Snu114 is the only GTPase required for mRNA splicing. As a homolog of elongation factor G, it contains three domains (III–V) predicted to undergo a large rearrangement following GTP hydrolysis. To assess the functional importance of the domains of Snu114, we used random mutagenesis to create conditionally lethal alleles. We identified three main classes: (1) mutations that are predicted to affect GTP binding and hydrolysis, (2) mutations that are clustered in 10- to 20-amino-acid stretches in each of domains III–V, and (3) mutations that result in deletion of up to 70 amino acids from the C terminus. Representative mutations from each of these classes blocked the first step of splicing in vivo and in vitro. The growth defects caused by most alleles were synthetically exacerbated by mutations in PRP8, a U5 snRNP protein that physically interacts with Snu114, as well as in genes involved in snRNP biogenesis, including SAD1 and BRR1. The allele snu114-60, which truncates the C terminus, was synthetically lethal with factors required for activation of the spliceosome, including the DExD/H-box ATPases BRR2 and PRP28. We propose that GTP hydrolysis results in a rearrangement between Prp8 and the C terminus of Snu114 that leads to release of U1 and U4, thus activating the spliceosome for catalysis.
PRE-mRNA splicing is catalyzed by the spliceosome, a large dynamic complex composed of five small nuclear RNAs (snRNAs) and >80 proteins (Burge et al. 1998; Jurica and Moore 2003). The chemistry of splicing comprises two sequential transesterification reactions (Moore et al. 1993). In the first reaction, the 5′ splice site is cleaved and a branched lariat structure is formed within the intron. In the second reaction, the 3′ splice site is cleaved and the two exons are joined together. During the splicing cycle, the RNA and protein components of the spliceosome undergo numerous rearrangements, which must be highly coordinated to ensure fidelity of the process (Staley and Guthrie 1998). Most of these rearrangements appear to be energy dependent and are correlated with the activity of individual ATPases of the DExD/H-box family. Eight known DExD/H-box proteins are required for the splicing cycle, and mutations in these proteins inhibit the ATP-dependent steps of splicing (Staley and Guthrie 1998). Additionally, splicing requires one GTPase, Snu114, which is an essential protein in Saccharomyces cerevisiae (Fabrizio et al. 1997). Notably, Snu114 is homologous to the ribosomal translocase elongation factor G (EF-G in prokaryotes/EF2 in eukaryotes), leading to the hypothesis that Snu114 may similarly use the energy of GTP hydrolysis to drive rearrangements of the spliceosome (Fabrizio et al. 1997).
Snu114 is packaged with other proteins and the U5 snRNA to form the U5 small ribonucleoprotein particle (snRNP). Prior to formation of the spliceosome, U5 snRNP interacts with the U4/U6 di-snRNP, in which U4 and U6 snRNAs are extensively base paired, thus forming U4/U6·U5 tri-snRNP (reviewed in Burge et al. 1998). According to the canonical model of spliceosome assembly, the tri-snRNP is then recruited to the prespliceosome, in which U1 snRNA is base paired with the 5′ splice site and U2 snRNA is base paired with the branchpoint sequence, an intronic consensus sequence near the 3′ splice site. Although the addition of tri-snRNP forms the complete spliceosome, this complex is catalytically inert. Activation requires that the U1/5′ splice site interaction and the base pairing between U4 and U6 be disrupted, such that U1 and U4 are no longer stably associated with the spliceosome. In contrast to the stepwise pathway of spliceosome assembly, recent evidence suggests that a holospliceosome containing all five snRNPs interacts as a complex with each intron (Stevens et al. 2002). Nonetheless, ordered rearrangements of the snRNPs must occur prior to catalysis.
Rearrangements that occur during the early stages of spliceosome activation are regulated by several components of the U5 snRNP (Brow 2002). The Prp28 ATPase is required to unwind the U1/5′ splice site duplex, possibly by destabilizing protein components of U1 snRNP (Staley and Guthrie 1999; J. Y. Chen et al. 2001), and the Brr2 ATPase is required to unwind the U4/U6 duplex (Laggerbauer et al. 1998; Raghunathan and Guthrie 1998a; Kim and Rossi 1999). Presumably, the activities of Prp28 and Brr2 must be tightly regulated to ensure that catalytic activation does not occur prematurely. Genetic studies have suggested that the activities of these ATPases are modulated by Prp8, which is a large (280 kDa), well-conserved U5 snRNP protein (Kuhn et al. 1999; Collins and Guthrie 2000; Kuhn and Brow 2000). Prp8 is believed to inhibit the activities of Prp28 and Brr2 until spliceosome formation has occurred (Kuhn et al. 2002). The mechanism by which Prp8 inhibits the ATPases is as yet unknown.
A strong physical interaction between Prp8 and Snu114 suggests that Snu114 may also play a regulatory role during spliceosome activation. Treatment of U5 snRNP from human cell extract with high concentrations of chaotropic salts disrupts the complex, but Prp8 (U5-220kD in human) and Snu114 (U5-116kD in human) remain associated as a heterodimer (Achsel et al. 1998). Indeed, deletion of the N-terminal 120 amino acids of Snu114, which compose the only domain that is not found in EF-G, causes a temperature-sensitive block to growth and inhibits the release of U1 and U4 from the spliceosome (Bartels et al. 2002). Similarly, a mutation in Snu114 that is predicted to convert GTP binding to XTP binding inhibits U1 and U4 release and also decreases cellular levels of U5 and tri-snRNP (Bartels et al. 2003). While the N-terminal and GTPase domains have been implicated in spliceosome activation, previous studies have not addressed a function for the other domains of Snu114.
We used a two-step strategy to elucidate the timing and mechanism of Snu114 activity. First, we generated conditionally lethal alleles of Snu114 by random mutagenesis. We identified mutations in all domains of the protein, including clusters of mutations in domains III–V and mutations within the GTPase domain. We also found that deletion of the C-terminal 70 amino acids (snu114-60) causes a growth defect at 16° and 37°. Second, we analyzed synthetic interactions between snu114 alleles and mutants of other proteins that function at distinct stages of splicing. We found strong genetic interactions between the snu114 alleles and mutations in factors involved in snRNP formation and spliceosome activation. In particular, synthetically lethal interactions with snu114-60 demonstrate a critical function for the C terminus of the protein during spliceosome activation.
