|Home | About | Journals | Submit | Contact Us | Français|
Nonsense-mediated mRNA decay (NMD) is an mRNA surveillance mechanism that in mammals generally occurs upon recognition of a premature termination codon (PTC) during a pioneer round of translation. This round involves newly synthesized mRNA that is bound at its 5’ end by the cap-binding protein (CBP) heterodimer CBP80-CBP20. Here, we show that precluding the binding of the NMD factor UPF1 to CBP80 inhibits NMD at two steps: the association of SMG1 and UPF1 with the two eukaryotic release factors (eRFs) during SURF complex formation at a PTC, and the subsequent association of SMG1 and UPF1 with an exon-junction complex. We also demonstrate that UPF1 binds PTC-containing mRNA more efficiently than the corresponding PTC-free mRNA in a way that is promoted by the UPF1-CBP80 interaction. A unifying model proposes a choreographed series of protein-protein interactions occurring on an NMD target.
Nonsense-mediated mRNA decay (NMD) is an mRNA surveillance mechanism that ensures the quality of gene expression (for recent reviews, see Isken and Maquat, 2008; Mühlemann and Lykke-Andersen, 2010; Rebbapragada and Lykke-Andersen, 2009; Shyu et al., 2008; Silva and Romão, 2009). By degrading mRNAs that prematurely terminate translation, NMD prevents the production of potentially deleterious truncated proteins and, as a consequence, human diseases. While NMD is also used by mammalian cells to properly regulate gene expression, many naturally occurring NMD targets are thought to derive from nonproductive alternative splicing (McGlincy and Smith, 2008).
Mammalian-cell NMD generally occurs when translation terminates more than ~50–55 nucleotides upstream of a post-splicing exon-exon junction during a pioneer round of translation. The pioneer round utilizes newly synthesized mRNA that is bound by the cap-binding protein heterodimer CBP80-CBP20 (CBC) and, depending on the translation initiation efficiency, more than one ribosome (Isken and Maquat, 2008). The role of an exon-exon junction during NMD is mediated by an exon-junction complex (EJC) of proteins that is deposited ~20–25 nucleotides upstream of the junction (Tange et al., 2004). Ultimately, CBC-bound mRNA is remodeled to mRNA that (i) is bound at its cap by eukaryotic translation initiation factor (eIF)4E instead of CBC, (ii) supports the bulk of cellular protein synthesis, and (iii) is not detectably targeted for NMD (Chiu et al., 2004; Ishigaki et al., 2001; Matsuda et al., 2007; Sato and Maquat, 2009; Woeller et al., 2008). NMD appears to be restricted to newly synthesized CBC-bound mRNA for at least three reasons. First, EJCs typify CBC-bound mRNA but not detectably eIF4E-bound mRNA (Kashima et al., 2006; Lejeune et al., 2002) since the majority of EJCs are removed during the pioneer round of translation (Dostie and Dreyfuss, 2002; Gehring et al., 2009; Sato and Maquat, 2009). Second, CBP80, unlike eIF4E, interacts with the UPF and SMG NMD factors either directly or indirectly, depending on the factor (Hosoda et al., 2005; Ishigaki et al., 2001; Kashima et al., 2006; Lejeune et al., 2002). Third, CBP80 enhances the efficiency of NMD by promoting the interaction of UPF1 and UPF2 (Hosoda et al., 2005).
Besides UPF1 and UPF2, NMD in mammalian cells involves UPF3 or UPF3X (also called, respectively, UPF3a and UPF3b), six SMG factors (Yamashita et al., 2009 and references therein), and hNAG and the DEAH-box protein hDHX34 (Longman et al., 2007). UPF3 and UPF3X direct NMD with different efficiencies, and one or the other appears to be a stable constituent of many if not most EJCs, as is UPF2 (Chan et al., 2007, 2009; Gehring et al., 2005; Gehring et al., 2003; Kunz et al., 2006; Lejeune et al., 2002). NMD appears to be triggered when translation terminates sufficiently upstream of an EJC (Kashima et al., 2006; Yamashita et al., 2009), i.e., so that the EJC is not removed by translating ribosomes (Dostie and Dreyfuss, 2002; Sato et al., 2009). Under such circumstances, the termination codon, which is often a premature termination codon (PTC), is recognized by the SURF complex that consists of the phosphoinositide 3-kinase-related protein kinase SMG1, UPF1, and translation termination factors eRF1 and eRF3 (Kashima et al., 2006). It has been proposed that the SMG1 and UPF1 constituents of SURF subsequently bind the PTC-distal EJC (Kashima et al., 2006), possibly via SURF bridging the PTC and EJC (Yamashita et al., 2009). Binding or bridging is thought to activate the SMG1-mediated phosphorylation of UPF1 that ultimately leads to translational repression and mRNA decay (Isken et al., 2008; Kashima et al., 2006; Wittmann et al., 2006; Yamashita et al., 2009).
