Complementation Analysis Demonstrates That Insulin Cross-links Both α Subunits in a Truncated Insulin Receptor Dimer*

The insulin receptor is a homodimer composed of two αβ half receptors. Scanning mutagenesis studies have identified key residues important for insulin binding in the L1 domain (amino acids 1-150) and C-terminal region (amino acids 704-719) of the α subunit. However, it has not been shown whether insulin interacts with these two sites within the same α chain or whether it cross-links a site from each α subunit in the dimer to achieve high affinity binding. Here we have tested the contralateral binding mechanism by analyzing truncated insulin receptor dimers (midi-hIRs) that contain complementary mutations in each α subunit. Midi-hIRs containing Ala14, Ala64, or Gly714 mutations were fused with Myc or FLAG epitopes at the C terminus and were expressed separately by transient transfection. Immunoblots showed that R14A+FLAG, F64A+FLAG, and F714G+Myc mutant midi-hIRs were expressed in the medium but insulin binding activity was not detected. However, after co-transfection with R14A+FLAG/F714G+Myc or F64A+FLAG/F714G+Myc, hybrid dimers were obtained with a marked increase in insulin binding activity. Competitive displacement assays revealed that the hybrid mutant receptors bound insulin with the same affinity as wild type and also displayed curvilinear Scatchard plots. In addition, when hybrid mutant midi-hIR was covalently cross-linked with 125I(A14)-insulin and reduced, radiolabeled monomer was immunoprecipitated only with anti-FLAG, demonstrating that insulin was bound asymmetrically. These results demonstrate that a single insulin molecule can contact both α subunits in the insulin receptor dimer during high affinity binding and this property may be an important feature for receptor signaling.

The metabolic actions of insulin are initiated by its binding to the insulin receptor, a transmembrane protein whose structure has been studied extensively and is now well characterized. The insulin receptor (IR) 2 is a homodimer composed of two identi-cal ␣␤ half receptors. A single disulfide bond covalently links the ␣␤ subunits in each half receptor, and the half receptors are also linked by multiple disulfide bonds between the ␣ subunits. The 719-aa ␣ subunit contains the insulin binding determinants, whereas the 619-aa ␤ subunit contains a 194-aa extracellular domain, a 23-aa transmembrane domain, and a 404-aa intracellular domain with intrinsic tyrosine kinase activity (for review, see Ref. 1).
A detailed description of insulin binding to the IR is not yet available. However, a number of studies have shown that the key determinants required for high affinity insulin binding are located in the ␣ subunit N-terminal L1 domain (aa 1-150) and the C-terminal region (aa 704 -719) (2)(3)(4)(5)(6). In particular, Whittaker et al. (7,8) have performed an alanine scanning mutagenesis of human IR ␣ subunit and have shown that alanine mutation of Arg 14 , Asn 15 , and Phe 64 in the L1 domain as well as mutations of Phe 705 , Glu 706 , Tyr 708 , His 710 , Asn 711 , Phe 714 , and Val 715 in the C-terminal region resulted in Ͼ10-fold loss in insulin binding affinity. Taken together, these results suggest that insulin interacts simultaneously with these two regions on the ␣ subunit to achieve high affinity binding. However, it has not been determined whether insulin binds to these regions derived from the same ␣ chain (cis binding mechanism) or from adjacent ␣ subunits in the IR dimer (contralateral binding). To address this issue, we have expressed truncated IR dimers (midi-hIR) with inactivating mutations at Arg 14 , Phe 64 , and Phe 714 . When expressed as homodimers, the mutant midi-hIRs did not bind insulin, but hybrid mutant midi-hIR formed by co-transfection of two midi-hIR containing complementary mutations exhibited high affinity insulin binding indistinguishable from wild type.