MATERIALS AND METHODS
Strains and plasmids:
Yeast strains are listed in Table 1 and plasmids are listed in Table 2. All strains are isogenic with BY4743 (Giaever et al. 2002), which is an S288C derivative, unless otherwise noted. Heterozygous diploids with KanMX deletions of SNU114, SUB2, PRP22, PRP43, BRR2, PRP28, and SNU66 and haploids with KanMX deletions of ISY1, SNT209, and BRR1 were obtained from Research Genetics (Huntsville, AL) (Giaever et al. 2002). A PRP8 deletion strain was created by transforming a wild-type diploid with the PRP8::LYS2 SacI/ApaI fragment from pJU224 (Umen and Guthrie 1996); integrants were confirmed by PCR. sad1-1 (BSY387) (Lygerou et al. 1999), prp2-1 (SS304), and prp19-1 (ts87) (Vijayraghavan et al. 1989) strains were backcrossed twice to snu114::KanMX strains; prp5-1 (SPJ 5.41) (Vijayraghavan et al. 1989) was crossed once to a snu114::KanMX strain. yTB136 was derived from GLS618 (Rader and Guthrie 2002), and yTV161 was derived from yS79 (MATα sister of yS78; Wang and Guthrie 1998). yTB2 was created by sporulating YPF5 (Fabrizio et al. 1997).
EcoRI/BglII sites (in boldface type) were inserted immediately following the AUG start codon of SNU114 by PCR amplifying pTB1 (pRS316/SNU114) (Fabrizio et al. 1997) with the primer pairs oTB1 (5′-GGGAACAAAAGCTGGGTACCGGGC-3′)/oTB2 (5′-GGAAGCGAATTCCATTTTGCTATGTTAGGAGCTATTG-3′) and oTB3 (5′-CCGACCGAATTCAGATCTGAAGGTGACGATTTATTCGATGA-3′)/oTB4 (5′-ATCCTCTCCGGAATGTTAGCCAT-3′). The oTB1/oTB2 PCR product was digested with KpnI and EcoRI and inserted into the same sites of pRS316. The resulting plasmid and the oTB3/oTB4 PCR product were digested with EcoRI and BamHI and ligated together. The 4.73-kb BspEI-NsiI fragment of the resulting plasmid was ligated with the 3.62-kb BspEI/NsiI fragment of pTB1 to create pTB3. The SNU114-containing XhoI/SacI fragments of pTB1 and pTB3 were inserted into the same sites in pRS314 to create pTB2 and pTB4, respectively.
A single myc epitope was placed immediately after the start codon of SNU114 by inserting the annealed oligos oKD140 (5′-AATTCCCAGAACAAAAATTGATTTCTGAAGAAGATTTGAATA-3′) and oKD141 (5′-GATCTATTCAAATCTTCTTCAGAAATCAATTTTTGTTCTGGG-3′), which have overhanging EcoRI/BglII sites, into the same sites of pTB4. The resulting plasmid was named pTB19. pTB19 was transformed into yTB2, the plasmid pTB1 was lost by passage on 5-FOA media, and the presence of the myc epitope was confirmed by Western blotting with the 9E10 antibody (BAbCO). The yeast strain carrying pTB19 grew at the same rate as the strain with pTB1 at 16°, 25°, 30°, and 37°.
The snu114-50 mutations E910G and C928R were separated by digesting pRS314/snu114-50 and pTB19 with PstI and NdeI and inserting the 356-bp fragment from each plasmid into the 7.89-kb fragment of the other plasmid.
To remove the myc tag and/or the restriction sites from the N terminus of snu114 alleles, the following restriction enzymes were used to clone the snu114 mutations into pTB2: BspEI and BstBI (snu114-12 and snu114-14), BstBI and SacI (snu114-30 and snu114-50), BstBI and PstI (snu114-40), and PstI and SacI (snu114-60). Plasmids pRS314/SNU114 and pRS314/snu114-12, -14, -15, -30, -40, -50, and -60 were named pTB95-102. The snu114 alleles were moved from pRS314 (pTB95-102) to pRS315 (pTB106-113) by transforming the PCR-amplified LEU2 marker from pRS315 with HindIII-cut pRS314/snu114 plasmids into a wild-type strain. Plasmids were recovered from LEU2+ transformants.
Mutagenesis of SNU114:
SNU114 was PCR amplified in two fragments. Fragment 1 was amplified with oTB7 (5′-CTTGCCAACGGCTGACGATTGC-3′; 67 bp upstream of start AUG) and oTB8 (5′-CAAACAGTCCATATACAGCTCTCC-3′; 1965 bp downstream of start AUG). Fragment 2 was amplified with oTB9 (5′-CCGCGACGTTGTACTCTGTAAAG-3′; 1720 bp downstream of start AUG) and oTB10 (5′-CCAGTGAATTGTAATACGACTCAC-3′; 3270 bp downstream of start AUG). Mutagenesis was performed under standard PCR conditions, using the natural error rate of Taq DNA polymerase. PCR reactions (100 μl volume) contained 1× Taq buffer (Roche), 200 μm each dNTP, 400 nm each oligo, 2.5 units of Taq (Roche), and 10 ng plasmid DNA. The PCR cycling parameters were the following: 2 min at 94°; 10 cycles of 1 min at 94°, 1 min at 55°, 2 min at 72°; and 4 min at 72°. After every 10 cycles of PCR amplification, reactions were diluted 1:100. Two separate sets of reactions were performed (set A/B and set C). For set A, a total of 20 PCR cycles were performed, and an additional 10 cycles gave set B. For set C, a total of 40 cycles using Taq were performed, followed by an additional 10 cycles using Expand Taq. At least two independent PCR reactions were performed for each set. pTB19 (pRS314/myc-SNU114) was the template for sets A and B and pTB4 (pRS314/SNU114) was the template for set C. Fragment 1 PCR products were transformed with the EcoRI/BstBI-cut vector of pTB19 or pTB4, and fragment 2 PCR products were transformed with the BstBI/NdeI-cut vector of pTB19 or pTB4. PCR products and gapped vectors were transformed into yTB2. Transformants were selected on SD-TRP media at 25° and replica plated to 5-FOA media at 25°. Lethality on 5-FOA was 8–13% for fragment 1 and 2–4% for fragment 2, depending on the number of PCR cycles. Strains were then tested for conditional lethality by replica plating to YPD and incubating at 16° for 3 days, 25° for 1–2 days, and 37° for 1 day. Approximately 25,000 colonies were screened for each fragment. Plasmids from strains that were growth impaired at 37° or 16° were recovered and transformed into yTB2 and yTB13. Plasmids that conferred conditional growth defects were sequenced by the Biomolecular Resource Center DNA sequencing facility at the University of California, San Francisco. The NCBI BLAST alignment server (http://www.ncbi.nlm.nih.gov/blast/bl2seq/wblast2.cgi) was used to identify mutations in the recovered snu114 alleles.
