Nuclear Import of Upf3p Is Mediated by Importin-/-ß and Export to the Cytoplasm Is Required for a Functional Nonsense-Mediated mRNA Decay Pathway in Yeast

Corresponding author: Michael R. Culbertson, R. M. Bock Labs, University of Wisconsin, 1525 Linden Dr., Madison, WI 53706., mrculber{at}facstaff.wisc.edu (E-mail)

Communicating editor: S. SANDMEYER

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Upf3p, which is required for nonsense-mediated mRNA decay (NMD) in yeast, is primarily cytoplasmic but accumulates inside the nucleus when UPF3 is overexpressed or when upf3 mutations prevent nuclear export. Upf3p physically interacts with Srp1p (importin-). Upf3p fails to be imported into the nucleus in a temperature-sensitive srp1-31 strain, indicating that nuclear import is mediated by the importin-/ß heterodimer. Nuclear export of Upf3p is mediated by a leucine-rich nuclear export sequence (NES-A), but export is not dependent on the Crm1p exportin. Mutations identified in NES-A prevent nuclear export and confer an Nmd^- phenotype. The addition of a functional NES element to an export-defective upf^- allele restores export and partially restores an Nmd⁺ phenotype. Our findings support a model in which the movement of Upf3p between the nucleus and the cytoplasm is required for a fully functional NMD pathway. We also found that overexpression of Upf2p suppresses the Nmd^- phenotype in mutant strains carrying nes-A alleles but has no effect on the localization of Upf3p. To explain these results, we suggest that the mutations in NES-A that impair nuclear export cause additional defects in the function of Upf3p that are not rectified by restoration of export alone.

EUKARYOTIC cells from a variety of organisms including yeast, nematodes, mice, and humans rapidly eliminate mRNAs that contain a premature termination codon (LOSSON and LACROUTE 1979 ; LEEDS et al. 1991 ; PULAK and ANDERSON 1993 ; PERLICK et al. 1996 ). Nonsense mRNA degradation occurs through a pathway called nonsense-mediated mRNA decay (NMD), which serves two purposes. One is in RNA surveillance in which nonsense mRNAs arising through errors in gene expression are rapidly eliminated to prevent the accumulation of deleterious truncated proteins (PULAK and ANDERSON 1993 ; CALI and ANDERSON 1998 ; LI and WILKINSON 1998 ; FRISCHMEYER and DIETZ 1999 ). A second purpose is to control the abundance of a subset of endogenous wild-type mRNAs containing built-in signals that can trigger premature termination of translation, which leads to faster decay as part of the normal repertoire of gene expression (LELIVELT and CULBERTSON 1999 ). NMD requires a mechanism to distinguish a premature nonsense codon from the normal wild-type termination signal. In yeast, this involves the presence of a degenerate downstream sequence element (DSE) located 3' of a premature nonsense codon (PELTZ et al. 1993 ; HAGAN et al. 1995 ; ZHANG et al. 1995 ; CZAPLINSKI et al. 1999 ). mRNAs containing a premature termination codon but lacking a DSE fail to be degraded by the NMD pathway (PELTZ et al. 1993 ; RUIZ-ECHEVARRIA et al. 1996 , RUIZ-ECHEVARRIA et al. 1998 ).

In yeast, the three factors Upf1p, Upf2p, and Upf3p are required for NMD (LEEDS et al. 1991 , LEEDS et al. 1992 ; CUI et al. 1995 ; HE and JACOBSON 1995 ; LEE and CULBERTSON 1995 ). Mutations in the UPF genes stabilize nonsense mRNAs, resulting in rates of decay similar to those of the corresponding wild-type mRNAs. In addition, the efficiency of translation termination at premature stop codons is decreased in strains carrying upf^- mutations while the overall efficiency of nonsense mRNA translation increases (MUHLRAD and PARKER 1999 ; BIDOU et al. 2000 ). These effects, combined with the increase in mRNA stability, contribute to the ability of upf^- mutations to suppress nonsense and frameshift mutations (CULBERTSON et al. 1980 ; LEEDS et al. 1991 , LEEDS et al. 1992 ; MADERAZO et al. 2000 ).

Homologs of the yeast Upf proteins have been identified in Schizosaccharomyces pombe, Caenorhabditis elegans, mice, and humans (PAGE et al. 1999 ; LYKKE-ANDERSEN et al. 2000 ; MENDELL et al. 2000 ; ARONOFF et al. 2001 ; SERIN et al. 2001 ). Human hUpf1p and hUpf2p were identified as homologs of yeast Upf1p and Upf2p (PERLICK et al. 1996 ; APPLEQUIST et al. 1997 ). A mutation in the conserved helicase domain of hUpf1p confers a dominant-negative phenotype in yeast (LEEDS et al. 1992 ) and partially inactivates the NMD pathway in monkey COS and human HeLa cells (SUN et al. 1998 ). Several homologs of yeast Upf3p were identified in humans (LYKKE-ANDERSEN et al. 2000 ; SERIN et al. 2001 ). These proteins are derived from two genes, each of which expresses several isoforms due to alternative splicing. These studies suggest that the function of the Upf proteins in identifying and targeting nonsense mRNAs for rapid decay is conserved among eukaryotes.

Translation is required for the rapid decay of nonsense mRNAs. Nonsense mRNAs are stabilized by the presence of nonsense tRNA suppressors (LOSSON and LACROUTE 1979 ), and they are recruited into polyribosomes (LEEDS et al. 1991 ; HE et al. 1993 ; ZHANG et al. 1997 ). In addition, a portion of the total cellular pool of Upf1p, Upf2p, and Upf3p cofractionates with polyribosomes (ATKIN et al. 1997 ). Physical interactions between the yeast Upf proteins suggest that they act in concert to promote NMD on polyribosomes (HE and JACOBSON 1995 ; HE et al. 1997 ). The Upf proteins copurify with release factor eRF3 and Upf1p also copurifies with release factor eRF1 (CZAPLINSKI et al. 1998 ; WANG et al. 2001 ).

Upf1p associates with polyribosomes in the absence of Upf2p or Upf3p, and it appears to facilitate the dissociation of Upf2p with polyribosomes. Upf3p is required for the association of Upf2p with polyribosomes (ATKIN et al. 1997 ). On the basis of these results, it was proposed that Upf3p recruits Upf2p to polyribosomes (ATKIN et al. 1997 ; CULBERTSON 1999 , CULBERTSON 2001 ). Upf2p may then facilitate NMD by interacting with Upf1p (HE et al. 1996 ). Upf1p could be recruited to polyribosomes through an association with release factors (HE et al. 1997 ; CZAPLINSKI et al. 1998 ).

It is still unclear how nonsense mRNAs are initially identified as substrates for NMD. The Upf proteins are not associated with all translating ribosomes (ATKIN et al. 1995 ; MADERAZO et al. 2000 ). Upf1p is at least 100-fold less abundant than ribosomes. Furthermore, Upf2p and Upf3p are 10- to 20-fold less abundant than Upf1p. The low abundance of the Upf proteins suggests the existence of a mechanism that serves to specifically recruit the Upf proteins to the site of premature translation termination of nonsense mRNAs.

