Distant N- and C-terminal Domains Are Required for Intrinsic Kinase Activity of SMG-1, a Critical Component of Nonsense-mediated mRNA Decay*

Phosphatidylinositol 3-kinase-related kinases (PIKKs) consisting of SMG-1, ATM, ATR, DNA-PKcs, and mTOR are a family of proteins involved in the surveillance of gene expression in eukaryotic cells. They are involved in mechanisms responsible for genome stability, mRNA quality, and translation. They share a large N-terminal domain and a C-terminal FATC domain in addition to the unique serine/threonine protein kinase (PIKK) domain that is different from classical protein kinases. However, structure-function relationships of PIKKs remain unclear. Here we have focused on one of the PIKK members, SMG-1, which is involved in RNA surveillance, termed nonsense-mediated mRNA decay (NMD), to analyze the roles of conserved and SMG-1-specific sequences on the intrinsic kinase activity. Analyses of sets of point and deletion mutants of SMG-1 in a purified system and intact cells revealed that the long N-terminal region and the conserved leucine in the FATC domain were essential for SMG-1 kinase activity. However, the conserved tryptophan in the TOR SMG-1 (TS) homology domain and the FATC domain was not. In addition, the long insertion region between PIKK and FATC domains was not essential for SMG-1 kinase activity. These results indicated an unexpected feature of SMG-1, i.e. that distantly located N- and C-terminal sequences were essential for the intrinsic kinase activity.

SMG-1 was first identified as a suppressor of morphogenetic effect on genitalia-1 in Caenorhabditis elegans. It is one of the critical components of the RNA surveillance pathway, termed nonsense-mediated mRNA decay (NMD), 4 which is conserved from worm to human (1)(2)(3)(4). NMD mediates rapid degradation of mRNAs bearing premature termination codons generated by genome mutations and by errors that occur during processes including transcription and splicing (5). NMD removes aberrant mRNAs containing premature termination codons from cells, thereby protecting them from accumulation of non-functional or potentially harmful truncated proteins encoded by aberrant mRNAs. In mammals, SMG-1 directly phosphorylates Upf1, a central component of NMD, at the serine/threonineglutamine-rich ((S/T)-Q-rich) motif in the C terminus of Upf1, followed by dephosphorylation by protein phosphatase 2A (PP2A) (3,6,7). This phosphorylation/dephosphorylation cycle of Upf1 is essential to promoting NMD. In addition to RNA surveillance, a recent study suggested the possible involvement of SMG-1 in genome surveillance in mammalian cells (8).
In this study, we investigated the structure-activity relationships of human SMG-1. For this purpose, we constructed a variety of point and deletion mutants and evaluated their kinase activities both in vitro and in vivo. Our analysis revealed that the FATC domain was important for SMG-1 kinase activity, and the essential amino acid residue for SMG-1 kinase activity was different from that of mTOR, but another conserved amino acid was required. Surprisingly, the N terminus of SMG-1, which is far from the catalytic domain, strongly affected SMG-1 kinase activity. On the other hand, the long insertion region between the PIKK domain and the FATC domain was not critical for SMG-1 kinase activity. Taken together, this first comprehensive analysis of relationships of kinase activity and structural features of SMG-1 suggested that SMG-1 activity was provided through both the N-terminal region and FATC domains.
Cell Culture and Transfection-HEK293T cells and HeLa TetOff (Clontech) cells were cultured in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum and antibiotics and were incubated at 37°C with 5% CO 2 . For the in vitro kinase assay experiments, HEK293T cells (2 ϫ 10 6 cells for the in vitro kinase assay and 1 ϫ 10 7 cells for the kinase amountdependent in vitro kinase assay) were transfected with the indicated FLAG-SMG-1 mutant plasmids using Polyfect (Qiagen) or FuGENE 6 (Roche Applied Science), and cells were collected after 36 -48 h.
Depletion of endogenous proteins by RNA interference was carried out by using siRNAs as described by Usuki et al. (21), except the transfection reagents. The following target sequences were used: non-silencing (5Ј-AAUUCUCCGAAC-GUGUCACGU-3Ј) and SMG-1 3Ј UTR (5Ј-GGAAGAUUUG-AUGCAUUCATTЈ-3Ј). Non-silencing siRNAs that were guaranteed to show no gene silencing activity were commercially prepared by Qiagen. HeLa TetOff cells were transfected with small interfering (siRNA) duplexes (as above) at a final concentration of 70 nM using RNAifect (Qiagen). At 36 h post-transfection, the cells were trypsinized and reseeded in 12-well plates. After a further 24 h, the cells were transfected with the same siRNAs together with 1 g of the indicated FLAG-SMG-1 constructs using Lipofectamine 2000 (Invitrogen). The cells were harvested 48 h after the second transfection.
