A Novel 160-kDa Phosphotyrosine Protein in Insulin-treated Embryonic Kidney Cells Is a New Member of the Insulin Receptor Substrate Family*

We have previously identified a 160-kDa protein in human embryonic kidney (HEK) 293 cells that undergoes rapid tyrosine phosphorylation in response to insulin (PY160) (Kuhné, M. R., Zhao, Z., and Lienhard, G. E. (1995) Biochem. Biophys. Res. Commun. 211, 190–197). The phosphotyrosine form of PY160 was purified from insulin-treated HEK 293 cells by anti-phosphotyrosine immunoaffinity chromatography, the sequences of peptides determined, and its cDNA cloned. The PY160 cDNA encodes a 1257-amino acid protein that contains, in order from its N terminus, a pleckstrin homology (PH) domain, a phosphotyrosine binding (PTB) domain, and, spread over the C-terminal portion, 12 potential tyrosine phosphorylation sites. Several of these sites are in motifs expected to bind specific SH2 domain-containing proteins: YXXM (7 sites), phosphatidylinositol 3-kinase; YVNM (1 site), Grb-2; and YIEV (1 site), either the protein-tyrosine phosphatase SHP-2 or phospholipase Cγ. Furthermore, the PH and PTB domains are highly homologous (at least 40% identical) to those found in insulin receptor substrates 1, 2, and 3 (IRS-1, IRS-2, and IRS-3). Thus, PY160 is a new member of the IRS family, which we have designated IRS-4.

The insulin receptor is a tyrosine kinase, which when activated by insulin binding phosphorylates cellular substrates. The most well characterized of these are two members of the IRS 1 family, IRS-1 and IRS-2, and the protein Shc. Tyrosine phosphorylation of the IRS proteins creates binding sites for SH2 domain-containing signaling molecules, including PI 3-kinase, the adapter molecule Grb-2, and the protein-tyrosine phosphatase SHP-2. Docking of these proteins in turn activates specific signal transduction pathways (reviewed in Refs. 1 and 2). Recently, we have identified, by purification and cloning, a third member of the IRS family, called IRS-3, which in insulintreated adipocytes is tyrosine-phosphorylated and associated with PI 3-kinase (3,4). All three IRS family members possess a common domain structure that includes PH and PTB domains at the N terminus and, C-terminal to these, a number of potential tyrosine phosphorylation sites (1,2,4,5). The presence of these features can therefore be viewed as defining an IRS. Previously, we have identified a 160-kDa protein in HEK 293 cells, termed PY160, which is rapidly tyrosine-phosphorylated in response to insulin but which is immunologically unrelated to IRS-1 (6). In the present study we have isolated PY160 from insulin-treated HEK 293 cells and cloned its cDNA. The predicted amino acid sequence shows that PY160 is a new member of the IRS family.

EXPERIMENTAL PROCEDURES
Cell Culture and Preparation of Lysates-HEK 293 cells were grown on 10-cm plates as described previously (6). Before use, confluent plates of cells were incubated in serum-free medium for 2 h and then incubated for 5 min further with either no addition or the addition of 1 M insulin to activate fully the insulin and IGF-1 receptors present on these cells (7). Each plate was rinsed with phosphate-buffered saline and lysed by the addition of 1 ml of 3% SDS, 10 mM dithiothreitol in Buffer A (50 mM Hepes, 100 mM NaCl, 2 mM EDTA, 1 mM sodium vanadate, pH 7.4, with protease inhibitors (10 M EP475, 10 M leupeptin, 10 g/ml aprotinin, 1 nM pepstatin A, 4 mM diisopropyl fluorophosphate)). The lysate was held at 100°C for 5 min, and the DNA in it was sheared by repeated passage through a syringe needle. Finally, the lysate was diluted by the addition of 5 ml of 3% Triton X-100 in Buffer A; free sulfhydryl groups were capped by the addition of N-ethylmaleimide to a final concentration of 6.7 mM; and the lysate was clarified by centrifugation at 150,000 ϫ g for 1 h.