We showed previously that Upf3p contains a functional nuclear export sequence, suggesting that Upf3p may function in an early step in nonsense mRNA recruitment (SHIRLEY et al. 1998 ). However, it was not known at that time whether nuclear entry and exit is required for NMD. Nuclear localization was recently demonstrated for the human homologs of Upf3p. Several isoforms of hUpf3p were shown to shuttle between the nucleus and the cytoplasm in cell culture heterokaryons (LYKKE-ANDERSEN et al. 2000 ; SERIN et al. 2001 ). In this article, we demonstrate that the import of yeast Upf3p into the nucleus is mediated by the importin-/ß heterodimer. We further demonstrate that the export of Upf3p from the nucleus via a leucine-rich nuclear export sequence is required for a fully functional NMD pathway. Our results support the hypothesis that Upf3p functions in one of the initial steps necessary to earmark a nonsense mRNA for recruitment as a substrate in the NMD pathway. This occurs prior to or during export of a nonsense mRNA from the nucleus to the cytoplasm.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Strains and plasmids:
Strain RSy5 (MAT ade2-1 leu2-1 tyr7-1 can1-100 upf3-1 trp^- ura^- his3^- GAL2⁺) was used for immunofluorescence microscopy and to assay suppression of the leu2-1, tyr7-1, and can1-100 nonsense mutations. Strain LRSy323 (MATa his4-38 SUF1-1 trp1-1 upf3-1 ura3-52 leu2-1) was used for the experiment shown in Fig 1C. Strain PJ69-4A (MATa trp1-901 leu2-3,112 his3-200 gal4 gal80 LYS2::GAL1-HIS3 GAL2-ADE2 met2::GAL7-lacZ; obtained from E. Craig) was used for the two-hybrid assay (JAMES et al. 1996 ). The localization of Upf3p-HA was examined in strain PSY730 (MATa srp1-31 leu2-3,112 his3^- ade2^- trp1^- ura3-52; TABB et al. 2000 ; strain obtained from P. Silver). The effects of CRM1 on Upf3p localization were examined in strains MNY12 (MATa CRM1::Kan^r leu2^- his3^- trp1^- ura3^- [pDC-CRM1T539C-GFP]) and MNY8 (MATa CRM1::Kan^r leu2^- his3^- trp1^- ura3^- [pDC-CRM1T539C-HA]; NEVILLE and ROSBASH 1999 ). The CRM1T529C allele confers sensitivity to leptomycin B.

View larger version (68K): In this window In a new window Download PPT slide   Figure 1. Srp1p mediates the import of Upf3p into the nucleus. (A) Amino acid segments of Upf3p fused in-frame to the Gal4 activation or binding domain used in all two-hybrid experiments. (B) Strain PJ69-4A was transformed with plasmids, resulting in strains with the genotypes indicated to the left. Growth was monitored using serial drop tests by plating 100, 101, 102, and 103 serial dilutions (left to right) of log phase cultures (see MATERIALS AND METHODS) on SD without leucine and uracil (left), SD without histidine (middle), and SD without histidine with 30 mM 3-AT (right). (Left) The relative growth under conditions that select for the presence of the plasmids carrying the SRP1-BD and UPF3-AD alleles is shown. (Middle and right) The relative growth of the same strain under conditions in which the extent of growth is proportional to the strength of the interaction between the proteins is shown. (C) Subcellular localization of Upf3p-HA expressed from a 2µ plasmid in the temperature-sensitive srp1-31 strain PSY730. Cells carrying srp1-31 were grown at 25° and then shifted to 37°. Cells were removed at 0, 15, 30, 60, and 180 min after the shift to 37°. The detection of FITC staining in representative cells from each time point is shown at the top. (D) Subcellular localization of epitope-tagged Upf3p-HA expressed from centromeric and 2µ plasmids (left and middle) and Upf3p-Triple-HA expressed from a centromeric plasmid in strain LRSy323 (right) is shown. (C and D, bottom) DAPI staining marks the nucleus. Bar, 2.5 µm.

Plasmids are listed in Table 1. All plasmids were created using techniques and reagents as described previously (SHIRLEY et al. 1998 ). pAF8 was constructed by ligating the XhoI-SacI fragment from pUZ178 (ATKIN et al. 1997 ) containing UPF2 into the same sites in pRS423. Plasmids expressing green fluorescent protein (GFP) fusion constructs were made as follows. Oligomers (Operon, Alameda, CA) were designed to amplify the UPF3 promoter and 5'-untranslated region (UTR), using pLS17 as a template. The product was subcloned into a shuttle vector using unique restriction sites engineered by PCR. Next, sequences encoding the GFP open reading frame (ORF) without the termination codon were amplified from plasmid pRSETB (courtesy of L. Robinson). Restriction sites were engineered into the PCR product to allow insertion immediately downstream of the UPF3 promoter. Finally, the UPF3 ORF and sequences corresponding to the UPF3 3'UTR were amplified via PCR using pLS17 as a template. Using unique restriction sites, the PCR product was subcloned downstream of the GFP ORF. The entire cassette was cloned into pRS316 and pRS426, creating pNE36 and pNE39, respectively. The plasmid expressing Upf3p-Triple-GFP was constructed by recombinational cloning. pNE36 was digested with HindIII and BseRI. Sequences encoding the upf3-Triple ORF were PCR amplifed from template pRLS125 and transformed into strain RSy5 along with linearized pNE36. Recombined plasmids were rescued and sequenced. The upf3-Triple-GFP cassette was subcloned into pRS426, creating pAF14. To create pAF51, upf3-Triple-REV-GFP, the BseRI-KpnI fragment of pRLS141 was purified and subcloned into the same sites of pAF14. The construction of additional plasmids is described below or was described previously (SHIRLEY et al. 1998 ).

Two-hybrid assay:
The two-hybrid vectors used were pGBDU-C1,C2 and pGBDU-C1,C2 (Table 1; JAMES et al. 1996 ). Most of the translational fusions between ORFs of interest and the Gal4 activation domains (AD) or binding domains (BD) were generally constructed as follows. Primers were designed to amplify the entire ORF of interest from plasmid DNA containing the gene sequence. The primers were used to engineer unique restriction sites and the 5' and 3' ends of the PCR products. After amplification using high-fidelity polymerase, the PCR products were digested with the specific enzymes and ligated into similar sites in multiple cloning sites in the two-hybrid vectors.

To construct plasmids expressing Upf3p-Gal4 fusions, internal restriction sites were used to delete certain sequences in UPF3. pAF25 and pAF30, expressing Upf3p(1-98)-BD and Upf3p(1-98)-AD, respectively, were constructed by digesting pAF10 with BamHI and PstI. The resulting fragment was ligated into the same sites of pGBDU-C2 and pGAD-C2. Upf3p(1-220) was fused to the Gal4 binding domain by digesting pAF10 with BamHI and PvuII. The resulting fragment was ligated into plasmid pGBDU-C2, which was first digested with BglII, blunt ended using T4 DNA polymerase, and digested with PstI. This created plasmid pAF27. The plasmids expressing Upf3p(99-387)-BD and Upf3p(99-387)-AD, pAF26 and pAF31, respectively, were created by digesting plasmid pAF10 with PstI. The resulting fragment was ligated into similar sites in pGBDU-C2 and pGAD-C2. Upf3p(99-220) was fused to Gal4-AD by digesting pAF10 with PstI and PvuII. The PstI-PvuII DNA fragment was ligated into pGAD-C2, which was digested with BglII, blunt ended using T4 DNA polymerase, and digested with PstI to generate plasmid pAF33.

To construct plasmid pAF12, which expresses a translational fusion between Upf2p and Gal4-AD, an XhoI-BglII fragment from pAF6 was ligated into SalI and BglII sites in pGAD-C1. This fragment contains the sequence coding for UPF2 but lacks the intron. pAF12, which carries UPF2, was digested with XhoI and BglII and ligated into SalI and BglII sites in pGBDU-C1 to create pAF23.

Translational fusions between Crm1p/Xpo1p and the Gal4p-BD were constructed by digesting pKW442 (CRM1; STADE et al. 1997 ) with EcoRI and BamHI. This fragment was ligated into pGAD-C1 and pGBDU-C1 to generate plasmids pAF41 and pAF21, respectively. As a positive control for expression of the Crm1p fusion proteins, translational fusions between Rip1p, Gal4p-BD, and Gal4p-AD were used.

