The Upf3 protein is a component of the surveillance complex that monitors both translation and mRNA turnover and affects viral propagation

Abstract

The nonsense-mediated mRNA decay pathway functions to degrade aberrant mRNAs that contain premature translation termination codons. In Saccharomyces cerevisiae, the Upf1, Upf2, and Upf3 proteins have been identified as trans-acting factors involved in this pathway. Recent results have demonstrated that the Upf proteins may also be involved in maintaining the fidelity of several aspects of the translation process. Certain mutations in the UPF1 gene have been shown to affect the efficiency of translation termination at nonsense codons and/or the process of programmed −1 ribosomal frameshifting used by viruses to control their gene expression. Alteration of programmed frameshift efficiencies can affect virus assembly leading to reduced viral titers or elimination of the virus. Here we present evidence that the Upf3 protein also functions to regulate programmed −1 frameshift efficiency. A upf3-Δ strain demonstrates increased sensitivity to the antibiotic paromomycin and increased programmed −1 ribosomal frameshift efficiency resulting in loss of the M₁ virus. Based on these observations, we hypothesize that the Upf proteins are part of a surveillance complex that functions to monitor translational fidelity and mRNA turnover.

Numerous studies have demonstrated that the processes of mRNA turnover and translation are directly linked (1–4). One example that couples these two processes is the fact that premature translation termination promotes rapid turnover of the transcript (1, 2, 4–6). This process has been termed nonsense-mediated mRNA decay (NMD). The NMD pathway has been observed in all eukaryotic systems examined and appears to have evolved as a surveillance mechanism to ensure that transcripts containing premature nonsense codons are rapidly degraded, thus preventing synthesis of incomplete and potentially deleterious proteins.

The NMD pathway has been extensively investigated in the yeast Saccharomyces cerevisiae (1, 2, 4, 5). Trans-acting factors involved in this process have been identified and characterized. Mutations in the UPF1, UPF2, and UPF3 genes were shown to selectively stabilize mRNAs containing early nonsense mutations without affecting the decay rate of most wild-type mRNAs (7–11). Studies using the two-hybrid system demonstrated that the Upf1p, Upf2p, and Upf3p interact and form a complex (10, 12–14). A human gene has been recently identified, RENT1, that is highly homologous to the UPF1 gene. A yeast/human hybrid of these two proteins is partially functional in yeast cells, suggesting that NMD is an evolutionarily conserved pathway (15, 16).

In addition to their role in the NMD pathway, recent results indicate that the Upf proteins may be also involved in modulating certain aspects of the translation process (12, 13, 17). Deletion of any one of the UPF genes results in a nonsense suppression phenotype (8, 9). Importantly, mutations in the UPF1 gene have been isolated that separate its ability to promote NMD from its function in modulating translation termination at nonsense codons (12, 13). A unique allele of the UPF1 gene was also isolated in a screen for mutations that enhance programmed −1 ribosomal frameshifting (17, 18). Taken together, these results suggest that the Upfp complex may be part of a surveillance complex that functions to monitor the accuracy of several processes in translation and in mRNA decay.

Programmed −1 ribosomal frameshifting, a unique alteration in translational fidelity, is used predominantly by certain RNA viruses to induce elongating ribosomes to shift reading frame in response to specific mRNA signals (19–22). The L-A double-stranded RNA (dsRNA) virus of yeast utilizes a programmed −1 ribosomal frameshift event for the production of a Gag-pol fusion protein. The M₁ satellite virus of L-A encodes a secreted killer toxin, and its dsRNA genome is encapsidated and replicated by using the L-A virus encoded gene products (20). Preservation of the appropriate efficiency of ribosomal frameshifting is critical for maintenance of the yeast M₁ dsRNA virus (23). Changes in the efficiency of programmed −1 ribosomal frameshifting by as little as 2- to 3-fold can promote loss of the M₁ virus. Therefore, the ability of the cell to maintain the M₁ virus can be used as an assay to monitor the ability of cells to preserve the accuracy of the translational reading frame. A screen for cellular mutations that increased the programmed −1 ribosomal frameshift efficiency identified mutations in nine chromosomal mof genes (for maintenance of frame; refs. 18, 20, and 23). As described above, characterization of the mof4-1 mutation demonstrated that it is allelic to UPF1 (17). Interestingly, although mof4-1 strains demonstrated increased programmed −1 ribosomal frameshifting efficiency and promoted loss of the M₁ virus, a upf1-Δ strain did not demonstrate any of these phenotypes (17). It is possible that mof4-1 represents a gain of function mutation or that it alters the Upfp complex such that it results in an increased efficiency of −1 ribosomal frameshifting.