Immunoadsorption of PY160 -Aliquots of lysates (1 ml) from basal and insulin-stimulated 293 cells were incubated with anti-Tyr(P) antibodies (20 l of 4G10 agarose from Upstate Biotechnology) for 4 h at 4°C. The beads were washed twice with a wash buffer (50 mM Hepes, 100 mM NaCl, 1.5% Triton X-100. 0.25% SDS, 1 mM sodium vanadate with protease inhibitors, pH 7.4), and the Tyr(P)-containing proteins were eluted by the addition of 135 l of 40 mM phenyl phosphate in the wash buffer. To estimate the recovery of PY160, samples containing the original extract, the depleted extract, and the phenyl phosphate eluate were immunoblotted for Tyr(P), as described (3). The yield of the Tyr(P) form of PY160 by immunoadsorption from the lysate of insulin-treated cells was approximately 15%.
Purification of PY160 and Sequencing of Peptides-PY160 was purified by anti-Tyr(P) affinity chromatography from an extract derived from a total of thirty-six 10-cm plates of insulin-stimulated HEK 293 cells. In a single purification, half of the extract (110 ml) was passed at 0.2 ml/min through a 1.0-ml column of goat IgG-agarose (Sigma) and then through a 1.5-ml column of immobilized anti-Tyr(P) antibody (4G10 agarose at 1 mg/ml). Once the extract was applied, the goat IgG column was disconnected, and the anti-Tyr(P) column was washed sequentially with (a) 30 ml of 1% Triton X-100, 0.25% SDS in wash buffer (20 mM Tris-HCl, 150 mM NaCl, 1 mM sodium vanadate, pH 7.4, * This work was supported by National Institutes of Health Grant DK 42816. 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) AF 007567.
¶ To whom correspondence should be addressed. Tel.: 603-650-1627; Fax: 603-650-1128; E-mail: gustav.e.lienhard@dartmouth.edu. 1 The abbreviations used are: IRS, insulin receptor substrate; HEK, human embryonic kidney; HPLC, high performance liquid chromatography; IGF, insulin-like growth factor; PCR, polymerase chain reaction; PH, pleckstrin homology; PI, phosphatidylinositol; PTB, phosphotyrosine binding; RACE, rapid amplification of cDNA ends; SH2, Src homology 2; bp, base pair(s); nt, nucleotide(s); aa, amino acids. with protease inhibitors (2 g/ml aprotinin, 2 M leupeptin, 0.2 nM pepstatin A)) at 1 ml/min, (b) 30 ml of 1% Triton X-100 in wash buffer with protease inhibitors at 1 ml/min, (c) 300 ml of 0.05% Triton X-100 in wash buffer with protease inhibitors at 0.3 ml/min, and (d) 7 ml of 0.015% sodium deoxycholate in wash buffer at 0.5 ml/min. Elution buffer (3 mM phenyl phosphate, 0.015% sodium deoxycholate in wash buffer) was run onto the column and the flow stopped for 10 min. Tyr(P)-containing proteins were then eluted at 0.5 ml/min, and 2-ml fractions were collected in low protein adsorption tubes (Nunc no. 443990). The purification was repeated with the other half of the lysate, and the two fractions from each preparation containing the bulk of the PY160 were concentrated by trichloroacetic acid precipitation as detailed in Ref. 3. The precipitates were resuspended in SDS sample buffer and separated in a single lane on a 7% acrylamide gel. Following transfer to ProBlott membrane (Applied Biosystems) and staining with Amido Black, the band corresponding to PY160 was excised (about 1.5 g (10 pmol)). The protein band was subjected to in situ digestion with LysC; the resultant peptides were separated by microbore HPLC; selected fractions were screened by MALDI-TOF mass spectrometry and microsequenced by the methods described previously (4). By UV absorbance, approximately 1-5 pmol of peptides were present in the HPLC separation.