In this communication, we examine the association of UPF1 with newly synthesized mRNP in view of two important findings. First, CBP80 of the pioneer translation initiation complex interacts directly with UPF1 and promotes NMD by augmenting the interaction of UPF1 with UPF2 – presumably EJC-bound UPF2 (Hosoda et al., 2005). Second, the co-immunoprecipitation (IP) of UPF1 and eRF1 and eRF3 is enhanced when the EJC component UPF2 or Y14 is downregulated, indicating that SURF complex formation precedes UPF1 joining to an EJC (Kashima et al., 2006; H.S. and L.E.M., unpub. data). Our goal was to investigate the molecular gymnastics of SMG1, UPF1, eRF1 and eRF3 on mRNP during the pioneer round of translation and NMD.
We first demonstrate that the interaction between cellular UPF1 and CBP80 is stabilized by cross-linking and partially susceptible to RNase A, consistent with the interaction being direct some of the time. We then map the regions of UPF1 and CBP80 that associate. With this information, we define a peptide from each protein that data indicate specifically inhibits the co-IP of UPF1 and CBP80 when expressed in HeLa cells. Using each interfering peptide, we provide evidence that inhibiting the association of UPF1 and CBP80 inhibits NMD not only by preventing the joining of UPF1 – notably together with SMG1 but not detectably with eRF1 or eRF3 – to an EJC but also by inhibiting the preceding step of SURF complex formation at a PTC. These results indicate that only SMG1 and UPF1 of SURF join an EJC. Moreover, we demonstrate that UPF1 binds PTC-containing mRNA an order of magnitude more efficiently than it binds PTC-free mRNA in a way that is promoted by the interaction of UPF1 and CBP80. UPF1 binding to PTC-containing mRNA is largely localized to the mRNA half that contains the PTC and downstream EJC. These and other findings indicate that the CBC of newly synthesized mRNAs associates with UPF1 during the pioneer round of translation so as to promote the decay of an NMD target at minimally two steps.
UPF1 is a key NMD factor that has been shown to interact directly with CBP80 using purified proteins and Far-western analysis (Hosoda et al., 2005). To corroborate that UPF1 binds CBP80 in vivo, HeLa CCL2 cells that stably express FLAG-UPF1 at less than 2-fold the level of endogenous UPF1 (Pal et al., 2001) were or, for comparison, were not cross-linked using formaldehyde prior to lysis (Gong et al., 2009). Lysates were immunoprecipitated using anti-FLAG or, to control for nonspecific immunoprecipitation (IP), mouse (m)IgG. RNase A, which degrades cellular RNA and helps to ensure that co-immunoprecipitating proteins reside within an RNase A-resistant complex, was included in half of the IPs that were performed without cross-linking. Subsequently, cross-links were undone using heat. Western blotting demonstrated that cross-linking, which would preserve any transient or weak direct interaction as well as any indirect interaction, increased the amount of CBP80 that co-immunoprecipitates with FLAG-UPF1 by ~2–3-fold (Figure 1A, upper). UPF2, SMG1 and eRF3, each of which is known to interact directly with UPF1, likewise co-immunoprecipitated with FLAG-UPF1 in a manner that was stabilized by cross-linking prior to cell lysis (Figure 1A, upper). The co-IP of CBP80, UPF2, SMG1 and eRF3 with FLAG-UPF1 was partially (20%; 60%, 80% and 75%, respectively) resistant to RNase A in the absence of cross-linking (Figure 1A, upper; see also Kashima et al., 2006; Luke et al., 2007; Schell et al., 2003), consistent with the existence of both direct and indirect interactions. RNase A treatment was effective as evidenced by the RT-PCR analysis of cellular SMG7 mRNA, which was degraded by RNase A (Figure 1A, lower), and the finding that the co-IP of PABPC1 with FLAG-UPF1 was abolished by RNase A (Figure 1A. upper). We conclude from these and previous results that UPF1 and CBP80 interact directly as well as indirectly in intact cells (see also below).
Since NMD targets CBC-bound mRNA, and the interaction of UPF1 with CBP80 promotes the association of UPF1 with UPF2 both in cells and using purified proteins (Hosoda et al., 2005), it is conceivable that the interaction of UPF1 with CBP80 functions at other steps of NMD. One way to test this hypothesis is to express a peptide of UPF1 that inhibits exclusively the interaction of UPF1 with CBP80 and subsequently assay for an effect on NMD. To define such a peptide, we generated four deletion variants of N-terminally MYC-tagged full-length UPF1, which consists of amino acids 1-1118. These variants are: MYC-UPF1(1-418) and MYC-UPF1(1-244), both of which contain the cysteine-histidine-rich (CH) domain that interacts with UPF2 and eRF3 (Chamieh et al., 2008; Gong et al., 2009; Ivanov et al., 2008); MYC-UPF1(115-914), which consists of the CH and RNA helicase domains; and MYC-UPF1(295-914), which contains the RNA helicase domain (Figure 1B).