The observation that the mof4-1 allele of the UPF1 gene affected programmed −1 ribosomal frameshifting and M₁ maintenance suggested that other proteins that are part of the surveillance complex may also be involved in modulating the maintenance of translational reading frame. The results presented here demonstrate that the Upf3 protein is involved in modulating −1 programmed frameshift efficiency. A upf3-Δ strain shows an increase in programmed −1 ribosomal frameshifting efficiency and a concomitant lose of the ability to maintain the M₁ virus. Based on these results, the role of Upfp as part of the surveillance complex that modulates translation and mRNA turnover will be discussed.

METHODS

Strains, Plasmids, Media, and General Methods.

The yeast strains used in this study are listed in Table 1. Plasmids pF8 and pTI25 were previously described (24) and are shown in Fig. 1. Plasmid pmof4BE carrying the mof4-1 allele was as described (9). Yeast media, transformations, cytoductions of L-A and M₁ into rho-o strains by using strains 3164 and 3165 as cytoduction donors, β-galactosidase (β-gal) assays and paromomycin sensitivity assays were performed as described (9, 17, 23, 25–27).

Table 1.

Strains used in this study

Strain	Genotype	Ref.
HFY1200	MATa ade2-1 his3-11, 15 leu2-3, 112 trp1-1	14
	ura3-1 can1-100 UPF1 NMD2 UPF3
HFY870	MATa ade2-1 his3-11, 15 leu2-3, 112 trp1-1	14
	ura3-1 can1-100 upf1∷HIS3 NM2 UPF3
HFY1300	MATα ade2-1 his3-11, 15 leu2-3, 112 trp1-1	14
	ura3-1 can1-100 UPF1 nmd2∷HIS3 UPF3
HFY861	MATa ade2-1 his3-11, 15 leu2-3, 112 trp1-1	14
	ura3-1 can1-100 UPF1 NMD2 upf3∷HIS3
HFY3000	MATα ade2-1 his3-11, 15 leu2-3, 112 trp1-1	14
	ura3-1 can1-100 upf1∷URA3 nmd2∷HIS3
	UPF3
HFY872	MATa ade2-1 his3-11, 15 leu2-3, 112 trp1-1	14
	ura3-1 can1-100 upf1-1∷URA3 NMD2
	upf3∷HIS3
HFY874	MATa ade2-1 his3-11, 15 leu2-3, 112 trp1-1	14
	ura3-1 can1-100 UPF1 nmd2∷URA3
	upf3∷HIS3
HFY883	MATa ade2-1 his3-11, 15 leu2-3, 112 trp1-1	14
	ura3-1 can1-100 upf1∷LEU2 nmd2∷URA3
	upf3∷HIS3
HYF870mof4	MATa ade2-1 his3-11, 15 leu2-3, 112 trp1-1	This study
	ura3-1 can1-100 upf1∷HIS3 NMD2 UPF3
	pmof4BE
HFY872mof4	MATa ade2-1 his3-11, 15 leu2-3, 112 trp1-1	This study
	ura3-1 can1-100 upf1-1∷URA3 NMD2
	upf3∷HIS3 pmof4BE
3164	MATa kar1-1 arg1L-AHN M1 K+	23
3165	MATα kar1-1 arg1 thr(1,x) L-AHN M1 K+	18
5X47	MATa/MATα his1/+trp/+ura3/+K−R−	23

Figure 1.