Protein Detection and Quantification-Samples were separated by 5.5-12% SDS-PAGE. Western blotting was performed using the indicated antibodies with a standard ECL system (GE Biosciences). Anti-SMG-1, Upf1, phospho-Upf1 (3B8), SMG-7 antibodies were generated as done previously (3,6), and the anti-FLAG M2 antibody was purchased from Sigma. Coomassie Brilliant Blue (CBB) staining was performed according to standard procedures. To measure the amounts of immunopurified mutant proteins from ϳ150 to 430 kDa, silver staining was used, because Western blotting efficiency was largely different for each molecular weight of mutants (supplemental Fig.  2). Silver staining was performed using the Silver Quest silver staining kit (Invitrogen) following the manufacturer's protocols. Signals were scanned by LAS 3000 (Fuji film) and quantified by MultiGauge (Fuji film). Phosphoproteins were detected by autoradiography using BAS 2500 (Fuji film), and signals were quantified as described above.
In Vitro Kinase Assay-FLAG-SMG-1 proteins were immunopurified from transiently transfected HEK293T cells. Cells were lysed in lysis buffer F (20 mM Tris-HCl at pH 7.5, 0.25 M sucrose, 1.2 mM EGTA, 20 mM ␤-mercaptoethanol, 1 mM sodium orthovanadate, 1 mM sodium pyrophosphate, 1 mM NaF, 1% Triton X-100, 0.5% Nonidet P-40, 150 mM NaCl, 100 nM okadaic acid (Calbiochem), and protease inhibitor mixture (Sigma)), and FLAG-SMG-1 proteins were collected on anti-FLAG affinity gels (Sigma). Immunoprecipitates were washed four times in lysis buffer F and twice in kinase reaction buffer (10 mM Hepes-KOH at pH 7.5, 50 mM ␤-glycerophosphate, 50 mM NaCl, 1 mM dithiothreitol, and 10 mM MnCl 2 ). Kinase reactions were performed for 15 min at 30°C in 50-l kinase reaction mixtures (kinase reaction buffer containing 5 g of substrate, 5 M ATP, and 5 Ci of [␥-32 P]ATP). As substrate, GST or GST fusion peptides containing Upf1 serine 1078 (3) were used. To measure the kinase amount shown in Fig. 1, immunopurified FLAG-tagged mutants were eluted twice with 50 l of kinase reaction buffer containing 1 mg/ml FLAG peptide for 30 min at 4°C. The eluted proteins and BSA (as control) were separated by 7% SDS-PAGE and visualized by silver staining. Amounts of eluted proteins were calculated from a BSA standard curve, and eluted proteins were used in an in vitro kinase assay, as described above.
To detect phosphorylated substrate, the reaction products were separated by 12% polyacrylamide gels, and the gels were stained with CBB. The gels were then dried, phosphoproteins were detected by autoradiography, and signals were quantified as described above. To measure the amount of SMG-1 proteins, the same reaction products were separated by 5.5% polyacrylamide gels, and Western blotting (Figs. 2 and 4B) or silver staining (Figs. 1, 3, and 4A) was performed to detect SMG-1 proteins. After that, the membranes or stained gels were subjected to autoradiography to measure autophosphorylation. The kinase activity or the autophosphorylation levels of SMG-1 proteins in Figs. 2-4 was corrected by phosphorylated signals of substrates or SMG-1 proteins themselves divided by the number of moles of SMG-1 proteins measured as descried above. Statistical analysis was performed by using Student's t test.

RESULTS
The FATC Domain Is Essential for SMG-1 Kinase Activity-To evaluate the intrinsic kinase activity of SMG-1, we immunopurified recombinant SMG-1 from HEK293T cells transfected with a SMG-1 expression construct, and in vitro kinase activity was evaluated using GST-Upf1 1072-1085 (3) as a substrate. Previous reports showed that SMG-1 associated with other NMD-related proteins such as Upf1 and SMG-7 (3, 7), but immunopurified recombinant SMG-1 did not show any signs of contaminations in silver staining (Fig. 1B, lanes 6 -10) and Western blotting (data not shown), indicating the high purity of the enzyme. As shown in Fig. 1, C and D, substrate phosphorylation depended on the amount of FLAG-SMG-1 WT. Furthermore, the fact that FLAG-SMG-1 D2331A, a kinase inactive point mutant in the PIKK catalytic domain (3), did not have any detectable kinase activity supported the validity of the assay. Similar results were also obtained when phosphate incorporation into SMG-1 was evaluated (data not shown). To simplify the assay system, we also performed a kinase assay without elution of FLAG-SMG-1 proteins (see "Experimental Procedures") ( Fig. 2).
