Cloning of a novel phosphatidylinositol kinase-related kinase: characterization of the human SMG-1 RNA surveillance protein.

We have cloned and characterized a new member of the phosphatidylinositol kinase (PIK)-related kinase family. This gene, which we term human SMG-1 (hSMG-1), is orthologous to Caenorhabditis elegans SMG-1, a protein that functions in nonsense-mediated mRNA decay (NMD). cDNA sequencing revealed that hSMG-1 encodes a protein of 3031 amino acids containing a conserved kinase domain, a C-terminal domain unique to the PIK-related kinases and an FKBP12-rapamycin binding-like domain similar to that found in the PIK-related kinase mTOR. Immunopurified FLAG-tagged hSMG-1 exhibits protein kinase activity as measured by autophosphorylation and phosphorylation of the generic PIK-related kinase substrate PHAS-1. hSMG-1 kinase activity is inhibited by high nanomolar concentrations of wortmannin (IC(50) = 105 nm) but is not inhibited by a FKBP12-rapamycin complex. Mutation of conserved residues within the kinase domain of hSMG-1 abolishes both autophosphorylation and substrate phosphorylation, demonstrating that hSMG-1 exhibits intrinsic protein kinase activity. hSMG-1 phosphorylates purified hUpf1 protein, a phosphoprotein that plays a critical role in NMD, at sites that are also phosphorylated in whole cells. Based on these data, we conclude that hSMG-1 is the human orthologue to C. elegans SMG-1. Our data indicate that hSMG-1 may function in NMD by directly phosphorylating hUpf1 protein at physiologically relevant sites.

The PIK 1 -related kinases are a subfamily of the phosphatidylinositol (PI) kinases based on homology to the core catalytic domain of phosphatidylinositol 3-kinase (PI3K) (1,2). Regions of homology between PI kinases and the protein kinase superfamily include the ATP-binding site and the catalytic/substrate-binding site (3). PIK-related kinases are distinct from PI kinases in that they are high molecular weight proteins that function as serine/threonine protein kinases, rather than lipid kinases (3). PIK-related kinases can be divided into three subgroups based on structural and functional similarities shared by certain family members. The ATM/ATR/RAD3 subgroup functions in DNA damage response pathways, and members contain regions called RAD3 homology domains (4,5). The targets of rapamycin or TORs (Saccharomyces cerevisiae TOR1 and TOR2 and human mTOR/FRAP/RAFT1/RAPT1) were originally identified as intracellular targets of the immunophilinimmunosuppressant complex FKBP12-rapamycin (6,7). TORs share sequence similarity within the FRB domain, which binds this complex (7). The TORs function in response to mitogenic signaling to regulate cap-dependent translation (8). Finally, the catalytic subunit of DNA-dependent protein kinase (DNA-PK cs ) shares no sequence similarity to other PIK-related kinases, aside from the kinase domain. DNA-PK functions in the repair of programmed DNA breaks generated by meiotic and V(D)J recombination, and those generated by genotoxic insults (9).
Recently, evidence has emerged for an essential PIK-related kinase in nonsense-mediated mRNA decay (NMD). NMD, or mRNA surveillance, is an evolutionarily conserved process by which mRNA species containing premature termination codons are preferentially degraded, thereby preventing accumulation of truncated proteins that might serve in a dominant negative or gain-of-function manner (11)(12)(13)(14)(15)(16)(17)(18). NMD also functions to regulate the level of a number of normal mRNAs (19 -22). NMD has been genetically analyzed in yeast and nematodes. Seven genes (SMG-1 to SMG-7) are involved in NMD in Caenorhabditis elegans (23). One of these genes, SMG-1 (ceSMG-1), is predicted by sequence to encode a PIK-related kinase (10). Biochemical studies of ceSMG-1 are lacking, and the protein has never been shown to have phosphotransferase activity. However, genetic evidence indicates that phosphorylation of SMG-2, an RNA helicase, is blocked in ceSMG-1 mutants (10). SMG-2 and its human homologue, human Upf1 (hUpf1), contain multiple potential PIK-related kinase (S/T)-Q and (S/T)-P phosphorylation site motifs (24). Furthermore, phosphorylation of hUpf1 is blocked by high concentrations of wortmannin (IC 50 ϭ 100 nM), consistent with the role of a PIK-related kinase (24). Such observations suggest that ceSMG-1 is a PIKrelated kinase that may phosphorylate SMG-2. However, direct * This work was supported by National Institutes of Health Grants R01 CA56869 (to A. P. F.) and DK33938 and GM59614 (to L. E. M.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) AY014957.