Lysates of HeLa cells that transiently expressed one of the MYC-UPF1 proteins were analyzed before or after IP using anti-CBP80 or, to control for nonspecific IP, normal rabbit serum (NRS) under conditions where the MYC-UPF1 proteins were expressed at similar levels (Figure 1C, Before IP, α-MYC) that were only ~2-fold above the level of endogenous UPF1 (Figure 1C, Before IP, α-UPF1). Results indicated that CBP80 binds MYC-UPF1(1-1118), MYC-UPF1(115-914) and MYC-UPF1(295-914) but not MYC-UPF1(1-244) or MYC-UPF1(1-418) (Figure 1C, After IP, α-MYC) or the negative-control protein p62 (Figure 1C, mAb414). None of the MYC-UPF1 proteins was immunoprecipitated using NRS. Therefore, the smallest MYC-UPF1 deletion variant that contains the CBP80 binding site is MYC-UPF1(295-914), and it binds CBP80 as efficiently as does MYC-UPF1(1–1118).
To further narrow the region of UPF1 that interacts with CBP80, we generated and analyzed three additional deletion variants: MYC-UPF1(295–418), MYC-UPF1(419–700) and MYC-UPF1(701–914), each of which forms discrete domains of the helicase region according to structural determinations (Figure 1D; Cheng et al., 2007). Results revealed that only MYC-UPF1(295–914) and MYC-UPF1(419–700) interact with CBP80 (Figure 1E, After IP, α-MYC). Thus, the smallest MYC-UPF1 deletion variant that contains the CBP80 binding site is MYC-UPF1(419-700), and it binds CBP80 as efficiently as does MYC-UPF1(295-914) and, therefore, MYC-UPF1(1-1118).
Since MYC-UPF1(419-700) contains the CBP80 interacting domain of UPF1, expressing MYC-UPF1(419-700) in HeLa cells is expected to inhibit the interaction of cellular UPF1 with cellular CBP80. To test this prediction, lysates of HeLa cells that transiently expressed MYC-UPF1(419-700) or, as a negative control, MYC alone were immunoprecipitated using anti-CBP80 or, as a control for nonspecific IP, NRS. As expected, anti-CBP80 immunoprecipitated MYC-UPF1(419-700) (Figure 2A, α-MYC). Also as expected, MYC-UPF1(419-700) reduced the co-IP of CBP80 and UPF1 to 20% of the level observed in the presence of MYC alone (Figure 2A, α-UPF1).
Important to our goal of disrupting the interaction of CBP80 and UPF1 but not other known interactions of CBP80, MYC-UPF1(419-700) did not detectably inhibit the co-IP of CBP80 with either eIF4G or CBP20 (Figure 2A, α-eIF4GII or α-CBP20, respectively), each of which is known to directly bind CBP80 (Calero et al., 2002; Izaurralde et al., 1994; Lejeune et al., 2004; Mazza et al., 2001). Furthermore, MYC-UPF1(419-700) did not inhibit the co-IP of CBP80 with GAPDH mRNA (Figure 2A), which is consistent with the finding that MYC-UPF1(419-700) does not compromise the translation of CBC-bound mRNA (Figure S1A,B). In related experiments, MYC-UPF1(419-700), in contrast to MYC-UPF1(1-1118), failed to co-IP with UPF2, eRF1, eRF3, SMG1 and eIF3a (Figure 2B), all but eRF1 is known to bind directly to UPF1 (Chamieh et al., 2008; Isken et al., 2008; Kashima et al., 2006; Singh et al., 2007). We conclude that MYC-UPF1(419-700) should inhibit the interaction of UPF1 with CBP80 without interfering with other interactions that are known to function during NMD.
To determine the effect of MYC-UPF1(419-700) on NMD, HeLa cells were transiently transfected with pmCMV-Gl and pmCMV-GPx1 test plasmids – either pmCMV-Gl Norm and pmCMV-GPx1 Norm, both of which lack a PTC, or pmCMV-Gl Ter and pmCMV-GPx1 Ter, both of which contain a PTC at, respectively, position 39 (Zhang et al., 1998) or position 46 (Moriarty et al., 1998) – and the phCMV-MUP reference plasmid (Belgrader and Maquat, 1994). These transfections also included either pCMV-MYC or pCMV-MYC-UPF1(295-418), each of which serves as a negative control, or pCMV-MYC-UPF1(419-700). Western blotting using anti-MYC demonstrated that the MYC-UPF1 proteins were expressed at comparable levels (Figure 2C, upper, α-MYC). RT-PCR revealed that MYC-UPF1(419-700), unlike MYC or MYC-UPF1(295-418), indeed inhibited NMD: the level of Gl Ter mRNA was ~7-8% of normal in presence of MYC or MYC-UPF1(295-418) but ~73% of normal in the presence of MYC-UPF1(419-700) (Figure 2C, middle); and the level of GPx1 Ter mRNA was ~5–7% of normal in presence of MYC or MYC-UPF1(295-418) but ~45% of normal in the presence of MYC-UPF1(419-700) (Figure 2C, lower). Notably, these RT-PCR results were corroborated using RT-coupled to real-time PCR (Table S2).
