that, within limits, is inversely proportional to the residual ATPase activity of Prp43p (Pandit et al. 2006). In this light, it is possible that the Spp382 and Prp43 proteins are components of the “discard pathway” for spliceosome dissociation predicted by the kinetic proofreading model of DExD/H-box protein function (Burgess et al. 1990; Konarska and Query 2005).

Recently, several groups made the surprising observation that the Prp43p splicing factor is also required for ribosome biogenesis. Mutations in PRP43 inhibit 35S pre-rRNA cleavage and limit downstream steps in this processing pathway (Lebaron et al. 2005; Combs et al. 2006; Leeds et al. 2006). Prp43p is ≥10-fold more abundant than the splicing-restricted DExD/H-box proteins (e.g., Brr2p, Prp2p, Prp5, Prp16, Prp22p, and Prp28p; see Ghaemmaghami et al. 2003) and, consistent with dual function in splicing and rRNA processing, nuclear Prp43p is enriched in the nucleolus (Huh et al. 2003). Furthermore, proteins and small RNAs acting exclusively in splicing or in rRNA biogenesis copurify with Prp43p, supporting its direct contribution to both processes (Ho et al. 2002; Lebaron et al. 2005; Combs et al. 2006; Gavin et al. 2006; Krogan et al. 2006; Leeds et al. 2006). While it is not certain how Prp43p is partitioned within the cell, its association with Spp382p appears critical for recruitment to the postcatalytic spliceosome and for stimulation of the intrinsic Prp43p helicase activity (Tsai et al. 2005; Boon et al. 2006; Pandit et al. 2006; Tanaka et al. 2007). A structurally related protein, Pxr1p, interacts with Prp43p (Lebaron et al. 2005) and is required for efficient rRNA processing (Guglielmi and Werner 2002). Pxr1p may serve a parallel role for Prp43p recruitment and activation within the rRNA processing apparatus but direct evidence for such function is lacking.

Here we describe a genetic approach to identify factors that interact with SPP382 and function in spliceosome dynamics. Specifically, we describe a dosage interference assay to find genes that when overexpressed inhibit the growth of a yeast strain in which the temperature-sensitive prp38-1 spliceosome assembly mutation (Xie et al. 1998) is suppressed by the spp382-1 suppressor allele (Pandit et al. 2006). We identify multiple GAL1-effector genes that preferentially inhibit growth of the prp38-1 spp382-1 double mutant compared with either single-mutant host or a wild-type yeast strain. Consistent with the goal of this screen, most genes cause splicing inhibition with galactose induction.

Among the recovered genes are SQS1 and CWC23, which encode proteins that purify from yeast in multisubunit protein complexes containing Prp43p but have unknown functions (Lebaron et al. 2005; Pandit et al. 2006). Sqs1p is a nonessential 87-kDa protein that, similar to the Prp43p interacting proteins Pxr1p and Spp382p, possesses the glycine-rich G-patch motif common to a subset of RNA binding proteins (Aravind and Koonin 1999). CWC23 encodes a 33-kDa protein with a canonical DnaJ motif characteristic of Hsp40-like activators of Hsp70 chaperones (Walsh et al. 2004; Vos et al. 2008). Cwc23p interacts with Spp382p in the two-hybrid assay and by affinity selection (Pandit et al. 2006). The genetic and biochemical results presented here establish Sqs1p and Cwc23p as Spp382p-interacting proteins with contributions to RNA processing distinct from what might be expected on the basis of their protein motif characteristics.

MATERIALS AND METHODS

RNA analysis:

Low-density yeast cultures (OD_{600 nm} <0.5) were collected by centrifugation and washed once in sterile water and twice in Li buffer (100 mm LiOAc, 10 mm Tris, pH 7.5, 1 mm EDTA). The cell pellets were then disrupted in one-fiftieth the original culture volume of Li buffer plus 100 μl of phenol:chloroform:isoamyl alcohol (PCI) at 50:49:1 by shaking with sterile glass beads for 4 min in a Mini-Beadbeater (Biospec Products). The RNA was purified by two additional PCI extractions followed by ethanol precipitation. Northern blot analysis of the pre-mRNA and spliceosomal snRNAs was conducted with random prime labeled probes (Promega, Madison, WI) prepared from cloned copies of the indicated DNAs that were hybridized and washed under standard conditions (Sambrook et al. 1989). A 5′-end-labeled oligonucleotide probe was used to detect the 35S rRNA precursor, 23S and 20S intermediates (oligo rRNA 004 5′-CGGTTTTAATTGTCCTA-3′), and the 27SA2 intermediate (oligo rRNA 003 5′-TGTTACCTCTGGGCCC-3′) (Kos and Tollervey 2005). Hybridization with the oligonucleotide probes was done overnight in 6× SSC, 5× Denhardt's solution, and 1% SDS at 43°. The blots were washed in 1× SSC, 0.5% SDS at 43° (rRNA 0003) or 36° (rRNA 004).

For RNase H digestion, 1 μm of oligo(dT) 15-mer (Integrated DNA Technologies) or a control oligo (SNR30, 5′-GAAACTGCTCGTAGTCTGACG-3′) was incubated in 1× digestion buffer [20 mm Tris-HCl (pH 7.5), 20 mm KCl, 10 mm MgCl₂, 0.1 mm EDTA, 0.1 mm DTT] at 55° for 10 min and then slowly cooled to 37°. Five units of Escherichia coli RNase H (United States Biochemical, Cleveland) were added and the reaction was incubated at 37° for 60 min prior to gel electrophoresis and Northern analysis. The membrane transfers were analyzed with a Typhoon 9600 Phosphorimager and ImageQuant 5.2 software (GE Corporation).