* phosphorylation of SMG-2 or hUpf1 protein by ceSMG-1 or an SMG-1-like kinase in mammalian cells has never been demonstrated.
In the present report, we describe the cloning and characterization of the human orthologue to ceSMG-1, which we term human SMG-1 (hSMG-1). hSMG-1 encodes a novel PIK-related kinase with significant homology to the TORs and ceSMG-1. Biochemical characterization of transiently expressed FLAGtagged hSMG-1 protein indicates that it is a protein kinase that exhibits both autophosphorylation and substrate-specific phosphorylation. FLAG-hSMG-1 directly phosphorylates hUpf1 protein at sites phosphorylated in whole cells, indicating that hSMG-1 is a physiologically relevant hUpf1 protein kinase.

EXPERIMENTAL PROCEDURES
hSMG-1 cDNA Cloning and Plasmid Construction-5Ј-RACE (rapid amplification of cDNA ends) was performed using K562 (human chronic myelogenous leukemia) Marathon Ready cDNA (CLONTECH Laboratories, Inc.) and gene-specific primers toward the 5Ј-end of KIAA0421 (NCBI gi: 2887416; accession number AB007881), a partial cDNA with significant sequence similarity to PIK-related kinases, to generate an additional 1000 base pairs of cDNA sequence. The remaining hSMG-1 cDNA sequence was obtained from a human prostate 5Ј-STRETCH (-gt10) cDNA library (CLONTECH Laboratories, Inc.) by PCR. The full-length cDNA sequence, derived from 12 overlapping clones, was verified by alignment of four independent full-length clones. Comparison of the sequence to the GenBank TM EST data base identified KIAA0220 (NCBI gi: 1504021; accession number D86974), which corresponds to amino acids 95-647 of the full-length hSMG-1 cDNA. The hSMG-1 coding sequence can be accessed via GenBank TM accession number AY014957 (NCBI gi: 372334).
Three kinase-deficient (KD) hSMG-1 cDNAs were generated by sitedirected mutagenesis at nucleotide 5114 (GAC to GCC), or nucleotide 5171 (GAT to GAG or GAT to GCT) within motif I (KD1) and motif II (KD2 and -3) of the catalytic domain. All hSMG-1 clones DNA were confirmed by sequencing. Sequence alignments were performed using the Vector NTI Suite II program (InforMax Inc.).
Northern Blot Analysis-Total cellular RNA from human HL60, K562, and HEK-293 cells was extracted using TRIzol ® reagent (Life Technologies, Inc.). Equal amounts (5 g) of RNA was resolved in formaldehyde-agarose gels, transferred to nitrocellulose, and hybridized with one of several probes spanning hSMG-1 as indicated in the legend to Fig. 2. The expression of hSMG-1 mRNA in the Human Tissue 12-Lane MTN ® Blot and Human Cancer Cell Line MTN ® Blot (CLON-TECH Laboratories, Inc.) was assessed by Northern blot hybridization. ␤-Actin mRNA served as a control for RNA loading using human ␤-actin cDNA as probe (CLONTECH Laboratories Inc.). Probes were 32 P-labeled by random priming using [␣-32 P]dCTP and the Megaprimer DNA labeling kit (Amersham Pharmacia Biotech).
Affinity-purified antibody was generated using purified GST-hSMG-1 resolved by SDS-PAGE in 10% acrylamide preparative gels, transferred to nitrocellulose, and visualized by staining with fast green. The area of nitrocellulose containing GST-hSMG-1 was excised, incubated with 1% bovine serum albumin in PBS containing 0.1% Tween 20 (1% BSA/PBST), and then incubated with a 1:1000 dilution of anti-hSMG-1 serum in 1% BSA/PBST for 1 h at room temperature. The nitrocellulose was washed twice for 10 min in PBST, once in PBS alone, and antibody eluted into a minimal volume of 0.5 M acetic acid at pH 2.5 for 2 min. Eluted antibody was immediately neutralized by addition of 3 volumes of 1% BSA/PBS and dialyzed three times for 30 min against ice-cold PBS. Purified antibody was concentrated using a Centricon-10 (Amicon Inc.) and stored at Ϫ20°C in PBS containing 10% glycerol.
