Genetic Interactions With CLF1 Identify Additional Pre-mRNA Splicing Factors and a Link Between Activators of Yeast Vesicular Transport and Splicing
Kevin Vincent, Qiang Wang, Steven Jay, Kathryn Hobbs, Brian C. Rymond

Abstract

Clf1 is a conserved spliceosome assembly factor composed predominately of TPR repeats. Here we show that the TPR elements are not functionally equivalent, with the amino terminus of Clf1 being especially sensitive to change. Deletion and add-back experiments reveal that the splicing defect associated with TPR removal results from the loss of TPR-specific sequence information. Twelve mutants were found that show synthetic growth defects when combined with an allele that lacks TPR2 (i.e., clf1Δ2). The identified genes encode the Mud2, Ntc20, Prp16, Prp17, Prp19, Prp22, and Syf2 splicing factors and four proteins without established contribution to splicing (Bud13, Cet1, Cwc2, and Rds3). Each synthetic lethal with clf1Δ2 (slc) mutant is splicing defective in a wild-type CLF1 background. In addition to the splicing factors, SSD1, BTS1, and BET4 were identified as dosage suppressors of clf1Δ2 or selected slc mutants. These results support Clf1 function through multiple stages of the spliceosome cycle, identify additional genes that promote cellular mRNA maturation, and reveal a link between Rab/Ras GTPase activation and the process of pre-mRNA splicing.

THE spliceosome is composed of five small nuclear ribonucleoprotein (snRNP) particles and an undetermined number of non-snRNP splicing factors (Collins and Guthrie 2000; reviewed in Brow 2002). Each snRNP contains small nuclear RNA (snRNA) and an associated set of proteins. SnRNP particles interact dynamically with one another and with the pre-mRNA to configure an active splicing enzyme. The U2 and U6 snRNAs are believed to function at the catalytic center of the enzyme while U5 snRNA provides a key substrate alignment role. The U1 and U4 snRNAs act in earlier steps of spliceosome assembly but are not essential for catalysis.

Proteins serve to promote, stabilize, and resolve RNA-based interactions within the spliceosome. A number of crosslinking experiments place selected spliceosomal proteins at or near the active site of the splicing enzyme (e.g., see Ismailiet al. 2001 and references in Reed and Chiara 1999) although it is unclear whether proteins take part in catalysis. Clf1 is an essential yeast splicing factor that resides in at least two distinct complexes, the Clf1-NTC and the Clf1-RNP (Wanget al. 2003). The Clf1-NTC is similar to the previously reported “nineteen complex” (NTC) implicated in a late stage of spliceosome maturation (Tarn et al. 1993a,b; Chenet al. 2002). The Clf1-RNP is reminiscent of a late-stage spliceosome and contains most Clf1-NTC proteins, the U2, U5, and U6 snRNAs, and a subset of the known snRNP proteins. While the functional relationship between the Clf1-NTC and Clf1-RNP complexes is still speculative, Clf1-based contacts are critical for completion of the spliceosome cycle in vitro. In the absence of Clf1 activity the U4/U6.U5 tri-snRNP particle is either no longer recruited to the assembled prespliceosome or, if recruited, no longer configured into an active state (Chunget al. 1999; Wanget al. 2003).

Clf1 is composed almost entirely of direct iterations of the 34-amino-acid TPR protein-binding motif. This repetitive structure and the abortive spliceosome assembly observed in Clf1-defective extracts led to speculation that Clf1 acts as a scaffold in spliceosome assembly (Wanget al. 2003). The Mud2, Prp40, Cef1, Isy1, Ntc20, Syf1, Syf2, Prp22, and Prp46 spliceosomal proteins have been shown to interact with Clf1 in two-hybrid or solution-binding assays (Chunget al. 1999; Ben-Yehudaet al. 2000; Chenet al. 2002; Ohi and Gould 2002). While identification of interacting proteins is consistent with a scaffold function, the selectivity of the poly-TPR platform is called into question by interactions that encompass identical or overlapping regions of Clf1.

Here a deletion approach is used to assay the Cfl1 TPR repeats for biological function and synthetic lethal and dosage suppressor screens are used to identify trans-acting factors that influence Clf1 activity. The results support the view that TPR-specific contacts promote Clf1 function, provide evidence for the involvement of several additional genes in the splicing pathway, and reveal a link between pre-mRNA splicing and Rab/Ras-GTPase activation in vesicular transport.

View this table:
TABLE 1

Oligonucleotides used in this study

MATERIALS AND METHODS

Plasmid and strain constructions: The clf1Δ TPR deletion mutants were made by inverse PCR with the paired oligonucleotides listed in Table 1 [TPR D1-1 (upstream), TPR D1-2 (downstream), etc.] and a 3-kb XbaI-SphI DNA fragment containing a TAP-tagged CLF1 allele inserted in pTZ19U (USB). The primers introduce a SmaI recognition site at the deletion endpoint and the PCR fragments were cleaved with SmaI prior to ligation. After confirmation of the DNA sequence, the deletion constructs were transferred into the yeast shuttle vector, YC-pLa22 (Gietz and Sugino 1988). The YCplac22 clf1Δ plasmids were transformed into the previously described yeast strain SY101 (a ade2-101 clf1::HIS3 his3200 leu21 lys2-801 trp1Δ1 ura3-52 pBM150 [URA3 GAL1::clf1(697)]; Chunget al. 1999). For the clf1Δ2 add-back experiments, PCR fragments composed of TPR1 (oligonucleotides AB1-1 and AB1-2), TPR2 (oligonucleotides AB2-1 and AB2-2), or TPR3 (oligonucleotides AB3-1 and AB3-2) were inserted at the SmaI deletion endpoint site.

