Novel Upf2p orthologues suggest a functional link between translation initiation and nonsense surveillance complexes

Abstract

Transcripts harboring premature signals for translation termination are recognized and rapidly degraded by eukaryotic cells through a pathway known as nonsense-mediated mRNA decay (NMD). In addition to protecting cells by preventing the translation of potentially deleterious truncated peptides, studies have suggested that NMD plays a broader role in the regulation of the steady-state levels of physiologic transcripts. In Saccharomyces cerevisiae, three trans-acting factors (Upf1p to Upf3p) are required for NMD. Orthologues of Upf1p have been identified in numerous species, showing that the NMD machinery, at least in part, is conserved through evolution. In this study, we demonstrate additional functional conservation of the NMD pathway through the identification of Upf2p homologues in Schizosaccharomyces pombe and humans (rent2). Disruption of S. pombe UPF2 established that this gene is required for NMD in fission yeast. rent2 was demonstrated to interact directly with rent1, a known trans-effector of NMD in mammalian cells. Additionally, fragments of rent2 were shown to possess nuclear targeting activity, although the native protein localizes to the cytoplasmic compartment. Finally, novel functional domains of Upf2p and rent2 with homology to eukaryotic initiation factor 4G (eIF4G) and other translational regulatory proteins were identified. Directed mutations within these so-called eIF4G homology (4GH) domains were sufficient to abolish the function of S. pombe Upf2p. Furthermore, using the two-hybrid system, we obtained evidence for direct interaction between rent2 and human eIF4AI and Sui1, both components of the translation initiation complex. Based on these findings, a novel model in which Upf2p and rent2 effects decreased translation and accelerated decay of nonsense transcripts through competitive interactions with eIF4G-binding partners is proposed.

FIG. 1

Sequence alignment of S. cerevisiae Upf2p (ScUpf2p), S. pombe Upf2p (SpUpf2p), and human rent2. A ClustalW alignment was performed using the MacVector 6.5.1 package of sequence analysis software. Identical residues shared between at least two of the three proteins are shaded in black; similar residues shared between at least two of the proteins are shaded in gray. The previously identified putative or documented functional domains of S. cerevisiae Upf2p are indicated by lines.

FIG. 2

Upf2p is required for NMD in S. pombe. (A) S. pombe UPF2 encodes a single ∼3.7-kb transcript which is absent in upf2Δ strains. Total RNA was isolated from logarithmically growing S. pombe strains and analyzed for UPF2 expression by Northern blotting. The blot was stripped and rehybridized with a YPT5 probe to control for loading differences. (B) Deletion of UPF2 specifically stabilizes nonsense transcripts in S. pombe. The complete UPF2 ORF was disrupted in S. pombe strains which were wild type at the nonessential ADE6 locus (wt), harbored two different nonsense mutations near the 5′ end of the ADE6 gene which give rise to transcripts degraded by NMD (ade6-M26 and ade6-M375; 5′ PTC1 and 5′ PTC2, respectively), contained a nonsense mutation near the 3′ end of ADE6 which is not a substrate for NMD (ade6-469; 3′ PTC), or contained a missense mutation (ade6-M216; mis). To measure the steady-state levels of wild-type or mutant ADE6 transcripts, total RNA was analyzed by Northern blotting with an ADE6 probe. The blot was stripped and rehybridized with a YPT5 probe to control for loading differences. (C) The decreased steady-state level of ade6 nonsense transcripts is due to a UPF2-dependent accelerated mRNA decay rate. The half-life (t_1/2) of wild-type (WT) ADE6 or mutant ade6 transcripts containing a nonsense mutation (5′PTC2; ade6-M375) was determined in UPF2 or upf2Δ strains. Transcription was inhibited with thiolutin, and transcript abundance was determined by Northern blot analysis at the indicated time points.

FIG. 3

Variation in RENT2 expression does not influence the efficiency of NMD. (A) Expression of RENT2 was assessed in adult human tissues by probing a multi-tissue Northern blot (Clontech) with a human RENT2 cDNA fragment. The blot was stripped and rehybridized with a β-actin cDNA probe to control for loading differences. (B) Expression of mouse Rent2 in adult tissues was analyzed by hybridizing a mouse multi-tissue Northern blot (Clontech) with a mouse Rent2 cDNA probe. β-Actin was again used as a loading control. (C) Northern blot analysis of poly(A) RNA from wild-type or homozygous mutant gus^mps mice. To measure steady-state Gus transcript levels, 2 μg of poly(A) RNA from the indicated adult tissues was probed with a Gus cDNA fragment. The blot was stripped and rehybridized with a G3PDH probe to control for loading differences.

