Nonsense‐mediated messenger RNA decay (NMD) generally degrades mRNAs that prematurely terminate translation as a means of quality control. NMD in mammalian cells targets newly spliced mRNA that is bound by the cap‐binding protein heterodimer CBP80/20 and one or more post‐splicing exon junction complexes during a pioneer round of translation. NMD targets mRNA that initiates translation using the encephalomyocarditis virus (EMCV) internal ribosome entry site (IRES), therefore NMD might target not only CBP80/20‐bound mRNA but also its remodelled product, eIF4E‐bound mRNA. Here, we provide evidence that NMD triggered by translation initiation at the EMCV IRES, similar to NMD triggered by translation initiation at an mRNA cap, targets CBP80/20‐bound mRNA but does not detectably target eIF4E‐bound mRNA. We show that EMCV IRES‐initiated translation undergoes a CBP80/20‐associated pioneer round of translation that results in CBP80/20‐dependent and Upf factor‐dependent NMD when translation terminates prematurely.
Nonsense‐mediated messenger RNA decay (NMD) in mammals is a translation‐dependent surveillance mechanism that controls the quality of gene expression by degrading the mRNAs of mutated genes that contain premature termination codons (PTCs; reviewed by Behm‐Ansmant et al, 2007; Chang et al, 2007; Isken & Maquat, 2007). NMD also downregulates the expression of natural genes, in some cases providing an autoregulatory circuit to inhibit the production of RNA‐binding proteins, including splicing factors (Mendell et al, 2004; Wittmann et al, 2006; Lareau et al, 2007; Ni et al, 2007). However, data indicate that many, if not most, naturally occurring NMD targets result from nonproductive alternative splicing (Pan et al, 2006).
mRNA that is bound by the cap‐binding protein heterodimer CBP80/20, which is mostly nuclear but also shuttles to the cytoplasm, is a precursor to mRNA that is bound at the cap by eIF4E, which is largely cytoplasmic (Lejeune et al, 2002). The NMD of nonsense‐containing mRNAs generally occurs when translation initiates in an mRNA cap‐dependent manner during a pioneer round of translation, which involves newly synthesized CBP80/20‐bound mRNA (Ishigaki et al, 2001; Sato et al, 2008). NMD does not detectably target mRNA once CBP80/20 has been replaced by eIF4E, which supports the bulk of cellular protein synthesis (Ishigaki et al, 2001; Chiu et al, 2004; Hosoda et al, 2005). This is in contrast to the situation in Saccharomyces cerevisiae, in which NMD targets not only Cbc1/2‐bound mRNA, which is orthologous to mammalian cell CBP80/20‐bound mRNA, but also eIF4E‐bound mRNA (Gao et al, 2005). The apparent restriction of NMD to CBP80/20‐bound mRNA in mammals is due to the roles of CBP80 and post‐splicing exon junction complexes (EJCs) in NMD, both of which are detectable on CBP80/20‐bound mRNA but not on eIF4E‐bound mRNA (Ishigaki et al, 2001; Lejeune et al, 2002; Hosoda et al, 2005; Kashima et al, 2006). EJCs are thought to be largely removed from newly synthesized mRNAs by translating ribosomes during the pioneer round of translation (Dostie & Dreyfuss, 2002).