Plasmids expressing two-hybrid fusion proteins were transformed into strain PJ69-4A, which contains GAL1-HIS3 (JAMES et al. 1996 ). Expression of the HIS3 reporter indicates an interaction. The strength of the interaction was assessed in the presence of 3-aminotriazole (3-AT), which inhibits His3p enzymatic activity. The specificity of the interactions between all the fusion proteins was tested by assaying for an interaction between each fusion protein and a fusion between the Gal4p activation domain and amino acids 615–753 of Exo84p or a fusion between the Gal4p binding domain and amino acids 756–931 of Prp8p (KUHN and BROW 2000 ).

Cellular localization:
To study the effects of srp1-31 on localization, we examined Upf3p-HA in strain PSY730 carrying plasmid pLS73. Cells were prepared for immunofluorescence as described previously (SHIRLEY et al. 1998 ) except for the addition of 3.5% formalin for 5 min prior to fixing with formaldehyde. To inhibit the function of Crm1p in strains MNY8 and MNY12, cells grown to OD₆₀₀ = 0.5 in YEP were pelleted and resuspended in YEP plus 100 ng/ml of leptomycin B (provided by M. Yoshida; KUDO et al. 1999 ) and incubated at room temperature for 0.5–2 hr. After formaldehyde fixation, the cells were washed with PBS containing 1% Triton X-100 and 0.5 µg/ml 4',6-diamidino-2-phenylindole (DAPI) followed by a final wash with PBS.

Assays for function:
The function of the UPF genes was assayed by suppression of nonsense mutations and accumulation of CYH2 pre-mRNA. Nonsense suppression monitors all effects of UPF mutations simultaneously, including function in translation termination and NMD, whereas CYH2 pre-mRNA accumulation monitors the function of the UPF genes only in NMD.

Suppression was monitored in strains carrying the leu2-1 (UAA) and tyr7-1 (UAG) nonsense mutations, which prevent growth on medium lacking leucine and tyrosine. Loss-of-function mutations in any of the UPF genes confer growth of a leu2-1 tyr7-1 strain on medium lacking leucine and tyrosine. The can1-100 (UAA) nonsense mutation confers resistance to canavanine in a Upf⁺ strain, whereas mutations in any of the UPF genes suppress can1-100 and prevent growth in the presence of canavanine. Growth was assayed by plating serial dilutions of cells (SHIRLEY et al. 1998 ).

NMD was monitored by measuring the accumulation of CYH2 pre-mRNA using Northern blotting. CYH2 (ribosomal protein L29) contains an intron that is inefficiently spliced from the pre-mRNA (KAUFER et al. 1983 ). The presence of an in-frame stop codon at position 19 in the intron triggers rapid decay of unspliced pre-mRNA, which is exported to the cytoplasm, associates with polyribosomes, and is degraded by the NMD pathway (HE et al. 1993 ). The pre-mRNA accumulates to a four- to sixfold higher level in Nmd^- strains due to a corresponding increase in the pre-mRNA half-life but without any change in the mature mRNA half-life. The accumulation of CYH2 pre-mRNA has routinely been used to monitor the activity of the NMD pathway (for example, see ZHANG et al. 1997 ; ZUK and JACOBSON 1998 ).

An antisense RNA probe complementary to nucleotides 572–959 of CYH2 pre-mRNA was used to detect pre-CYH2 and mature mRNA (LELIVELT and CULBERTSON 1999 ). To assess the relative accumulation, the CYH2 pre-mRNA/mRNA ratio in the strain to be analyzed was normalized to the ratio in the UPF3 strain to calculate the fold change in the accumulation of the pre-mRNA. The fold changes were averaged across all trials and standard deviations (SD) were derived. To calculate P values, the statistical similarities of the average fold change in mRNA accumulation between pairwise sets of strains were assessed using a two-tailed t-test assuming equal variances at = 0.05.

PCR/oligonucleotide-directed mutagenesis and selection for mutations:
Multiround PCR was used to generate mutations in sequences coding for residues L88, I90, L92, and L93 of the Upf3p NES-A. Oligonucleotide RR03 [caaccgagaatgaaggatttaag (a,c,g)(a,c,t,g)(a,c) gtt (a,c,g)(a,c,t,g)(a,c) aga (a,c,g) (a,c,t,g)(a,c) (a,c,g)(a,c,t,g)(a,c) cctccaaatttgactgcagatg; mutated codons underlined] contains a degenerate sequence for the 12 bases that specify each of the 4 amino acids, allowing each of the three base codons to specify 15 different amino acids without resulting in a nonsense codon. In the first round of PCR, oligonucleotides RRO3 and LSO183 and template pLS17 generated DNA that contained mutations in NES-A and included a recognition site for SnaBI. The fragment generated in the first round of PCR was purified and used as a megaprimer in the second round. The megaprimer and oligonucleotide T7 (taatacgactcactataggg) and template pLS17 generated DNA that contained mutations in NES-A and included recognition sites for SnaBI and SacII. The third round of PCR amplified the DNA isolated from the second round. This fragment was digested with SacII and SnaBI and used to replace the SacII-SnaBI fragment from pRR2 that contained the wild-type NES-A element. Plasmid pRR2 contains a translational fusion between UPF3 and a loxP site from bacteriophage P1 (UPF3-loxP) that is separated from another loxP site fused to sequences coding for the wild-type HIV-1 Rev NES (loxP-Rev) by the ADE2 gene. The above ligation mixture was used to transform XL10-Gold ultracompetent cells (Stratagene, La Jolla, CA) and plated onto LB medium containing 100 µg/ml ampicillin. DNA was isolated from a pool of 8600 transformants.

A two-step strategy was devised to select for mutations that confer loss of Upf3p function followed by restoration of function after a functional NES was recombined en masse into a pool of PCR-mutagenized genes. In step 1, strain RSy5 was cotransformed with a pool of pRR2 plasmids containing mutations in NES-A (pRR2Mut) in UPF3-loxP and 2µ plasmid pRLS207, which carries the cre recombinase gene under the control of the GAL1 promoter. Expression of the cre recombinase promotes recombination between the loxP sites, resulting in a translational fusion between UPF3 and loxP and a functional HIV-1 Rev NES. Transformants were plated on SD medium without uracil, histidine, adenine, leucine, and tyrosine to select for the presence of the plasmids pRR2Mut and pRLS207 and for mutations that suppress leu2-1 and tyr7-1. Plates were incubated 4–5 days at 30°. In step 2, colonies were replica plated to SD-uracil, histidine-containing galactose to induce expression of the cre recombinase gene. Colonies were replica plated to SD-uracil, histidine-containing canavanine to select for canavanine resistance due to loss of suppression of the can1-100 nonsense allele. A total of 84,000 transformants were plated in step 1 and 389 survived step 2.

The survivors potentially include alleles in which the function of Upf3p is restored by the addition of the HIV Rev1-1 NES or they could include alleles that confer partial suppression where the addition of the NES makes no phenotypic difference. Upon further testing of a subset of alleles, we found that none of them exhibited improved function when the NES was added and all of them conferred partial suppression of nonsense mutations. DNA fragments from five transformants each containing the mutated NES-A region were generated by PCR using template DNA from yeast spheroplasts and oligonucleotides T7 and RSO62 gactgaggctgagaggagttg (KLEBANOW and WEIL 1999 ). The fragments were sequenced to characterize the mutations. Fragments were digested with SpeI and either SnaBI or BseRI and ligated into the same sites in pLS17 and pRLS134, replacing the wild-type NES and creating plasmids pRRNES2, pRRNES3, pRRNES5, pRRNES6, pRRNES7 and plasmids pRRNES2-Rev, pRRNES3-Rev, pRRNES5-Rev, pRRNES6-Rev, pRRNES7-Rev, respectively (Table 1).