Schematic diagram of the vectors used to measure programmed −1 ribosomal frameshift efficiencies in vivo. Transcription is driven from the PGK1 promoter and uses the PGK1 translation initiation codon. In pTI25, the bacterial lacZ gene is in the 0-frame with respect to the start site. In plasmid pF8, the lacZ gene is positioned 3′ of the L-A virus frameshift signal and in the −1 frame relative to the translation start site.

Cloning of UPF3.

The UPF3 gene was cloned following the strategy used to clone the UPF2 gene (9). Briefly, a yeast strain harboring the upf3-Δ, his4-38 and SUF1-1 alleles was transformed with a yeast genomic library and screened for cells carrying a plasmid that could overcome the allosuppressor phenotype of the upf3-Δ. Subsequent subcloning and sequence analysis identified the sequence of the UPF3 gene that was previously reported (11).

Killer Assays, Frameshifting Assays, and Extraction and Analysis of Total Nucleic Acids.

The killer assay was carried out as described (23) by replica plating colonies onto 4.7MB plates newly seeded with a lawn of 5 × 47 killer indicator cells. Killer activity was observed as a zone of growth inhibition around the colonies. To quantitate loss of killer activity, colonies that had been identified as killer⁺ were restreaked for single colonies and the percentage of killer⁻ colonies were determined. The efficiencies of −1 frameshifting were determined as described (17, 28) by using the 0-frame control (pTI25) and −1 reporter (pF8) plasmids.

The abundance of L-A and M₁ (+) RNAs were monitored by RNA blotting analysis as described (17, 26). RNA abundance of the lacZ −1 frameshift reporter mRNA and U3 small nuclear RNA (snRNA) was determined by ribonuclease protection assays essentially as described (29).

RESULTS

A upf3-Δ Strain Demonstrates an Increased Efficiency of Programmed −1 Ribosomal Frameshifting.

mof4-1 is a unique allele of the UPF1 gene that specifically increases programmed −1 ribosomal frameshifting efficiency and promotes loss of the M₁ satellite virus (17). A upf1-Δ strain, however, does not demonstrate these phenotypes. We hypothesized that other factors of the putative surveillance complex, including the Upf2 or Upf3 proteins, could also affect programmed −1 ribosomal frameshifting. Therefore, we investigated whether isogenic strains harboring deletions of the UPF genes demonstrated increased ribosomal frameshifting efficiencies.

Methods to measure efficiencies of programmed ribosomal frameshifting in vivo have been described previously (17, 24, 28). A series of lacZ reporter plasmids were used (Fig. 1) in which a translational start codon is followed by the Escherichia coli lacZ gene. Plasmid pTI25 serves as the 0-frame control because lacZ is in the 0-frame with respect to the translational start site (Fig. 1). In plasmid pF8, an L-A derived programmed −1 ribosomal frameshift signal is cloned into the polylinker and the lacZ gene is in the −1 frame with respect to the translational start site (Fig. 1). Therefore, in this construct the lacZ gene will be translated only if the ribosome shifts frame in the −1 direction. The +1 frameshift reporter plasmid, pJD104 (Fig. 1), contains the lacZ gene inserted 3′ of a programmed +1 ribosomal frameshift signal derived from the Ty1 retrotransposable element of yeast. In this construct, the lacZ gene will be translated only if the ribosome shifts frame in the +1 direction. The efficiency of −1 ribosomal frameshifting is calculated by determining the ratio of β-gal activities measured in cells harboring the −1 frameshift reporter plasmid, pF8, to those harboring the 0-frame control plasmid, pTI25, and multiplying by 100%. Similarly, the +1 ribosomal frameshift efficiency is calculated based on the pJD104 to pTI25 β-gal ratios.