CWC23 and SQS1 mutagenesis and gene overexpression:

Site-specific mutations were introduced by inverse PCR using mutagenic oligonucleotides (sequences available upon request) with Expand Long Template polymerase (Roche Biochemicals) on CWC23 in a YCplac22 backbone (Gietz and Sugino 1988) containing the open reading frame and ∼300 bp 5′- and 100 bp 3′-flanking sequence or as a cloned open reading in the pACT and PAS2 two-hybrid vectors (Durfee et al. 1993). SQS1 was mutated similarly except that the GAL1 fusion clone in BG1805 (Gelperin et al. 2005) was used for expression. To facilitate downstream analysis, a unique Not1 recognition sequence was included at the mutagenesis site. All mutant constructs were sequenced prior to use. Functional analysis of the cwc23 mutants was conducted after removal of a wild-type CWC23 allele on a URA3-based vector by plasmid shuffle. Two-hybrid analysis was conducted in the PJ69-4A yeast strain (James et al. 1996) and growth scored at 30° on medium lacking adenine or on histidine-deficient medium supplemented with 20 mm 3-aminotrizaole.

Overexpression studies were conducted using GAL1 fusion constructs (Open Biosystems) in the URA3-based plasmid BP1805 (Gelperin et al. 2005) transformed into strain MGD353 46D (α ura3-52 trp1-289 leu2-112, 113 his cyh2^r), ts192 (α prp38-1 ura3-52 trp1-289 leu2-112, 113 his cyh2^r), SP302 (a ura3Δ0 trp1-289 leu2 Δ0 his3 Δ1 spp382∷KAN pYCplac111-spp382-1), or the double mutant (α prp38-1 ura3-52 trp1-289 leu2-112 113 his spp382-1 cyh2^r) as described in Pandit et al. (2006). The plate cultures were grown on synthetic defined medium or nutrient-rich, yeast extract peptone (YP) agar (Kaiser et al. 1994). Liquid cultures were grown in synthetic defined medium with 2% glucose to saturation at 23°. The cultures were then collected by centrifugation, washed with sterile water, and diluted 1/100 to 1/1000 in synthetic defined medium with 2% galactose. The cultures were grown for an additional 6 hr at room temperature followed by 2 hr at 37° (or as indicated). The PRP43 coexpression assays were performed with plasmid p358-PRP43 (Martin et al. 2002) transformed into wild-type yeast previously transformed with GAL1-SQS1. The plate growth assays were performed at 30° for 3–15 days as indicated in the figure legends.

RESULTS

Dosage interference of spp382-1-based suppression:

Prp38p is an essential component of the U4/U6.U5 tri-snRNP particle (Blanton et al. 1992; Xie et al. 1998). At restrictive temperatures, the prp38-1 temperature-sensitive (ts) mutation allows for efficient snRNP recruitment but greatly impedes spliceosome activation at the stage of U4 snRNP release (Xie et al. 1998). This splicing defect and corresponding ts growth defect can be partially relieved by the recessive spp382-1 suppressor allele (Pandit et al. 2006). To identify potential interacting factors we sought genes that when overexpressed specifically inhibit growth in this genetically sensitized prp38-1 spp382-1 suppressor background on nutrient-rich YP medium (Figure 1 and see materials and methods). In principle, such dosage effectors might either act by antagonizing the suppressor function of spp382-1 or more directly inhibit splicing by sequestration of other splicing factors. We scored a candidate set of 140 GAL1-promoter fusion genes that encode most splicing-related factors in this library collection (Gelperin et al. 2005), other RNA-associated proteins, and randomly selected controls (supporting information, Figure S1).

Figure 1.—

Schematic of the dosage interference assay. This diagram illustrates two possible fates of a spliceosome, completion of the splicing reaction and dissociation through Spp382p-stimulated Prp43p disassembly. Prp38-1p temperature inactivation is envisioned to favor complex dissociation (Kdis) over splicing (Ksp). The spp382-1 suppressor allele is shown to restore balance toward splicing in the prp38-1 mutant. This mutation also reduces the effectiveness of Prp43p in spliceosome disassembly, resulting in the accumulation of excess excised intron RNA. We screened a collection of yeast genes expressed by the robust GAL1 promoter to identify factors (large stars) that inhibit growth through interference with Spp382-1p-based suppression of prp38-1 or the natural Spp382p function in splicing.