In PIKKs, several point mutations in the region other than the PIKK catalytic domain affect their kinase activity (14 -16,22). Therefore, we first tested point mutants in which conserved amino acids between SMG-1 and PIKKs were changed using the above kinase assay system. We constructed the W1962F point mutant in which tryptophan at residue 1962 in the TS domain was replaced with phenylalanine. The corresponding residue in mTOR (Trp-2027) was critical for kinase activity (supplemental Fig. 1A) (22); however, FLAG-SMG-1 W1962F retained kinase activity and autophosphorylation, and kinase activity was ϳ50% compared with FLAG-SMG-1 WT (Fig. 1, C and D, and Fig. 2). This suggested that tryptophan 1962 in the TS domain was not critical for SMG-1 kinase activity, which was different from the results for mTOR.
The next set of point mutants that we used were point mutants (L3646A and W3653F) in the FATC domain, in which conserved leucine or tryptophan between FATC domains in PIKKs were substituted with alanine or phenylalanine essential for the basal/regulatory kinase activity of mTOR (16)/ATM (17) (supplemental Fig. 1A). The L3646A mutant showed greatly reduced kinase activity (7.9% compared with WT). On the other hand, the W3653F mutant retained significant activity (50% compared with WT) (Fig. 1, C and D, and Fig. 2B, black bar). Importantly, similar results were obtained with a different substrate (GST-p53 11-30 ) (data not shown), and with SMG-1 autophosphorylation (Fig. 2B, white bar.) These results suggested that mutation at the conserved FATC domain strongly affected SMG-1 activity and that the critical amino acid for kinase activity in SMG-1 was different from those of mTOR and ATM. Because data for kinase activity with or without the elu-tion step were almost similar to each other (Figs. 1D and 2B), we thereafter performed a kinase assay without an elution step.
The N-terminal End of SMG-1 Is Required for the Interaction with SMG-7-To evaluate the role of the N-terminal region of SMG-1, we next examined the interaction between SMG-1 and its known partner proteins Upf1, Upf2, and SMG-7 (7). Upf1 can independently and directly interact with both the N-terminal half (amino acids 1-2223) and C-terminal half (amino acids 2068 -3657) of SMG-1, whereas Upf2 can interact directly with the C-terminal half of SMG-1 (7). Upf2 connects the link between the SMG-1 and exon junction complex, which are required for phosphorylation of Upf-1 in vivo. On the other hand, SMG-7, which recruits the SMG-5-SMG-7-Upf3aS complex to phosphorylated Upf-1, can interact with the N-terminal half of SMG-1 (7). As for Upf1 and Upf2, we have overexpressed them with SMG-1 in HEK293T cells and evaluated the interaction by immunoprecipitation. As shown in supplemental Fig. 3, all of the mutants showed the ability to interact with Upf1 and Upf2. These results and our previous study (7) suggested that both the N-and C-terminal regions of SMG-1 are involved in the interaction with Upf1, and the C-terminal region (except the insertion region) might be important for interaction with Upf2 in vivo.