Frameshift efficiency was measured in isogenic yeast strains harboring deletions of different UPF genes (Table 2). The results of these experiments demonstrated that the apparent efficiency of programmed −1 ribosomal frameshifting were slightly greater in upf1-Δ and upf2-Δ strains than in wild-type cells (Table 2). As discussed below, the small increase in −1 programmed frameshifting was not sufficient to promote loss of the M₁ virus. In contrast, the efficiency of programmed −1 ribosomal frameshifting in upf3-Δ cells was 3.4-fold higher than wild-type cells (Table 2) and was sufficient to promote loss of the M₁ virus (see below). None of the upf-Δ strains demonstrated a significant increase in the apparent efficiency of programmed +1 ribosomal frameshifting, as measured by the levels of β-gal activity (data not shown). Taken together, these results indicated that, analogous to a mof4-1 strain, the upf3-Δ strain specifically alters −1 ribosomal frameshifting.

Table 2.

Programmed −1 ribosomal frameshifting and M₁ virus maintenance of strains harboring a single deletion of a UPF gene

Strain (genotype)	% −1 ribosomal frameshifting*	Killer maintenance†
UPF+ (HFY1200)	2.5	+
upf1-Δ (HFY870)	4.5	+
upf2-Δ (HFY1300)	3.9	+
upf3-Δ (HFY861)	8.4	−

^*The −1 ribosomal frameshift efficiency (%) was determined by the ratio of β-gal activity in a strain harboring the −1 ribosomal frameshifting reporter plasmid to the activity in the same strain harboring the 0-frame control plasmid. 

^†L-AHN and M₁ were introduced into the strains by cytoduction and the maintenance (+) or loss (−) of M₁ dsRNA was analyzed by the killer plate assay and Northern blot analyses as described. 

The Abundance of the Frameshift Reporter Transcript Is Equivalent in the upf-Δ Strains.

The −1 frameshift reporter transcripts used in these assays have short protein coding regions 5′ of the frameshift site followed by sequences that code for a reporter protein that is out of frame with the translation initiation site of the 5′ ORF. The apparent changes in ribosomal frameshifting efficiencies could result from changes in the abundance of the LacZ −1 frameshift reporter mRNA that the translational machinery may recognize as a nonsense-containing mRNA (see ref. 17). Deletion of the UPF genes could lead to stabilization of the −1 frameshift reporter transcript, resulting in increased synthesis of the β-gal reporter protein. To address whether a upf3-Δ strain accumulates the reporter transcript to a greater extent than upf1-Δ or upf2-Δ strains, the abundance of the LacZ −1 frameshift reporter mRNA was determined by RNase protection analysis. As a loading control, we also determined the abundance of the U3 snRNA. The results indicated that the abundances of the LacZ frameshift reporter mRNA, normalized to the U3 snRNA, were equivalent in isogenic wild-type, upf1-Δ, upf2-Δ, and upf3-Δ strains (Fig. 2). Therefore, the increased programmed −1 ribosomal frameshifting efficiency observed in a upf3-Δ, when compared with the upf1-Δ or upf2-Δ strains, was not a consequence of stabilizing the reporter transcript to a greater extent than in the other upf-Δ strains. Therefore, a upf3-Δ strain also demonstrates a mof phenotype in that it increases the efficiency of −1 ribosomal frameshifting independently of its ability to stabilize nonsense mRNAs.

Figure 2.

A upf3-Δ strain increases programmed −1 ribosomal frameshifting independently of its ability to promote stabilization of nonsense-containing transcripts. The abundance of the PGK1–LacZ −1 reporter mRNA in the different upf deletion strains was determined by RNase protection analysis. The abundance of the U3 snRNA was used as an internal control for loading. The abundance of the reporter transcript in the wild-type strain was taken arbitrarily as 1.0.

The M₁ Killer Virus Is Not Maintained in a upf3-Δ Strain.