Predictably, complementation by wild-type SPP382 blocks growth of the prp38-1 spp382-1 double mutant at 35° as this reveals the underlying temperature sensitivity of the unsuppressed prp38-1 mutation (Figure 2A). Of the remaining GAL1-gene fusions tested, 10 show strong growth impairment on galactose medium when compared to the empty vector control transformant. These encode the established splicing factors Prp8p, Prp9p, Prp11p, Prp39p, and Yhc1p; a DExD/H-box protein involved in mRNA decapping, Dhh1p; mitochondrial proteins Yml6p and Mrp13p; and the product of an uncharacterized open reading frame, YDR230W. Two other genes with possible links to splicing, SQS1 and CWC23 (see below), also showed markedly impaired growth. Overexpression of PRP38 or the known prp38-1 dosage suppressor, SPP381 (Lybarger et al. 1999), caused modest growth inhibition in the prp38-1 spp382-1 background, similar to that seen with GAL1-PRP43 or with GAL1-BET2 that encodes the Bet2p subunit of the yeast geranylgeranyltransferase. Other than the exceptions noted below, none of these genes significantly reduce the growth of wild-type yeast or the spp382-1 or prp38-1 single mutants on the nutrient-rich galactose medium.

Figure 2.—

Impact of gene overexpression on yeast growth. (A) The spp382-1 suppressed prp38-1 spliceosome assembly mutant strain (prp38-1 spp382-1), wild-type yeast (PRP38 SPP382), and the spp382-1 and prp38-1 single-mutant stains were transformed with the indicated GAL1-driven constructs and assayed for growth on rich nutrient medium in galactose (YP Gal) or glucose (YP Glu) at the indicated temperatures. The unsuppressed prp38-1 allele supports growth at 32° but not at 35°. Spp382-1 is growth impaired at 35°. To facilitate comparison between culture conditions, the plates were incubated for 2–6 days until the fastest growing strain in each set reached a colony size similar to the vector-only control transformant (Vector) in the prp38-1 spp382-1 background. (B) A growth assay of transformants in the wild-type PRP38 SPP382 (WT) background on defined synthetic medium containing galactose (−Ura Gal) or glucose (−Ura Glu) at 23° and 35°.

Consistent with a dosage effect, downregulation of the GAL1 promoter on glucose medium largely reverses the GAL1-PRP9, GAL1-PRP11, and GAL1-MRP13 growth defects in the prp38-1 spp382-1 background at 35° (Figure 2A). This is not true for most transformants, however, suggesting that even modestly elevated expression of the effector gene may impair growth. Of all genes testing positive, only MRP13 and SQS1 greatly inhibit growth of the prp38-1 spp382-1 suppressor strain at room temperature on YP-galactose medium. As expected, increased expression of PRP38 and the previously described SPP381 dosage suppressor of prp38-1 (Lybarger et al. 1999) permits growth of the ts lethal prp38-1 single mutant at the restrictive growth temperature. Increased levels of a Spp382p-associated protein, Ntr2p (Tsai et al. 2005), had no obvious effect on the prp38-1 spp382-1 double mutant although it did improve the slow growth of the spp382-1 single mutant (Figure 2A). Finally, we find that all transformants revert to the original growth state after plasmid removal (data not shown).

Unlike what is seen here, overexpression of SQS1 was previously shown to inhibit the growth and cell cycle progression of wild-type yeast on synthetic defined medium (Stevenson et al. 2001). We explored this apparent discrepancy and found that in contrast to the results obtained on nutrient-rich medium, SQS1 overexpression does indeed impair wild-type yeast growth on synthetic defined medium (Figure 2B). Growth studies conducted in liquid medium indicate a four- to fivefold increase in generation time with GAL1-SQS1 expression (data not shown). The medium-dependent difference in yeast sensitivity is not limited to SQS1 expression, however, as overexpression of PRP8, PRP9, CWC23, YDR230W, DHH1, MRP13, YML6, and SPP382 also reduces growth of wild-type yeast on defined medium at 23° or 35°, although inhibition in the prp38-1 spp382-1 background remains comparatively more severe. The basis of this media-dependent growth difference was not investigated.

Enhanced gene expression inhibits pre-mRNA splicing:

Although the GAL1-effector genes might interfere with any essential cellular process, the hypersensitivity of the prp38-1 spp382-1 background suggests specificity to the splicing pathway. To investigate this possibility we assayed the splicing efficiency of prp38-1 spp382-1 transformants after growth in galactose-based defined medium. The yeast cultures were shifted to 37° for the final 2 hr of an 8-hr galactose induction to increase the sensitivity for detecting splicing defects. A Northern transfer of RNA isolated from each of these strains was hybridized with probes against the intronless U2 snRNA to show equal sample loading and general RNA integrity in each sample (Figure 3A). Equivalent results were obtained after hybridization with a probe specific for the intronless ADE3 mRNA (data not shown).

Figure 3.—

Splicing impairment with GAL1 gene expression. (A) RNA isolated from the indicated prp38-1 spp382-1 transformants after growth on galactose medium for 8 hr (the initial 6 hr at room temperature and the final 2 hr at 37°). The top panel shows hybridization for the intron-bearing RPS17A pre-RNA (P) and processed mRNA (M). The mRNA/pre-mRNA indicator of splicing efficiency is in parentheses (the values for SQS1 were too low to measure and are reported as not done, ND). Normalization for loading and general RNA destabilization is provided by hybridization for the intronless U2 snRNA. (B) RNA was isolated as in A except that yeast were grown continuously at 23°. The mRNA/pre-mRNA ratio is noted above the lanes. (C) RNA from yeast that express GAL1-PRP43 was resolved before (no oligo) and after treatment with RNase H and an oligonucleotide complementary to the poly(A) tail [oligo(dT)] or a sequence not present in the RPS17A transcript (control oligo). The unprocessed RPS17A precursor dimer (Pre-mRNA, brackets) and mRNA are shown at the left.