DISCUSSION
Analyses of kinase activity of SMG-1 mutants in vitro and in vivo revealed the specific structural features of SMG-1 and the common conserved features for PIKKs. First, we found that a single point mutation (L3646A) in the FATC domain strongly affected SMG-1 kinase activity (Fig. 1, C  and D, and Fig. 2), suggesting that the FATC domain probably locates close to the catalytic domain spatially. Furthermore, the fact that the large insertion region between the PIKK domain and the FATC domain was not critical for SMG-1 kinase activity in vitro (Fig. 3) may indicate that the large insertion region is not critical for the orientation between the PIKK and the FATC domains. This is consistent with the fact that lengths of insertion regions of SMG-1 are not conserved among species (supplemental Fig. 1A) (1,3,20). The substrate-induced conformational change between the PIKK and FATC domains was also suggested for the case of DNA-PKcs based on the electron microscopy observations (23,24). Other reports show that the FATC domain is necessary for basal kinase activity in mTOR and DNA-PKcs, although not in ATM (15)(16)(17). A drastic effect of the single amino acid mutation in the FATC domain of SMG-1 on an intrinsic kinase activity of SMG-1 suggests that the role of the FATC domain for SMG-1 kinase activity is similar to those for mTOR and DNA-PKcs. Among the conserved amino acids in the FATC domains of PIKKs, Leu-3646 was essential for SMG-1 kinase activity, but Trp-3653 was not ( Figs.  1 and 2). The L3646A mutation of SMG-1 might disrupt the folding of the FATC domain itself, because the corresponding leucine in yeast TOR1 (Leu-2459) is required to make the correct folding of the FATC domain together with other hydrophobic and aromatic residues (25). The essential role of corresponding leucine in ATM (Leu-3045) for its ionized radiation-inducing activation (17) suggests that the possible contribution of Leu-3646 resides in the regulation of SMG-1 kinase activity by extracellular stimulation, such as ionized radiation (8), besides basal kinase activity. Another conserved amino acid   N-terminal region of SMG-1 is essential for kinase activity. A, kinase activities of FLAG-SMG-1 sequential N-terminal deletion mutants. The left panel is a schematic structure of deletion mutants. The right panel represents the kinase activity levels of deletion mutants from an in vitro kinase assay, as described in the legend to Fig. 2. Recombinant SMG-1 levels were determined by silver staining, and substrates were detected by autoradiography followed by CBB staining. Values represent mean Ϯ S.D. from three independent trials. B, kinase activities of FLAG-SMG-1 mutants that lack short amino acid sequences in the N-terminal region. The left panel is a schematic structure of deletion mutants, and the right panel represents the kinase activity levels of deletion mutants from an in vitro kinase assay, as described above. Recombinant SMG-1 levels were determined by immunoblotting, and substrates were detected as described above. The black bars indicate the kinase activity of SMG-1 proteins for GST-Upf1 1072-1085 , and the white bars indicate autophosphorylation levels of SMG-1 proteins. Values represent mean Ϯ S.D. from three independent trials. aa, amino acid(s). (Trp-1962) in the TS domain weakly affected SMG-1 kinase activity (Fig. 2), whereas the corresponding amino acid in mTOR, i.e. Trp-2027, was critical for kinase activity (22). There is no evidence that SMG-1 interacts with FKBP12-rapamycinbinding complex, which interacts with mTOR in the TS domain; however, the TS domain might regulate SMG-1 kinase activity similar to mTOR by binding to other unknown proteins or lipids.
Second, we showed that N-terminal sequences were required for the kinase activity of SMG-1. The ⌬1-617 mutant, which lacks most of OCR1, showed significantly reduced kinase activity compared with WT (Fig. 4A). Although the ⌬1-617 mutant, which contains all sequences of the isoform as reported by Denning et al. and Brumbaugh et al. (amino acids 626 -3657; 340 kDa) (2,8), showed detectable kinase activity, its autophosphorylation activity was only 4.5% of WT, which was significantly higher than that of D2331A (1.5% of WT) (p Ͻ 0.005, Student's t test) (Fig. 4A, white bar). We could not detect any signals of endogenous SMG-1 except for 430-kDa (Fig. 5 A, black arrow-head) and 400-kDa (Fig. 5 A, white arrowhead) bands, using our antibody against four independent regions of SMG-1 in any cell lines and tissues we tested (3), and the ⌬1-617 band was significantly smaller than 400 kDa (Fig. 5A). Thus, we suggest that the isoform reported previously is not the 400-kDa isoform of SMG-1 and is a minor isoform that possesses little kinase activity, if any exists. However, we did not exclude the possibility that the splice variant might be expressed in some specific cell phases and/or after stimulation.
Detailed truncation and internal deletion analyses of the N-terminal region of SMG-1 revealed that the whole N-terminal region was required for SMG-1 kinase activity. N-terminal halves of PIKKs comprise tandem helical repeats (supplemental Fig. 1B) (18,19), and these helical repeats mediate interactions with distinct binding partners (26 -28). However, there is little evidence whether this region is involved in the intrinsic kinase activity of PIKKs. N-terminal truncation analyses of mTOR and ATM showed that deletion of Ͼ1000 amino acids far from the catalytic domain, containing tandem helical repeats, did not seriously affect kinase activity (22,29). However, in SMG-1, a small deletion in the N-terminal region strongly affected kinase activity in vitro and in vivo (Fig. 4). Because immunopurified SMG-1 proteins used for the kinase assay showed no signs of contamination by other proteins (Fig. 1B), the cause of inactivation observed for many of the SMG-1 mutants did not seem to involve failure in association with unknown accessory proteins. Therefore, we suggest that the N-terminal region of SMG-1 is required for maintaining the proper conformation of the catalytic PIKK domain. Note that all of the mutants tested, except for ⌬1-152 and ⌬1-617, could associate with known direct partner proteins Upf1, Upf2, and SMG-7, supporting the idea that they retain the ability to interact with these proteins.