Changing the efficiency of −1 ribosomal frameshifting alters the ratio of Gag to Gag-pol proteins available for viral particle assembly, consequently interfering with viral propagation (17, 18, 23, 28). Therefore, we next asked whether a upf3-Δ strain was able to maintain the L-A and/or the M₁ satellite virus as an independent assay to monitor changes in the −1 ribosomal frameshift efficency. The L-A and M₁ viruses were introduced by cytoduction into isogenic wild-type, upf1-Δ, upf2-Δ, and upf3-Δ strains, and these cells were grown and replica plated onto a lawn of cells sensitive to the killer toxin. Cells maintaining the M₁ virus secrete the killer toxin, creating a ring of growth inhibition (17, 23, 28). The results of this assay demonstrated that the wild-type, upf1-Δ and upf2-Δ strains maintained the killer phenotype, whereas the upf3-Δ strain lose the ability to maintain the killer phenotype (Fig. 3A; Table 2). Consistent with previous results, cells harboring the mof4-1 allele of the UPF1 gene were also unable to maintain the killer phenotype (17).

Figure 3.

A upf3-Δ strain cannot maintain the M₁ killer virus. (A) Killer assay of upf mutant strains. Colonies of these strains were grown onto a lawn of cells that are sensitive to the secreted killer toxin produced by the M₁ virus. Killer activity was observed as a zone of growth inhibition around the colonies. (B) Total RNAs were isolated from the same strains and analyzed by Northern blotting for the presence of L-A and M₁ viral RNAs.

To determine whether lack of the killer phenotype was a consequence of a virus maintenance defect rather than interference with production of the killer toxin, total nucleic acids were extracted from the wild-type, upf1-Δ, upf2-Δ, and upf3-Δ strains and analyzed by Northern blot analysis by using L-A and M₁ (+)-strand RNA-specific probes. The results demonstrated that the 1.8-kb M₁ dsRNA was absent in the mof4-1 and upf3-Δ cells but present in upf1 and upf2 mutant and the wild-type strains (Fig. 3B). These results support the hypothesis that deleting the UPF3 gene alters the efficiency of −1 ribosomal frameshifting interfering with the propagation of M₁ satellite virus.

The upf3-Δ Strain Demonstrates Increased Sensitivity to Paromomycin.

Strains harboring mutations that diminish translational fidelity are hypersensitive to the aminoglycoside antibiotic paromomycin, a drug that is thought to increase the frequency of misreading in yeast (30, 31). Previous results demonstrated that cells harboring the mof4-1 allele of the UPF1 gene showed an increased sensitivity to paromomycin relative to the isogenic wild-type strain (17). We determined whether a upf3-Δ strain also demonstrates increased sensitivity to this antibiotic. Paromomycin sensitivity was monitored in isogenic wild-type and upf3-Δ strains by determining the zone of growth inhibition around a disc containing 1 mg of paromomycin placed onto a lawn of cells. The results demonstrated that, analogous to a mof4-1 strain, a upf3-Δ strain was more sensitive to paromomycin than the isogenic wild-type strain (Fig. 4A). Neither the upf1-Δ nor upf2-Δ strains demonstrate hypersensitivity to paromomycin (ref. 17 and data not shown).

Figure 4.

Effects of paromomycin in a upf3-Δ strain. (A) A upf3-Δ strains shows increased sensitivity to paromomycin. Sensitivity to this drug was determined in isogenic wild-type and upf3-Δ strains. Cells were plated in yeast extract/peptone/dextrose (YPD) medium and a disc containing 1 mg of paromomycin was placed on the lawn of cells. The diameter of the zone of growth inhibition was determined after the plates were incubated at 30°C for 2 days. (B) Isogenic wild-type and upf3-Δ strains were transformed with either −1 frameshift tester or 0-frame control plasmids. Increasing concentrations of paromomycin were added to cells inoculated at 0.1 OD₆₀₀/ml and grown at 30°C for 4 h. The efficiency of −1 ribosomal frameshifting was calculated by determining the ratio of β-gal activities measured on cells harboring the −1 frameshift reporter plasmid to those harboring the 0-frame control plasmid, grown under the same concentration of paromomycin, and multiplying by 100%.