At 37°, the prp38-1 spp382-1 double mutant shows residual splicing impairment and unprocessed RPS17A pre-mRNA is clearly visible in the vector-only control transformant (Figure 3 and see Pandit et al. 2006), consistent with its slow growth compared with wild-type yeast (Pandit et al. 2006 and data not shown). The most severe reduction in RNA abundance is seen with SQS1 expression as both mRNA and pre-mRNA are almost undetectable after 8 hr of gene induction, preventing an estimate of splicing efficiency (but see below). Overexpression of PRP8, CWC23, PRP9, PRP11, BET2, YDR230W, DHH1, NTR2, MRP13, YHC1, PRP39, and SPP382 showed clear evidence of splicing inhibition compared with the vector control (measured as decreased mRNA/pre-mRNA ratios). Expression of the GAL1-driven PRP43, YML6, and SPP381 genes or the complementing PRP38 gene reduced splicing less severely or not at all. Similar results were obtained when an ACT1 probe was used to monitor splicing efficiency (data not shown). Other than YML6, which shows a very strong growth defect but comparatively modest splicing impairment, the splicing and growth inhibition assay results correlate reasonably well. Splicing is much more efficient for all strains when monitored at temperatures between 23° and 30° where only SQS1, BET2, and MRP13 show obvious reductions of fully processed mRNA after 8 hr of GAL1 induction (Figure 4B). While inhibited splicing may contribute to reduced mRNA levels at lower temperatures, other factors, such as lower rates of transcription or mRNA turnover due to direct or indirect affects of GAL1-effector gene expression, have not been ruled out.

Figure 4.—

GAL1-SQS1 expression inhibits rRNA processing. Hybridization of the membrane transfer described in Figure 3A is shown with oligonucleotides specific for the 35S rRNA precursor and the 27SA2, 23S, and 20S rRNA processing intermediates. An unidentified background band seen with the 27SA2 probe provides an additional normalization control.

The pre-mRNA that accumulates with splicing inhibition often appears as broad band or doublet. The upper band of the doublet is generally of low abundance but can be ≥25% of the total pre-mRNA (e.g., PRP8, PRP43, DHH1, MRP13, YHC1, and the empty vector control). We used RNase H and oligo(dT) to investigate whether differential poly(A) tail length contributes to this pre-mRNA spread. After treatment, a downward shift is seen with the ∼400-nt mature RPS17A mRNA in the GAL1-PRP43 sample when compared with the untreated sample or when an unrelated control oligonucleotide is used (Figure 3C). In contrast to the comparatively uniform shift of the RPS17A mRNA band with poly(A) tail removal, RNase H treatment converts the broad unspliced RPS17A precursor into a much more sharply focused ∼800-nt band indicating significant poly(A) tail length heterogeneity in this pre-mRNA. Taken together, these data show that multiple genes inhibit yeast growth and pre-mRNA processing when overexpressed in the splicing sensitized prp38-1 spp382-1 genetic background and reveal construct-specific poly(A) tail length differences after GAL1 induction.

SQS1 overexpression inhibits rRNA processing:

While the function of Sqs1p is unknown, its presence in pre-rRNA processing complexes (Gavin et al. 2006; Krogan et al. 2006) suggests possible involvement in this pathway. An ethidium bromide stain of the membrane presented in Figure 3A shows comparable levels of mature 25S and 18S rRNA after 8 hr of GAL1-effector gene expression (Figure 4A). We could not rule out a rRNA processing defect based on the mature 25S and 18S rRNA levels, however, since the inhibited growth observed after GAL1-SQS1 induction will limit dilution of the stable rRNA by cell division even if rRNA biogenesis is blocked. To address this possibility, we rehybridized the Northern blot membrane with oligonucleotide probes specific for the unprocessed pre-rRNA and diagnostic rRNA intermediates (Kos and Tollervey 2005); a diagram of the processing pathway and the probe description is found in Figure S2). Many of the samples show modest general reductions in the 35S rRNA precursor and processing intermediates relative to the vector control, possibly due to reduced rRNA transcription in response to impaired cell growth (Figure 4B). In contrast, overexpression of SQS1 causes 23S intermediate levels to increase while the levels of the 23S-derived 20S rRNA intermediate and the independently derived 27SA2 intermediate are more severely reduced than seen with the other GAL1-effector constructs. Similar changes are reported when the Dbp4p DExD/H-box protein is depleted or when components of the small ribosomal subunit processome (SSU) are removed and are consistent with inhibition of cleavages at pre-rRNA sites A0, A1, and A2 (Gallagher et al. 2004; Kos and Tollervey 2005).