If the inhibition of NMD were due to the failure of cellular UPF1 and CBP80 to interact, then expressing increasing levels of FLAG-CBP80 should restore NMD. In fact, this was observed: while the level of Gl Ter mRNA was ~5% of normal in the presence MYC regardless of the concentration of FLAG-CBP80, the ~72% of normal level in the presence of UPF1(419-700) was decreased to ~62% and ~52% of normal with increasing amounts of FLAG-CBP80 (Figure 2D; the level was not restored to ~5% of normal because cells could not be transfected with a sufficient amount of pFLAG-CBP80 without toxicity). Notably, MYC-UPF1(419-700) does not inhibit NMD by blocking mRNA export from the nucleus to the cytoplasm (Figure S1C). In summary, all results are consistent with the idea that the interaction of UPF1 with CBP80 contributes to the efficiency of NMD.
Since UPF1 functions during NMD as a constituent of the SURF complex that forms at a PTC, it is conceivable that MYC-UPF1(419-700) could inhibit NMD by preventing SURF complex assembly. To test this hypothesis, HEK 293T cells, which unlike HeLa cells readily express HA-SMG1, were transiently transfected with pcDNA3.1-HA-SMG1 (Brumbaugh et al., 2004) and either pCMV-MYC or pCMV-MYC-UPF1(419-700). Cells were lysed without cross-linking (Figure 3) or after cross-linking (Figure S2), the latter of which allows for more quantitative yet stringent analyses since co-IP efficiencies are higher after cross-linking yet likely to be devoid of interactions that form as an experimental artifact after cell lysis. Notably, HA-SMG1 was expressed at only ~0.3-fold the level of cellular SMG1 (data not shown). IPs of cell lysates were performed using anti-HA or, as a negative control, rat IgG in the presence of RNase A, which degrades cellular RNA (Figure S2A,B). When MYC was expressed, the co-IP of the HA-SMG1 with the remaining three SURF constituents was detected in the presence of RNase A (Figure 3A; Figure S2C), consistent with their constituting a complex that does not require intact RNA for detection (Kashima et al., 2006). In contrast, MYC-UPF1(419-700) expression significantly reduced the RNase A-resistant co-IP of HA-SMG1 with eRF1 and eRF3 but not with UPF1 (Figure 3A; Figure S2C). This result indicates that the interaction of UPF1 with CBP80 promotes SURF complex formation by promoting the association of SMG1 and UPF1 with eRF1-eRF3, the latter two of which form a heterodimer in cells (Funakoshi et al., 2007; Stansfield et al., 1995). Consistent with this conclusion, MYC-UPF1(419-700) expression reduced the RNase A-resistant co-IP of FLAG-UPF1 with eRF1 and eRF3 but not SMG1 (Figure 3B; Figure S2D; where mIgG served to control for nonspecific IP). Remarkably, CBP80 also co-immunoprecipitated with HA-SMG1 in the presence of MYC but not appreciably in the presence of MYC-UPF1(419-700) (Figure 3A; Figure S2C), suggesting that CBP80 physically joins SURF during PTC recognition while chaperoning SMG1-UPF1 binding to eRF1-eRF3 at the PTC.