PY160 cDNA-Total RNA was obtained from HEK 293 cells using the Trizol reagent (Life Technologies, Inc.), and mRNA was subsequently purified from it using the Fast-Track kit (Invitrogen). A 10-cm plate of confluent HEK 293 cells yielded about 3 g of mRNA. A Marathon Ready TM cDNA library from human fetal kidney was obtained from CLONTECH. The nucleotide sequence encoding the central portion of peptide a (see "Results and Discussion" for peptides) was obtained by PCR amplification. The sequences of the N and C termini of peptide a were used to design degenerate oligonucleotides; restriction sites for EcoRI (sense) and BamHI (antisense) were incorporated to facilitate cloning (5ЈGCGAATTCYTNGARACNGCNGA3Ј and 5ЈGAG-GATCCGCRTTYTCRTARTA3Ј, where Y is C or T, R is A or G, and N is A, C, G, or T; restriction sites are underlined). cDNA, produced by random hexamer primed reverse transcription of HEK 293 mRNA, was used as the template. A PCR product of the expected size (63 bp) was obtained and cloned into the EcoRI/BamHI site of pBluescript II (SKϪ). Several clones were sequenced and found to encode the middle portion of the peptide (APARLE; nt 390 -408). The 5Ј-end of the cDNA was obtained in two separate 5Ј-RACE reactions. In the first, a degenerate antisense primer derived from the C terminus of peptide a (5ЈTT-NCGNGCRTTYTC3Ј and 5ЈTTYCTNGCRTTYTC3Ј mixed in a 2:1 molar ratio, respectively) was used to reverse transcribe HEK 293 mRNA. The resulting cDNA was tailed with dCTP and amplified by PCR with an antisense primer derived from the sequence encoding the middle portion of peptide a (5ЈTCRTARTATTCCAGCCGAGCT3Ј; exact sequence underlined, see above) and the abridged anchor primer of a 5Ј-RACE kit, according to the manufacturer's instructions (Life Technologies, Inc.). The products were reamplified using a nested antisense primer (derived from the N terminus of peptide a (5ЈTGGGGCGTC-NGCNGTYTC3Ј)) and the abridged universal amplification primer of the kit. A 330-bp product was purified and cloned into pCR-Script (Stratagene), and the sequences of five clones were determined (nt 90 -378). In the second 5Ј-RACE, the Marathon Ready TM cDNA library from human fetal kidney was amplified with an antisense primer (nt 352-371) and the AP1 primer of the kit according to the manufacturer's instructions. The reaction mixture was reamplified using a nested antisense primer (nt 270 -290) and the AP2 primer of the kit. A mixture of PCR products with a size range of 300 -330 bp was purified and sequenced directly (nt 1-267). Upstream of nt 1 the sequence was a mixture; this may reflect heterogeneity in the start point of the PY160 mRNA.
The 3Ј-portion of PY160 cDNA was obtained in a number of 3Ј-RACE reactions. cDNA was synthesized from HEK 293 mRNA using either degenerate antisense primers derived from the sequences of peptide b (5ЈXACDATNACYTGCCANACRT3Ј) and peptide c (5ЈXTTNCCRAAR-TARCTNC3Ј and 5ЈXTTNCCRAARTANGANC3Ј in a 1:2 molar ratio, respectively) or using oligo(dT) (X(T) 17 , where X is an adapter sequence 5ЈGGCCACGCGTCGACTAGTAC3Ј and D is A, G, or T). A 550-bp PCR product (produced from amplification of peptide b-primed cDNA using a sense primer (nt 243-260) and the adapter primer) and a 1200-bp PCR product (produced from amplification of peptide c-primed cDNA using a sense primer (nt 754 -774) and the adapter primer) were cloned into pCR-Script. The inserts of several clones were sequenced and were found to encode nt 266 -774 (peptide b primed) and nt 775-1914 (peptide c primed). The remainder of the sequence was obtained by 3Ј-RACE using oligo(dT)-primed cDNA. This cDNA was amplified using either of two upstream sense primers (nt 1762-1783 and nt 2956 -2978) and the adapter primer. A 1300-bp PCR product was obtained with the nt 1762-1783 primer, while the nt 2956 -2978 primer yielded a 980-bp product. These two products which are shorter than the sizes expected from the size of the PY160 mRNA (see "Results and Discussion") most likely arise due to internal priming by the oligo(dT) primer. The 1300and 980-bp PCR products were sequenced directly (nt 1840 -3052 and nt 2979 -3939, respectively). To confirm the DNA sequence, overlapping fragments were generated by PCR amplification of cDNA obtained by random hexamer-primed reverse transcription of total RNA from HEK 293 cells and sequenced directly on both strands (nt 170 -637, 598 -1124, 1076 -1704, 1651-2308, 2139 -2889, 2722-3302, and 3200 -3898). Sequencing was performed on the Applied Biosystems 373 DNA sequencing system using the Perkin-Elmer DNA sequencing kit; data were analyzed with the Applied Biosystems software. Homology searches were performed with the BLAST program (8).