Clf1::LEU2 was made by blunt-end ligation of a 1.6-kbp LEU2 restriction fragment from YdLEU2 (Berbenet al. 1991) into an SmaI-cleaved clf1 derivative in which all TPR coding sequences were removed by inverse PCR with oligonucleotides TPR Dall-1 and TPR Dall-1. The clf1::LEU2 allele was excised from the vector and used to replace the endogenous CLF1 gene in strain YCH125 (a ade2 ade3 ura3 leu2 trp1) transformed with p102 (YCplac22 clf1Δ2). The resulting strain was transformed with plasmid p101 (YCp50 containing URA3, ADE3, and CLF1) to create BRY556.

A diploid yeast strain heterozygous for the rds3::KanR disruption was obtained from the ATCC. The GAL1::RDS3 fusion was prepared by insertion of a BamHI-digested PCR fragment from yeast genomic DNA (strain MGD35346D; α cyhR leu2-3,113 his trp1-289 ura3-52; primersRDS1-1 and RDS1-2) into the BamHI site of pBM150 (Johnston and Davis 1984). Yeast that exclusively express the GAL1 fusion gene were obtained from the meiotic offspring of diploid transformant. The GAL1::rds3-1 strain was prepared similarly with DNA isolated from the slc6-1 mutant.

Isolation of slc mutants and dosage suppressors: Strain BRY556 (a ade2 ade3 ura3 leu2 trp1 clf1::LEU2 YCplac22 [clf1 Δ2 TRP1] p2965 [CLF1 URA3]) was mutagenized with ethyl methanesulfate (EMS) to 40% viability as previously described (Blantonet al. 1992). The clf1Δ2 allele used for synthetic lethal selection was not marked with a TAP or HA epitope. Approximately 100,000 yeast colonies were plated on CSM-tryptophan plates and incubated 4-5 days at 23° or 30°. Non-sectoring (i.e., solid red) colonies were scored for growth on 5-fluoroorotic acid (5-FOA) medium at 23° and for temperature sensitivity on YPD medium at 37°. The temperature-sensitive (ts) strains were backcrossed to the wild-type strain, MGD 35346D, and slc mutants isolated from the meiotic offspring free of the clf1::LEU2 knockout and plasmids present in the parent. A YCp50-based yeast genomic library (Roseet al. 1987) was used to obtain the wild-type alleles of the slc mutants by complementation at 37°. The complementing genes were identified by subclone and linkage analyses.

Synthetic lethality between the slc mutants and alternative clf1 deletion alleles was tested after transformation of the YC-plac22-based clf1 deletion derivative into strain BRY555 (a ade2 ade3 ura3 leu2 trp1 p2965 [CLF1 URA3]). Gene knockout mutants obtained from the ATCC were first mated with BRY556 followed by selection of diploid strains that lost plasmid p2965 (CLF1 URA3) on 5-FOA medium. The offspring from at least 40 tetrads were then assayed for the kanR gene on G418 medium and on selective plates for the nutritional markers present in the clf1::LEU2 knockout and the YCplac22 (TRP1, clf1Δ2) plasmid. A heterozygous PRP19/prp19-1, CLF1/ clf1Δ2 diploid was generated as a cross between the clf1Δ2 mutant and JM796 (α ade2 his3Δ ura3 prp19-1 tyr1). The meiotic offspring were scored by crossing all ts isolates back to the mutant parents and assaying for complementation at 37°.

Figure 1.

—Clf1 TPR deletion mutants vary in splicing impairment. (A) Individual or adjacent TPR elements were deleted from the CLF1 gene (boxes) and the resultant mutations were assayed for function at 30° and 37°. NTD and CTD refer to the non-TPR amino (N) and carboxyl (C) terminal domains, respectively. Growth was considered wild type (WT) if colony diameter was within 50% of that found with the full-length construct at both temperatures. Temperature-sensitive (TS) colonies showed little or no growth at 37° after 4 days of incubation. (B) Northern analysis of pre-mRNA splicing. RNA was extracted from the wild-type strain (CLF1) and various deletion mutants were grown continuously at 30° or after a 2-hr shift to 37° and resolved on a 1% agarose formaldehyde gel. Hybridization probes were prepared from the intron-containing RPS17A gene and the intronless SNR20 (U2 snRNA) gene.

High-copy-number dosage suppressors were isolated by transformation of YKH101 (a clf1::HIS3 trp1-289 leu2-3,112 ura3-52 YCplac33 [clf1 Δ2 URA3]) with a YEp13-based genomic DNA library (ATCC stock 37323) and screening for enhanced colony size at 34° on CSM-leucine medium (Kaiseret al. 1994). To rule out library recovery of CLF1, colonies were counter-screened on 5-FOA medium, which selects against the YC-plac33-clf1Δ2 plasmid. Plasmids were recovered from yeast that were 5-FOA- and showed enhanced growth at 34°. The suppressor genes were identified by subclone analysis with vector YEplac118 (Gietz and Sugino 1988). Tests of dosage suppression by SSD1, BTS1, and BET4 were conducted on YPD medium at the semipermissive growth temperature of 35° (slc1-1, slc2-1, slc2-2, slc4-1, and clf1 Δ2) or at 37° (slc3-1, slc5-1, slc6-1, and slc7-1). For segregation analysis, the yeast URA3 gene was placed directly upstream of SSD1 integrative transformation using a 2.6-kb EcoRI fragment of yeast DNA blunt end ligated into the HindIII and EcoRI sites of YIp211 (Gietz and Sugino 1988). This plasmid was cleaved with BglII prior to yeast transformation.

Analysis of pre-mRNA splicing: Total cellular RNA extracted from yeast cultures was resolved on a 1% agarose/formaldehyde gel, transferred to a nylon membrane, and hybridized as previously described (Blantonet al. 1992). Splicing efficiency was evaluated using a Typhoon phosphoimager (Molecular Dynamics, Sunnyvale, CA) to estimate the mRNA to pre-mRNA ratio, a more reliable indicator of splicing impairment than absolute changes of mRNA or pre-mRNA since mutants can show decreased pre-mRNA stability (Rymondet al. 1990).