Immunoblot analysis of hSMG-1 was performed using a 1:10 dilution of affinity-purified antibody in 1% BSA/PBST. Total cell lysates from HEK-293 cells were separated in a 7.5% acrylamide Tris-HCl gel (Bio-Rad), transferred to nitrocellulose and blocked in 5% nonfat dry milk/ PBST for 1 h prior to incubation with affinity-purified hSMG-1 antibody for 2 h. Membrane was washed three times in PBST, incubated in a 1:10,000 dilution of goat anti-rabbit IgG (Kirkegaard & Perry Laboratories) in 5% milk/PBST for 1 h, and washed two times in PBST and once in PBS. Chemiluminescent detection of antigen-antibody complexes was performed using SuperSignal ® West Pico substrate (Pierce).
Protein kinase assays were performed in kinase buffer containing 100 M ATP, 10 mM MnCl 2 , 1 g of recombinant PHAS-1 (Calbiochem) or immunopurified FLAG-hUpf1 protein, and 5 Ci of [␥-32 P]ATP (PerkinElmer Life Sciences) in a final volume of 20 l. Reactions were performed at 30°C and terminated after 30 min by the addition of 20 l of 2ϫ SDS-PAGE sample buffer followed by boiling for 5 min. Samples were analyzed in 4 -20% acrylamide Tris-HCl (Bio-Rad) or NuPAGE 3-8% acrylamide Tris acetate gels (Invitrogen) and transferred to nitrocellulose. Radioactivity was detected using an InstantImager (Packard) and autoradiography on X-Omat film (Eastman Kodak Co.). For wortmannin treatment, immunopurified FLAG-hSMG-1 was incubated with the indicated amount of wortmannin or solvent (Me 2 SO) in a final volume of 30 l at room temperature for 1 h, washed twice with kinase buffer, and the kinase assay performed as indicated above. For rapamycin treatment, immunopurified FLAG-hSMG-1 was incubated with 10 M rapamycin or 100 M FK506 in the presence or absence of 10 g GST-FKBP12 in a final volume of 50 l. After 1 h, beads were washed twice with kinase buffer, and kinase assay performed as indicated above. Immunoblot analysis of FLAG-tagged proteins was performed using anti-FLAG M2 ® monoclonal antibody peroxidase conjugate (Sigma) at a 1:1000 dilution in PBST followed by chemiluminescent detection.
As a positive control for rapamycin inhibition, mTOR was immunopurified from HEK-293 cells using mTAb1 (27). Briefly, 1000 g of clarified cell lysate was incubated with 5 g of mTAb1 antibody conjugated to 20 l of protein A-Sepharose 4B beads (Sigma) for 2 h at 4°C. Beads were washed as described above and treated with or without 10 M rapamycin and 10 g of FKBP12 for 1 h. After washing twice with kinase buffer, kinase activity was assessed using 1 g of PHAS-1 as substrate as described above. Immunoblot analysis of immunopurified mTOR was performed using 1:1000 dilution of mTAb1 antibody in 5% nonfat dry milk/PBST for 1 h at room temperature. Membranes were washed three times in PBST, incubated with a 1:10,000 dilution of goat anti-rabbit IgG (Kirkegaard & Perry Laboratories) in 5% nonfat dry milk/PBST for 1 h, washed, and subjected to chemiluminescent detection.