Three-dimensional analyses of DNA-PKcs also revealed that the N-terminal regions of DNA-PKcs were located near the  C-terminal region containing the PIKK domain and the FATC domain (23,24,30), and this conformation changed in response to binding to the DNA and Ku subunit. In addition, the x-ray crystal structure of the catalytic subunit of the class IB ␥ isoform of PI3K (PI3K␥), whose catalytic domain is similar to PIKK domain, revealed that the helical domain akin to the HEAT repeat provides a core scaffold, the surface of which interacts with other domains including catalytic domain (31,32). In this case, such a folding manner makes an N-terminal Ras-binding domain, which is distantly positioned from the catalytic domain in the amino acid sequence position, a close contact to the catalytic domain, enabling a Ras-induced PI3K activation. A mutation of the helical domain of the class Ia ␣ isoform of PI3K (PI3K␣), the structure of which is similar to PI3K␥, affects the kinase activity of PI3K␣ (33). Taken together with our biochemical study and the three-dimensional analyses of DNA-PKcs and PI3K, we suggest that the N-terminal region of SMG-1 would locate near the catalytic domain, and such specific conformation might be required for an intrinsic kinase activity of SMG-1. The catalytic domains of most protein kinases transit between an active closed state and an inactive open state, which are formed by the N-and C-terminal lobes of the catalytic domain (34 -36). We suggest that the N-terminal region of SMG-1 would modulate the folding structure of the catalytic domain. Our findings that the N-terminal region is required for SMG-7 and presumably for Upf1 bindings support the notion that the large N-terminal domain forms a large scaffold for other accessory proteins and regulates the catalytic activity of SMG-1. SMG-7 is required for recruiting the SMG-5-SMG-7-PP2A complex to the phosphorylated Upf1 (6). Our recent study has also revealed that SMG-7 can interact with SMG-1-Upf1 complex in the absence of the phosphorylation of Upf1 (7). Thus, it is possible that SMG-7 can have dual activities, recruitment of PP2A phosphatase complex to phosphorylated Upf1, and suppression of SMG-1 catalytic activity through the interaction with the N-terminal domain.
A recent study (37) reports that the N-terminal region of mTOR is required for multimerization in response to nutrients. We also observed multimerization of SMG-1 via both the Nand C-terminal regions (supplemental Fig. 4, A and C). However, this multimerization became undetectable in the presence of 1% Triton X-100 (supplemental Fig. 4B), a similar condition for our immunopurification and kinase assays. Thus, multimerization may not be related to SMG-1 kinase activity as far as our assay conditions were concerned, although we could not exclude the possibility that multimerization might regulate SMG-1 kinase activity or affinity to other proteins in vivo. Although autophosphorylations of ATM and DNA-PKcs have been reported to be involved in regulation of their activities (9,38,39), a role of autophosphorylation of SMG-1 observed in this study still remains unclear, because accelerated autophosporylation of ⌬2513-3490 did not significantly affect the kinase activity.
We found that all mutants tested for Upf1 phosphorylation in vivo showed kinase activity in vitro (Fig. 5). We could not compare the kinase activity between in vivo and in vitro accurately, because the amount of SMG-1 mutants in Fig. 5 could not be measured accurately due to the difference of molecular weight, although purified SMG-1 proteins could be measured in an in vitro kinase assay (Figs. 1-4). However, the mutants, which possessed kinase activity in vitro, retained the ability to phosphorylate Upf1 more than the kinase-inactive mutant did. We failed to obtain SMG-1 mutants with intrinsic kinase activity but without interaction with other essential components. For example, Upf2 binds to the C-terminal region of SMG-1 (7), and our previous study established that in vivo phosphorylation of Upf1 required interaction between SMG-1 and Upf2 (7). However, all mutants retained interaction with Upf2 (supplemental Fig. 3C).
In conclusion, our extensive mutational analysis of SMG-1 revealed a novel aspect into nature of the basal kinase activity of SMG-1, the newest member of the PIKK family. Characterization of SMG-1 kinase activity will provide important clues for understanding the regulation of SMG-1 kinase activity and mRNA surveillance.