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

The nuclear import of Upf3p is mediated by the importin-/importin-ß heterodimer:
The nuclear import of proteins containing either an SV-40-like nuclear localization signal (NLS) element (DINGWALL and LASKEY 1991 ) or a bipartite NLS element (ROBBINS et al. 1991 ) involves binding of the NLS-bearing protein to a heterodimer composed of importin- (Imp) and importin-ß (Impß; YANO et al. 1992 , YANO et al. 1994 ; ENENKEL et al. 1995 ; GORLICH et al. 1995 ). Imp binds to the NLS-containing proteins through recognition of the nuclear localization signal and serves as an adapter between the import cargo and Impß.

Three NLS motifs in Upf3p that were shown previously to direct reporters to the nucleus are referred to as NLS1, NLS2, and NLS3. They correspond to amino acids 15–31, 58–74, and 284–300, respectively (Fig 1A; LEE and CULBERTSON 1995 ; SHIRLEY et al. 1998 ). The presence of NLS motifs in Upf3p prompted us to test whether Imp/ß heterodimer mediates the import of Upf3p into the nucleus. To accomplish this, we determined whether Upf3p interacts with the Saccharomyces cerevisiae homolog of Imp, Srp1p, and whether the localization of Upf3p is altered in a temperature-sensitive srp1-31 strain.

The two-hybrid system was used to assess whether Srp1p interacts with Upf3p (Fig 1B). A 2µ plasmid carrying the SRP1-BD allele codes for a fusion protein that contains the Gal4p DNA binding domain and full-length Srp1p. This plasmid was cotransformed into yeast strain PJ69-4A (JAMES et al. 1996 ) with a 2µ plasmid expressing UPF3-AD, which codes for a fusion protein containing full-length Upf3p and the Gal4p activation domain. Coexpression of SRP1-BD and UPF3-AD resulted in robust growth on medium lacking histidine, indicating an interaction. The growth on selective medium was specific to transformants expressing the Upf3p and Srp1p fusion proteins. These results indicate that Upf3p and Srp1p interact in the two-hybrid system.

To characterize the regions of Upf3p that contribute to the interaction, we generated 2µ plasmids expressing Gal4p-AD fused in-frame to different fragments of Upf3p (Fig 1A). UPF3(1-98)-AD codes for a fusion protein that includes both NLS1 and NLS2. UPF3(99-387)-AD codes for a fusion protein that contains NLS3. UPF3(99-220)-AD codes for a fusion protein that does not contain NLS1, NLS2, or NLS3.

Transformants carrying SRP1-BD and UPF3(1-98)-AD grew more robustly on medium lacking histidine than did transformants expressing the full-length Upf3 protein (Fig 1B). To determine the strength of the interaction between Srp1p-BD and Upf3p(1-98)-AD, we tested growth of the transformants on media lacking histidine and containing varying concentrations of 3-AT. The expression of full-length UPF3-AD and SRP1-BD failed to promote growth on SD medium without histidine that contained 3 mM 3-AT. In contrast, transformants expressing SRP1-BD and UPF3(1-98)-AD grew robustly on SD medium without histidine with up to 30 mM 3-AT, indicating that the region of Upf3p containing NLS1 and NLS2 interacts strongly with Srp1p.

Coexpression of SRP1-BD and UPF3(99-387)-AD promoted growth on SD medium lacking histidine to the same extent as the transformants expressing SRP1-BD and full-length UPF3-AD (Fig 1B). Transformants carrying SRP1-BD and the UPF3(99-220)-AD fusion failed to grow on medium lacking histidine. Taken together, these results indicate that regions of Upf3p that include the NLS1 and NLS2 or NLS3 motifs contribute to the interaction of Upf3p with Srp1p. The internal region of Upf3p from residue 99 to 220 that does not contain an identifiable NLS motif does not interact.

To determine whether the nuclear import of Upf3p is mediated by the Imp/ß heterodimer, we examined the localization of epitope-tagged Upf3p (Upf3p-HA) in a strain carrying the srp1-31 temperature-sensitive allele (Fig 1C). Other studies show that the nuclear import of substrates in the srp1-31 strain diminishes with time following a shift to 37° with 95% of protein import blocked after 6 hr at the nonpermissive temperature (TABB et al. 2000 ). A 2µ plasmid expressing UPF3-HA was transformed into strain PSY730 (srp1-31). The localization of Upf3p-HA was determined by indirect immunofluorescence microscopy at several time points following a shift to 37° and compared to DAPI staining. As controls (Fig 1D), we assessed the localization of wild-type Upf3p-HA expressed from centromeric and 2µ plasmids and a mutant version of Upf3p-HA expressed from a centromeric plasmid carrying a defective nuclear export sequence (Upf3p-Triple-HA). Overexpressed Upf3p-HA and export-defective Upf3p-HA expressed at a normal level accumulate in the nucleolus, whereas wild-type Upf3p-HA expressed from a centromeric plasmid accumulates only in the cytoplasm (SHIRLEY et al. 1998 ).

At the permissive temperature, Upf3p-HA accumulates in a distinct area of the nucleus adjacent to the area stained by DAPI as well as throughout the cytoplasm (Fig 1C, RT). The area of staining within the nucleus corresponds to the nucleolus (SHIRLEY et al. 1998 ) and serves as an indicator of import into the nucleus. After 15 min following the shift of the srp1-31 strain to 37°, the localization of Upf3p-HA was indistinguishable from its localization at room temperature. However, after 30 min Upf3p-HA was not detected in the nucleolus and was visible only in the cytoplasm. We also examined the distribution of Upf3p-HA expressed from a 2µ plasmid in an SRP1 strain. Even after 180 min following the shift to 37°, Upf3p-HA localized to the nucleolus and cytoplasm in a pattern indistinguishable from its localization at room temperature. The gradual decrease in the nucleolar accumulation of Upf3p-HA at the nonpermissive temperature in the srp1-31 strain indicates that Srp1p is necessary for the import of Upf3p into the nucleus. These results indicate that nuclear import of Upf3p is mediated by the Imp/ß heterodimer.

Export of Upf3p via a leucine-rich nuclear export sequence is required for a fully functional NMD pathway. We showed previously that the export of Upf3p from the nucleus requires a leucine-rich signal sequence called NES-A (Fig 1A; SHIRLEY et al. 1998 ). The substitution of alanine for two leucine and one isoleucine residue in NES-A (upf3-Triple-HA) causes a redistribution of the protein from the cytoplasm to the nucleolus (Fig 1D). This allele impairs the function of Upf3p-Triple as indicated by allosuppression of the NMD-sensitive his4-38 frameshift mutation and by increased accumulation of nonsense mRNAs. The insertion of the wild-type NES from HIV-1 Rev at the C terminus of Upf3p-Triple-HA (Upf3p-Triple-HA-Rev) restores the export of Upf3p-Triple-HA-Rev from the nucleus; however, we found that strains carrying the upf3-Triple-HA-Rev allele were phenotypically indistinguishable from a upf3-Triple-HA strain when assayed by the his4-38/SUF1-1 allosuppression assay (SHIRLEY et al. 1998 ). These initial results precluded the demonstration that retaining Upf3p in the nucleus directly affects the function of Upf3p.

Since allosuppression of the his4-38 frameshift mutations is complex and requires the presence of a tRNA frameshift suppressor, we reasoned that direct suppression of a nonsense mutation is likely to be a more sensitive indicator of perturbations in the function of Upf3p. We used nonsense suppression assays to reexamine the function of upf3-Triple-HA and upf3-Triple-HA-Rev in strains carrying leu2-1 (UAA), tyr7-1 (UAG), and can1-100 (UAA). In the presence of wild-type UPF genes, strains carrying these nonsense mutations fail to grow on SD medium without leucine and tyrosine but are able to grow when canavanine is added. Mutations that impair the function of the Upf proteins confer growth on SD medium without leucine and tyrosine and prevent growth on SD medium plus leucine, tyrosine, and canavanine.