RESULTS

Clf1 is differentially sensitive to TPR motif deletions: Clf1 contains 15 direct iterations of the TPR motif flanked by 30- and 109-amino-acid non-TPR segments at the amino and carboxyl ends, respectively. To investigate this domain organization, individual or multiple TPR elements were deleted and the resulting constructs were assayed for biological activity in yeast on single-copy plasmids expressed by the natural CLF1 promoter. The host strain contains a chromosomal gene disruption complemented by the functional but nutritionally regulated GAL1::clf1(679) allele (Chunget al. 1999). Downstream characterizations were facilitated by the insertion of the TAP affinity tag (Puiget al. 2001) into the nonessential C-terminal coding sequence. No differences in growth rate or splicing efficiency were observed when the wild-type CLF1 allele was modified in this way (Chunget al. 1999; Wanget al. 2003).

The clf1Δ mutant transformants were scored for colony formation on glucose-based medium [to repress GAL1::clf1(679) transcription] at 30° and 37°. The results show that while CLF1 is quite tolerant of mutation, not all TPR elements function equivalently (Figure 1A; see Chunget al. 1999; Wanget al. 2003). The amino terminus of Clf1 is particularly sensitive to deletion, as removal of either TPR1 or TPR2 imparts a tight ts growth defect while deletion of elements TPR1-TPR3 is lethal. In contrast, deletion of individual or grouped TPR elements through the central TPR domain or a 76-amino-acid segment of the non-TPR carboxyl terminal domain (CTD; amino acids 579-653) has little or no impact on Clf1 function. The exception is a deletion of TPR elements 7-8, which causes a leaky ts growth phenotype. Improper protein folding may account for the clf1Δ7-8 defect, as a larger deletion (i.e., clf1Δ6-8) supports growth at near wild-type levels. We previously showed that the terminal 40% of Clf1 is nonessential, as yeast harboring a frameshift mutation at TPR11 are viable albeit growth and splicing impaired (Chunget al. 1999). In contrast, N-terminal expansion of this deletion from TPR10 (beginning at amino acid 357) through amino acid 653 within the CTD is lethal. The Clf1Δ10-CTD protein is stable in yeast (data not shown) and individual deletions of TPR10, the CTD segment (Figure 1A), or the removal of all coding sequence downstream of TPR12 (Chunget al. 1999) do not impair growth.

RNA isolated from wild-type and clf1Δ mutant cultures was probed with the intron-bearing RPS17A gene to assay for splicing inhibition (Figure 1B). U2 snRNA was used as a control for RNA loading and transfer efficiencies. For all nonlethal mutations, the GAL1::clf1(679)-bearing plasmid was removed by 5-FOA selection (Boekeet al. 1987) prior to assay. At 30° the functional mutant alleles of CLF1 showed mRNA/pre-mRNA ratios within 2.5-fold of that of the wild-type allele (Figure 1B; mRNA/pre-mRNA ratio for wild type is ∼20; Rymondet al. 1990). At 37°, however, splicing is greatly impaired in the three ts deletion mutants (clf1Δ1, clf1Δ2, and clf1Δ7-8; mRNA/pre-mRNA values of 1-2) and reproducibly less efficient in the clf1Δ10 strain (mRNA/pre-mRNA value of 3.5). The remaining nonlethal deletions showed more modest splicing defects or were indistinguishable from wild type. The lethal mutant constructs were assayed 6 hr after GAL1::clf1(679) repression in glucose-based medium. The mRNA/pre-mRNA ratio for each lethal mutant was indistinguishable from that of the untransformed GAL1::clf1(679) strain (i.e., 0.75-1), indicating that the products are nonfunctional. Primer extension with RPS17A and ACT1 exon II oligonucleotides established a step 1 splicing block for each mutant (data not shown). Together these data show that the Clf1 TPR elements are not functionally equivalent in splicing and implicate the N-terminal region as contributing critical intra- or intermolecular contacts.

TPR elements are not interchangeable: TPR elements are often clustered and likely function in groups to support ligand association (reviewed in Blatch and Lassle 1999). In principle, deletion of a TPR element might impair function by the reduction in number of functionally equivalent repeats or by the loss of TPR-specific sequence information. At a superficial level the deletion analysis supports the second model since Clf1 activity is impaired only after removal of certain repeats. It is possible, however, that sensitive elements reside within in a spatially restricted domain (delimited by an uncharacterized feature of the Clf1 structure) and therefore deletions elsewhere in the protein are irrelevant with respect to the mutant phenotype. To address this, we assayed the activity of add-back constructs in which PCR products encoding TPR1, TPR2, or TPR3 are inserted at the site of the clf1Δ2 deletion (Figure 2). The original clf1Δ2 allele and each add-back construct support growth at 30°, showing that none of the reintroduced TPRs create a dominant-negative mutation. At 37°, however, the clf1Δ2 strain produces no colonies while the reconstituted TPR2 construct displays wild-type growth. Neither the TPR1 nor the TPR3 add-back construct showed any rescue at 37°. Similar results were obtained when TPR4 was inserted into clf1Δ2 (data not shown), leading us to conclude that TPR2 contains sequence-specific information critical for Clf1 activity.

Figure 2.

—TPR2 contains unique sequence information. Yeast strains that express a wild-type CLF1 gene (WT), the clf1ΔTPR2 mutation (Δ2), and the clf1ΔTPR2 mutation modified by the insertion of TPR1 (Δ2+1), TPR2 (Δ2+2), or TPR3 (Δ2+3) were assayed for growth on YPD medium at 30° and 37°.