The importance of EJCs to NMD is supported by reports that downregulating the EJC constituents Y14, MNL51/BTZ, eIF4AIII, Upf2 or Upf3X (also called Upf3b) inhibits NMD (reviewed by Isken & Maquat, 2007). Furthermore, mRNAs that are derived from intronless genes and, thus, lack EJCs are immune to NMD when they harbour PTCs (Maquat & Li, 2001; Neu‐Yilik et al, 2001; Brocke et al, 2002; Matsuda et al, 2007). In classical NMD, EJCs are thought to function by recruiting the NMD factor Upf1 through the EJC constituent Upf2 once translation terminates sufficiently upstream from the EJC (Lykke‐Andersen et al, 2001; Serin et al, 2001; Kim et al, 2005; Kashima et al, 2006). The importance of CBP80 to NMD is evident from the finding that CBP80 promotes the interaction of Upf2 with Upf1, the latter being stably associated with the CBP80/20 cap‐binding complex (CBC; Hosoda et al, 2005). Consistent with this interpretation, tethering Upf1 sufficiently downstream from a nonsense codon bypasses the need for CBP80 so that mRNA decay is extended to eIF4E‐bound mRNA (Hosoda et al, 2005). As there is evidence that Upf1 is the last of the Upf proteins to join the EJC (Lykke‐Andersen et al, 2001; Hosoda et al, 2005; Kim et al, 2005; Singh et al, 2007), it follows that CBP80 is no longer required to promote NMD once Upf1 joins an EJC (Hosoda et al, 2005; Kim et al, 2005).
EJCs, and possibly CBP80/20, also function in fail‐safe NMD, which does not seem to depend on a post‐splicing EJC situated downstream from a nonsense codon (Zhang et al, 1998; Bühler et al, 2006; Matsuda et al, 2007). However, fail‐safe NMD requires at least one EJC upstream from the nonsense codon and, similar to classical NMD, targets CBP80/20‐bound mRNA but does not detectably target eIF4E‐bound mRNA (Matsuda et al, 2007).
On the basis of the finding that translation initiation from the encephalomyocarditis virus (EMCV) internal ribosome entry site (IRES) supports NMD, it was recently suggested that translation initiation from the mRNA cap is not essential for NMD (Holbrook et al, 2006). It follows that the recruitment of ribosomes to an mRNA by CBP80/20 and, possibly, the presence of CBP80/20 on mRNA would be dispensable for NMD. We set out to test this possibility by considering that, although never examined, CBP80/20 would probably constitute newly synthesized messenger ribonucleoprotein when translation initiates for the first time using the EMCV IRES. Thus, for the reasons mentioned above, NMD resulting from EMCV IRES‐dependent translation initiation might also be restricted to CBP80/20‐bound mRNA.
Here, we show that CBP80/20 is a constituent of EMCV IRES‐containing messenger ribonucleoprotein. Furthermore, we show that NMD resulting from EMCV IRES‐dependent translation initiation involves Upf1, Upf2, Upf3X and CBP80/20. We also used two complementary approaches to show that NMD resulting from EMCV IRES‐initiated translation targets CBP80/20‐bound mRNA but does not detectably target eIF4E‐bound mRNA.
Results And Discussion
EMCV IRES: CBP80/20 and Upf1 factors function in NMD
As noted above, classical NMD generally depends on CBP80/20 and the NMD factors Upf1, Upf2 and Upf3X. To determine whether CBP80/20 and Upf proteins function in NMD that results from translation initiation at the EMCV IRES, HeLa cells were transiently transfected with a short interfering RNA (siRNA) that downregulates CBP80/20, Upf1, Upf2 or Upf3X or, to control for nonspecific effects, with a control siRNA (Hosoda et al, 2005; Kim et al, 2005). After 2 days, cells were transfected with a pRLuc‐Gl or php‐EMCV IRES‐RLuc‐Gl test plasmid, either PTC‐free (Norm) or PTC‐containing at Gl codon 39 (Ter), and the pMUP reference plasmid (phCMV‐MUP in Belgrader & Maquat, 1994). Each set of test plasmids (Fig 1A) encodes Renilla (R) luciferase (Luc) fused to β‐globin (Gl). pRLuc‐Gl encodes RLuc‐Gl mRNA that initiates cap‐dependent translation. php‐EMCV IRES‐RLuc‐Gl encodes RLuc‐Gl mRNA that initiates EMCV IRES‐dependent translation, as it contains a stable hairpin (hp) structure 12 nucleotides downstream from the cap that inhibits cap‐initiated ribosome scanning (De Gregorio et al, 1999). Consistent with the hairpin structure blocking ribosome scanning, php‐EMCV IRES(CCCC)‐RLuc‐Gl Norm, in which the GCGA tetraloop was mutated to CCCC so as to inactivate the IRES (Robertson et al, 1999; Holbrook et al, 2006), did not produce an appreciable level of RLuc activity (supplementary Fig 1 online). By contrast, php‐EMCV IRES‐RLuc‐Gl Norm was active, producing approximately 21‐fold greater RLuc activity than php‐EMCV IRES(CCCC)‐RLuc‐Gl (supplementary Fig 1 online).