RESULTS AND DISCUSSION
Cloning of hSMG-1 cDNA-In a search for sequences with homology to ceSMG-1, we used the GenBank TM data base of ESTs to identify a partial cDNA, termed KIAA0421 (NCBI gi: 2887416; accession number AB007881) with homology to PIKrelated kinases (30). Both 5Ј-RACE and PCR screening were used to clone the full-length hSMG-1 cDNA as described under "Experimental Procedures." The starting methionine codon was identified using two criteria: 1) the presence of upstream in-frame stop codons in three independent clones and 2) a Kozak consensus context. The open reading frame is encoded by 9096 base pairs and predicts a protein of 3031 amino acids with a molecular mass of 340,674 daltons and a pI of 6.00. The full nucleotide and amino acid sequence of hSMG-1 can be accessed through GenBank TM accession number AY014957 (NCBI gi: 372334). Interestingly, a fourth independent clone contained a unique sequence upstream of the designated start codon, which extended the open reading frame, suggesting that alternative splicing variant(s) of hSMG-1 may exist.
hSMG-1 Shares Sequence Homology with Members of the PIK-related Kinase Family-Comparison of the deduced amino acid sequence of hSMG-1 with the protein data base revealed high sequence homology with the PIK-related kinase family of protein kinases, including 1) the ataxia telangiectasia gene product, ATM, and the related human ATR and yeast RAD3; 2) the targets of rapamycin (yeast TOR1, TOR 2, and the human homologue mTOR/FRAP/RAFT1/RAPT1); and 3) the catalytic subunit of DNA-dependent protein kinase (DNA-PK cs ). hSMG-1 contains several highly conserved motifs found in all PIK-related kinases, including a conserved ATP-binding site (Lys 1525 ), and motif I (D 1705 XXXXN 1710 ) and motif II (D 1724 XX) sites within the catalytic domain (Fig. 1C). In addition, hSMG-1 contains a short C-terminal region at amino acids 3001-3031 (termed FATC for FRAP, ATM, TRRAP at C-terminus), found in the majority of the PIK-related kinases (Fig. 1A) (31). Among the well characterized PIK-related kinases, hSMG-1 exhibits the highest sequence homology to the TORs, which extends beyond the kinase domain to the FRB (FKBP12rapamycin binding) domain (Fig. 1, A and B). Mutational analysis of the FRB domain of mTOR has defined it as the site of binding for rapamycin, a selective inhibitor of TOR (8,32). Several residues within this domain are also required for mTOR kinase activity (8). Sequence alignment indicates that hSMG-1 lacks a critical serine residue that is required for binding of the FKBP12-rapamycin complex to mTOR (Fig. 1B) (32). However, the FRB domain of hSMG-1 retains a conserved tryptophan residue, which when mutated in mTOR abolishes kinase activity (Fig. 1B) (8). Conservation of this tryptophan residue suggests that it is important for hSMG-1 kinase activity.
The overall structure of hSMG-1 differs from that of other PIK-related kinases, since the kinase domain is not located at the extreme C terminus (Fig. 1A). Rather, a large region is located between the kinase domain and the FATC region that is 100% identical to LIP (lambda-interacting protein). LIP was identified by yeast two-hybrid screening as a protein that interacts with the zinc finger domain of PKC/ (33). LIP cDNA was reported to be 2142 base pairs, coding for a protein of 713 amino acids with a predicted molecular mass of 79.7 kDa (NCBI gi: 5542015; accession number U32581) (33). The observation that LIP forms part of hSMG-1 raised the possibility that LIP is a splicing variant of hSMG-1. However, our Northern and immunoblot analyses argue against this hypothesis (see below).
hSMG-1 Shares Sequence Homology to ceSMG-1, a C. elegans Protein Required for Nonsense-mediated mRNA Decay-In addition to homology to mTOR, hSMG-1 exhibits significant homology to ceSMG-1 in three major regions: an N-terminal ϳ200-amino acid region of 44% similarity, a ϳ1000-amino acid region in the middle of the coding sequence of 46% similarity that includes the FRB-like and kinase domains, and a C-terminal ϳ88-amino acid region of 61% similarity encompassing the FATC domain (Fig. 1A). Two additional regions of homology between ceSMG-1 and hSMG-1 (termed SMG-1 homology Domains 1 and 2: SD1 and SD2) have not previously been described in other PIK-related kinases and may constitute unique functional domains characteristic of ceSMG-1 and hSMG-1.