The mutations K146I and T147N were created using the QuikChange site-directed mutagenesis method (Stratagene, La Jolla, CA) with the template pTB2. The mutations P216N and H218R were created using site-directed mutagenesis of the template pTB19. Mutations were confirmed by sequencing and were subcloned into fresh plasmids.
Integration of snu114 alleles:
An integrating snu114-12 plasmid (pTB126) was made by inserting the 3.3-kb PvuII fragment of pTB107 into the PvuII-cut vector of pRS306. Integrating snu114-40 and snu114-60 plasmids (pTB130 and pTB132) were created by ligating the 4.25-kb NheI/NgoMIV fragment of pTB126 with the 3.63-kb NheI/NgoMIV fragments of pTB111 and pTB113. Integrating plasmids pTB126, pTB130, and pTB132 were linearized with NheI (pTB126) or Eco47III (pTB130/132), transformed into yTB128, and selected on SD-URA media at 25°. Following growth on 5-FOA, transformants were streaked to YPD and grown at 16°, 25°, and 37° to select for thermal-sensitive (ts) or ts/cold-sensitive (cs) integrants.
For primer extensions, the following strains were grown in liquid YPD media to OD 0.5–1.0: yTB128 (SNU114), yTB165 (snu114-60), and yTB171 (snu114-12), all grown at 25°; and yTB23 carrying either pTB106 (SNU114) or pTB111 (snu114-40), grown at 30°. Cells were spun down, resuspended in YPD media preincubated at 37° or 16°, and grown in water baths at either 37° or 16°; 10-ml aliquots were removed at the indicated times. Cultures were diluted during the time course to maintain an OD between 0.5 and 1.0. prp16-2 [yS79 (Wang and Guthrie 1998) + BHM110] was shifted to 37° for 3 hr as a control for a mutant that blocks the second step of splicing. RNA was isolated (Schmitt et al. 1990), and primer extension was performed as described (Boorstein and Craig 1989). Ten micrograms of RNA was used per reaction. The following oligos were used: U3, 5′-CCAAGTTGGATTCAGTGGCTC-3′; RPS17/RP51, 5′-CTTAGAAGCACGCTTGACGG-3′; PGK1, 5′-ATCTTGGGTGGTGTTCC-3′; and U14, 5′-ACGATGGGTTCGTAAGCGTACTCCTACCGT-3′. Data were quantitated by phosphorimager analysis (Molecular Dynamics, Sunnyvale, CA).
In vitro splicing assays:
Liquid cultures of yTB23 in which pTB1 had been replaced by pTB106 (SNU114), pTB107 (snu114-12), pTB111 (snu114-40), or pTB113 (snu114-60) were grown at 30° to OD 1.2–1.4. Splicing extracts were prepared and actin pre-mRNA was spliced as described (Umen and Guthrie 1995). Extracts were preincubated for 20 min at 37° or on ice in the presence of splicing buffer components (2.5 mm MgCl2, 60 mm potassium phosphate pH 7, 3% PEG 8000) prior to addition of radiolabeled actin and ATP (final concentrations of 0.5 nm and 2 mm, respectively). Splicing reactions were performed at 25°.
Testing genetic interactions:
To test genetic interactions, we created strains deleted for SNU114 (snu114::KanMX) in combination with deletion or mutation of a second gene. Deletions were covered by wild-type plasmids marked with URA3. Plasmids containing mutant alleles of snu114 (and, in some cases, of a second gene) were transformed, and the ability to lose the URA3-marked wild-type plasmid(s) on 5-FOA was tested.
The following strains (mutant allele/wild-type allele) were transformed with pTB106-113: yTB102/yTB23 (snu66Δ), yTB103/yTB105 (brr2-1), yTB106/yTB107 (prp28-1), yTB117/yTB118 (prp5-1), yTB133/yTB23 (brr1D), yTB139/yTB23 (snt309Δ), yTB142/yTB143 (sad1-1), yTB144/yTB145 (prp19-1), and yTB146/yTB145 (prp2-1). The strains yTB111/yTB15 (isy1Δ) were transformed with pTB95-102. yTB108 (prp8Δ) was transformed with combinations of either pTB95-102 and pJU204, pAK338, pJU206, and pCC18 or pTB106-113 and pCC11 and pCC121. yTB134 (prp22Δ) was transformed with pTB106-113 and pTB115-117. yTB135 (prp43Δ) was transformed with pTB106-113 and pTB118-121. yTB136 (prp24Δ) was transformed with pTB106-113 and pPR113, pSR53, pSR70, and pSR39. yTB148 (sub2Δ) was transformed with pTB106-113 and pAK354-356. yTV161 (prp16Δ) was transformed with pTB106-113 and pSB58, BHM108-110, and BHM115. Additionally, snu114ΔN (Bartels et al. 2002) was cloned into pRS315 and transformed into yTB23, yTB102, yTB103, and yTB105-107. In all cases, corresponding empty vectors were also transformed as negative controls. Transformants were selected on the appropriate selective media at 25°. Between 4 and 14 transformants were streaked onto 5-FOA-containing media and incubated at 25° for up to 6 days. If no colonies grew on 5-FOA, the combination of alleles was considered synthetically lethal. Viable strains were streaked to YPD media at 25°. Liquid cultures of each strain in duplicate were grown overnight at 25° and diluted to OD 0.1. Tenfold serial dilutions were grown on YPD plates at 16° for 8 days and at 25°, 30°, and 37° for up to 4 days. For prp8 strains, six 5-fold serial dilutions were made; the middle three dilutions are shown in Figure 7. For weak genetic interactions, we cannot distinguish between synthetic enhancement and additive enhancement.
A high number of transformants of snu114-12, snu114-14, and snu114-15 in combination with prp28-1 or snu66Δ (and snu114-50 with snu66Δ) either were not viable on 5-FOA or produced very few colonies. For example, for the combination of snu114-12 and snu66Δ, 4/14 transformants were not viable on 5-FOA, and 9/14 transformants produced only a small number of colonies on 5-FOA. To avoid the possibility that the viable colonies represented suppressors or revertants, we crossed integrated snu114-12 (yTB171), snu114-40 (yTB163), snu114-60 (yTB165), and SNU114 (yTB28) to snu66::KanMX (yTB100) and prp28-1 (yEJS51) (Strauss and Guthrie 1991) strains and dissected tetrads. Genotypes of the spores were determined by replica plating to G418-containing media (for snu66::KanMX) and to YPD plates that were incubated at 16° and 37° (for the other ts/cs mutations).