By using the level of p62 or calnexin to control for variations in protein loading, it was shown by western blotting that the level of Upf1, Upf2, Upf3X or the combination of CBP80 and CBP20 (CBC) was downregulated to 5%, 10%, 12% or 5% and 12% the level in the presence of control siRNA, respectively (Fig 1B–E, upper panels). siRNA‐mediated effects on processes other than NMD were controlled for by comparing PTC‐containing mRNA with PTC‐free mRNA. CBC siRNA seems to retard the nucleocytoplasmic export of spliced mRNA (C.F.W. and L.E.M., unpublished data), and the degree to which it inhibits NMD because of the direct effect of CBC on NMD or on mRNA export is uncertain. Semiquantitative reverse transcription–PCR (RT–PCR) showed that NMD reduced the level of RLuc‐Gl Ter and hp‐EMCV IRES‐RLuc‐Gl Ter mRNAs in the presence of control siRNA to 6–7% and 13–14% the level of the corresponding Norm mRNA, respectively (Fig 1B–E, lower panels). Furthermore, downregulating each protein inhibited the NMD of RLuc‐Gl Ter and hp‐EMCV IRES‐RLuc‐Gl Ter mRNAs about two‐ to sixfold, depending on the particular siRNA (Fig 1B–E, lower panels). This variation probably reflects (i) differences in siRNA efficiencies and differences in the cellular fraction of each targeted protein that functions in NMD, and (ii) less efficient translation initiation from the IRES compared with that from the 5′ cap, which results in less efficient PTC recognition and, as a consequence, NMD. Notably, the reliability of using semiquantitative RT–PCR was validated by using real‐time PCR for samples involving Upf1 siRNA (supplementary Fig 2 online). We conclude that Upf1, Upf2, Upf3X and CBP80/20 function in NMD that results from translation initiation at either an mRNA cap or the EMCV IRES.
EMCV IRES: 4E‐BP1 does not detectably inhibit NMD
To determine whether NMD that results from EMCV IRES‐initiated translation targets CBP80/20‐bound mRNA or eIF4E‐bound mRNA or both, cells were transfected with the pACTAG2‐HA3‐4E‐BP1 effector plasmid, which produces 4E‐BP1 (Gingras et al, 1998) or, as a control, pACTAG2‐HA3. 4E‐BP1 inhibits the translation of eIF4E‐bound mRNA but not CBP80/20‐bound mRNA (Chiu et al, 2004). After 1 day, cells were transfected with (i) pRLuc‐Gl or php‐EMCV‐IRES‐RLuc‐Gl, either Norm or Ter, and (ii) pMUP.
HA3‐4E‐BP1 expression significantly reduced the abundance of the secreted major urinary protein (MUP; Fig 2A) and the activity of RLuc that derived from cap‐dependent pRLuc‐Gl expression but not cap‐independent php‐EMCV‐IRES‐RLuc‐Gl expression (Fig 2B). This demonstrates that HA3‐4E‐BP1 inhibited the cap‐dependent translation of eIF4E‐bound mRNA, which supports the bulk of cellular protein synthesis. As expected, RT–PCR showed that HA3‐4E‐BP1 was of no importance to the NMD of RLuc‐Gl Ter mRNA that involves the cap‐dependent translation of CBP80/20‐bound mRNA (Fig 2C), consistent with previous results (Chiu et al, 2004; Matsuda et al, 2007). HA3‐4E‐BP1 was also of no consequence to the NMD of hp‐EMCV IRES‐RLuc‐Gl Ter mRNA, which initiates translation at the EMCV IRES (Fig 2C). These data indicate that translation initiation from the EMCV IRES, similar to translation initiation from the mRNA cap, supports the NMD of newly synthesized CBP80/20‐bound mRNA but not detectably its remodelled product, eIF4E‐bound mRNA.