hSMG-1 Is Widely Expressed as a High Molecular Weight RNA and Protein-To assess hSMG-1 expression, RNA blot hybridization was performed using total RNA from K562, HL60, and HEK-293 cells (Fig. 2, A-C). hSMG-1 is expressed as an RNA of ϳ11.6 kb in HL60 and K562 cells as determined using a probe spanning nucleotides 5937-6890 (probe 1) of hSMG-1 ( Fig. 2A). This probe overlaps with the first 150 nucleotides of the reported LIP sequence, suggesting that the reported ϳ7.5-kb LIP mRNA (33) is not expressed in these cells. To confirm this observation, RNA blot hybridization was performed using a LIP-specific probe (probe 2, nucleotides 6741-7812) and a hSMG-1-specific probe (probe 3, nucleotides 3521-4654). Like probe 1, these two probes detect an ϳ11.6-kb RNA in HL60, K562, and HEK-293 cells, but no smaller RNA(s) corresponding to LIP (Fig. 2, B and C). RNA blot hybridization of cancer cell lines and human tissues using probe 1 (Fig. 2, D  and E) detected hSMG-1 RNA in each of the cancer cell lines with relatively low levels in lung carcinoma (A459) and melanoma (G361) cell lines. hSMG-1 RNA was also detected in the majority of human tissues at varying levels. Therefore, hSMG-1 RNA is widely expressed in multiple tissues and cell lines.
Immunoblot analysis using an affinity-purified hSMG-1 antibody detected a single immunoreactive band in HEK-293 cells with an apparent molecular mass consistent with the predicted molecular mass of 340 kDa. Importantly, no lower molecular mass bands of ϳ80 kDa consistent with LIP (33) were detected. The antigen used to produce our hSMG-1 antibody consisted of the C-terminal region of the reported LIP cDNA. Therefore, it is likely that our antibody would detect LIP if it were expressed in these cells. We have also failed to detect LIP by immunoblot analysis in K562 and HL60 cells (data not shown).
An identifying feature of PIK-related kinases is sensitivity to wortmannin at high nanomolar concentrations (IC 50 ϳ50 -500 nM) (34, 37), but not at low nanomolar concentrations that inhibit PI3K (IC 50 ϭ 5 nM) (38). Incubation of immunopurified FLAG-WT-hSMG-1 with increasing concentrations of wort-mannin leads to dose-dependent inhibition of both autophosphorylation and PHAS-1 phosphorylation with an IC 50 of 105 nM (Fig. 3, C and D). hSMG-1 is also inhibited by LY294002 at low micromolar concentrations comparable with LY294002 inhibition of PI3 kinase and other PIK-related kinases (data not shown) (37).
The finding that hSMG-1 contains an FRB-like domain was of interest, since to date the FRB domain has been a unique feature of the TORs (39). Mutational analysis identified a critical serine residue within the FRB domain of mTOR (Ser 2035 ) that is required for binding of FKBP12-rapamycin complex and mTOR kinase inhibition (8,32). Multiple sequence alignments of hSMG-1, ceSMG-1, and mTOR indicate that hSMG-1 and ceSMG-1 lack this critical Ser residue (Fig. 1B). To test whether hSMG-1 kinase activity is inhibited by rapamycin, immunopurified FLAG-WT-hSMG-1 was incubated with either rapamycin alone, rapamycin and FKBP12, FK506 alone, or FK506 and FKBP12 prior to assay for autophosphorylation and phosphorylation of PHAS-1 (Fig. 3E). hSMG-1 kinase activity was not affected by any of the combinations of rapamycin, FKBP12, or FK506. In contrast, mTOR exhibited the expected inhibition by FKBP12-rapamycin complex (27,40). Treatment of pCI-FLAG-WT-hSMG-1 HEK-293 cell transfectants with rapamycin prior to isolation failed to inhibit immunopurified FLAG-hSMG-1 kinase activity (data not shown).