Since studies using purified proteins indicate that CBP80 promotes the association of UPF1 with UPF2, and given that SURF complex formation at a PTC is thought to precede the binding of any SURF constituents to an EJC during NMD, it is possible that MYC-UPF1(419-700) would also prevent the RNase A-resistant co-IP of the EJC constituent Y14 with SMG1 and UPF1. In support of this, data indicated that MYC-UPF1(419-700) expression, unlike MYC expression, indeed inhibits the co-IP of Y14 with SMG1 and UPF1 but not with the UPF2 and eIF4AIII EJC constituents (Figure 3C). Since neither eRF1 nor eRF3 detectably co-immunoprecipitated with Y14 under any circumstances (Figure 3C; Figure S2E), we conclude that the only SURF constituents to detectably join the EJC after PTC recognition are SMG1 and UPF1. This conclusion is consistent with the finding that hyperphosphorylated UPF1 does not detectably co-immunoprecipitate with eRF1 or eRF3 but does co-immunoprecipitate with Y14 and UPF2 (Kashima et al, 2006). Notably, CBP80 also co-immunoprecipitated with Y14 in the presence of MYC but significantly less so in the presence of MYC-UPF1(419-700) (Figure 3C; Figure S2E), suggesting that CBP80 physically joins the EJC after PTC recognition while chaperoning SMG1-UPF1 binding to the EJC. As expected of EJC constituents, the co-IP of UPF2 and eIF4AIII with Y14 was largely resistant to RNase A treatment (Kashima et al., 2006; Lejeune et al., 2004; Yamashita et al., 2009), whereas the co-IP of PABPC1 with Y14 was not (Figure 3C; Figure S2E). We conclude that MYC-UPF1(419-700) inhibits NMD first by inhibiting SURF complex formation at a PTC – specifically, by inhibiting the CBP80-mediated binding of SMG1 and UPF1, presumably as a heterodimer, to eRF1-eRF3 – and subsequently by inhibiting the CBP80-mediated joining of SMG1-UPF1 to a downstream EJC (see Discussion).
Since inhibiting the interaction of UPF1 with CBP80 using MYC-UPF1(419-700) inhibits NMD, it follows that inhibiting this interaction by expressing a region of CBP80 that binds UPF1 should also inhibit NMD and corroborate the importance of the interaction. Thus, two deletion variants of FLAG-CBP80 were generated (Figure 4A), neither of which was predicted to disrupt the binding of CBP20 to CBP80 (Mazza et al., 2001). FLAG-CBP80(664-790) was found to co-immunoprecipitate with UPF1 but not eIF4G (Figure 4B, α-UPF1 and α-eIF4GII, respectively), and it reduced the co-IP of UPF1 with CBP80 to ~20% of normal (Figure 4C). However, FLAG-CBP80(664-790) did not inhibit the co-IP of CBP20 with either CBP80 or GAPDH mRNA (Figure 4D), consistent with its having no detectable effect on the translation of CBC-bound mRNA (Figure S1D). Therefore, the finding that FLAG-CBP80(664-790) abrogates the NMD of Gl Ter mRNA (Figure 4E) appears to be attributable specifically to the FLAG-CBP80(664-790)-mediated reduced binding of UPF1 to CBP80. As for MYC-UPF1(417-900), this failure results in an inhibition of SURF complex formation without or with cross-linking prior to cell lysis: the RNase A-resistant co-IP of HA-SMG1 with eRF1 and eRF3 but not UPF1 was inhibited (Figure 5A; Figure S2F; Figure S2A,B for RNase A analyses). The inhibition of SURF complex formation was also evident by a FLAG-CBP80(664-790)-mediated decrease in the co-IP of UPF1 with eRF1 and eRF3 but not SMG1 (Figure 5B; Figure S2G). This inhibition was accompanied by decreased SMG1 and UPF1 binding to EJCs. As in Figures 3C and S2E, our failure to detect eRF1 or eRF3 in the anti-Y14 IP (Figure 5C; Figure S2H) indicates that the two translation termination factors apparently do not move with SMG1 and UPF1 to a PTC-distal EJC.
Our data indicate that UPF1 binds PTC-containing mRNA and, possibly, PTC-free mRNA. To determine how the level of UPF1 binding to PTC-containing mRNA compares to the level of UPF1 binding to PTC-free mRNA, HeLa cells were transiently transfected with the pmCMV-Gl Norm or pmCMV-Gl Ter test plasmid and the phCMV-MUP reference plasmid. Two days later, nuclear lysates (Figure S3A), which are enriched in Gl Ter mRNA that is undergoing NMD, were generated and immunoprecipitated using anti-UPF1 or, to control for nonspecific IP, NRS. Lysates subjected to Western blotting prior to or after IP demonstrated successful and specific IPs (Figure 6A, upper). RT-PCR quantitations revealed that the level of Gl Ter mRNA prior to IP was ~9% of normal, whereas the level of UPF1-bound Gl Ter mRNA after IP was ~100% of normal (Figure 6A, middle). Thus, Gl Ter mRNA binds UPF1 ~10-fold more efficiently than does Gl Norm mRNA. Notably, UPF1 binding to Gl mRNA is not attributable to binding after cell lysis as determined by comparing lysates of cells that express both FLAG-UPF1 and Gl mRNA to a mixture of two lysates: one of cells expressing FLAG-UPF1 but not Gl mRNA, and the other of cells expressing Gl mRNA but not FLAG-UPF1 (Figure S3B). Analyses of GPx1 Ter mRNA using total-cell lysates also revealed that PTC-containing mRNA binds UPF1 ~17-fold more efficiently than does its PTC-free counterpart (Figure 6A). Since Gl and GPx1 Ter mRNAs metabolically differ from, respectively, Gl and GPx1 Norm mRNAs because of their PTC, and because PTC recognition requires the process of translation, the enhanced binding of UPF1 to PTC-containing mRNAs relative to PTC-free mRNAs must dependent on translation (see Discussion).