Strain RSy5 (leu2-1 tyr7-1 can1-100 upf3-1) was transformed with a centromeric plasmid containing two tandem copies of upf3-Triple-HA (pRLS125). The presence of two gene copies compensates for the approximately twofold underexpression of protein expressed from upf3-Triple (SHIRLEY et al. 1998 ). To create the upf3-1, UPF3, and UPF3-HA strains, RSy5 was transformed with an empty vector (pRS316) and with centromeric plasmids containing UPF3 (pLS17) and UPF3-HA (pLS51).

The upf3-1 transformant grew on SD medium without leucine and tyrosine but not in the presence of canavanine (Fig 2). The transformants expressing wild-type UPF3 or epitope-tagged UPF3-HA failed to grow on SD medium without leucine and tyrosine but grew on canavanine-containing medium. Transformants carrying upf3-Triple-HA grew on SD medium without leucine and tyrosine and on the canavanine-containing medium but not as robustly as either the upf3-1 strain or the UPF3 strain, respectively. These results indicate that the level of suppression in the upf3-Triple-HA strain is reduced compared to that of the upf3-1 strain. Reduced suppression of leu2-1, tyr7-1, and can1-100 indicates that the three alanine substitutions in NES-A cause decreased function of the Upf3p-Triple protein.

View larger version (47K): In this window In a new window Download PPT slide   Figure 2. Effect of the HIV-1 Rev nuclear export sequence on the function of Upf3p-Triple-HA. Suppression of leu2-1, tyr7-1, and can1-100 was used as an indicator of Upf3p function in which growth was monitored on SD-leucine, tyrosine, and SD medium containing canavanine, using serial drop tests (Fig 1; MATERIALS AND METHODS). Strain RSy5 (leu2-1 tyr7-1 can1-100 upf3-1) was transformed with centromeric plasmids, resulting in strains with the genotypes indicated at the left. Relative growth rates were compared as described in Fig 1 on SD-uracil (left), SD-leucine, tyrosine (middle), and SD medium containing canavanine (right). (Left) Relative growth under conditions that select for the presence of the URA3 plasmids carrying the upf3 alleles is shown. (Middle) Relative growth of the same strains under conditions in which the extent of growth is proportional to the extent of impairment of Upf3p function is shown. (Right) Relative growth in the presence of canavanine in which growth is inversely proportional to the extent of impairment of Upf3p function is shown.

The above transformants were also assayed for the activity of the NMD pathway by determining the relative levels of accumulation of CYH2 pre-mRNA (HE et al. 1993 ; Fig 3; Table 2). The amount of CYH2 pre-mRNA was 5.29 ± 0.57-fold more abundant in a upf3-1 strain than in a transformant carrying wild-type UPF3. The relative accumulation in transformants carrying upf3-Triple-HA is 3.44 ± 0.30. This accumulation was significantly less than that observed in a upf3-1 strain (P < 0.005), indicating that the alanine substitutions in NES-A seriously impair but do not completely abolish the function of Upf3p.

View larger version (40K): In this window In a new window Download PPT slide   Figure 3. Effect of the wild-type and M10 Rev NES on the function of Upf3p-Triple-HA in NMD. Northern blotting was used to assess the effect of inserting sequences coding for the wild-type and M10 Rev NES in upf3-Triple-HA on the accumulation of CYH2 pre-mRNA. Strain RSy5 was transformed with centromeric plasmids expressing the genes indicated at the top of each lane. The plasmids expressing the upf3-Triple alleles indicated in the figure contained duplicate copies of each gene. The representative hybridization signal specific to the precursor and mature forms of CYH2 RNA is shown. The relative accumulation of CYH2 pre-mRNA expressed as the fold increase and the associated SD was calculated on the basis of five experiments. The calculation of relative accumulation and the statistical treatment of the data are described in MATERIALS AND METHODS.

The Upf3p-Triple-HA protein exhibits impaired nuclear export, and as a result it redistributes from the cytoplasm to the nucleolus (Fig 1D). To further examine the export of this protein, we transformed strain RSy5 with a plasmid containing either two tandem copies of upf3-Triple-HA-Rev (pRLS145) or tandem copies of upf3-Triple-HA-M10 (pRLS144). The upf3-Triple-HA-Rev and upf3-Triple-HA-M10 alleles code for Upf3-Triple-HA that contains either the wild-type Rev NES or the export-defective M10 Rev NES at the C terminus (MEYER and MALIM 1994 ; FISCHER et al. 1995 ). Consistent with previous results (SHIRLEY et al. 1998 ), we found that the wild-type Rev NES but not the export-deficient M10 NES restores the nuclear export of Upf3p-Triple-HA (Fig 1D).

To examine the effect that restoration of nuclear export has on the function of the mutant Upf3-Triple-HA-Rev protein, we compared the level of nonsense suppression in transformants expressing upf3-Triple-HA-Rev with those expressing upf3-Triple-HA. The growth of the upf3-Triple-HA-Rev strain on SD medium without leucine and tyrosine was less than that of the upf3-Triple-HA strain (Fig 2). When canavanine was present, the upf3-Triple-HA-Rev strain grew more robustly than the upf3-Triple-HA strain (Fig 2). These results demonstrate that the level of suppression of the leu2-1, tyr7-1, and can1-100 nonsense mutations is lower in the transformants expressing upf3-Triple-HA-Rev than in those expressing upf3-Triple-HA. The growth of transformants carrying upf3-Triple-HA-M10 was indistinguishable from the growth of a upf3-Triple-HA strain on both types of selective media (Fig 2). These data indicate that only the export-competent Rev NES element improves the function of the Upf3-Triple protein.

A comparison of the relative levels of CYH2 pre-mRNA accumulation in transformants carrying upf3-Triple-HA-Rev (2.98 ± 0.29) and upf3-Triple-HA (3.44 ± 0.30) indicates that the addition of the exogenous Rev NES significantly decreases the accumulation of CYH2 pre-mRNA (P < 0.05; Fig 3; Table 2). This decrease was specific to the wild-type Rev NES since the relative level of CYH2 pre-mRNA accumulation in the upf3-Triple-HA-M10 strain (3.56 ± 0.29) was statistically indistinguishable from the upf3-Triple-HA strain (P > 0.90). These results indicate that rescuing the export of Upf3p-Triple-HA by insertion of the Rev NES improves the ability of Upf3p to function in NMD. We conclude, therefore, that impairing the export of Upf3p causes an impairment of NMD. This indicates that the export of Upf3p is required for a fully functional NMD pathway.

Nonconservative amino acid substitutions in NES-A impair Upf3p function more severely than alanine substitutions:
We isolated five new upf3-nes alleles containing less conservative amino acid changes in NES-A than the alanine substitutions encoded by the upf3-Triple allele. Transformants carrying the mutant upf3-nes alleles were assayed for suppression of the leu2-1, tyr7-1, and can1-100 nonsense mutations and for the accumulation of CYH2 pre-mRNA. The cellular distribution of two of the mutant proteins was also examined.

The growth of transformants carrying centromeric plasmids expressing the upf3-nes alleles on SD medium without leucine and tyrosine and the canavanine-containing medium was indistinguishable from the growth of a strain expressing upf3-Triple-HA (Table 3). Similarly, the accumulation of CYH2 pre-mRNA in the five upf3-nes strains was similar to that found in transformants carrying upf3-Triple-HA. These results indicate that the amino acid substitutions in the Upf3-nes proteins cause impaired function. The staining patterns of epitope-tagged Upf3p-nes2-HA and Upf3p-nes7-HA were indistinguishable from the nucleolar distribution of Upf3p-Triple-HA, indicating that export of the mutant proteins is impaired (data not shown).