EMCV IRES: no detectable NMD of eIF4E‐bound mRNA
In a second approach to determine whether NMD that is supported by EMCV IRES‐initiated translation targets CBP80/20‐bound mRNA or eIF4E‐bound mRNA or both, the extent to which a PTC reduces the level of CBP80/20‐bound hp‐EMCV IRES‐Gl‐RLuc mRNA compared with eIF4E‐bound hp‐EMCV IRES‐Gl‐RLuc mRNA was assessed. CBP80/20‐bound mRNA and eIF4E‐bound mRNA are distinguishable by immunoprecipitation using anti‐CBP80 and anti‐eIF4E, respectively (Ishigaki et al, 2001; Lejeune et al, 2002; Chiu et al, 2004; Hosoda et al, 2005; Matsuda et al, 2007). If the NMD of hp‐EMCV IRES‐Gl‐RLuc Ter mRNA is restricted to CBP80/20‐bound mRNA, then a PTC should reduce the levels of CBP80/20‐bound hp‐EMCV IRES‐RLuc‐Gl Ter mRNA and eIF4E‐bound hp‐EMCV IRES‐RLuc‐Gl Ter mRNA to a comparable percentage of the corresponding Norm mRNA. If NMD targets both CBP80‐bound and eIF4E‐bound hp‐EMCV IRES‐RLuc‐Gl Ter mRNA, then the level of eIF4E‐bound mRNA should be reduced below the level of CBP80/20‐bound mRNA when each is presented as a percentage of the corresponding Norm mRNA. Cells were transfected with pRLuc‐Gl or php‐EMCV IRES‐RLuc‐Gl test plasmid, either Norm or Ter, and pMUP. After 2 days, cells were collected, and protein and RNA were prepared before and after immunoprecipitation using anti‐CBP80, anti‐eIF4E or, to control for nonspecific immunoprecipitation, normal rabbit serum or mouse IgG, respectively.
Western blotting showed that CBP80 but not eIF4E was immunoprecipitated using anti‐CBP80 but not normal rabbit serum (Fig 3A, upper panel), and eIF4E but not CBP80 was immunoprecipitated using anti‐eIF4E but not mouse IgG (Fig 3B, upper panel). RT–PCR of samples after immunoprecipitation showed that the level of CBP80/20‐bound RLuc‐Gl Ter mRNA was 12±4% the level of CBP80/20‐bound RLuc‐Gl Norm mRNA, and the level of eIF4E‐bound RLuc‐Gl Ter mRNA was 13±4% the level of eIF4E‐bound RLuc‐Gl Norm mRNA, which were comparable with levels before immunoprecipitation (Fig 3A,B, lower panels). This is consistent with our previous finding that NMD resulting from cap‐dependent translation initiation targets CBP80/20‐bound mRNA. RT–PCR also showed that the level of CBP80/20‐bound hp‐EMCV IRES‐RLuc‐Gl Ter mRNA was 20±4% the level of CBP80/20‐bound hp‐EMCV IRES‐RLuc‐Gl Norm mRNA, and the level of eIF4E‐bound hp‐EMCV IRES‐RLuc‐Gl Ter mRNA was 23±5% the level of eIF4E‐bound hp‐EMCV IRES‐RLuc‐Gl Norm mRNA (Fig 3A,B, lower panels). These data, similar to the results obtained by blocking eIF4E‐bound mRNA translation using 4E‐BP1, indicate that NMD triggered by EMCV IRES‐initiated translation is restricted to CBP80/20‐bound mRNA and does not detectably target eIF4E‐bound mRNA.