RESULTS AND DISCUSSION
Identification and Purification of PY160 -Treatment of HEK 293 cells with insulin elicits the tyrosine phosphorylation of a protein of approximately 160 kDa, which is immunologically distinct from IRS-1 (6). Immunoblotting of HEK 293 cell lysates with antibodies to IRS-2 detected a protein larger than PY160; this result indicated that PY160 was also not IRS-2 (data not shown).
To assess the feasibility of isolating PY160 by anti-Tyr(P) immunoaffinity chromatography, we performed immunoadsorptions with anti-Tyr(P) immobilized on agarose beads. Detergent extracts were prepared from basal and insulin-stimulated HEK 293 cells, incubated in the presence or absence of phenyl phosphate (a ligand competing with Tyr(P)), and then immunoadsorbed with anti-Tyr(P) beads. The adsorbed proteins were eluted with phenyl phosphate and analyzed by immunoblotting and protein staining. Fig. 1 (lanes 1-4) shows the eluted Tyr(P) proteins as detected by anti-Tyr(P) immunoblotting. Two major insulin-elicited Tyr(P) proteins were present. One had the size expected for the Tyr(P) form of PY160. The other, based on its size of about 100 kDa, is most likely a mixture of the tyrosine-phosphorylated ␤-subunits of the insulin and IGF-1 receptors; both receptors are present in HEK 293 cells, and the latter would be expected to be activated at 1 M insulin (7). Protein staining of the eluates showed a single major protein, which co-migrated with the Tyr(P) form of PY160 (Fig. 1, lanes 5-8). The recovery of this protein from extracts of basal and insulin-treated cells paralleled the recovery of the Tyr(P) form of PY160; this indicates that the protein FIG. 1. Immunoadsorption of PY160 from basal and insulinstimulated HEK 293 cells. HEK 293 cells were incubated in the absence (Ϫ) or presence (ϩ) of 1 M insulin, lysates were prepared, and the Tyr(P) proteins were immunoadsorbed, in either the absence (Ϫ) or presence (ϩ) of 40 mM phenyl phosphate, on anti-Tyr(P) beads, as described under "Experimental Procedures." Bound proteins were eluted with phenyl phosphate, and samples were separated by SDS-gel electrophoresis, transferred to nitrocellulose, and either immunoblotted for Tyr(P) (lanes 1-4) or stained for protein with colloidal gold (lanes 5-8). Lanes 1-4 and 5-8 contained samples derived from 3 and 12% of a 10-cm plate, respectively. The failure of the protein stain to detect receptors at 100 kDa suggests that these are more heavily tyrosinephosphorylated than in PY160. was PY160 (Fig. 1, compare lanes 5 and 6 with lanes 1 and 2). The binding to the anti-Tyr(P) beads was specific; no proteins were present in the eluates from anti-Tyr(P) immunoprecipitates of lysates preincubated with phenyl phosphate (Fig. 1,  lanes 3 and 4 and 7 and 8).