We determined whether increased doses of paromomycin can modulate the −1 frameshifting efficiency of the upf3-Δ strain. Isogenic wild-type and upf3-Δ strains harboring plasmids pF8 (−1 frameshift reporter construct) or pTI25 (0-frame control), were grown in liquid media in the presence of different concentrations of the drug, and the β-gal activity was determined, normalizing to the number of cells used in the assay. The results indicated that the β-gal activity from upf3-Δ cells carrying pF8 (−1 frameshift reporter construct) increased continuously with increased concentrations of paromomycin. However, the β-gal activity was unaffected in wild-type cells containing pF8 or in any of the strains carrying pTI25 (0-frame control construct) (Fig. 4B).

The Increased −1 Programmed Frameshifting and Killer Virus Maintenance Defect Phenotypes of upf3-Δ and upf3-Δ mof4-1 Strains Are Equivalent.

The results described above indicated that a upf3-Δ strain has similar phenotypes as mof4-1 cells. Because the mof4-1 allele of the UPF1 gene, but not deletion of the UPF1 gene, affected programmed −1 ribosomal frameshifting and M₁ maintenance, we hypothesized that the mof4-1p could alter the function of the Upf3p. Thus, a mof4-1 upf3-Δ strain should have the same programmed −1 frameshifting and killer phenotypes as a upf3-Δ strain. We monitored both the programmed −1 ribosomal frameshifting efficiency and virus maintenance phenotypes in isogenic mof4-1, upf3-Δ, and mof4-1 upf3-Δ strains as described above. The programmed −1 ribosomal frameshifting efficiencies observed in mof4-1, upf3-Δ and mof4-1 upf3-Δ strains were equivalent, and all these strains lacked the killer phenotype (Table 3).

Table 3.

Programmed −1 ribosomal frameshifting and M₁ virus maintenance of strains harboring multiple mutations of UPF genes

Genotype (strain)	% ribosomal frameshifting*	Killer maintenance
UPF+ (HFY1200)	2.5	+
upf3-Δ (HFY861)	8.4	−
mof4-1 (HFY870mof4)	7.0	−
mof4-1 upf3-Δ (HFY872mof4)	8.0	−
upf1-Δ upf2-Δ (HFY3000)	3.2	+
upf1-Δ upf3-Δ (HFY872)	7.2	−
upf2-Δ upf3-Δ (HFY874)	9.2	−
upf1-Δ upf2-Δ upf3-Δ (HFY883)	8.0	−

^*Programmed −1 ribosomal frameshifting efficiency and M₁ virus maintenance was determined as described in the legend of Table 2. 

The Programmed Frameshifting and Killer Phenotypes of a upf3-Δ Allele Are Independent of the Other upf-Δ Alleles.

The epistatic relationships between upf1-Δ, upf2-Δ and upf3-Δ were examined with regard to both −1 ribosomal frameshifting efficiencies and killer maintenance. Both programmed −1 ribosomal frameshifting and killer phenotypes were monitored as described above in isogenic UPF⁺, upf1-Δ upf2-Δ, upf1-Δ upf3-Δ, upf2-Δ upf3-Δ, and upf1-Δ upf2-Δ upf3-Δ strains. All of the strains harboring the upf3-Δ, independent of the presence or absence of the other UPF genes, had increased −1 ribosomal frameshifting efficiencies, equivalent to that harboring deletion of the UPF3 gene (Table 3). Conversely, upf1-Δ UPF3⁺, upf2-Δ UPF3⁺, and upf1-Δ upf2-Δ UPF3⁺ strains did not demonstrate an increase in programmed −1 frameshifting efficiencies sufficient to promote loss of the killer phenotype (Tables 2 and 3). Taken together, these results suggest that the Upf3p is the central component of the Upfp complex that modulates programmed frameshifting.

DISCUSSION

The Upf Proteins Are Part of the Surveillance Complex That Monitors Both mRNA Turnover and Translation.