Our finding that UPF1 binds PTC-containing Gl mRNA more efficiently than it does its PTC-free counterpart is consistent with findings using S. cerevisiae (Johansson et al., 2007) and C. elegans (Johns et al., 2007) that UPF1 selectively “marks” mRNAs that are targeted for NMD. However, it remained to be determined for mammals if the preferential binding of UPF1 to PTC-containing mRNA is attributable to UPF1 binding to CBP80. Thus, HeLa cells were transiently transfected with pCMV-MYC or pCMV-MYC-UPF1(419-700), the pmCMV-Gl Ter test plasmid, and the phCMV-MUP reference plasmid. Nuclear lysates were immunoprecipitated with anti-UPF1 or, as a control for nonspecific IP, NRS. Using samples in which a comparable amount of UPF1 had been immunoprecipitated (Figure 6B, upper; notably, anti-UPF1 does not immunoprecipitate MYC-UPF1(419-700)), RT-PCR demonstrated that MYC-UPF1(419-700) expression decreased the efficiency of NMD by increasing the level of Gl Ter mRNA ~11-fold prior to IP (Figure 6B, middle). Considering this decrease, MYC-UPF1(419-700) expression was accompanied by a ~6-fold decrease in the level of UPF1-bound Gl Ter mRNA (Figure 6B, middle).
Since Gl Norm mRNA naturally binds a low level of UPF1 (i.e., one-tenth the level observed with Gl Ter mRNA), and this level is decreased in the presence of MYC-UPF1(419-700), it is reasonable to assume that GAPDH mRNA, which like Gl Norm mRNA lacks a PTC, also binds a low level of UPF1 that is decreased in the presence of MYC-UPF1(419-700). This assumption appears to be correct: while MYC-UPF1(419-700) expression did not affect the level of GAPDH mRNA – either CBP80-bound GAPDH mRNA (Figure 2A) or GAPDH mRNA in the nuclear fraction (Figure 6B) – it did decrease the level of UPF1-bound GAPDH mRNA ~0.7-fold (Figure 6B, lower). It follows that MYC-UPF1(419-700) expression will also slightly reduce the level of UPF1-bound MUP mRNA, which is likewise PTC-free. Thus, the MYC-UPF1(419-700)-mediated decrease in UPF1 binding to Gl Ter mRNA is a slight overestimate since it was normalized to MUP mRNA. Nevertheless, we can conclude that inhibiting the interaction of UPF1 and CBP80 significantly inhibits the binding of UPF1 to mRNA undergoing NMD.
To roughly map the region of Gl Ter mRNA that most stably associates with UPF1, the above-described transfections were repeated. However, nuclear lysates were incubated with RNase H in the absence or presence of an antisense DNA oligonucleotide that cleaves Gl mRNA upstream of the two EJCs and the PTC (Figure 6C; Sato and Maquat, 2009). Lysates were subsequently immunoprecipitated using anti-UPF1 or, to control for nonspecific IP, NRS. RT-PCR revealed that UPF1 co-immunoprecipitated with the 3’-cleavage product but very little of the 5’-cleavage product of Gl Ter mRNA in the presence of the oligonucleotide (Figure 6D). Furthermore, MYC-UPF1(419-700) expression inhibited the co-IP of UPF1 with full-length Gl Ter mRNA (compare 952 ± 58% to 197 ± 35%) and both the 5’- and 3’-cleavage products of Gl Ter mRNA (Figure 6D), consistent with dependence of UPF1 binding to the 3’-cleavage product on UPF1 binding to the 5’-cleavage product. Thus, for PTC-containing mRNA, UPF1 binds to cap-bound CBP80 significantly less efficiently than to the PTC and/or an EJC. Nevertheless, UPF1 binding to cap-bound CBP80 promotes UPF1 binding to the PTC and/or an EJC (see Discussion).
In control IPs that did not involve either MYC or MYC-UPF1(419-700) expression, CBP80 co-immunoprecipitated with the 5’-cleavage product of Gl Norm mRNA in the absence or presence of oligonucleotide but with the 3’-cleavage product only in the absence of oligonucleotide (Figure S3C) as expected since CBP80 is primarily bound to the mRNA cap structure. In contrast, HA-UPF3X immunoprecipitated with 3’-cleavage product of Gl Norm mRNA in the absence or presence of oligonucleotide but with the 5’-cleavage product only in the absence of oligonucleotide (Figure S3D) as expected since HA-UPF3X is primarily bound to EJCs. We conclude that a relatively transient or weak interaction between UPF1 and CBP80 that occurs on both PTC-free and PTC-containing mRNAs promotes the binding of SMG1-UPF1 to eRF1-eRF3 to form the SURF complex and the subsequent binding of SMG1-UPF1, presumably from SURF, to a PTC-distal EJC during NMD.