Prp43p is a target for Sqs1p-based cytotoxicity:

In addition to Sqs1p, four other G-patch proteins are annotated in the MIPS Comprehensive Yeast Genome Database: Spp382p, Pxr1p, Spp2p, and the product of the uncharacterized open reading frame, YLR271W (Figure S3). The position of this motif within each protein varies greatly and there is limited sequence homology within this protein set outside of the G-patch domain. Of these genes, only GAL1-SQS1 expression strongly inhibits splicing within 8 hr of transfer to galactose medium in wild-type yeast (Figure 5A). The strikingly elevated pre-mRNA levels seen here contrast with the loss of both pre-mRNA and mRNA signal in prp38-1 spp382-1, reflecting the greater sensitivity of the splicing sensitized double mutant. This sensitivity extends to the rRNA processing pathway as well since, in contrast to the prp38-1 spp382 mutant, we detect little or no change in the rRNA processing with GAL1-SQS1 expression (Figure 5A and data not shown), strongly suggesting that inhibited splicing most likely accounts for the growth impairment observed with GAL1-SQS1 expression in wild-type yeast. Although too much Sqs1p is toxic and blocks splicing, the absence of Sqs1p appears phenotypically neutral; haploid yeast bearing a sqs1∷KAN null allele grow as well as wild-type yeast and efficiently process pre-mRNA at temperatures between 19° and 37° (Figure 5B and data not shown).

Figure 5.—

GAL1-SQS1 inhibits pre-mRNA splicing in wild-type yeast and interacts with Prp43p. (A) Northern analysis of RNA isolated from wild-type yeast transformed GAL1-driven copies of the five yeast G-patch protein genes (SPP2, SPP382, SQS1, PXR1, and YLR271W) and an empty vector. The cultures were grown on galactose medium for 8 hr (the initial 6 hr at room temperature and the final 2 hr at 37°; equivalent results were obtained when the cultures were grown continuously at 30°). The positions of the RPS17A precursor (P) and spliced mRNA (M) and the mRNA/pre-mRNA ratio are noted (top panel) and the 23S and 20S ribosomal RNA intermediates are shown (bottom panel). (B) Northern analysis of RNA isolated from wild-type yeast (WT), the sqs1∷KAN strain, and the splicing-impaired spp382-6 mutant (Pandit et al. 2006) grown continuously at 23° (−) or after a 2-hr shift to 37° (+). The positions of RPS17A pre-mRNA, mRNA, and ADE3 mRNA are indicated at the left. The mRNA/pre-mRNA ratio is in parentheses. (C) Growth of yeast transformed with the indicated constructs on medium that induces (galactose) or represses (glucose) GAL1 expression for 3 days at 30° (except for the GAL1-SQS1 pPRP43 double transformant and its matched GAL1-SQS1 single transformant that incubated for 15 days). (D) Two-hybrid assays conducted on medium lacking histidine supplemented with 20 mm 3-aminotriazole for reporter gene activation (Y2H Reporter) or simply lacking leucine and tryptophan to score for the two-hybrid plasmids (Plasmid Marker). pACT is the Gal4 activator construct, pAS2 is the Gal4 DNA binding construct, and empty refers to an empty vector control.

It seemed likely that the inhibition of splicing and ribosomal RNA biogenesis caused by GAL1-SQS1 expression results from the sequestration of one or more essential processing factors into inactive complexes with Sqs1p. Prp43p appeared an obvious candidate since this protein can be purified with Sqs1 (Gavin et al. 2006; Krogan et al. 2006) and is required for both splicing and rRNA biogenesis. If Prp43p is a major target for Sqs1p, then increased Prp43p abundance might be expected to mitigate the toxic impact of elevated Sqs1p levels. Indeed, transformation of yeast with a second copy of PRP43 expressed from its natural promoter on a single-copy plasmid improves growth of the GAL1-SQS1 yeast (Figure 5C). We note, however, that the enhanced growth is modest and that additional increases in PRP43 expression did not further reduce Sqs1p toxicity, suggesting that other cellular targets of Sqs1p may exist. Artificially increased levels of the Prp43p activator, Spp382p or the Pxr1 protein do not reduce GAL1-SQS1 toxicity (M. Chen and B. C. Rymond, unpublished results).

Spp382p is proposed to interact with Prp43p through its G-patch motif (Tanaka et al. 2007). Since Spp382p, Pxr1p, and Sqs1p all possess the G-patch domain and interact with Prp43p, the simplest interpretation of the splicing and rRNA processing defects seen with GAL1-SQS1 expression is that Sqs1p competes with Spp382p and Pxr1p for Prp43p association through this common element (i.e., the G-patch domain). Consistent with direct binding competition, we find that full-length constructs of Prp43p and Sqs1p interact in the two-hybrid assay; Sqs1p also interacts with Spp382p in this assay (Figure 5C). However, the simple model of competitive binding of Prp43p through the Sqs1p G-patch motif appears incorrect since a complete deletion of this motif from Sqs1p does not detectably inhibit its association with Prp43p (or with Spp382p) (Figure 5D). Furthermore, the GAL1-sqs1ΔG-patch construct is as inhibitory to growth (Figure 5C) and splicing (M. Chen and B. C. Rymond, unpublished results) as GAL1-SQS. And, as seen with GAL1-SQS1, growth inhibition due to GAL1-sqs1ΔG induction is reduced by increased Prp43p expression (data not shown). Thus, while Prp43p interacts with Sqs1p and is inhibited by excess Sqs1p, the Prp43p–Sqs1p association does not require the G-patch domain in common with the Prp43p binding partners Pxr1p and Spp382p.