Sequence alignment and structure modeling:
An alignment between Snu114 and Eft1 (S. cerevisiae EF2) was created using ClustalW (Thompson et al. 1994) and modified by hand, and the structure of Snu114 was modeled on the structure of apo-EF2 (PDB code 1N0V) and sordarin-bound EF2 (PDB code 1N0U) using the program MODELLER (Sali and Blundell 1993). Structures were visualized with PyMOL (DeLano 2002). Coordinates of the models are available upon request.
The alignment of domain IVb was shaded using BOXSHADE (version 3.2; K. Hofmann and M. Baron). Swiss-Prot accession numbers are as follows: Snu114 from S. cerevisiae (P36048), Schizosaccharomyces pombe (O94316), Arabidopsis thaliana (Q9LNC5), Homo sapiens (Q15029), Caenorhabditis elegans (Q23463), Drosophila melanogaster (Q9VAX8); EF2 from S. cerevisiae (P32324), H. sapiens (P13639), C. elegans (P29691); EF-G from Thermus thermophilus (P13551), Escherichia coli (P02996).
Modeling the structure of Snu114:
To analyze how mutations in Snu114 might affect its function, we used the program MODELLER (Sali and Blundell 1993) to model the structure of Snu114 onto the crystal structure of S. cerevisiae EF2 (Jorgensen et al. 2003), which was possible because of the high sequence similarity between the two proteins (26% identity, 46% similarity). By homology with EF2, Snu114 contains five structural domains (Figure 1A), as well as a 120 -amino-acid N-terminal extension that is not conserved in EF2 and thus could not be modeled. The C terminus of the protein (IVb) folds back onto domain IVa and so is considered part of domain IV. For clarity, we refer to the two portions of domain IV as IVa and IVb. Figure 1, B and C, show the structure of Snu114 modeled onto nucleotide-free EF2, and Figure 1D shows the structure of Snu114 modeled onto EF2 bound to the translation inhibitor sordarin, which is believed to block EF2 on the ribosome in a post-translocation state (Jorgensen et al. 2003). The two structures demonstrate the flexibility of the protein. In particular, domains I and II appear as a rigid body, while domain III rotates around a linker between domains II and III, and domains IV and V rotate as a rigid unit.
Screen for conditionally lethal alleles of SNU114:
To generate conditionally lethal alleles of SNU114, we used the error-prone polymerase Taq to amplify the gene in two pieces: fragment 1 spans domains N, I, and II, and fragment 2 spans domains III–V (Figure 1A). Each fragment was transformed in combination with an appropriately gapped plasmid containing SNU114 and the TRP1 marker into a yeast strain in which the chromosomal copy of SNU114 was deleted and wild-type SNU114 was present on a counterselectable URA3-marked plasmid (Muhlrad et al. 1992). For each fragment, we screened ∼25,000 transformants. When colonies were replica plated to 5-FOA-containing media to select against the wild-type SNU114 plasmid, >87% of the colonies were viable. We then screened for colonies that were unable to grow at 37° or 16°. From the transformation of mutagenized fragment 1, we isolated 37 ts strains and one strain that was both cs and ts. Mutagenesis of fragment 2 yielded 54 ts and three cs/ts strains. Plasmids were recovered from 20 ts strains for each of the fragments and from all of the cs/ts strains and were retransformed into the starting strain. Five of the plasmids from fragment 1 did not retest and were discarded. The remaining plasmids were sequenced.
The sequenced alleles contained an average of two mutations per gene, with a range of one to six mutations (Tables 3 and 4). Despite the presence of multiple mutations per allele, many of the mutations clustered within small stretches of highly conserved amino acids, highlighting regions that are important for Snu114 function (Figure 1A). Individual amino acids within these clusters were often mutated in multiple independent clones.
Fragment 1 mutations:
Domain I contains the conserved motifs, G1–G5, which are present in all GTPases and are necessary for GTP binding and hydrolysis (Bourne et al. 1991). Of the 16 alleles from fragment 1 that were sequenced, 6 contain a mutation in the G1 motif, and 4 contain a mutation in the G3 motif. Concomitant with the PCR mutagenesis, we designed two mutations in the G1 motif of SNU114 that are expected to decrease nucleotide binding (Ogg et al. 1998); these mutations, K146I (snu114-15) and T147N, confer a ts growth phenotype (Table 5). To study single point mutations in the G3 motif of SNU114 that are predicted to affect GTP hydrolysis, we created the individual mutations P216N (snu114-12) and H218R, which had arisen in combination with other mutations during PCR mutagenesis. Each of these mutations alone causes a ts growth defect (Table 5).
While the structure of the G domain of most GTPases is similar, an insertion termed the G′′ domain is found in EF2 and Snu114, but not in the other translational GTPases (Aevarsson 1995). Although no function has been assigned to this domain, it has been postulated to act as a guanine exchange factor (GEF), since EF2 is not known to have an extrinsic GEF (Aevarsson et al. 1994; Czworkowski et al. 1994). In Snu114, a single mutation in this domain, L381P (snu114-14), causes slow growth and a weak ts/cs phenotype.
Fragment 2 mutations:
Over half of the mutations identified in domains III–V are found within three 10- to 20-amino-acid stretches, which are underlined in Figure 1A and highlighted in Figure 1, C and D. The 23 sequenced fragment 2 alleles contain a total of 49 mutations. Thirteen of the mutations fall within a short stretch of domain III (residues 645–664), 8 mutations are within a small region of domain IVa (residues 842–851), and 7 mutations are clustered in domain V (residues 909–928). While a majority of the alleles that arose from mutagenesis of fragment 2 contain multiple mutations, 6 alleles contain a single point mutation. Four of these mutations, including snu114-30, are in the domain III cluster, and one, snu114-40, is in the domain IVa cluster. That single-amino-acid changes cause a conditional growth defect emphasizes the functional significance of these regions. Furthermore, in comparing Snu114, EF2, and EF-G, the amino acid cluster in domain III contains the most conserved stretch of amino acids outside of domain I (Cammarano et al. 1992). As shown in Figure 1, C and D, the clusters in domain III and V are in close proximity to each other and to the G3 motif. The cluster of residues in domain IVa forms a helix that faces domain IVb.