It has long been appreciated that constituents of newly synthesized mRNPs in mammalian cells assemble progressively during the various steps of mRNA biogenesis (for recent reviews see Glisovic et al., 2008; Moore and Proudfoot, 2009). Data presented here reveal previously unappreciated insights into mRNP rearrangements that occur during the process of NMD, which requires a pioneer round of translation.
Several lines of evidence indicate that UPF1 binding to the CBC of newly synthesized mRNA contributes to the process of NMD. First, inhibiting the interaction of UPF1 and CBP80 by expressing MYC-UPF1(419-700) abrogates NMD in a way that can be restored by expressing exogenous CBP80 (Figure 2). Second, inhibiting the interaction of UPF1 and CBP80 using FLAG-CBP80(664-790) also abrogates NMD (Figure 4). Third, each interfering peptide inhibits NMD by inhibiting the joining of SMG1 and UPF1 to eRF1 and eRF3 to form the SURF complex (Figures 3 and and5),5), which is critical for PTC recognition. Fourth, each interfering peptide additionally inhibits NMD by inhibiting the binding of SMG1 and UPF1 to an EJC (Figures 3 and and5).5). Fifth, inhibiting the interaction of UPF1 and CBP80 inhibits the enhanced binding of UPF1 to an NMD target (Figure 6), which is largely localized to the region of mRNA harboring the PTC and PTC-distal EJC (Figure 6). Notably, the enhanced binding of UPF1 to PTC-containing mRNA relative to PTC-free mRNA must depend on translation, which provides the mechanism that distinguishes the two mRNAs. Consistent with this idea, existing models of NMD indicate that UPF1 binding to a PTC or EJC depends on translation, although it is possible that UPF1 could transiently interact with an EJC without physiological consequence until translation terminates sufficiently upstream of the EJC.
All of our findings provide insight into the dynamics of UPF1 binding to mRNA during the process of NMD (Figure 7). We propose that UPF1 transiently and weakly binds the CBC of newly synthesized mRNA prior to or during the pioneer round of translation. Should the mRNA be targeted for NMD, SURF forms upon translation termination and before SMG1 and UPF1 associate with an EJC since the co-IP of SMG1 with UPF1, eRF1 and eRF3 increases as the abundance of intact EJCs is experimentally decreased using UPF2 or Y14 siRNA (Kashima et al., 2006; H.S. and L.E.M., unpub. data). Significantly, there must be a feature of the pioneer round of translation that nucleates SURF so as to distinguish this round from steady-state translation given that UPF1 and SMG1 do not detectably associate with eIF4E-bound mRNA (Hosoda et al., 2005; Kashima et al., 2006). We propose that this feature is CBP80.
The binding of UPF1 to an EJC, which we show here is promoted by CBP80, appears to involve a direct interaction with EJC-associated UPF2 since a UPF1 variant that fails to bind UPF2 associates abnormally efficiently with eRF1 and eRF3 (Kashima et al., 2006). Competition between eRF3 and UPF2 for binding to UPF1 has never been demonstrated even though both proteins bind to the UPF1 cysteine-histidine-rich region (Clerici et al., 2009; Gong et al., 2009; Ivanov et al., 2008). However, our data are consistent with the possibility of competition: the only SURF constituents to detectably bind to an EJC are SMG1 and UPF1 and, remarkably, they do so along with CBP80. Thus, CBP80 appears to chaperone SMG1 and UPF1 to an EJC after chaperoning SMG1 and UPF1 to PTC-associated eRFs to form SURF. UPF1 and SMG1 binding to an EJC may involve a direct interaction of not only UPF1 but also SMG1 with UPF2 based on the finding that the FLAG-tagged C-terminal domain of SMG1 purified from mammalian cells interacts with in vitro-synthesized UPF2 (Kashima et al., 2006). Furthermore, Y14 immunoprecipitates with UPF2 and SMG1 in an RNase A-insensitive fashion (Kashima et al., 2006). UPF1 and SMG1 binding to UPF2 triggers the SMG1-mediated phosphorylation of UPF1 (Kashima et al., 2006; Ohnishi et al., 2003; Wittmann et al., 2006; Figure 7). Phosphorylated UPF1 binds to eIF3 of the 43S pre-initiation complex that is poised at the translation initiation codon of the NMD target, thereby repressing further translation by inhibiting 60S ribosomal subunit joining (Isken et al., 2008). Phosphorylated UPF1 also promotes mRNA decay by recruiting mRNA degradative activities (Isken et al., 2008). mRNA is then degraded from either or both ends (Lehner and Sanderson, 2004; Lejeune et al., 2003; Lykke-Andersen, 2002; Unterholzner and Izaurralde, 2004; Yamashita et al., 2005a).