The results in Fig. 1 showed that it would be possible to purify PY160 from HEK 293 cells by anti-Tyr(P) chromatography in an amount sufficient to obtain peptide sequences. A large scale purification was performed by a slight modification of a method that we previously used to purify Tyr(P) proteins from insulin-treated adipocytes (3) (see "Experimental Procedures"). This yielded sufficient PY160 to allow determination of the sequences of five peptides: a, LETADAPARLEYYENARK; b, DVWQVIVK; c, RSYFGK; d, FLGRGLDK; and e, EVSYN-WDPK (see Fig. 2). A search of the data base using the BLAST program revealed that peptides a and b had significant homology with sequences in IRS-1.
cDNA Encoding PY160 -PCR products representing PY160 were obtained by 5Ј-and 3Ј-RACE reactions using HEK 293 mRNA or a human fetal kidney cDNA library, and their sequences were determined. The nucleotide sequence and predicted amino acid sequence of PY160 are presented in Fig. 2. An open reading frame extending from nt 79 to 3852 encodes a 1257-amino acid polypeptide that contains the sequences of all five PY160 LysC peptides. It is not certain that the ATG codon at nt 79 -81 initiates translation, since an alternative downstream ATG codon at nt 238 -240 also conforms to the human Kozak consensus sequence for initiation (9). Although there is no in-frame stop codon upstream of the first ATG codon, the murine gene for PY160 has an in-frame stop codon 117 nt upstream of the ATG codon corresponding to nt 79 -81 in the human sequence (data not shown). It is therefore likely that initiation occurs at either of these two ATG codons. If the first ATG codon is used, the predicted molecular mass of PY160 is 133.8 kDa, a value that is smaller than the size of approximately 160 kDa for the Tyr(P) form estimated by SDS gel electrophoresis. The explanation for the difference is most likely an aberrantly low mobility on electrophoresis, as is the case for IRS-1 (10). The fact that no consensus polyadenylation sequence is present at the 3Ј-end of the cDNA shown indicates that the 3Ј-untranslated sequence extends further. In agreement with this conclusion, Northern analysis of HEK 293 mRNA using either of two probes derived from the PY160 cDNA (nt 104 -369 and 1428 -1895) detects messages of 6 and 10 kilonucleotides (data not shown).
Two lines of evidence show that the predicted protein is PY160. First, as described below, the structure of the protein is that expected for a substrate of the insulin receptor. Second, we have prepared affinity-purified antibodies against the C-terminal peptide (16 amino acids) of the predicted protein and used these to show its identity with PY160. Lysates from untreated and insulin-treated 293 cells were immunoprecipitated with the antibodies against the C-terminal peptide, and then immunoprecipitates were immunoblotted with antibodies against Tyr(P) as well as those against the C-terminal peptide. The immunoprecipitates from the untreated and insulin-treated cells contained equal amounts of a 160-kDa protein, as detected with the antibodies against the peptide, and its tyrosine phosphorylation was markedly enhanced in insulin-treated cells (data not shown).
Domains and Tyr(P) Motifs in PY160 -A comparison of the amino acid sequence of PY160 with the data base and an examination for the presence of potential tyrosine phosphorylation sites revealed that PY160 contains, in order from its N terminus, a PH domain, a PTB domain, and spread over the C-terminal portion, 12 potential tyrosine phosphorylation sites (Fig. 3A). This architecture is the same as that of the three known members of the IRS family (see the Introduction). Therefore, PY160 is a new member of the IRS family, and henceforth we refer to it as IRS-4.
IRS-4 is of a similar length (1257 aa) to both IRS-1 and IRS-2 (1242 aa for human IRS-1 and 1321 aa for mouse IRS-2 (5)). Overall IRS-4 displays limited sequence identity with IRS-1 and IRS-2 (27 and 29%, respectively). However, significant homology is found in the PH and PTB domains.
The PH domain of IRS-4 consists of 120 amino acids (residues 78 -197) and exhibits a high degree of homology with the domain in IRS-1, IRS-2, and IRS-3 (49, 50, and 43% identity, respectively) (Fig. 3B). This high degree of conservation suggests a common function for the PH domain in IRS family members. In IRS-1, the PH domain is necessary for efficient tyrosine phosphorylation by the insulin receptor in vivo, al- though it does not appear to interact directly with the receptor (11)(12)(13).