Isolation of mutants synthetically lethal with clf1Δ2: The clf1Δ2 mutation provides a sensitized background with which to identify genes that act in support of Clf1 function. Using the red/white sectoring (Kranz and Holm 1990) and 5-FOA sensitivity (Boekeet al. 1987) plasmid-loss assays, we identified synthetic lethal mutants from an EMS-mutagenized culture as colonies that are inviable in the absence of CLF1. The putative synthetic lethal mutants were then screened for those that are growth impaired at 37° in the presence of the wild-type CLF1-bearing plasmid. Eight ts slc mutants were identified that also met additional screen criteria (see below). The mutants were backcrossed to obtain slc progeny in a CLF1 chromosomal background. Each slc mutation segregated as a single Mendelian trait unlinked to CLF1. The complementation pattern obtained from crosses within this mutant set showed that two slc mutations are tightly linked and likely reside in the same gene (slc2-1 and slc2-2) while the rest reside in unique genes.

In comparison to the wild-type strain and a ts but splicing competent control strain (K46), splicing in the clf1Δ2 and slc backgrounds is 3- to 30-fold less efficient at the restrictive temperature and, at least for slc1-1, slc5-1, slc7-1, and the clf1Δ2 strains, somewhat impaired at the permissive temperature (Figure 3). The ts growth and splicing defects cosegregated in eight of eight offspring tested from each slc mutant backcross. Primer extension showed most of the intron-bearing RNA that accumulates in the slc1-1 mutant is lariat intermediate (data not shown). All other mutants show predominantly enhanced pre-mRNA levels relative to the wild-type control. With one exception, no reproducible differences in mRNA length or abundance were detected with the intronless, heat-shock-inducible TSF1 gene probe (Figure 3). For slc4-1, however, the induced TSF1mRNA levels were always much lower than those of the wild-type control. Lower levels of the intronless yet constitutively expressed ADE3 mRNA were also observed with slc4-1 at the restrictive temperature while no differences were observed in the levels of rRNA or spliceosomal snRNAs (data not shown). These results indicate that the ts slc1-slc3 and slc5-slc7 mutations reside within genes that support pre-mRNA splicing while the slc4-1 mutation influences splicing as well as the levels of certain intronless pol II RNAs.

Figure 3.

Slc mutant inactivation inhibits pre-mRNA splicing. RNA was isolated from CLF1 (WT) and the indicated mutant yeast strains were grown continuously at 23° (-) and after a 2-hr shift to the restrictive temperature of 37° (+). A Northern transfer of this RNA was hybridized with probes from the intron-containing RPS17A gene and from the intronless (but heat-inducible) TSF1 gene. K46 is a ts mutant without an obvious splicing defect.

Figure 4.

—Two DExD/H-box proteins genetically interact with clf1ΔTPR2. (A) A schematic of Prp22 domain organization. The conserved DEAD-box motifs I-IV are shown with the relevant Prp22-specific indicated. The protein changes caused by the slc2-1 mutation (V505A and G692S) and from the slc2-2 mutation (G692D) are presented above the schematic. (B) Western blot of Clf1-TAP complexes isolated from sequential affinity purification on protein A agarose and calmodulin Sepharose (Clf1-TAP). An equivalent sample prepared from an extract that lacks the TAP epitope is presented as a negative control (Clf1). The blot was incubated with polyclonal antibodies specific for Prp22 and Prp16.

Identification of the SLC genes: A yeast genomic DNA library assembled on a single-copy plasmid vector was used to recover genes that complement the slc mutations. DNA sequence analyses showed that all plasmids recovered for a given slc mutant defined identical or overlapping regions of the yeast genome. Subsequent subcloning experiments determined the identities of the effector genes. SLC1 encodes Prp16, a DExD/H-box ATPase originally identified as a suppressor of mutant branchpoint sequences (Coutoet al. 1987) and subsequently shown to induce an ATP-dependent conformational change within the spliceosome required for the second RNA cleavage/ligation event in splicing (Schwer and Guthrie 1992). SLC2 encodes Prp22, a related DExD/H-box protein that is required for the ATP-dependent mRNA release from the spliceosome (Companyet al. 1991) and, at least in vitro, for a less well-defined ATP-independent step prior to exon ligation (Schwer and Gross 1998). The slc2-1 and slc2-2 alleles of PRP22 were isolated and sequenced (Figure 4A). slc2-1 contains two mutations that result in a V505A substitution at an often conserved residue immediately upstream of the DExD/H-box motif I (or Walker A box) and a G692S substitution within the core NTPase domain. The slc22-2 allele also contains a mutation at codon 692, resulting in a G692D substitution. Both Prp16 and Prp22 copurify with Clf1-TAP complexes isolated from cell extracts through two rounds of affinity enrichment (Figure 4B; see Wanget al. 2003). Neither protein is recovered with a control extract that lacks a TAP-tagged protein. Thus, two enzymatic effectors of the spliceosome cycle, Prp16 and Prp22, interact genetically and physically with the Clf1-bearing complexes.

SLC3 encodes Cwc2, an essential yeast protein with C3H1 zinc finger (amino acids 67-94) and RRM (amino acids 136-210) motifs shared with a number of likely homologs [Homo sapiens (gi|8922328), Drosophila melanogaster (gi|16769690), Arabidopsis thaliana (gi|15227567), and Schizosaccharomyces pombe (gi|19114249)]. When likely homologs are excluded, Cwc2 sequences best match with the N terminus of the yeast Hsh49 U2 snRNP protein (amino acids 10-175), showing 44% sequence similarity with Cwc2 residues 175-328. Although likely an RNA-binding protein, Cwc2 was not reported to copurify with yeast snRNP particles (Brow 2002). The slc3-1 allele contains a single nucleotide change resulting in a glycine-to-aspartic-acid substitution at amino acid 79, a conserved position within the zinc finger domain. Cwc2 copurifies with Clf1 complexes (Ohiet al. 2002; Wanget al. 2003).