Cwc23p is needed for efficient splicing but does not act as a canonical Hsp40 protein in the requisite step of pre-mRNA processing:

CWC23 encodes a 283-amino-acid protein that interacts with Prp43p by the two-hybrid assay and is the only spliceosomal protein observed to biochemically copurify with Spp382p at high stringency (Pandit et al. 2006). Cwc23p was previously shown to possess a strong match to the DnaJ consensus motif (or J domain) roughly positioned between amino acids 12 and 93. This motif is used by bacterial DnaJ (or eukaryotic Hsp40) proteins to activate the ATPase activity of a companion Hsp70 protein during macromolecular assembly and disassembly events (Walsh et al. 2004; Vos et al. 2008). Null mutants of CWC23 were previously described as lethal (Giaever et al. 2002) or viable but temperature and cold sensitive (Tizon et al. 1999). We were able to isolate slow-growing yeast without CWC23 by 5-fluorootic acid (5-FOA) selection against a plasmid-borne wild-type allele in a cwc23∷KAN chromosomal background lacking the entire CWC23 protein coding sequence (Figure 6A). While growth is impaired at 30°, the cwc23∷KAN mutant does not show much greater growth inhibition at lower (19°) or higher (37°) temperatures (data not shown) relative to wild-type yeast.

Figure 6.—

Cwc23p is required for efficient pre-mRNA splicing but its DnaJ motif is not critical. (A) Growth of wild-type yeast (CWC23), the cwc23 null mutant (cwc23∷KAN), and yeast bearing the indicated J-motif mutations on synthetic defined medium at 30°. (B) Northern blot of RNA isolated from the yeast described in A probed to detect RPS17A pre-mRNA (P) and RPS17A mRNA (m) and intronless ADE3 mRNA (top panel) or the excised intron from the ACT1 gene (bottom panel). (C) Two-hybrid interaction between a full-length construct of Spp382p and full-length Cwc23p or the deletion derivative removing much of the DnaJ domain (cwc23Δ50-122). Rev indicates a control construct in which the CWC23 open reading frame was inserted in the opposite orientation to the GAL4 sequence of the two-hybrid vector. The yeast strains were scored for reporter gene activation on medium lacking adenine.

Given its tight physical association with the Spp382 splicing factor, a role for Cwc23p in pre-mRNA processing seemed likely. Indeed, we find that splicing is very poor in the cwc23∷KAN null mutant compared with an otherwise isogenic control strain (Figure 6B). Primer extension analysis of RNA isolated from this mutant confirms that the slowly migrating RNA band labeled pre-mRNA in Figure 6B is unspliced transcript and not the similarly sized lariat intermediate (S. Pandit and B. C. Rymond, unpublished observation). In addition to reduced splicing efficiency, we note that the level of excised intron product increases in the cwc23∷KAN deletion mutant. This apparent inconsistency (less RNA spliced but more intron present) is characteristic of mutations in the splicing apparatus defective in intron release from the spliceosome [e.g., spp382 (Pandit et al. 2006)] or subsequent degradation of the released intron [e.g., dbr1 (Chapman and Boeke 1991)].

The only known role of the DnaJ motif is in the activation of Hsp70 chaperones during substrate presentation. Since at least two Hsp70 proteins, Ssa2 and Ssa4, can be recovered with Cwc23 in the native yeast spliceosome (e.g., Wang et al. 2003), it seemed possible that the splicing and intron release defects of the cwc23∷KAN mutant might result from reduced protein chaperone activity. To test this, we mutated the canonical J domain by removal of the HPD tripeptide required for chaperone function (Walsh et al. 2004). Cwc23p is unusual in having a pair of nearly adjacent HPD repeats within the J domain (i.e., 50HPDKHPD56) as well as a third HPD sequence 29 amino acids downstream of the J domain (120HPD122). We removed each HPD element singly (Δ50–52, Δ54–56, Δ120–122), removed the first two HPD elements together (Δ50–56), or removed all protein coding sequence between the first and last HPD elements (Δ50–122) and assayed each construct for the ability to support growth and splicing. For each mutant, three alanine codons replaced the deleted sequence. Surprisingly, each Cwc23p derivative supported efficient growth and pre-mRNA splicing (Figure 6, A and B). Only the largest deletion, which removes much of the DnaJ motif and 25% if the entire CWC23 protein coding sequence, shows a minor decrease in splicing efficiency. This Cwc23Δ50-122 derivative continues to interact with Spp382p in the two-hybrid assay although with somewhat decreased efficiency (Figure 6C). On the basis of these observations, we conclude that similar to Spp382p, Cwc23p is critical for efficient cellular splicing and intron metabolism but that these functions do not require Cwc23p to act as a typical Hsp40-like chaperone.

Finally, while they are capable of exacerbating the growth of the prp38-1 spp382-1 double mutant, we find no evidence for direct interaction of Cwc23p or Sqs1p with Prp38p. Neither protein was reported in the U4/U6.U5 tri-snRNP particle with Prp38p (Stevens and Abelson 1999; Stevens et al. 2001) and we find that full-length constructs of Cwc23p and Sqs1p fail to interact with Prp38p by the two-hybrid method (data not shown). In addition, when selected under conditions that release Prp38p from the tri-snRNP particle (Xie et al. 1998), Prp38p copurifies as a three-component complex with the Spp381p and Snu23p splicing factors without Cwc23p or Sqs1p (Figure S4).