Recent results indicate that the factors involved in the NMD pathway play additional roles in modulating several aspects of the translation process. Genetic and biochemical studies of the Upf1p suggest that it is a multifunctional protein that acts both in NMD and in modulating the translation termination process (12, 13, 32). Furthermore, mutations in the UPF genes affect programmed frameshift. The mof4-1 allele of the UPF1 gene demonstrates an increase in programmed −1 ribosomal frameshifting efficiency and is unable to maintain the M₁ killer virus (17). In addition, mof2-1 mutants manifest increased programmed −1 ribosomal frameshift efficiency (33). The mof2-1 mutant is allelic to the SUI1 gene (33), which was previously shown to play a role in translation initiation start site selection. Interestingly, mof2-1 mutant strains also demonstrate accumulation of nonsense-containing transcripts (Y. Cui and S.W.P., unpublished results). These results suggest that the surveillance complex, including factors involved in NMD, may also be involved in monitoring other steps in the translation process. The results presented here indicate that the Upf3p, in addition to its role in NMD, is also probably part of the putative surveillance complex involved in maintaining appropriate translational reading frame.

The Upf3p Is the Key Factor That Links the Upfp Complex to Programmed −1 Ribosomal Frameshifting.

The results presented here indicate that the Upf3p has a function in ensuring appropriate maintenance of translational reading frame. Monitoring the programmed ribosomal frameshifting and M₁ virus maintenance profiles of cells harboring deletions of the UPF1, UPF2, or UPF3 genes demonstrated that a upf3-Δ strain affected programmed −1 frameshift efficiency and virus maintenance (Tables 2 and 3). The notion that the Upf3p is the central component of the Upfp complex that modulates programmed frameshifting is supported by the observation that a mof4-1 upf3-Δ strain has the same programmed −1 ribosomal frameshift and killer phenotypes as a mof4-1 strain (Table 2). In addition, the mof4-1 allele of the UPF1 gene, but not a upf1-Δ allele, affected programmed −1 ribosomal frameshifting and killer maintenance, suggesting that Upf1p does not directly influence the maintenance of the translational reading frame. We hypothesize that the mof4-1 allele of the Upf1p alters the Upfp complex such that the Upf3p can no longer function to help maintain the correct translational reading frame, resulting in an increased programmed −1 ribosomal efficiency which consequently leads to loss of the M₁ satellite virus. Although we strongly favor this model, our results do not completely rule out the possibility that mof4–1p functions independently of Upf3p. The fact that a mof4-1 upf3-Δ strain shows the same phenotype as a strain harboring a single mutant could also be explained if both proteins function by affecting the same target that is located downstream of both, mof4p and Upf3p.

The Upfp Complex May Be Part of a Surveillance Complex That Function as a Translational Checkpoint.

As described above, mutations in the UPF genes can result in altered translation termination phenotypes, increased programmed frameshifting and stabilization of nonsense-containing transcripts (refs. 12, 13, and 17; reviewed in refs. 2 and 5). This result suggests that the Upfp complex, perhaps as part of a larger surveillance complex, functions as a “translational checkpoint.” Analogous to cell cycle control checkpoints, the UPF genes are not essential, but ensure that the processes that they are involved in occur with high fidelity.

An important question is what triggers the activity of the surveillance complex. We hypothesize that a paused ribosome may be a key event that promotes assembly of this complex, which can subsequently monitor these processes. Both programmed frameshifting and translation termination involve a ribosomal pause (refs. 34 and 35; reviewed in ref. 36). We suggest that the interaction of the Upfp with the translation termination release factors eRF1 and eRF3 of a paused ribosome helps promote the assembly of the Upfp complex (32, 37). We suggest that this interaction leads to enhanced translation termination and subsequent degradation of nonsense-containing transcripts (32). In the case of programmed −1 ribosomal frameshifting, the RNA pseudoknot following the slippery site promotes a ribosomal pause (36, 38), which may also trigger assembly of the surveillance complex. This complex, or a subset of the Upf proteins, may help the ribosome maintain the appropriate translational reading frame. The goal of our future experiments will be to determine how these factors interact with the translational machinery to promote decay of nonsense-containing mRNAs, enhance termination, and help maintain the ribosome in the appropriate translational reading frame.