SLC4 encodes the Cet1 RNA triphosphatase, an integral component of the yeast mRNA capping enzyme (Tsukamotoet al. 1997 and references within). Cet1, the Ceg1 guanylyltransferase, and the Abd1 methyltransferase serve to replace the initiating 5′ triphosphate with the N7 methyl cap structure common to pol II transcripts. The cet1-1 allele contains two mutations that result in D422N and L495 amino acid substitutions within the catalytic domain (Lehmanet al. 1999; Schweret al. 2001). Although snRNA levels often decrease after splicing factor inactivation (Blantonet al. 1992), snRNA remains constant in the slc4-1 mutant (and all other slc mutants) for at least 2 hr after temperature shift. Consequently, improper pre-mRNP organization rather than decreased snRNA stability may account for the rapid splicing impairment noted with slc4-1 inactivation (see discussion).

SSD1 was found to relieve the ts growth defect of the slc5-1 mutant in two independent experiments. However, unlike the other SLC genes, SSD1 incompletely abated the slc5-1 phenotype. For instance, as shown in Figure 5, ectopically expressed SLC1/PRP16 restores splicing and growth at 37° to wild-type levels in the slc1-1 mutant whereas SSD1 expression enhanced slc5-1 splicing only weakly (≤1.5-fold). The enhanced growth with SSD1 expression appears more than proportionate to the improvement in RPS17A splicing (2- to 3-fold vs. 1.5-fold), suggesting that the processing of this transcript is not rate limiting under these conditions. The comparatively weak effect of SSD1 expression might reflect semidominance by the slc5-1 mutation or SSD1 function as a low-copy suppressor. Semidominance was ruled out by the observation that a heterozygous slc5-1/SLC5 diploid grows as well at 37° as an otherwise isogenic wild-type control (data not shown).

Ssd1 has been shown to bind RNA (Uesonoet al. 1997), suppress certain splicing mutants (Luukkonen and Seraphin 1999), and in someway function as a general post-transcriptional regulator of gene activity (Kaeberlein and Guarente 2002 and references within). If SSD1 functions as a dosage suppressor of slc5-1, one should be able to genetically separate the SSD1 and SLC5 loci. To test this, the SSD1 locus in the slc5-1 mutant was marked with URA3 (see materials and methods), this strain was backcrossed, and the genetic linkage of URA3 and SLC5 was scored. One-half of the offspring from 40 tetrads were temperature sensitive (due to slc5-1) and these were evenly divided between ura+ and urastrains. Thus, SSD1 is unlinked to SLC5 and acts as a single-copy dosage suppressor of slc5-1. Repeated attempts to isolate SLC5 proved unsuccessful.

Figure 5.

—Incomplete recovery of the slc5-1 splicing defect with SSD1 overexpression. (A) Northern analysis with the intron-bearing RPS17A gene on RNA from yeast grown continuously at 23° (-) or after a 2-hr shift to 37° (+). The cultures included yeast wild type for splicing related genes (WT) and the slc1-1 and slc5-1 mutants with and without plasmids containing genes that fully relieve (PRP16) or partially relieve (SSD1) the ts growth and splicing defects. (B) Growth of the same strains incubated at 37° for 3 days on YPD medium.

SLC6 is defined by open reading frame YPR094W, named RDS3 (for regulators of drug sensitivity) in a recent study of zinc cluster protein function (Akacheet al. 2001; Akache and Turcotte 2002). A RDS3 knockout was reported to show increased sensitivity to ketoconazole and cyclohexamide and reductions in PDR5 and SNQ2 (drug transporter gene) mRNA, leading to the suggestion that Rds3 is a transcription factor in the drug transport pathway. Slc6/Rds3 is a 12-kD protein with five CxxC zinc fingers and a strongly basic carboxyl terminus (Figure 6). This is an exceptionally highly conserved protein with the fly and human homologs sharing 95% identity and the yeast/human homologs sharing 56% identity. The slc6-1 mutation causes an aspartic acid for glycine substitution at position 20, a conserved residue three amino acids upstream of the first zinc finger.

RDS3 has been reported as an essential gene (Giaeveret al. 2002) and as a nonessential gene (Akacheet al. 2001) in yeast. We obtained the heterozygous diploid strain used in both studies and confirmed that at 30° on rich agar medium the rds3::KanR knockout is lethal. All four meiotic offspring germinate, but two of the four haploid strains arrest as tiny colonies, which cannot be propagated on fresh medium (Figure 7A and data not shown). None of the large colonies are resistant to the antibiotic G418, showing that all encode the wild-type RDS3 allele. The rds3::KanR knockout is complemented by ectopic expression of a GAL1::RDS3 or a GAL1::rds3-1 promoter fusion gene on a URA3-marked plasmid (Figure 7B). The GAL1::rds3-1 strain exhibits temperature-sensitive, galactose-dependent growth. The GAL1::RDS3 strain is not temperature sensitive and forms colonies on glucose-based medium, presumably due to residual transcription of the fusion gene (data not shown). The GAL1::RDS3 and GAL1::rds3-1 strains do not form colonies on 5-FOA glucose (or galactose) medium (Figure 7B), confirming the Rds3 requirement for mitotic growth.

When GAL1::rds3-1 transcription is repressed at 37°, pre-mRNA levels elevate within 2 hr and mRNA levels drop to near minimal levels 8 hr later (Figure 7C). In contrast to the 10- to 20-fold reduction in splicing efficiency observed after 10 hr with the GAL1::rds3-1 culture, only a minor splicing impairment (∼2-fold; characteristic wild-type cultures) is observed with the RDS3 strain. GAL1::rds3-1 culture growth largely ceases after 10 hr of incubation under these conditions. This time course of splicing and growth impairment is very similar to what has been reported for the depletion of other essential pre-mRNA splicing factors (e.g., see Brown and Beggs 1992; Lockhart and Rymond 1994). Splicing inhibition is also seen if the GAL1::rds3-1 culture is grown on galactose at 37° or on glucose at 30° although in both cases the extent of inhibition is less complete, presumably due to residual Rds3 activity.