DISCUSSION

Spp382p recruits and activates the Prp43 DExD/H-box protein to recycle splicing factors upon completion of the spliceosome cycle. In addition, diminished Spp382p or Prp43p activity suppresses splicing defects associated with the prp38-1, prp38-2, prp19-1, prp8-1, and prp8-2 spliceosome assembly mutants (Pandit et al. 2006; S. Pandit and B. C. Rymond, unpublished results), suggesting that this complex may also stimulate the disassembly of defective splicing complexes that differ in subunit composition from the postcatalytic spliceosome. Here we describe a genetic approach to find modifiers of Spp382p activity by seeking genes that antagonize the growth of a spp382-1 suppressed prp38-1 spliceosome assembly mutant when overexpressed. While the gene set examined was purposely biased toward splicing factors, most had no impact on growth. However, ∼10% of the GAL1-effector plasmids did show growth inhibition upon induction and, consistent with the goals of this screen, most showed enhanced splicing defects when expressed at elevated temperature. Furthermore, two of the identified genes, SQS1 and CWC23, are shown to have physical and genetic properties linked to Spp382p-dependent Prp43p activation and function.

While we have not studied the modes of splicing inhibition in detail, it seems likely that many of the splicing factors testing positive in this assay (i.e., Prp8, Prp9p, Prp11p, Yhc1p, and Prp39p) act by sequestering other spliceosomal components in nonproductive complexes. This situation might be exacerbated in the prp38-1 spp382-1 background as this strain has both a primary defect in spliceosome maturation due to prp38-1 and a second mutation that inhibits the recycling of spliceosomal complexes (i.e., spp382-1). The sequestration of splicing factors might antagonize Spp382-1p suppressor function or directly inhibit mRNA processing in this splicing compromised background. For instance, GAL1-SPP382 presumably acts by direct complementation of spp382-1, revealing the conditional lethal defect of prp38-1. Increased Prp43p levels might also antagonize spp382-1 suppression although the constitutively high levels of Prp43p (Ghaemmaghami et al. 2003) may diminish the magnitude of GAL1-PRP43 contribution. Other genes may antagonize spp382-1 suppression by actively promoting spliceosome disassembly. For instance, it is conceivable that excessive amounts of an early-acting splicing factor (e.g., the U1 snRNP protein, Yhc1p) might stimulate the dissociation of a late-acting factor (e.g., the U6 snRNP) through competition for a common protein or RNA binding sequence [the pre-mRNA 5′-splice site (Du and Rosbash 2002)]. Dhh1p, which stimulates mRNA decapping and decay (Coller and Parker 2005), presumably acts by a mechanism related to RNA stability.

Four genes without obvious connections to RNA processing were found to inhibit growth when overexpressed in the double-mutant background. This might reflect indirect exacerbation of cellular processes rendered inefficient by the intrinsically poor splicing in this mutant and would include mitochondrial protein synthesis or function (YML6, MRP13, and perhaps YDR230) (Dimmer et al. 2002) and vesicular transport (BET2). Such indirect contributions would not explain the greater splicing defect observed after GAL1 induction, however. An alternative possibility is that the artificially high levels of some effector proteins may reduce the expression or function of an authentic splicing factor. In this light, the identification of the Bet2p β-subunit of geranylgeranyltransferase is especially curious since two other components of this pathway, the Bts1p geranylgeranyl diphosphate synthase and the Bet4p α-subunit of the geranylgeranyltransferase, were previously found as dosage suppressors with another splicing mutant (i.e., clf1Δ2) (Vincent et al. 2003). Geranylgeranyl substrates include various Ras, Rho, Rab, and Rac signaling proteins (Dogbo et al. 1988; Kuzuguchi et al. 1999). While the biochemical basis of these genetic interactions needs additional study, it is enticing to speculate that geranylgeranyl addition may influence signal transduction pathway(s) shown to promote changes in splicing efficiency (Pleiss et al. 2007; see Lynch 2007).

Beyond splicing defects, the expression of some GAL1 promoter fusion constructs alters pre-mRNA tail lengths, suggesting mechanistic differences in the RNA processing defects, e.g., short poly(A) tails with GAL1-CWC23 expression and equal amounts of short and long with GAL1-PRP43. This observation is consistent with the hyperadenylation of nuclear retained unprocessed pre-mRNA (Hilleren et al. 2001; Jensen et al. 2001; Hammell et al. 2002), although the possibility of inhibited poly(A) addition or enhanced metabolism for pre-mRNA with shorter poly(A) tails has not been ruled out. Finally, while we present these effects as due to simple overexpression, each construct is expressed with a 19-kDa C-terminal affinity tag (Gelperin et al. 2005) that might interfere with normal function and inhibit splicing. But even in such cases, the elucidation of the underlying mode of splicing inhibition may add to our understanding of spliceosome homeostasis and function.

Like the Prp43p-binding proteins, Spp382p and Pxr1p, Sqs1p has one copy of the ∼45-amino-acid G-patch motif that is common to a subset of RNA processing factors (Aravind and Koonin 1999). Of the five G-patch proteins annotated in the MIPS Comprehensive Yeast Genome Database (http://mips.gsf.de/genre/proj/yeast/), three are essential (Spp2p, Spp382p) or nearly essential (Pxr1p) and assist DExD/H-box factors in RNA processing. Spp2p functions with Prp2p in splicing (Silverman et al. 2004) while Spp382p and Pxr1p act with Prp43p in splicing and ribosome biogenesis, respectively (Guglielmi and Werner 2002; Tsai et al. 2005; Boon et al. 2006; Pandit et al. 2006; Tanaka et al. 2007). Pxr1p and Spp382p also influence telomerase activity and genomic stability (Lin and Blackburn 2004; Herrmann et al. 2007), suggesting additional contributions to nuclear function. The biological roles for Sqs1p and the Ylr271W product are unknown but because single (Giaever et al. 2002) or combined (S. Pandit and B. C. Rymond, unpublished results) null mutants of these genes are viable, neither one makes a critical contribution to cellular biochemistry under standard conditions.