To determine if the wild-type Rev NES can restore function to these proteins, the sequence coding for the Rev NES was inserted in the upf3-nes2, -nes3, -nes5, -nes6, and -nes7 alleles. Transformants carrying centromeric plasmids expressing these alleles were assayed for suppression of the leu2-1, tyr7-1, and can1-100 nonsense mutations and for the accumulation of CYH2 pre-mRNA. The growth of transformants carrying the upf3-nes-Rev alleles on SD medium without leucine and tyrosine and on the canavanine-containing medium resembled the growth of strains containing the identical mutations but without the insertion of HIV-1 Rev. Similarly, the insertion of the HIV-1 Rev NES did not affect the accumulation of CYH2 pre-mRNA (Table 4). These results indicate that, unlike Upf3p-Triple-HA, more drastic amino acid changes in NES-A impair Upf3p function in a manner that is not corrected by HIV-1 Rev-directed restoration of nuclear export.

Overexpression of UPF2 partially restores impaired function of mutant Upf3p:
To further examine mutations in NES-A, we tested whether overexpression of UPF1 or UPF2 modifies the phenotypes of the NES-A mutations. Overexpression of UPF1 in a strain carrying upf3-Triple-HA had no phenotypic effect as measured by the level of nonsense suppression (data not shown). However, coexpression of the upf3-Triple-HA allele with a 2µ plasmid carrying UPF2 led to observable changes in the suppression of leu2-1, tyr7-1, and can1-100 and in the accumulation of CYH2 pre-mRNA (Fig 4; Table 4).

View larger version (86K): In this window In a new window Download PPT slide   Figure 4. Effects of Upf2p overexpression in transformants expressing upf3-Triple-HA and upf3-Triple-HA-Rev. (A) Suppression of leu2-1, tyr7-1, and can1-100 was assayed by serial drop tests as described in Fig 1 and MATERIALS AND METHODS. Strain RSy5 was transformed with sets of plasmids, resulting in strains with the genotypes indicated to the left. (B) Northern blotting was used to assess the accumulation of CYH2 pre-RNA. Strain RSy5 was transformed with sets of plasmids, resulting in strains with the genotypes indicated at the top. The representative hybridization signal specific to the precursor and mature forms of CYH2 RNA is shown. The relative accumulation of CYH2 pre-mRNA expressed as the fold increase and the associated SD was calculated on the basis of six experiments. The calculation of relative accumulation and the statistical treatment of the data are described in MATERIALS AND METHODS.

Transformants carrying both upf3-Triple-HA and 2µ UPF2 grow slightly less than the upf3-Triple-HA strain on SD medium without tyrosine and leucine (Fig 4A). The same transformants grew better than the upf3-Triple-HA strain on the canavanine-containing medium. These results suggest that increasing the expression of UPF2 in a upf3-Triple-HA strain decreases the extent of nonsense suppression, which indicates improved function. The effect caused by the overexpression of UPF2 was specific to the upf3-Triple-HA allele because overexpression of UPF2 had no effect on nonsense suppression in a strain carrying a upf3-1 disruption (data not shown). The decrease in nonsense suppression in a upf3-Triple-HA strain caused by the overexpression of UPF2 was approximately the same as the decrease observed upon addition of the Rev NES element to Upf3p-Triple-HA.

The strains were assayed for the effect of overexpressing UPF2 on NMD (Fig 4B). The relative accumulation of CYH2 pre-mRNA was 3.87 ± 0.52 in a transformant carrying upf3-Triple-HA. Overexpression of UPF2 in a upf3-Triple-HA strain caused a significant decrease in the relative accumulation of CYH2 pre-mRNA to 3.03 ± 0.43 (P < 0.05; Fig 4; Table 5). By comparison, the relative accumulation of CYH2 pre-mRNA in the upf3-Triple-HA strain carrying overexpressed UPF2 was not significantly different from the 3.12 ± 0.39-fold accumulation observed in a upf3-Triple-HA-Rev strain (P > 0.5). The results indicate that the overexpression of UPF2 suppresses the Nmd^- phenotype in a upf3-Triple-HA strain by the same magnitude as the addition of Rev NES to Upf3p-Triple-HA.

When UPF2 is overexpressed, nonsense suppression was decreased in strains carrying upf3-nes2, upf3-nes3, upf3-nes5, upf3-nes6, and upf3-nes7 (Table 4). Overexpression of UPF2 caused a significant decrease in the relative accumulation of CYH2 pre-mRNA in transformants carrying upf3-nes5, upf3-nes6, and upf3-nes7. The overexpression did not cause a statistically significant change in the accumulation of CYH2 pre-mRNA in the transformants carrying upf3-nes2. Although the insertion of the Rev NES has no effect on the function of the Upf3-nes5, Upf3-nes6, and Upf3-nes7 proteins in NMD, the overexpression of UPF2 partially suppressed the Nmd^- phenotype in strains carrying these upf3-nes alleles. Taken together, these data demonstrate that the effects of the NES-A substitutions on the function of the mutant Upf3 proteins can be partially suppressed by overexpression of UPF2.

We asked whether simultaneously adding the Rev NES and overexpressing UPF2 fully restores function to Upf3p-Triple. To accomplish this, centromeric plasmids expressing upf3-Triple-HA-M10 and upf3-Triple-HA-Rev were separately cotransformed with the 2µ UPF2 plasmid into strain RSy5. The transformants were assayed for nonsense suppression and for the accumulation of CYH2 pre-mRNA (Fig 4; Table 5).

Overexpression of UPF2 caused reduced suppression of nonsense mutations to the same extent in strains carrying either upf3-Triple-HA-M10 or upf3-Triple-HA (Fig 4A). In a strain carrying upf3-Triple-HA-Rev, overexpression of UPF2 caused a slight reduction in suppression of nonsense mutations compared to a strain carrying upf3-Triple-HA-Rev without 2µ UPF2 and a strain carrying upf3-Triple-HA with 2µ UPF2. These results show that the insertion of the Rev NES in combination with the overexpression of UPF2 causes only slightly more nonsense suppression than that which occurs due to the Rev NES or UPF2 overexpression alone.

Transformants carrying upf3-Triple-HA-M10 and 2µ UPF2 displayed a relative accumulation of CYH2 pre-mRNA of 3.13 ± 0.52 (Fig 4B). By comparison, the accumulation level was 3.93 ± 0.77 in the absence of overexpressed UPF2. We expected these values to be different because the M10 NES is nonfunctional, but in this experiment the difference was of borderline statistical significance (P > 0.05; Table 5). Transformants carrying upf3-Triple-HA-Rev and 2µ UPF2 displayed a relative accumulation of 2.82 ± 0.41. The accumulation level was 3.12 ± 0.39 in the absence of 2µ UPF2. These values are not significantly different. Similarly, the accumulation of CYH2 pre-mRNA in transformants carrying upf3-Triple-HA and 2µ UPF2 (3.03 ± 0.43) was not significantly different from the upf3-Triple-HA-Rev strain carrying 2µ UPF2. Taken together, these data demonstrate that while insertion of the Rev NES and overexpression of UPF2 separately improve the function of Upf3p-Triple, insertion of the Rev NES in combination with overexpression of UPF2 does not further restore function. The effects are nonadditive.