FIG. 4

rent2 localizes to the cytoplasmic compartment of mammalian cells despite possessing functional NLSs. The subcellular localization of N-terminal HA-tagged full-length rent2 (HA-rent2), a full-length C-terminal rent2-GFP fusion (rent2-GFP), or the N-terminal 120 amino acids of rent2 fused to GFP [rent2(1–120)GFP] was determined in transiently transfected HeLa cells in the presence or absence of the inhibitor of nuclear export leptomycin B. As a positive control for leptomycin B treatment, the subcellular localization of cyclin B1 was determined in the presence or absence of the drug.

FIG. 5

Conserved rent1 and rent2 interactions. (A) Schematic representation of rent1–GAL4 activation domain (GAL4AD) and rent2–GAL4 DNA-binding domain (GAL4BD) fusion constructs. Predicted functional domains of rent1 and rent2 based on homology to S. cerevisiae Upf1p and Upf2p, respectively, are shaded. Overlying numbers indicate amino acid position. (B) Stringent growth assay for interaction. The indicated GAL4AD and GAL4BD fusion constructs were co-introduced into strain AH109 and plated on minimal medium lacking Trp and Leu (left) to select for cotransformants or plated on medium lacking Trp, Leu, His, and Ade (right) to select for interacting proteins. As a positive control, SV40 large T antigen (SV40-T)–GAL4AD and p53-GAL4BD fusion constructs were used to test a known interaction between these proteins. The lack of interactions with a lamin C-GAL4BD fusion peptide served as a negative control. (C) Reduced-stringency growth assay for interaction. rent2-GAL4BD fusions which failed to interact with rent1-GAL4AD fusions were tested on minimal media lacking Trp, Leu, and His. (D) Coimmunoprecipitation of rent1 and rent2. Cell extracts from HeLa cells overexpressing rent1 alone (lane 1) or rent1 plus either HA-tagged luciferase (lane 2), residues 1 to 1095 of rent2 with a C-terminal HA tag (lane 3), or residues 757 to 1272 of rent2 with a C-terminal HA tag (lane 4) were precipitated with an anti-HA monoclonal antibody (Covance) and analyzed by Western blotting. Duplicate blots from the same experiment are shown, probed with anti-rent1 antiserum (upper panel) or an anti-HA monoclonal antibody (lower panel). The intense lower band in all lanes is the heavy chain of the anti-HA antibody used for immunoprecipitation.

FIG. 6

The 4GH domains of S. pombe Upf2p are required for NMD. (A) Alignment of all known 4GH domains. With the exception of S. cerevisiae Upf2p (ScUpf2p) and S. pombe Upf2p (SpUpf2p), all proteins shown are human. Arrows indicate residues mutated in the upf2-4GH1 and upf2-4GH2 S. pombe Upf2p expression constructs. (B) Effect of 4GH mutations on the function of S. pombe Upf2p. A upf2Δ S. pombe strain harboring a nonsense mutation in ADE6 (ade6-M375) was transformed with an empty expression vector or vector containing wild-type S. pombe UPF2 (UPF2), UPF2 mutated at positions 3 and 6 of 4GH1 (upf2-4GH1), UPF2 mutated at positions 3 and 6 of 4GH2 (upf2-4GH2), or UPF2 mutated at positions 3 and 6 of both 4GH domains (upf2-4GH1,2). The steady-state levels of the ade6 nonsense transcript were determined by quantitative Northern blot analysis. Following hybridization with an ADE6 probe, the blot was reprobed for YPT5 to standardize for loading differences. The ade6/YPT5 ratio was calculated for each sample and normalized to the ratio measured in the upfΔ strain.

FIG. 7

Two-hybrid analysis of rent2-translation initiation factor interactions. (A) Schematic representation of the rent2–GAL4 DNA-binding domain (GAL4BD) fusion constructs used in this study. (B) Growth assay for interaction. Full-length human eIF4AI [eIF4AI(1–407)], truncated human eIF4AI [eIF4AI(1–325)], human Sui1, and human Prt1 were fused to the GAL4AD and cointroduced into strain AH109 with the indicated GAL4BD fusions. Yeast strains were plated on minimal medium lacking Trp and Leu (left) to select for cotransformants or plated on medium lacking Trp, Leu, and His (right) to select for interacting proteins. SV40 large T antigen (SV40-T)–GAL4AD and p53-GAL4BD fusion constructs were used as a positive control, and a lamin C-GAL4BD fusion peptide served as a negative control.

FIG. 8

Proposed model of 4GH domain-mediated transcript destabilization. Stable, efficiently translated transcripts are believed to adopt a closed-loop conformation mediated by protein-protein interactions which bridge the 5′ cap (m7G) and the 3′ poly(A) tail. The model suggests that the 4GH domains of rent2 and NAT1 may compete with the corresponding domain in eIF4G (and perhaps PAIP-1) for interacting partners eIF3 and/or eFI4A. Such associations (dashed lines) may occur in cis when rent2 is a component of a mature surveillance complex (partnered with rent1 and rent3) but in trans upon induction of NAT1 expression.