The PTB domain of IRS-4 consists of 101 amino acids (residues 231-331) and exhibits a high degree of homology with this domain in IRS-1, IRS-2, and IRS-3 (66, 62, and 43% identity, respectively). The crystal structure of the IRS-1 PTB domain complexed with a 9-residue Tyr(P) peptide modeled on the residues surrounding Tyr-960 in the insulin receptor has been determined (14). Of interest, 13 out of the 15 amino acids in IRS-1 that interact with the bound peptide are identical in IRS-4, including the two arginines whose guanidinium groups contact the phosphate of the Tyr(P) residue (Fig. 3B). Thus, it is likely that the PTB domain in IRS-4 will also be found to bind to the activated insulin receptor by association with the Tyr(P) 960 segment.
Both IRS-1 and IRS-2 contain non-PTB domains, which are important for interaction with the insulin receptor. These are a domain immediately downstream of the PTB domain, referred to as the SAIN domain (residues 313-462 in human IRS-1 and by homology residues 349 -535 in mouse IRS-2), and a second domain C-terminal to the SAIN domain that is present only in IRS-2 (residues 591-786) (13,(15)(16)(17). There is little identity between these domains and the corresponding regions in IRS-4. It will be of interest to determine if these regions in IRS-4 also interact with the insulin receptor.
Among the 12 potential tyrosine phosphorylation sites in IRS-4, seven are in YXXM motifs (Tyr-487, -700, -717, -743, -779, -828, and -921) (Fig. 3A). This motif, in its Tyr(P) state, binds to the SH2 domains of the PI 3-kinase 85-kDa subunit (18). IRS-4 also contains a potential tyrosine phosphorylation site (Tyr-921) in a motif that is expected to bind the SH2 domain of Grb-2, the adapter for Sos (the GDP-releasing factor for Ras), and another site (Tyr-1015) in the motif expected to bind the N-terminal SH2 domain of either the Tyr(P) phosphatase SHP-2 or phospholipase C␥ (18,19). In fact, we have recently found that PI 3-kinase and Grb-2 co-immunoprecipitate with IRS-4 from lysates of HEK 293 cells, with more of each associated with IRS-4 after insulin treatment of the cells. 2 It remains to be determined if IRS-4 associates with SHP-2 or phospholipase C␥. IRS-1, IRS-2, and IRS-3 also contain the motifs for binding PI 3-kinase, Grb-2, and SHP-2 (4, 5), and the similar arrangement of these motifs in IRS-1, IRS-2, and IRS-4 ( Fig. 3A and Ref. 5) is an additional feature of the homology among these proteins.
Implications-This study, in conjunction with our recent discovery of IRS-3 (4), raises the question of whether there are additional members of the IRS family to be discovered. A major challenge now is to define the roles of each IRS in signaling from the insulin and IGF-1 receptors and possibly other receptors as well.

FIG. 3. Structure of IRS-4 (PY160) and homology with IRS family members.
A, comparison of the organization of human (h) IRS-4 with that of human IRS-1, showing the PH and PTB domains and the potential sites of tyrosine phosphorylation. The latter are those tyrosine residues that have one or more acidic residues within the immediate five upstream amino acids (20). Potential tyrosine phosphorylation sites that are found at similar positions in both IRS-4 and IRS-1 and are predicted to bind the same SH2 domain-containing proteins (18,19) are indicated by dashed lines. Tyrosines in human IRS-1 that correspond to known tyrosine phosphorylation sites in rat IRS-1 are marked with asterisks (21). B, the PH and PTB domains from human (h) IRS-1, mouse (m) IRS-2, rat (r) IRS-3, and human IRS-4 were aligned with the PILEUP program (GCG, Wisconsin). Positions where at least 3 out of 4 amino acid residues are identical are on a black background, and positions where 2 out of 4 amino acids residues are either identical or are conserved are on a gray one. The percentage values at the end of each sequence are the percentage of identical amino acids in a pairwise comparison of each to the IRS-4 sequence. The PH and PTB domains of IRS-1 and IRS-2 are 69 and 75% identical to each other, respectively (5). The two conserved Arg residues of the IRS-1 PTB domain that contact the Tyr(P) of the bound peptide similar in sequence to that around Tyr-960 of the insulin receptor (14) are marked with an asterisk.