We explored the possibility that improving the interaction between Upf2p and Upf3p-Triple might increase the efficiency of export of Upf3p-Triple and therefore improve the function of the mutant Upf3 protein. If Rev NES and Upf2p have synonymous or related functions, this hypothesis would account for the lack of additivity when Rev NES and Upf2p are combined. We examined the distribution of Upf3p-Triple-HA using indirect immunofluorescence in a strain overexpressing UPF2, but saw no change in the distribution of the mutant protein. Upf3p-Triple-HA was still trapped in the nucleolus (data not shown). Furthermore, the distribution of Upf3p-HA did not change in upf2-1 cells (data not shown). These results suggest that overexpression of Upf2p improves the function of the Upf3p-Triple by a mechanism unrelated to the ability of the mutant protein to export from the nucleus to the cytoplasm.

Effects of nesA^- alleles on the physical interaction between Upf3p and Upf2p:
Amino acids 78–278 of Upf3p, which include the NES-A element, are necessary for the interaction between Upf3p and Upf2p (HE et al. 1997 ). Since one way to compensate for an altered interaction between two proteins is by overexpressing one of the interacting components (PHIZICKY and FIELDS 1995 ), we speculated that overexpression of UPF2 might improve an impaired interaction between Upf2p and Upf3p-Triple.

The Gal4p DNA binding domain was fused in-frame to full-length, wild-type Upf3p and the Gal4p DNA activation domain was fused in-frame with Upf2p. The plasmids were separately transformed into strain PJ69-4A. Coexpression of UPF3-BD and UPF2-AD resulted in robust growth on SD medium without histidine containing up to 5 mM 3-AT (Fig 5). Next we made plasmids carrying genes coding for Upf3p-Triple or Upf3p-nes fused to the Gal4p binding domain. Coexpression of upf3-Triple-BD and UPF2-AD failed to confer growth on selective medium. In addition, coexpression of the five upf3-nes-BD alleles with UPF2-AD also failed to promote growth on selective medium (data not shown). RSy5 expressing upf3-Triple-BD suppresses the leu2-1, tyr7-1, and can1-100 alleles to the same extent as the upf3-Triple-HA allele (data not shown). The lack of interaction between Upf3-Triple-BD and Upf2p is therefore not due to gross misfolding and/or severe instability of the fusion protein. These results suggest that amino acid substitutions in the NES-A region of Upf3p diminish the interaction between Upf3p-Triple-BD and Upf2p-AD.

View larger version (51K): In this window In a new window Download PPT slide   Figure 5. Contribution of the NES-A element of Upf3p to the Upf3p/Upf2p interaction. Strain PJ69-4A was transformed with the plasmids expressing the genes indicated on the left. Growth was monitored by serial drop tests (Fig 1; MATERIALS AND METHODS). The interaction between the Gal4 activation domain and Gal4 binding domain fusion proteins was monitored by growth on SD-histidine (middle) and SD-histidine with 5 mM 3-AT (right).

We also examined two-hybrid interactions using fusions in which the Gal4p activation and DNA binding domains were reversed. Similar to the results described above, coexpression of UPF3-AD and UPF2-BD conferred growth on selective medium (Fig 5). However, a stronger interaction was detected since growth was observed on SD-histidine medium in the presence of up to 25 mM 3-AT (data not shown). In contrast to the opposite orientation of the fusions, coexpression of upf3-Triple-AD and UPF2-BD conferred growth in the presence of 5 mM 3-AT (Fig 5) and 25 mM 3-AT (data not shown). Increasing the concentration of 3-AT failed to result in detectable differences between transformants expressing UPF3-AD or upf3-Triple-AD with UPF2-BD. Similar results were observed when the five Upf3-nes-AD proteins were tested. Upf3p-Triple-AD suppresses the tyr7-1 and can1-100 nonsense alleles to the same extent as Upf3p-Triple, indicating that the fusion does not have a gain-of-function mutation allowing for an interaction between Upf2p-BD and Upf3p-Triple-AD (data not shown). More likely, the difference between the two orientations is due to the weaker interaction observed between the Upf3p-BD and Upf2p-AD fusion proteins. The weaker interaction in this orientation may provide for a more sensitive assay that allows detection of smaller changes in the binding affinities between Upf3p-Triple and Upf2p.

To clarify the situation, we determined the extent to which the NES-A region contributes to the interaction with Upf2p. We made plasmids expressing fusions between the Gal4p binding domain and amino acids 1–98 or amino acids 99–387 of Upf3p (see Fig 1A). Upf3p(1-98) contains only the first 21 N-terminal amino acids, including those of NES-A, which were previously identified as part of the Upf2p interacting domain. Upf3p(99-387) contains the majority of the fragment that was previously identified as interacting with Upf2p except for the NES domain (HE et al. 1997 ). Plasmids expressing UPF3(1-98)-BD and UPF3(99-387)-BD were separately cotransformed with a 2µ plasmid expressing UPF2-AD into strain PJ69-4A. Coexpression of UPF3(1-98)BD and UPF2-AD did not confer growth on selective media, indicating that the NES-A region alone does not support an interaction between Upf3p and Upf2p (Fig 5). Coexpression of UPF3(99-387)-BD and UPF2-AD promoted growth but to a lesser extent than transformants expressing full-length UPF3-BD and UPF2-AD.

We reversed the fusion constructs and found that transformants expressing UPF3(1-98)-AD and UPF2-BD again did not support growth on selective media. The strain expressing UPF3(99-387)-AD and UPF2-BD grew less robustly in the presence of 5 mM 3-AT than did transformants expressing UPF3-AD and UPF2-BD (Fig 5). Therefore, consistent with our results using fusions to the Gal4p binding domain, these data indicate that NES-A alone is not sufficient for an interaction with Upf2p, but removing the NES-A region from Upf3p weakens the interaction. The NES-A region is necessary but not sufficient for an interaction.

The export of Upf3p does not require the function of Crm1p/Xpo1p exportin:
Crm1p/XpoIp is a nuclear export receptor that binds to leucine-rich NES sequences (FORNEROD et al. 1997 ; STADE et al. 1997 ; IZAURRALDE and ADAM 1998 ; HO et al. 2000 ). We asked whether Upf3p physically interacts with Crm1p and whether Upf3p export depends on the function of this exportin. Using the two-hybrid system, we confirmed the known interaction between Crm1p-AD and Rip1p-BD (NEVILLE et al. 1997 ; HO et al. 2000 ), but were unable to detect an interaction between Crm1p and Upf3p (data not shown). To determine whether Crm1p mediates the nuclear export of Upf3p, we localized Upf3p-HA in a strain carrying a temperature-sensitive xpo1-1 mutation. In this strain, the export of proteins by Crm1p ceases 5–15 min following a shift to the nonpermissive temperature (STADE et al. 1997 ). We were unable to detect nucleolar accumulation of Upf3p-HA expressed from a centromeric plasmid at the nonpermissive temperature (data not shown).

Proteins up to 60 kD have the potential to passively diffuse through the yeast nuclear pore complex (NPC; BORER et al. 1989 ; PANTE and AEBI 1995 ). Since Upf3p is a 44.9-kD protein (LEE and CULBERTSON 1995 ), it could potentially cross the NPC by passive diffusion. To test the passive diffusion model, we increased the size of Upf3p by fusing GFP to the Upf3p N terminus generating an 72-kD fusion protein. The distribution of Upf3p-GFP expressed from a 2µ plasmid in the strain carrying the xpo1-1 mutation was examined. Consistent with previous results, the distribution of Upf3p-GFP did not change upon shift to the nonpermissive temperature (data not shown). We also examined the distribution of Upf3p-GFP expressed from a 2µ plasmid in three different strains containing viable mutant alleles of crm1: crm1-1, crm1-2, and crm1-3 (NEVILLE et al. 1997 ). However, we did not observe an increase in the nucleolar accumulation or a decrease in cytoplasmic staining of Upf3p-GFP in the crm1 strains (E. NEENO-ECKWALL and M. CULBERTSON, unpublished data).