In addition to the mutations in domains III, IVa, and V, an additional class of mutations was found in domain IVb. Domain IVb is larger in Snu114 than in its homologs: the domain is 76 residues in Snu114, but only 44 residues in yeast EF2 and 20 residues in E. coli EF-G (Figure 1E). In comparison with EF2, Snu114 domain IVb has several insertions as well as a C-terminal extension. Only three sets of mutations were found in this domain, and all result in early truncation of the protein. The weak ts allele snu114-62 causes a deletion of 23 amino acids, while the ts/cs alleles snu114-60 and snu114-61 result in the deletion of 70 and 68 amino acids, respectively (Table 4).
The severity of the growth defect of the C-terminal deletions was affected by the presence of a myc epitope that had been placed at the N terminus of the gene prior to mutagenesis. Although myc-snu114-60 and snu114-61 have similar deletions, the growth defect of myc-snu114-60 is much stronger. The allele snu114-61 also contains mutations in domains II and IV and does not contain the myc epitope, perhaps due to homologous recombination with chromosomal SNU114. We tested whether any of these differences could suppress the growth defect of myc-snu114-60 and found that removing the myc epitope from this allele greatly reduces the severity of the growth defect at both high and low temperatures (Figure 2A). The presence or absence of the myc epitope did not affect the growth phenotype of the other snu114 alleles that were further characterized.
Where possible, we focused subsequent experiments on a set of alleles with single mutations in each of the domains that were identified above as important. For the work described below, we used the alleles snu114-15 (G1 motif), snu114-12 (G3 motif), snu114-14 (G′′ domain), snu114-30 (domain III), snu114-40 (domain IVa), snu114-50 [domain V; snu114-50 contains two mutations, because the single mutations did not cause a growth phenotype (Table 5)], and snu114-60 (truncation of domain IVb; see Tables 3 and 4). All alleles are ts. snu114-14 and snu114-60 are only weakly ts as well as weakly cs (Figure 2B).
snu114 mutants exhibit in vivo and in vitro splicing defects:
We tested whether a subset of the snu114 mutations inhibit splicing. To monitor in vivo splicing, we purified RNA from cells that had been shifted to the nonpermissive temperature for various times and performed primer extensions on the intron-containing transcripts U3 and RPS17/RP51. U3 is a nucleolar snoRNA, and RPS17 encodes a ribosomal protein. Mutations that block the first chemical step of splicing are expected to increase the level of precursor mRNA, while mutations that affect the second step of splicing should cause an accumulation of lariat intermediate. Strains containing the integrated snu114-12 allele grow slowly even at 25° and show a constitutive accumulation of U3 precursor (Figure 3A). The level of pre-U3 in snu114-60 cells, which are both ts and cs, increases following a shift to either 37° or 16° (Figure 3, B and C). Primer extension of the less stable RPS17 transcript reveals that shifting snu114-40 to 37° causes a rapid increase in pre-mRNA, as well as a rapid and dramatic decrease in the level of mature mRNA (Figure 3D). While primer extension with U3 allows only the differentiation of precursor from mature RNA, the lariat intermediate of RPS17 can be resolved. In contrast to the second-step mutant prp16-2 (Burgess and Guthrie 1993), none of the snu114 mutants that we tested accumulate lariat intermediate (Figure 3D and data not shown). Together, pre-mRNA accumulation and a lack of lariat intermediate accumulation show that the snu114 mutants block splicing at or before the first chemical step.
The data from in vitro splicing in extracts made from snu114 mutant strains grown at the permissive temperature were consistent with the in vivo splicing data. To monitor splicing in vitro, radiolabeled actin pre-mRNA was incubated with extracts that had been preincubated either at 37° or on ice. The splicing intermediates and products were then separated by PAGE and visualized by phosphorimaging. The snu114-12 extract splices poorly even at 25°, consistent with a constitutive defect (Figure 4A). Splicing in both snu114-40 and snu114-60 extracts can be inactivated by preincubation at high temperature (Figure 4, B and C). While the kinetics of splicing in snu114-60 extract at 16° in vitro are slow, they are not strongly inhibited (data not shown). For all snu114 alleles, the levels of all splicing intermediates are decreased in the mutant extracts, indicating an early block to splicing.
Genetic interactions of snu114 mutants:
Inhibition of the first step of splicing can arise for many reasons, including defects in snRNP levels, defects in U1 or U2 addition, defects in tri-snRNP addition, and defects in activation of the spliceosome. Synthetic enhancement has proven to be a powerful tool for elucidating functions of and interactions between factors of large multicomponent complexes (Guarente 1993; Doye and Hurt 1995). To gain an understanding of the timing of Snu114 function(s), we examined genetic interactions between the snu114 alleles and mutations or deletions of 16 splicing factors known to be required at different stages of splicing (Figure 5). We tested each combination for viability and for conditional growth defects. Taking into consideration changes in growth rates at 16°, 25°, 30°, and 37°, the overall strength of synthetic enhancement for each interaction was rated on a scale of 0 to 5, where 0 indicates no interaction and 5 indicates a synthetically lethal interaction (Table 6). Interactions that were synthetically lethal or synthetically sick are summarized in Figure 5 (solid stars and shaded stars, respectively).
Interactions with ATPases:
Each of the eight spliceosomal DExD/H box ATPases acts at a clearly defined and distinct stage of splicing (Staley and Guthrie 1998). Strikingly, mutations in the ATPases Prp28 and Brr2, which are needed for activation of the spliceosome, exhibit strong genetic interactions with snu114 mutations. Although snu114-60 has only a weak conditional growth defect, it is inviable in combination with prp28-1, a cs mutation of Prp28 that disrupts the exchange of U1 for U6 at the 5′ splice site (Staley and Guthrie 1999) (Figure 6A). prp28-1 is also synthetically sick in combination with mutations in domain I (snu114-12, snu114-14, and snu114-15) (data not shown). brr2-1, a cs mutation that decreases the U4/U6 unwinding activity of Brr2 (Raghunathan and Guthrie 1998a), is synthetically lethal with snu114-60 at 16° and 37° (Figures 6A and 7B). In contrast to prp28-1, brr2-1 interacts strongly only with snu114-60, although snu114-14 and snu114-40 enhance the cs growth defect (Figure 7B and data not shown).
snu114-60 is not synthetically lethal with any of the other ATPase mutations that were tested (Figure 6A and data not shown). The growth defects of the other snu114 alleles are moderately enhanced by mutations in Sub2, which functions during a U2 snRNP addition (Kistler and Guthrie 2001; Libri et al. 2001; Zhang and Green 2001); in Prp2, which acts immediately before the first chemical step (Kim and Lin 1996); and in Prp22, which helps to disassemble the postcatalytic spliceosome (Company et al. 1991). Mutations in Prp5, Prp16, and Prp43, which are involved in U2 addition, second-step catalysis, and spliceosome recycling, respectively, exhibit no genetic interactions with snu114 alleles (Schwer and Guthrie 1991; Ruby et al. 1993; Arenas and Abelson 1997). Together, the pattern of genetic interactions with the spliceosomal ATPases suggests that Snu114, and especially domain IVb, is specifically involved in spliceosome activation.