Data demonstrating that UPF1 binds PTC-containing mRNA more efficiently than its PTC-free counterpart (Figure 6) are consistent with studies of S. cerevisiae and C. elegans demonstrating that UPF1 selectively marks PTC-containing mRNAs (Johansson et al., 2007; Johns et al., 2007). Marking in C. elegans requires neither UPF1 phosphorylation or dephosphorylation factors (Johns et al., 2007), and cellular UPF1 is primarily hypophosphorylated in C. elegans (Johns et al., 2007; Page et al., 1999) as it is in mammals (Isken et al., 2008; Pal et al., 2001). The fact that NMD in S. cerevisiae and C. elegans occurs independently of an EJC and, at least in S. cerevisiae (Gao et al., 2005) if not also in C. elegans, is not restricted to CBC-bound mRNA raises questions of how the three UPF NMD factors are recruited to an NMD target to trigger mRNA decay in these organisms.
Our findings indicating that CBP80 at an mRNA 5’ end can chaperone SMG1-UPF1 to eRF1-eRF3 poised at a PTC and, subsequently, to a PTC-distal EJC add to examples of protein-protein interactions that span a surprisingly large number of nucleotides so as to choreograph steps of mRNA metabolism. These examples suggest that mRNAs fold back on themselves in more than one configuration. Of other established long-range interactions that bring together the 5’ and 3’ ends of mRNAs via protein-protein interactions, PABPC1, which associates with an mRNA 3’-poly(A) tail, is thought to also bind eIF4G at the 5’ end of the same mRNA so as to enhance translation initiation, and PABPC1 additionally binds eRF3 at the termination codon so as to enhance the efficiency of translation termination (see, e.g., Amrani et al., 2008; Ivanov et al., 2008). Somewhat analogously to the PABPC1-eIF4G interaction, stem-loop binding protein (SLBP)-interacting protein 1 appears to bridge SLBP at the 3’ end of replication-dependent histone mRNAs, which lack a poly(A) tail, and eIF4G at the mRNA 5’ end (Cakmakci et al., 2008). Additionally, there are many examples of proteins that bind the 3’ UTR of an mRNA so as to affect the recruitment of 43S preinitiation complexes at the 5’ end of that mRNA (Duncan et al., 2008 and references therein). Future studies are expected to unravel more about the causes and consequences of known and yet-to-be-appreciated mRNP dynamics.
HeLa CCL2 or HEK 293T cells were grown in Dulbecco’s Modified Eagle’s Medium supplemented with 10% fetal bovine serum and, when specified, transiently transfected with plasmid DNA using Lipofectamine 2000 (Invitrogen) (Ishigaki et al., 2001; Lejeune et al., 2003). Total-cell or nuclear lysates were prepared (Hosoda et al., 2005; Ishigaki et al., 2001). Where specified, cells were cross-linked using 1% formaldehyde (Sigma) in 1X PBS, pH 7.4 (Gong et al., 2009) prior to lysis.
Nuclear extracts were incubated with 0.1 U/ml of RNase H (New England BioLabs) and 3 mM MgCl2 in the absence or presence of 0.25 mg/ml of 5’-CACGTTCACCTT-3’ (antisense) at 30°C for 30 min (Sato and Maquat, 2009). This oligonucleotide anneals 30- to 41-nucleotides upstream of the exon 1-exon 2 junction of Gl mRNA.
Samples for the analysis of protein and RNA were generated before and after IP in the presence or absence of RNase A (Sigma) as reported (Chiu et al., 2004; Ishigaki et al., 2001; Gong et al., 2009). Notably, IPs were performed in 150 mM NaCl when cross-linking was not performed and in 500 mM NaCl when cross-linking was performed. IPs utilized anti-UPF1 (Lykke-Andersen et al., 2000) or antibodies that are described in Supplemental Information for Western blotting.
We are grateful to Michael Gleghorn for help with designing the UPF1 deletion variants; Jens Lykke-Andersen for anti-UPF1; Diane Otterness and Bob Abraham for pcDNA3.1-HA-SMG1; Elisa Izaurralde for anti-CBP80; Nahum Sonenberg for anti-eIF4AIII and anti-eIF4GII; Oliver Jean-Jean for anti-eRF1 and anti-eRF3, and Olaf Isken for comments on the manuscript. This work was supported by NIH R01 GM59614 to L.E.M. H.S. is supported in part by a Fellowship from the Japan Society for the Promotion of Science.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.