GAL1-SQS1 induction is shown here to poison the Prp43p-dependent processes of rRNA biogenesis and pre-mRNA splicing. When Sqs1p is expressed at physiological levels, it copurifies with Prp43p in complexes containing splicing and rRNA biogenesis factors (Gavin et al. 2006; Krogan et al. 2006) and a recent screen for interactions among nonessential yeast genes provides evidence for possible Sqs1p function in ribosome biogenesis or activity (Decourty et al. 2008). These observations, the Prp43p–Sqs1p two-hybrid interaction and the buffering effect of increased Prp43p expression in Sqs1p toxicity, make it likely that Prp43p is an intracellular binding partner of Prp43p. If so, direct competition between Sqs1p and the structurally similar Spp382p or Pxr1p for Prp43p association is an attractive possibility to account for the observed Sqs1p toxicity and RNA processing defects. Splicing rather than rRNA processing appears to be the primary target of inhibition when GAL-SQS1 is expressed in wild-type yeast, consistent with Sqs1p acting primarily as an rRNA biogenesis factor, with excess Sqs1p limiting the Prp43p pool available for splicing. Sqs1p might normally act to recruit Prp43p to a processing site where it serves a nonessential auxiliary function in rRNA maturation. Alternatively, Sqs1p might serve to restrict premature Prp43p activation in either splicing or rRNA biogenesis. Precedent exists for the negative regulation of a DExD/H-box protein in the inhibition of eIF4III ATPase by the MAGOH/Y14 components of the exon junction complex (Nielsen et al. 2009).

Mutational studies suggest that the G-patch element in Spp382p is needed to stimulate Prp43p RNA helicase activity (Pandit et al. 2006; Tanaka et al. 2007) and in Spp2p to recruit Prp2p to the spliceosome (Silverman et al. 2004). While these observations suggest a protein-recognition interface, the G-patch motif has also been proposed to serve as an RNA-binding function (Bauerova-Zabranska et al. 2005) and such function has not been ruled out for any of the yeast G-patch proteins. Surprisingly, the removal of the Sqs1p G-patch motif did not reduce its ability to bind Prp43p or change the growth inhibition seen when overexpressed, ruling out a simple G-patch competition model for Prp43p association. As we were not able to detect convincing conservation outside of G-patch domains, Sqs1p likely binds Prp43p in a manner distinct from Spp382p or Pxr1p. Nevertheless, it remains likely that association of Spp382p, Pxr1p, and perhaps Sqs1p with Prp43p contributes to the functional partitioning of Prp43p between the splicing and rRNA processing pathways.

Cwc23p was first found as a copurifying protein with the Cef1p splicing factor (Ohi et al. 2002) and was subsequently identified in other spliceosomal preparations (Wang et al. 2003; Chen et al. 2007). Alternatively reported as essential (Giaever et al. 2002) or nonessential (Tizon et al. 1999), we find the cwc23∷KAN null mutant to be viable but sickly. This stain is also genetically unstable, giving rise to more rapidly growing suppressors, an observation perhaps relevant to the diversity of cwc23 phenotypes reported (S. Pandit, M. Chen and B. C. Rymond, unpublished results). Cwc23p interacts by the two-hybrid assay with Spp382p and is the only splicing factor seen to copurify with Spp382p at 450 mm NaCl (Pandit et al. 2006), raising the possibility that Cwc23p may assist Spp382p in the recruitment, activation, or dissociation of Prp43p. Indeed, we find that the cwc23∷KAN mutant not only is splicing defective but also accumulates excess excised introns, a characteristic of impaired spliceosome disassembly that is observed with certain spp382 and prp43 mutants (Martin et al. 2002; Pandit et al. 2006).

Twenty-two Hsp40p/DnaJ-like proteins are present in yeast (Walsh et al. 2004), with Cwc23p, a class C or III subfamily member (Stirling et al. 2006; Vos et al. 2008) and the only DnaJ protein observed in the yeast spliceosome. At least two DnaJ proteins, DNAJC8 and DNAJC13, copurify with the mammalian spliceosome but neither one has been studied for function in splicing (Jurica et al. 2002; Chen et al. 2007). Removal of much of the DnaJ motif from yeast Cwc23p has little obvious consequence on splicing, growth, or this protein's interaction with Spp382p. Consequently, while important for efficient splicing, Cwc23p does not provide an obligate Hsp40-like cochaperone function with spliceosome-associated Hsp70 proteins [e.g., Ssa2p and Ssa4p (Wang et al. 2003)] or use this canonical J domain in an unexpected manner to stimulate another ATPase activity necessary for splicing (e.g., Prp43p). It remains possible, however, that Cwc23p stimulates a cellular Hsp70p needed for the processing of pre-mRNA from only certain genes or under special environmental conditions when splicing efficiency is stressed (Yost and Lindquist 1991; Vogel et al. 1995; Bracken and Bond 1999).