Interactions with PRP8 alleles:
Distinct mutations of the core U5 protein Prp8 affect several stages of spliceosome assembly and both steps of splicing. While none of the snu114 mutations have strong genetic interactions with prp8-101, which inhibits the second step of splicing (Umen and Guthrie 1995), many of the alleles have synthetic interactions with prp8 mutations that cause earlier splicing defects (Figures 6B and 7A). The mutations prp8-1 and prp8-brr impair the formation of U5 and tri-snRNP (Brown and Beggs 1992; Collins 2001) and exhibit strong genetic interactions with snu114 mutations. Notably, prp8-1 and prp8-brr are synthetically lethal with snu114-40 and snu114-60, indicating that Snu114 domain IV is important for interacting with Prp8 during snRNP formation. Additionally, many of the snu114 mutations are synthetically sick or lethal in combination with prp8-201, an allele that affects spliceosome activation (Kuhn et al. 1999).
Interactions with factors involved in tri-snRNP addition, snRNP biogenesis, and snRNP recycling:
Because snu114 mutations cause a block prior to the first step of splicing, we also tested genetic interactions with factors that are involved in the addition of tri-snRNP to the spliceosome and in snRNP biogenesis and recycling. In mammalian extract, depletion of either of the tri-snRNP proteins Snu66 or Sad1 inhibits tri-snRNP addition (Makarova et al. 2001). In yeast, deletion of SNU66 causes a cs growth defect and inhibits the first step of splicing (Gottschalk et al. 1999; Stevens et al. 2001), and the ts allele sad1-1 blocks splicing and decreases the formation of U4/U6 di-snRNP (Lygerou et al. 1999). Deletion of SNU66 is synthetically lethal with snu114-60 and enhances the growth defects of snu114-12, -14, -15, and -50 (Figures 6A and 8 and data not shown). All of the snu114 alleles are synthetically lethal or sick in combination with sad1-1.
The recycling factor Prp24 assists in the base pairing of U4 and U6 snRNAs (Raghunathan and Guthrie 1998b). It has also been proposed to influence spliceosome activation, as alleles of prp24 are synthetically lethal with prp28-1 and with U4-cs1, a mutation in U4 that blocks spliceosome activation (Strauss and Guthrie 1991; Kuhn and Brow 2000). The ts mutation prp24-RRM3sub, which is believed to disrupt the ability of Prp24 to bind RNA (Vidaver et al. 1999), enhances the growth defect of snu114-60. However, none of the other snu114 mutations exhibit genetic interactions with prp24 mutations.
Deletion of the nonessential gene BRR1 causes a cs growth defect, affects the processing of newly transcribed snRNAs, and causes a decrease in snRNA and snRNP levels (Noble and Guthrie 1996; Inada 2004). The deletion of BRR1 is synthetically lethal with snu114-60 at 16° and 37°, and the combination of brr1Δ with the other snu114 mutations impairs growth at 30° (Figure 7C and data not shown).
Interactions with the NTC:
Finally, we tested components of the prp-nineteen complex (NTC), a complex that interacts with the spliceosome immediately prior to the first step of catalysis (Tarn et al. 1993; Chan et al. 2003). prp19-1, a ts mutation that blocks the first step of splicing (Vijayraghavan et al. 1989), enhances the growth defects of all of the snu114 alleles. Notably, Prp19 is an E3 ubiquitin ligase, and the prp19-1 mutation decreases its enzymatic activity (Ohi et al. 2003). Deletions of the nonessential NTC proteins ISY1 and SNT309 (C. H. Chen et al. 1998, 2001; Dix et al. 1999) do not interact genetically with the snu114 alleles.
Snu114 is a GTPase with homology to the ribosomal translocase EF-G, suggesting that it may mediate conformational rearrangements in the spliceosome. On the basis of comparison with EF-G, Snu114 can be divided into five structural domains (I–V), as well as a nonconserved N-terminal domain. In this study, we screened for conditionally lethal alleles of Snu114. We recovered a large number of mutations within the GTPase domain and in three small clusters of amino acids in domains III–V. Additionally, we found that snu114-60, which causes an almost complete deletion of the C-terminal domain IVb, causes a conditional growth defect. Domain IVb is larger in Snu114 than in its ribosomal homologs, suggesting that this domain may be important for an interaction specific to splicing. snu114-60 is synthetically lethal with mutations in prp8 and in other factors involved in activation of the spliceosome. We propose that domain IVb interacts with Prp8 to influence the activities of the DExD/H box ATPases Prp28 and Brr2 during spliceosome activation.
Studies of EF-G and EF2 indicate that GTP hydrolysis causes a substantial movement of domain IV with respect to domains I and II (Stark et al. 2000; Jorgensen et al. 2003) (see Figure 1, C and D). This conformational rearrangement of EF-G/EF2 is predicted to drive movement of tRNA within the ribosome (Rodnina et al. 2000). By analogy, GTP hydrolysis by Snu114 may cause rearrangements of the spliceosome.
All members of the GTPase superfamily share a similar architecture of the G domain and contain conserved motifs (G1–G5) that interact with GTP/GDP (Bourne et al. 1991; Sprang 1997). Our screen for conditionally lethal alleles of Snu114 identified mutations in the G1 and G3 motifs, which are predicted to cause defects in GTP binding and hydrolysis (Sigal et al. 1986; Anborgh et al. 1989; Zeidler et al. 1995). In multidomain G proteins, the GTP/GDP binding status is communicated to other domains via rearrangements of the G2 and G3 motifs (Sprang 1997). The clusters of mutations that we identified in domains III and V are located near points of contact with the G3 motif (Figure 1, C and D). Although the G2 motif is disordered in all EF-G/EF2 structures, comparison with the structure of the GTPase elongation factor Tu (EF-Tu) suggests that it is also close to the domain III cluster (Laurberg et al. 2000). Therefore, the mutations that we found in domains III and V may impair a conformational rearrangement of the protein that normally results from GTP hydrolysis.