Leptomycin B (LMB), a potent inhibitor of Crm1-mediated transport, binds to mammalian and S. pombe Crm1p (FORNEROD et al. 1997 ; ASKJAER et al. 1998 ). Although S. cerevisiae Crm1p is insensitive to LMB, a T539C substitution renders Crm1p sensitive to LMB and inhibits Crm1-mediated export within 15 min after exposure to the drug (NEVILLE and ROSBASH 1999 ). We examined the localization of Upf3p-GFP before and after addition of the drug in strain MNY8, which carries the LMB-sensitive allele CRM1T539C-HA. The distribution of Upf3p-GFP expressed from a 2µ plasmid was compared to DAPI staining (Fig 6). Upf3p-GFP remained primarily cytoplasmic for 15 min to 2 hr following the addition of LMB to the medium, suggesting that the export of Upf3p is not dependent on Crm1p. The localization of Upf3p-HA was also examined using indirect immunolocalization in strain MNY12, which carries the LMB-sensitive allele CRM1T539C-GFP. Upf3p-HA remained primarily cytoplasmic even after 2 hr of incubation in the presence of LMB.

View larger version (39K): In this window In a new window Download PPT slide   Figure 6. Loss of Crm1p/Xpo1p function does not affect the export of Upf3p. The leptomycin-sensitive strain MNY8 was transformed with 2µ plasmids pNE39, pAF14, and pAF51 expressing UPF3-GFP, upf3-Triple-GFP, and upf3-Triple-Rev-GFP, respectively. Cells were incubated without and with leptomycin B (left and right, respectively; see MATERIALS AND METHODS). GFP fluorescence was compared with DAPI staining as indicated. Bar, 2.5 µm.

Since the Rev NES is known to direct nuclear export through interaction with Crm1p (FORNEROD et al. 1997 ; STADE et al. 1997 ; IZAURRALDE and ADAM 1998 ), we compared the distribution of Upf3p-Triple-GFP, which exports independently of Crm1p, with Upf3p-Triple-Rev-GFP. Leptomycin-sensitive cells of strain MNY8 expressing upf3-Triple-GFP from a 2µ plasmid were incubated in the presence of LMB for 30 min to 2 hr prior to fixation with formaldehyde and visualization. As expected, Upf3p-Triple-GFP localized to a nuclear region that most likely corresponds to the nucleolus, and its localization did not change even after 2 hr of exposure to LMB (Fig 6). However, Upf3p-Triple-Rev-GFP, which is primarily cytoplasmic, redistributed to the nucleolus as early as 30 min after incubation with LMB. After 60 min, a majority of the cells showed relocalization of the protein to the nucleolus. These results indicate that wild-type Upf3p is not a substrate for export via Crm1p.

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Nuclear import of Upf3p:
We showed previously that Upf3p is found primarily in the cytoplasm in wild-type cells (SHIRLEY et al. 1998 ). However, elevated expression of the UPF3 gene results in accumulation of Upf3p in the nucleolus without causing suppression of nonsense mutations or impairment of NMD. The export-defective protein product of the upf3-Triple allele used in this study accumulates in the nucleolus with concomitant suppression of nonsense mutations and impairment of NMD. The product of upf3-Triple is 50% less abundant than the wild-type protein (SHIRLEY et al. 1998 ). We used plasmids that carry two copies of upf3-Triple because this was shown to result in a protein level that approximates the level of the wild-type protein. We therefore believe that our results showing a role for the export of Upf3p in NMD are physiologically meaningful and are not simply the consequence of overexpression. It is not currently known why Upf3p concentrates in the nucleolus vs. the nucleoplasm. Since ribosomal subunits assemble in the nucleolus, there may be ligands in the nucleolus to which Upf3p can bind when excess protein is present in the nucleus. Although we do not attach undue significance to accumulation in the nucleolus vs. the nucleoplasm, we view the nucleolar phenotype as a valid cytological indicator of perturbations in the nuclear/cytoplasmic shuttling of Upf3p.

Upf3p contains three sequence elements that resemble a classical bipartite NLS (LEE and CULBERTSON 1995 ; SHIRLEY et al. 1998 ). NLS1 and NLS2 direct a reporter to the nucleoplasm, whereas NLS3 directs the same reporter to the nucleolus. Our results show that the importin-/ß heterodimer, which mediates the import of many NLS-containing proteins (GORLICH et al. 1994 , GORLICH et al. 1995 ; WEIS et al. 1995 ), is required for the nuclear import of Upf3p. The nucleolar distribution of overexpressed Upf3p diminishes with time after shift of a strain carrying srp1-31 (Imp) to a restrictive temperature. Srp1p and Upf3p physically interact in the two-hybrid system. The strength of the interaction is comparable to the strength of the Upf3p/Upf2p interaction. Presumably Srp1p binds to the NLSs since only fragments of Upf3p that include at least one of the NLS motifs interact with Srp1p. Assuming that Upf3p imports into the nucleus by a single mechanism regardless of expression level, our results suggest that Upf3p is actively imported by a mechanism in which Srp1p(Imp) binds to Upf3p and interacts with the transport receptor Kap95(Impß) to mediate the import of the complex into the nucleus.

Nuclear export of Upf3p is required for NMD:
A leucine-rich nuclear export sequence (NES-A) mediates the export of Upf3p from the nucleus to the cytoplasm. NES-A is a functional nuclear export signal based on the observation that an allele containing three alanine-for-leucine/isoleucine substitutions in NES-A (upf3-Triple) alters the distribution of Upf3p from a primarily cytoplasmic localization to a nucleolar localization (SHIRLEY et al. 1998 ). In addition to impairing export, mutations in NES-A impair the function of Upf3p leading to suppression of nonsense mutations and changes in the accumulation of nonsense mRNAs.

In our previous studies (SHIRLEY et al. 1998 ), we were unable to make a direct connection between the export of Upf3p and its role in NMD. It was therefore not clear at that time whether the movement of Upf3p into and out of the nucleus is required for NMD or whether it serves some purpose unrelated to NMD. In this article, we show that the insertion of a heterologous NES from HIV-1 Rev to Upf3p-Triple restores the ability of the mutant protein to export from the nucleus and this results in partial restoration of the function of Upf3p in the NMD pathway. By comparison with wild-type and null upf3 strains, the partially reduced levels of suppression of the leu2-1, tyr7-1, and can1-100 nonsense mutations correlate with intermediate levels of CYH2 pre-mRNA. These results suggest that the export of Upf3p from the nucleus is necessary for Upf3p to promote NMD and associated processes such as termination of translation at premature stop codons. Several possible explanations for the lack of full restoration of NMD by exportable mutant versions of Upf3p are discussed further below.

An alternative export pathway for Upf3p:
The exportin Crm1p/Xpo1p does not mediate the export of Upf3p despite the resemblance between NES-A and the leucine-rich sequences known to serve as binding sites for Crm1p (FORNEROD et al. 1997 ; STADE et al. 1997 ). Upf3p does not interact with Crm1p in two-hybrid assays and fails to be retained in the nucleus in cells where Crm1p has been inactivated by the presence of mutations in the CRM1 gene. Leptomycin B, which blocks the function of a leptomycin-sensitive Crm1 protein, fails to disrupt the export of wild-type Upf3p but does prevent the export of a version of Upf3p carrying an HIV-1 Rev NES sequence known to bind to Crm1p. Additional results indicate that Upf3p does not passively diffuse through the NPC.

We suggest that Upf3p is actively exported to the cytoplasm by an alternative Crm1p-independent mechanism. We define NES-A as a nuclear export sequence because mutations in the sequence impair export. However, the superficial resemblance of this NES to the canonical NES exemplified by that found in HIV-1 Rev may be simply fortuitous. Further studi