SLC7 encodes Bud13, a 30.5-kD basic protein recently identified in a screen for mutants defective in bud site selection (Ni and Snyder 2001). Diploid null mutants of BUD13 display a unipolar budding pattern with bud scars restricted to a region near the birth scar at the proximate pole. Homologs of Bud13 are apparent in many organisms [e.g., H. sapiens (gi|14249338), Anopheles gambiae (gi|21301692), D. melanogaster (gi|21355513), and S. pombe (gi|19075968)] and show considerable length variation with the highest level of primary sequence conservation present in the carboxyl half of the protein. All Bud13 homologs have 5-7 consecutive lysine or arginine residues present within the highly charged amino terminus and the human and fly homologs have lengthy, highly basic internal insertions. Slc7-1 introduces a nonsense codon within the conserved carboxyl segment (codon 232).

Allele specificity and directed tests for synthetic lethality: A plasmid shuffle approach was used to score four alternative clf1 alleles for synthetic interactions with the slc mutants (Table 2). clf1Δ1 shows a pattern of interaction equivalent to that of clf1Δ2. This is understandable given the close proximity of TPR1 and TPR2 repeats and the similar Clf1 complex defects observed with both mutants (Wanget al. 2003). Deletion of TPR7-8 imparts a ts growth defect and shows strong synthetic interaction with slc1-1 (prp16), slc3-1 (cwc2), slc4-1 (cet1), slc5-1, and slc7-1 (bud13). Synthetic lethality is also observed between the weak splicing mutant clf1Δ6-8 and slc7-1 (bud13) and greatly impaired growth is observed with double mutants of clf1Δ6-8 and slc1-1 (prp16), slc3-1 (cwc2), and slc5-1. The clf1ΔCTD mutant supports near normal pre-mRNA splicing but shows a modest synthetic growth defect with slc7-1 (bud13), slc3-1 (cwc2), and slc4-1 (cet1). By this assay, the slc2-1, 2-2 (prp22), and slc6-1 (rds3) interactions display the greatest allele specificity, producing strong synthetic growth defects only with TPR1 or TPR2 deletions.

Figure 6.

—Multiple sequence alignment of Rds3. Clustl-WP alignment of putative Rds3 homologs highlighted with the Blosum scoring matrix with 100, 80, and 60% sequence similarity presented in boxes of decreasing darkness. The sequences are from H. sapiens (gi|14249398), D. melanogaster (gi|21711783), A. thaliana (gi|25406958), S. pombe (gi|19114704), and S. cerevisiae (gi||6325351). The G20D amino acid substitution encoded by slc6-1 (rds3) is shown above the alignment.

Figure 7.

—RDS3 encodes an essential pre-mRNA splicing factor. (A) Tetrad dissection of a diploid strain heterozygous for the rds3::KanR mutation. The spores were arrayed on YPD agar at 30° for 5 days. All large colonies were confirmed to contain the wild-type RDS3 allele. (B) Growth of wild-type yeast (RDS3) and the rds3::KanR mutant transformed with the GAL1::RDS3 gene or the GAL1::rds3-1 gene. The cultures were streaked on 5-FOA medium and complete medium that lacks uracil with galactose as a sugar source (-ura galactose) and incubated for 4 days at 30°. (C) Pre-mRNA splicing efficiency was monitored in wild-type yeast transformed with the GAL1::rds3-1 gene (RDS3) and in the rds3::KanR mutant transformed with the GAL1::rds3-1 gene (GAL1::rds3-1). RNA was isolated from yeast growing continuously on galactose (Gal) at 30° and after the indicated time in glucose (Glu)-based broth at 37°. RPS17A DNA was used as a probe.

Prp19-1 and several viable knockout mutants with genetic or biochemical links to Clf1 were also scored for synthetic interactions. Synthetic lethality is observed when clf1Δ2 is combined with prp19-1, ntc20::kanR, prp17::kanR, or syf2::kanR. In addition, clf1Δ2, mud2::kanR double mutants show greatly impaired growth (Figure 8). No synthetic defects were observed between clf1Δ2 and a randomly selected gene knockout (fus1::kanR) or with a knockout of a gene unrelated to splicing that nevertheless interacts with Clf1 in the two-hybrid assay (gpx2:: kanR; Uetzet al. 2000).

Dosage suppressors link splicing and GTPase activation in vesicular transport: Dosage suppression was used as a complementary approach to identify genes that interact with clf1Δ2. From ∼150,000 yeast transformants, 16 plasmids were recovered that enhanced growth at the semirestrictive temperature of 34° but which did not contain CLF1. Two of the plasmids contained overlapping sequences from the left arm of chromosome XVI (group 1) while 14 contained identical or overlapping regions of the left arm of chromosome X (group 2). Neither suppressor type supports growth of a clf1::HIS3 null mutant.

The group 1 suppressor, BTS1, encodes the yeast geranylgeranyl diphosphate synthase (Jianget al. 1995) while the group 2 suppressor, BET4, encodes the α-subunit of the Rab/Ypt-protein geranylgeranyltransferase (GGTase; Jianget al. 1993). All known or predicted GGTase substrates are nucleotide-binding proteins involved in signal transduction events (see discussion). We believe that the BTS1/BET4 dosage suppressors act downstream of the primary clf1Δ2 pre-mRNA splicing defect since growth is enhanced without detectable improvement in pre-mRNA splicing (tested for ACT1, SNR17, and RPS17A; data not shown). In addition, a viable bts1::KanR knockout mutant shows no obvious splicing defect (Q. Wang and B. C. Rymond, unpublished observations).