Domain IV of EF-G is necessary for transmitting a rearrangement within the protein, arising from GTP hydrolysis, into a conformational change of the ribosome, as deletion of IVa or IVa + IVb decreases translocation by ∼2000-fold without affecting GTP hydrolysis (Rodnina et al. 1997; Martemyanov and Gudkov 1999; Savelsbergh et al. 2000). Notably, deletion of six amino acids in EF-G that correspond to the cluster of mutant amino acids that we identified in Snu114 domain IVa, including snu114-40, causes a 300-fold reduction in translocation of the ribosome (Savelsbergh et al. 2000). Thus, the mutations in Snu114 domain IVa may prevent the transmission of a conformational rearrangement that normally results from GTP hydrolysis or nucleotide exchange. Studies of EF-G/EF2 have not directly addressed the role of domain IVb alone (although this domain is only 20 amino acids in E. coli). On the basis of our finding that deletion of domain IVb (snu114-60) causes a conditional growth defect and a splicing defect, we predict that domain IVb could be an interaction domain that communicates conformational changes of Snu114 to other splicing factors.
The βαββαβ structure of domains III, IVa, and V of EF-G is characteristic of the RNA recognition motif (RRM) (Aevarsson et al. 1994; Murzin 1995; Laurberg et al. 2000). In RRM domains, hydrophobic residues in two of the β-sheets interact directly with RNA (Varani and Nagai 1998). In our mutagenesis of Snu114, we found very few mutations in residues that would be predicted to interact with RNA; thus, our data do not provide evidence for a direct interaction between Snu114 and RNA through the RRM motifs. However, it is possible that we failed to obtain such mutations because they cause lethality or because our screen was not saturating.
Biochemistry and genetics suggest that Snu114 functions during spliceosome activation:
We have found that mutations in Snu114 block splicing prior to the first chemical step both in vivo and in vitro. A first-step block to splicing could reflect defects at a number of stages, including snRNP biogenesis and stability, tri-snRNP addition to the spliceosome, and spliceosome activation. In fact, we found that snu114 mutants exhibit genetic interactions with splicing mutants that affect each of these stages (Figure 5). However, snu114 mutants do not interact genetically with any factors that function after the first step of catalysis, in agreement with the biochemical data showing a first-step block.
A number of proteins that we tested affect multiple stages of splicing. For example, Prp24 and Sad1 are both implicated in tri-snRNP addition/reorganization as well as in formation of U4/U6 snRNP (Raghunathan and Guthrie 1998b; Lygerou et al. 1999; Kuhn and Brow 2000; Makarova et al. 2001). The snRNP biogenesis factor BRR1 exhibits genetic interactions not only with snRNP biogenesis factors, but also with factors that affect spliceosome assembly. Interestingly, brr1Δ and the snu114 alleles are synthetically lethal with many of the same mutations, including prp24-RRM3sub, sub2 mutations, sad1-1, and snu66Δ (Inada 2004). The genetic interactions observed between snu114 mutations and prp24-RRM3sub, sad1-1, and brr1Δ could arise because (a) all of the proteins are involved in the addition and rearrangement of tri-snRNP, (b) the function of Snu114 during spliceosome activation is particularly sensitive to low levels of snRNPs, or (c) Snu114 is also necessary for snRNP biogenesis. However, snu114-60, which has the strongest interactions with prp24, sad1, and brr1 mutants, contains wild-type snRNP levels (T. J. Brenner and C. Guthrie, unpublished data).
Snu114 alleles exhibit strong genetic interactions with factors involved in spliceosome activation, including prp28-1, brr2-1, and prp8-201 (Guthrie 1998a; Kuhn et al. 1999; Raghunathan and Staley and Guthrie 1999). We also found strong genetic interactions with deletion of SNU66 and mutation of SAD1, factors involved in tri-snRNP addition (Makarova et al. 2001). All of these genetic interactions may be related, as mutations that block spliceosome activation, including prp-28-1 and brr2-1, decrease the interaction of tri-snRNP with the spliceosome (Raghunathan and Guthrie 1998a; Staley and Guthrie 1999), indicating that rearrangements within the tri-snRNP are necessary for its stable association with the spliceosome. Together, these genetic interactions strongly implicate Snu114 in spliceosome activation.
The N- and C-terminal domains of Snu114 are involved in similar functions:
Snu114 contains a 120-amino-acid N-terminal domain that is not found in EF-G/EF2. Previous studies showed that deletion of this domain (snu114ΔN) causes a ts growth defect and a block to the release of U1 and U4 from the spliceosome (Bartels et al. 2002). Similar to the biochemical phenotype of snu114ΔN, genetic interactions suggest that snu114-60 causes defects in spliceosome activation. Furthermore, we found that the presence of a 15-amino -acid epitope tag at the N terminus of Snu114 is synthetically lethal with snu114-60 at 16° and 37° (Figure 2A), which suggests that both the N-terminal domain and domain IVb are necessary for the same process. If the N and C termini were involved in the same function, we predicted that they would exhibit the same genetic interactions. Similarly to snu114-60, snu114ΔN is synthetically lethal with prp28-1 and snu66Δ and sick with brr2-1 (T. J. Brenner and C. Guthrie, unpublished data). We conclude that both domains are necessary for tri-snRNP addition/spliceosome activation. It is possible that the N-terminal domain is necessary for the activation of GTPase, while the C-terminal domain is necessary to transmit a conformational change driven by GTP hydrolysis. Alternatively, it is possible that both domains must interact with other factors, such as Prp8, for a conformational change in Snu114 to be transmitted.
A model for the activity of Snu114:
Snu114 and Prp8 physically interact with each other, and both proteins can be crosslinked to U5 snRNA (Achsel et al. 1998; Dix et al. 1998). Thus, interactions between Snu114, Prp8, and U5 snRNA form the core of U5 snRNP. Domain IVa/IVb of Snu114 may be particularly important for this interaction, as snu114-40 and snu114-60 are synthetically lethal with prp8-1 and prp8-brr.