Our structural and biochemical analyses demonstrate that hSMG-1 is a bona fide PIK-related kinase. Like other PIKrelated kinases, hSMG-1 is a high molecular weight protein with a conserved kinase domain closely related to that of the phosphatidylinositol-kinases. Like other PIK-related kinases, hSMG-1 phosphorylates itself and the generic PIK-related kinase substrate, PHAS-1. hSMG-1 kinase activity is inhibited by wortmannin and LY294002 at concentrations that inhibit other PIK-related kinases. However, hSMG-1 does not exhibit , and FLAG-hSMG-1 protein kinase activity toward immunopurified FLAG-hUpf1 ([ 32 P]hUpf1p) was assessed. Immunoblot analysis using FLAG M2 monoclonal antibody demonstrates the amount FLAG-hSMG-1 (anti-FLAG(hSMG-1)) and FLAG-hUpf1 (anti-FLAG(hUpf1p)) present in each reaction. B, HEK-293 cells were transfected with pCI-FLAG-hUpf1, labeled for 24 h with [ 32 P]orthophosphate and 32 P-labeled FLAG-hUpf1 immunopurified. 32 P-Labeled FLAG-hUpf1 phosphorylated by FLAG-hSMG-1 in vitro (in vitro) and 32 P-labeled FLAG-hUpf1 from radiolabeled HEK-293 cells (whole cells) were subjected to comparative two-dimensional tryptic phosphopeptide mapping. 32 P-Labeled phosphopeptides from whole cell labeling and in vitro phosphorylation were mixed prior to analysis to confirm the identity of phosphopeptides 1 and 2 (mix).
hSMG-1 Phosphorylates hUpf1 Protein at Sites Phosphorylated in Whole Cells-hUpf1, the human orthologue to C. elegans SMG-2 and the S. cerevisiae Upf1 protein, is an ATPdependent helicase required for NMD in mammalian cells (10,26,41,42). hUpf1, like SMG-2, is a phosphoprotein whose phosphorylation is inhibited by high concentrations of wortmannin (IC 50 ϭ 100 nM) (24). These observations, and the fact that hUpf1 is rich in (S/T)-Q and (S/T)-P motifs that serve as phosphorylation sites for PIK-related kinases, suggest that a PIK-related kinase may mediate hUpf1 protein phosphorylation (24). This hypothesis is supported by the observation that phosphorylated SMG-2 is not detected in a ceSMG-1 mutant background (10). These biochemical and genetic data suggest that hSMG-1 may be a physiologic hUpf1 protein kinase.
To directly test this hypothesis, immunopurified FLAG-hUpf1 protein was used as a substrate for immunopurified FLAG-WT-hSMG-1 or FLAG-KD-hSMG-1 (Fig. 4A). FLAG-WT-hSMG-1 directly phosphorylates hUpf1 protein in vitro, whereas FLAG-KD-hSMG-1 is incapable of doing so. To determine whether hSMG-1 phosphorylates physiologically relevant sites on hUpf1, the sites of hSMG-1-mediated phosphorylation were compared with those phosphorylated in whole cells. HEK-293 cells were transiently transfected with pCI-FLAG-hUpf1 and incubated with [ 32 P]orthophosphate for 12 h, and 32 P-labeled FLAG-hUpf1 was immunopurified. Comparative two-dimensional tryptic phosphopeptide mapping (Fig. 4B) demonstrates that hUpf1 phosphorylated by hSMG-1 contains two major phosphopeptides (Fig. 4B, in vitro, labeled 1 and 2). A similar phosphopeptide pattern consisting of two major phosphopeptides was observed with hUpf1 phosphorylated in whole cells (Fig. 4B, whole cells, labeled 1 and 2). Analysis of a mixture of these digests revealed that phosphopeptides 1 and 2 are identical. Thus, hSMG-1 phosphorylates hUpf1 at the same two major sites phosphorylated in whole cells.
We conclude that hSMG-1 is a physiologically relevant hUPF1 kinase. The physiologic role of hSMG-1-mediated phosphorylation of hUpf1 is unknown. One possible role is to regulate hUpf1 function in NMD. We have been unable to generate direct evidence of a critical role for hSMG-1 in NMD, likely due to the low levels of KD-hSMG-1 expression that we can achieve in transiently transfected cells (data not shown). It is unclear whether this low level of expression is due to a cytotoxic effect of KD-hSMG-1. Future studies aim to directly assess the importance of hSMG-1 and hSMG-1-mediated phosphorylation of hUpf1 protein in NMD in human cells.