To learn how general the dosage suppression patterns are, clf1Δ2 and the slc mutants were transformed with plasmids bearing SSD1, BTS1, and BET4 and assayed for enhanced growth (Table 3). At the semirestrictive temperature used, all mutants are viable but colony size is reduced compared with wild type. The growth of four mutants, slc4-1 (cet1), slc2-2 (prp22), slc6-1 (rds3), and slc5-1 improved with SSD1 overexpression. Suppression is not correlated simply with the tightness of the mutant allele as slc4-1 (cet1) is ts lethal at 37° but suppressed by SSD1 while slc2-1 (prp22), slc3-1 (cwc2), slc7-1 (bud13), and the uncharacterized K46 mutant are slow growing at 37° but not suppressed with elevated SSD1 expression.

View this table:
TABLE 2

Allele specificity of clf1 mutant interactions

BTS1 and BET4 showed an identical suppression pattern, consistent with joint contribution to the pathway of protein geranylgeranylation. In addition to clf1Δ2, the slc6-1 (rds3) mutant is enhanced by BTS1/BET4 overexpression. Of the nine ts yeast strains assayed, only slc6-1 (rds3) was suppressed by all three genes. The distinct SSD1 and BTS1/BET4 suppression patterns suggest that at least two distinct pathways exist to reduce splicing-related growth defects.

DISCUSSION

Clf1 interactions: The poly-TPR structure of Clf1 presents a potential docking surface for multiple splicing factors. The deletion analysis presented here supports a scaffold function for Clf1 by showing that the individual Clf1 TPR elements are biologically functional and differ in contribution to splicing. This and other recent studies have identified interactions between Clf1 and factors that act from the earliest stages of spliceosome assembly through product release (Table 4). For instance, U1-Prp40 and Mud2 bind Clf1 and promote U1 snRNP recruitment in the commitment complex while U2 snRNA, Hsh155, and Rse1 are present in Clf1 complexes and snRNP components of prespliceosome. Rds3 interacts genetically with Clf1 and, as described below, is also associated with U2 snRNP proteins. Such “early factor” associations are reinforced by the observation that wild-type RPS17A (RP51A) prespliceosomes fail to mature into stable snRNP-complete complexes in the absence of Clf1 (Chunget al. 1999).

The NTC appears to contain a dozen or fewer proteins that act after prespliceosome formation (Tarnet al. 1994; Wanget al. 2003). Clf1Δ2 shows synthetic lethal interactions with NTC members Prp19, Ntc20, Syf2, and Cwc2 and interacts with Ntc20, Syf2, and other NTC components in two-hybrid and protein-binding assays. Although it is attractive to imagine the NTC as a multi-subunit splicing factor, there is no evidence to support a requirement for NTC preassembly in splicing. Nevertheless, Clf1 is an essential NTC protein that is implicated with other members of this structure in events that occur during and after the prespliceosome to spliceosome transition.

The NTC proteins also reside in RNP structures that contain additional proteins and the U2, U5, and U6 snRNAs (the Clf1-RNP and related complexes; Ohiet al. 2002; Wanget al. 2003) or the U1, U2, U4, U5, and U6 snRNAs (the penta-snRNP; Stevenset al. 2002). A number of the RNP-specific proteins interact with Clf1. For instance, Snu114, a U5 snRNP-associated protein that contributes to dissociation of the U4/U6 helices (Bartelset al. 2002), shows conditional association with ClfΔ2 (Wanget al. 2003) while overexpression of U2 snRNA partially suppresses certain CFL1 (also called SYF3) mutations (Ben-Yehudaet al. 2000). Several second-step splicing factors are present in the U2, U5, and U6 snRNA-bearing complexes (but not in the penta-snRNP). Of these, Prp16, Prp17, and Prp22 were found as synthetic lethal mutants with clfΔ2. Like the slc2 mutants, most other PRP22 helicase domain mutants show a step 1 block to splicing in vivo. Interestingly, the Prp22 G692D mutant described here was also identified as an intragenic suppressor of an SAT → AAT motif III change that uncouples the Prp22 ATPase activity from the RNA helicase and mRNA release activities (Schwer and Meszaros 2000; Campodonico and Schwer 2002). Prp22 copurifies with Clf1 complexes and interacts with the Clf1 in the two-hybrid assay—although perhaps at a binding site well removed from TPR2 (Ben-Yehudaet al. 2000). Splicing inhibition by either slc2 inactivation or clf1Δ2 coexpression is consistent with joint Prp22/Clf1 contribution to a late (and likely ATP-dependent) step in the spliceosome cycle.

Newly defined splicing-relevant genes: Mutations within BUD13, CET1, CWC2, RDS3, and SLC5 were found to be synthetically lethal with clf1Δ2 and splicing defective when expressed in a wild-type CLF1 background. Cet1 phosphatase activity is required for cap formation on pol II RNAs, including pre-mRNA and snRNA. The slc4-1 (cet1) mutations create amino acid substitutions at sites believed important for enzyme structure (residue 422) and activity (amino acid 495; Peiet al. 1999; Bisaillon and Shuman 2001). That the slc4-1 (cet1) splicing defect is due principally to improper cap formation is reinforced by the observation that CEG1 guanylyltransferase mutants are also splicing impaired (Schwer and Shuman 1996). Cap structure enhances snRNA stability and, at least in mammals, facilitates nuclear import of snRNP particles (Will and Luhrmann 2001). In vitro, the nuclear cap-binding complex promotes the initial interaction between the U1 snRNP and the splicing substrate (Colotet al. 1996; Lewiset al. 1996) and functions during 5′ splice site rearrangements that lead to the displacement of U1 snRNA by U6 snRNA (O’Mullane and Eperon 1998). While the clf1Δ2/slc4-1 (cet1) arrest point is unknown, we note the cap-sensitive U6 for U1 snRNP swap occurs at or near the time of Clf1 (and Prp19) function in the spliceosome cycle.

Figure 8.

—Synthetic growth defect of the mud2::kanR, clf1Δ2 double mutant. Yeast colonies containing the mud2:: kanR, clf1Δ2, or both mutations are shown after 4 days of growth on YPD agar at 23°.