Synthetic lethality between snu114-60 and mutations in factors involved in spliceosome activation indicates a pivotal role for domain IVb in this process. Previous genetic data have suggested that an allosteric interaction between Prp8 and the ATPases Prp28 and Brr2 regulates the timing of spliceosome activation (Kuhn and Brow 2000; Kuhn et al. 2002). The mutations prp28-1 and brr2-1 are synthetically lethal with the U4-cs1 mutation, which increases the base pairing between U4 and U6 snRNAs and inhibits the release of U1 and U4 from the spliceosome (Li and Brow 1996; Kuhn et al. 1999). A large number of prp8 alleles, including prp8-201, can suppress the cold sensitivity of U4-cs1, and a subset of these prp8 alleles can suppress prp28-1 or brr2-1. Thus, it has been hypothesized that Prp8 inhibits the activity of the ATPases until spliceosome formation has occurred, while certain mutations of prp8 may relax this inhibition.
On the basis of the strong genetic interactions between SNU114 and PRP8, PRP28, and BRR2, we suggest that Snu114 may regulate Prp8's inhibition of Prp28 and Brr2 (Figure 9). We hypothesize that Snu114 is bound to GTP when tri-snRNP binds to the spliceosome. GTP hydrolysis could be triggered by interactions with U1 snRNP, since Snu114 and the U1 proteins Prp39 and Prp40 interact with neighboring regions of Prp8, according to yeast two-hybrid analyses (Abovich and Rosbash 1997; Dix et al. 1998; van Nues and Beggs 2001; Grainger and Beggs 2005). Thus, successful interaction with the spliceosome would induce GTP hydrolysis by Snu114 and a concomitant structural rearrangement of domains III–V with relation to the N terminus. In particular, this may modify the interaction between domain IVb and Prp8, which in turn could alter the conformation of Prp8 and lead to the unwinding of the U1/5′ splice site duplex by Prp28 and the U4/U6 duplex by Brr2. It is possible that Prp8 sequesters the RNA helices, and an altered conformation of Prp8 would allow Brr2 and Prp28 access to their substrates. Alternatively, Prp8 could inhibit the ATPase activity of Prp28 and Brr2 via direct physical interactions with the DExD/H-box proteins, and altering the conformation of Prp8 could relieve the inhibition by severing these associations (Kuhn and Brow 2000; Kuhn et al. 2002).
Previous biochemical studies support the model that Snu114 hydrolyzes GTP during spliceosome activation. A mutation that is expected to convert the nucleotide specificity of Snu114 from GTP to XTP causes a temperature-sensitive block to U1 and U4 release, which can be partially overcome by addition of XTP (Bartels et al. 2003). Nonhydrolyzable XTP did not stimulate snRNA release, suggesting that GTP hydrolysis, and not just GTP binding, is important (Bartels et al. 2003).
For most GTPases, GTPase-activating proteins (GAPs) are necessary to stimulate GTP hydrolysis. The ribosome itself serves as the GAP for EF-G and EF-Tu, which delivers aminoacyl tRNA to the ribosome (Mohr et al. 2002). In the case of EF-Tu, a cognate codon:anticodon interaction is required to trigger GTP hydrolysis (Rodnina et al. 2005). Because the structure of EF-Tu bound to aminoacyl-tRNA resembles EF-G (Nissen et al. 1995), and because EF-G and Snu114 are homologous, it is tempting to speculate that interactions between spliceosomal components and Snu114 domain IV, which would correspond to the anticodon arm of tRNA, may be necessary to trigger GTPase activity. Just as GTP hydrolysis by EF-Tu functions as a checkpoint to ensure that the appropriate aminoacyl tRNA is retained in the ribosome, GTP hydrolysis by Snu114 after addition of tri-snRNP could serve as a checkpoint to ensure proper formation of the spliceosome. It will be informative to investigate what components of the spliceosome function as a GAP for Snu114. Additionally, we are interested in testing whether mutations in domain IV of Snu114 directly affect GTP binding and hydrolysis.
A link to ubiquitination:
Strong genetic interactions between Snu114 and proteins that are involved in ubiquitin metabolism suggest that a cycle of ubiquitination could affect splicing. In a large-scale proteomics study, Snu114 was one of only three splicing proteins, including Sad1, that were found to be ubiquitinated (Peng et al. 2003). Interestingly, Sad1, which has strong genetic interactions with Snu114, contains a ubiquitin hydrolase domain (Costanzo et al. 2000). Prp19, which we also found to interact genetically with Snu114, is a member of the U-box family of E3 ubiquitin ligases (Hatakeyama et al. 2001). The prp19-1 mutation specifically disrupts the fold of the U-box domain and greatly decreases ubiquitin ligase activity in vitro (Ohi et al. 2003). Finally, the Snu114-interacting protein Prp8 contains a Jab/MPN domain, which is typically associated with ubiquitin removal (Maytal-Kivity et al. 2002; Verma et al. 2002). Thus, the genetic interactions with sad1-1, prp-19-1, and prp8 mutants could reflect the requirement for a cycle of ubiquitination and deubiquitination of Snu114. Currently, the timing of Snu114 ubiquitination is not known. One possibility is that after U5 snRNP is released from the postcatalytic spliceosome in complex with the NTC (Makarov et al. 2002), ubiquitination of Snu114 by Prp19 induces dissociation of the two particles. Ubiquitin removal by Sad1 during the following round of spliceosome assembly could promote tri-snRNP binding or spliceosome activation. The particularly strong genetic interactions between sad1-1 and snu114 mutations suggests that deubiquitination of Snu114 is essential for splicing. Ubiquitination of Snu114 may also affect its interactions with Prp8, as Prp8 is predicted to bind ubiquitin via its Jab/MPN domain. We are most interested in determining when during splicing Snu114 is ubiquitinated and how this is affected by mutations in PRP19, SAD1, and PRP8.
We thank Beate Schwer, Patrizia Fabrizio, Scott Stevens, and Dave Brow for strains and plasmids, Jean Beggs for communicating results prior to publication, Marcy Diaz for assistance with genetic crosses, and Michael Dinglasan and Jorge Mendoza for excellent technical assistance. Thanks go to John Abelson, Elizabeth Blackburn, Hiten Madhani, Mike Springer, Miri VanHoven, and members of the Guthrie lab for discussions and comments on the manuscript. T.J.B. is a Howard Hughes Medical Institute predoctoral fellow. C.G. is an American Cancer Society Research Professor of Molecular Genetics. This work was supported by National Institutes of Health grant GM21119.
Communicating editor: T. Stearns
- Received February 15, 2005.
- Accepted April 13, 2005.
- Genetics Society of America