NMD also results when translation of a PTC‐containing mRNA initiates using the type I poliovirus IRES (Wang et al, 2002). We suggest that NMD resulting from IRES‐mediated translation initiation would generally fail to detectably target eIF4E‐bound mRNA. An exception is the cricket paralysis virus IRES, for reasons unrelated to the nature of the cap‐binding protein (Isken et al, 2008). Thus, although translation initiation from the mRNA cap is not essential for NMD, NMD seems to involve the presence of CBP80 and CBP20, the latter of which binds directly to both the 5′ cap and CBP80 (Calero et al, 2002; Mazza et al, 2002). As noted above, although eIF4E‐bound mRNA can be engineered to be a target of NMD by, for example, tethering Upf1 downstream from a nonsense codon (Hosoda et al, 2005), the usual limitation of NMD to CBP80/20‐bound mRNA might reflect (i) the ability of CBP80 to interact directly with Upf1 and promote the interaction of Upf1 with Upf2 (Hosoda et al, 2005) and (ii) the detectable presence of EJCs and, thus, Upf2 and Upf3 or Upf3X on CBP80‐bound mRNA but not on eIF4E‐bound mRNA (Lejeune et al, 2002; Kashima et al, 2006).
Plasmid constructions. See the supplementary information online.
Cell transfections. HeLa CCL2 and Cos‐7 cells were cultured in Dulbecco's modified Eagle's medium (Invitrogen, Carlsbad, CA, USA) supplemented with 10% fetal bovine serum (Invitrogen). Cos cells (3–4 × 106 per 60‐mm dish) were transiently transfected with 4.0 μg of pACTAG2‐HA3 or pACTAG2‐HA3‐4E‐BP1 together with 0.4 μg of a pRLuc‐Gl test plasmid or 0.4 μg of a php‐EMCV IRES‐RLuc‐Gl test plasmid and 0.1 μg of the pMUP reference plasmid. When immunoprecipitations were performed, HeLa cells (1 × 107 per 150‐mm dish) were transfected with 10 μg of a pRLuc‐Gl test plasmid or 10 μg of a php‐EMCV IRES‐RLuc‐Gl‐RLuc test plasmid and 2 μg of pMUP by using Lipofectamine 2000.
In experiments that used siRNA, HeLa cells (3–4 × 106 per 60‐mm dish) were cultured as described above and transiently transfected with 100 nM of the specified in vitro‐synthesized siRNA, 1.0 μg of a pRLuc‐Gl test plasmid or 1.0 μg of a php‐EMCV IRES‐RLuc‐Gl test plasmid and 0.3 μg of pMUP. Upf1, Upf2 and Upf3X siRNAs were as described previously (Kim et al, 2005). CBP80 and CBP20 siRNAs were, respectively, 5′‐r(GCUGAUCUUCCUAACUACA)d(TT)‐3′ and 5′‐r(GGGAACCUCUCUAAAUAAUUU)d(TT)‐3′.
Immunoprecipitations. Immunoprecipitations were performed as described previously (Ishigaki et al, 2001).
RNA and protein isolation. Total‐cell RNA and protein were isolated using Trizol reagent (Invitrogen) and passive lysis buffer (Promega, Madison, WI, USA), respectively (Ishigaki et al, 2001).
Western blot analyses, luciferase activity assays and RT–PCR. See the supplementary information online.
Supplementary information is available at EMBO reports online (http://www.emboreports.org).
Conflict of Interest
The authors declare that they have no conflict of interest.
We thank E. Izaurralde for anti‐CBP80, H. Baumann and B. Held for anti‐MUP and N. Sonenberg for anti‐4E‐BP1 and the pACTAG2 plasmids. This work was supported by National Institutes of Health Grant GM074593 to L.E.M.
- Copyright © 2008 European Molecular Biology Organization