Cwc2 is one of >30 proteins present in Clf1-containing complexes (Ohiet al. 2002; Wanget al. 2003). While this work was in progress the Gould lab provided evidence that targeted degradation of Cwc2 blocks pre-mRNA splicing (Ohi and Gould 2002). The mutational data presented concur and extend this observation to show that the zinc finger domain of Cwc2 is essential for biological activity. Cwc2 interacts through its N-terminal region with Prp19 but does not appear to bind Clf1 directly (Ohi and Gould 2002). Cef1 binds both Clf1 and Prp19. Under restrictive conditions, the clf1Δ2 mutation causes the release of Prp19 and Cef1 from Clf1 complexes and blocks recruitment of Prp19 to the spliceosome (Wanget al. 2003). As the slc3-1 (cwc2) mutation resides within the proposed Prp19-binding domain (Ohi and Gould 2002), the clf1Δ2/slc3-1 (cwc2) lethality most likely results from disruption of the Clf1/Cef1/Prp19/Cwc2 organization within the splicing complex.

The bud site defect of the bud13 (slc7-1) (Ni and Snyder 2001) is most likely an indirect consequence of impaired pre-mRNA splicing as disruption of numerous biological pathways (e.g., lipid metabolism, RNA transport and processing, translation, and vesicular transport) impairs the budding process. Consistent with this, mutations in the genes encoding the NTC splicing factor Isy1, the U2 snRNP protein Ist3/Snu17, and the U6 core snRNP protein Lsm6 also result in bud defects (Ni and Snyder 2001). Although little is known of Bud13 function in other organisms, RNAi experiments show that it is required for Caenorhabditis elegans embryogenesis (Jianget al. 2001), consistent with a conserved role in pre-mRNA splicing.

View this table:
TABLE 3

Enhanced colony growth with dosage-suppressor expression

Rds3 rivals the core histone proteins in its exceptional level of sequence conservation. Given the apparent stringent constraints on its sequence, it is not surprising that RDS3 is essential for yeast viability. Turcotte and colleagues previously reported RDS3 as a nonessential gene (Akacheet al. 2001; Akache and Turcotte 2002). A possible explanation for this discrepancy is provided by the observation that when rds3::KanR yeast are cultured at 23° (rather than at 30°), cell division occurs but at a very slow rate (i.e., a doubling time >12 hr). Rare bypass suppressors occasionally arise under these conditions and rapidly overgrow the culture. It is possible that such a bypass mutant was isolated by the random spore selection used to acquire the rds3::KanR strain for the drug sensitivity study. The decreased expression of the PDR5 and SNQ2 drug transporter genes reported in the putative rds3::KanR background might be an indirect consequence of splicing impairment although we note that neither gene contains an intron. Q. Wang and B. C. Rymond (unpublished observations) found that Rds3 is a stable component of the spliceosome and acts, in vitro, to promote splicing. Although functional data are not available, the human homolog of Rds3 was recently found to be associated with the 17S U2 snRNP particle (Willet al. 2002). Yeast Rds3 also interacts with several U2 snRNP proteins, including Clf1-RNP components, providing possible targets for the synthetic lethal interaction (Q. Wang and B. C. Rymond, unpublished observations).

View this table:
TABLE 4

Summary of Clf1 genetic and biochemical interactions

Suppression of pre-mRNA splicing defects: Mutations in the PRP2 and RSE1 splicing factor genes were shown previously to cause vesicular transport defects (Chenet al. 1998) that could be suppressed by enhanced expression of the SAR1 GTPase that promotes ER to Golgi sorting. Thus, under some but perhaps not all (Bigginset al. 2001; Burnset al. 2002) conditions, splicing inhibition renders vesicular transport limiting for yeast cell growth. The substrate for Bet4 modification is not clear, however, since Sar1 lacks the consensus site for geranylgeranylation and none of the predicted Bet4 substrates contain introns. In the absence of GGTase activity, the small Ypt1 and Sec4 GTPases are not lipid modified, their membrane localization is impaired, and vesicular transport is disrupted (Jiang and Ferro-Novick 1994). Importantly, other known or suspected substrates for C-terminal geranylgeranyl modification are also rab-GTPases or ras-like proteins and casein kinase (i.e., Cdc42, Rho1, Rho2, Rho3, Rho4, Rsr1, Sec4, Vps21, Yck1, Yck2, Ypt1, Ypt6, and Ypt7; Costanzoet al. 2001) that function in signal transduction pathways involving vesicular transport or related membrane-associated events. Although direct involvement of certain secretory proteins in splicing has been suggested (Awasthiet al. 2001), it seems likely that the BTS1/BET4 dosage suppression works by increasing the pool of modified and hence active, membrane-bound signal transduction molecules, thus driving downstream events in the secretory pathway.

A recent analysis of global gene expression suggests differences in the set of pre-mRNAs most affected by the inactivation of specific splicing factors (Burnset al. 2002). Presumably dosage suppression depends on which RNAs become limiting for cellular growth. In the case of the RNA-binding protein, Ssd1, suppression is likely RNA based (for instance, through enhanced stabilization of pre-mRNA, mRNA, or snRNA). In contrast, for BTS1, BET4 suppression most likely acts post-translationally to improve the efficiency of splicing-dependent biological pathways, most critically vesicular transport, which involves the activity of numerous intron-bearing genes (e.g., BET1, BOS1, ERV41, GOT1, MSB4, NYV1, SFT1, SNC1, and others).

Acknowledgments

We are grateful to Jay Dunn and Brandon Thomas for assistance in the preliminary characterization of CLF1. Beate Schwer is thanked for generously providing antibodies against Prp16 and Prp22, for subclones of both genes, and for helpful comments while this work was in progress. The prp19-1 yeast strain was provided by John Woolford. This work was supported by the National Institutes of Health award GM-42476.

Footnotes

  • Communicating editor: M. Johnston

  • Received January 17, 2003.
  • Accepted March 25, 2003.

LITERATURE CITED

View Abstract