EXPERIMENTAL PROCEDURES
Materials-Oligonucleotide primers were prepared using an Applied Biosystems DNA Synthesizer. PCR amplification and ligation of DNA fragments, preparation of plasmid DNA, and DNA sequence analyses were performed using standard techniques. Cultured 293H cells were purchased from Invitrogen. 125 I(A14)-insulin was obtained from Amersham Biosciences. All chemicals were molecular biology or reagent grade.
Construction and Expression of midi-hIR-The midi-hIR cDNA corresponds to mIR.Fn0/Ex10 as described by Brandt et al. (6) and was constructed by PCR amplification and ligation of * This work was supported by National Institutes of Health Grants R01 DK013914 and P60 DK020595 and by the Howard Hughes Medical Institute. 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. 1  DNA fragments using a cloned hIR cDNA available in our laboratory as the template. The resultant WTϩFLAG and WTϩMyc cDNAs encoded hIR signal peptide and residues 1-601 fused to residues 650 -719 together with an epitope tag sequence FLAG (DYKDDDK) or Myc (EQKLISEEDL) at the C terminus and were inserted into mammalian expression vector pcDNA 3.1 (Invitrogen). DNA sequencing confirmed the plasmids contained the canonical hIR sequence (9) except for coding mutations H144Y and K224R due to allelic variations present in our cloned hIR cDNA (sequences available upon request). Mutant receptors R14AϩFLAG (CGG mutated to GCG at residue 14), F64A ϩFLAG (TTC mutated to GCC at residue 64), and F714GϩMyc (TTC mutated to GGC at residue 714) were generated by in vitro mutagenesis of the WT midi-hIR plasmid using the QuikChange site-directed mutagenesis kit from Stratagene. The entire cDNA insert in each mutant midi-hIR was verified by DNA sequencing. The midi-hIR constructs were transiently expressed by transfection into 293H cells using Lipofectamine 2000 reagent (Invitrogen). Medium was collected 72 and 144 h post-transfection, concentrated 7-fold by ultrafiltration (Amicon Ultra 15; Millipore), adjusted to 50 mM Hepes, pH 7.5, 0.02% NaN 3 , and stored at 4°C until used.
Insulin Binding Assay-Binding of 125 I(A14)-insulin by the midi receptors was assayed by the polyethylene glycol precipitation method (10). Samples containing midi-hIR were incubated in 200 l of binding buffer (100 mM Hepes, pH 7.5, 100 mM NaCl, 10 mM MgCl 2 , 0.25% (w/v) bovine serum albumin, 0.025% (w/v) Triton X-100) containing 30,000 cpm (ϳ10 pM) 125 I(A14)-insulin overnight at 4°C. 500 l of 0.2% ␥-globulin and 500 l of 30% (w/v) polyethylene glycol M r 8000 were added to co-precipitate the insulin-receptor complex, the sample was microfuged for 10 min, the supernatant was removed by aspiration, and the pellet was counted in a ␥ counter. Specific binding was determined by subtracting background radioactivity obtained with only binding buffer. For competitive displacement assays, unlabeled human insulin was added to the samples in the concentrations indicated.
Chemical Cross-linking and Immunoprecipitation-Disuccinimidyl suberate (DSS) was used to covalently cross-link 125 I(A14)-insulin to the midi-hIR. After an overnight incubation in binding buffer at 4°C, samples containing the insulinreceptor complex were treated with 10 mM DSS (freshly dissolved in Me 2 SO) added to a final concentration of 0.2 mM and held on ice for 30 min. The reaction was stopped by the addition of 1 M Tris-HCl, pH 7.5, to 20 mM, and an aliquot was mixed with 4ϫ loading buffer for SDS-PAGE analysis. For immunoprecipitation, the cross-linked 125 I(A14)-insulin-receptor complex was diluted in binding buffer containing 0.1% Triton X-100, 0.1% sodium deoxycholate, and 3 g of anti-Myc or anti-FLAG antibody. After 2 h of incubation, 30 l of protein A-Sepharose (GE Healthcare) was added, and the mixture was rocked on a mixing platform for an additional 2 h at 4°C. The immune complexes were then pelleted by centrifugation in a microfuge for 2 min, rinsed with ice-cold buffer containing 20 mM Tris-HCl, pH 7.5, 150 mM NaCl, resuspended in 1ϫ Laemmli loading buffer (11), and analyzed by SDS-PAGE.

SDS-PAGE and
Immunoblotting-Samples were dissolved in Laemmli loading buffer in the presence (reducing) or absence (non-reducing) of 2% ␤-mercaptoethanol, heated at 90°C for 3 min, and electrophoresed on 6% acrylamide slab gels. For autoradiography, gels were fixed for 20 min in 10% acetic acid/25% isopropanol (v/v), dried onto Whatman 3MM paper, and exposed to x-ray film (Kodak Biomax MS) or a phosphor screen (Packard Cyclone) for autoradiography. For immunoblotting, proteins in gels were electrophoretically transferred onto a polyvinylidene fluoride membrane (Immobilon P; Millipore). The membrane was blocked for 30 min with 5% nonfat dry milk in PBST (10 mM NaHPO 4 , 150 mM NaCl, 0.2% Tween 20, pH 7.4) and probed with rabbit anti-human IR␣ N-20 (Santa Cruz Biotechnology), which recognizes residues 1-20 in the IR ␣ subunit, mouse monoclonal antibody anti-FLAG M2 (Sigma), or rabbit polyclonal anti-Myc (Santa Cruz). The membrane was washed two times with PBST, reacted with secondary antibodies anti-rabbit or anti-mouse IgG conjugated to horseradish peroxidase, and the immunoreactive proteins were visualized with ECL reagent (Amersham Biosciences).

RESULTS
The expressed midi-hIR sequence corresponded to the mIR. Fno/Ex10 construct as described by Brandt et al. (6) and contained the hIR signal peptide plus aa 1-601 fused to aa 650 -719 of the ␣ subunit. In addition, a Myc (EQKLISEEDL) or FLAG (DYKDDDDK) epitope tag sequence was fused to the C terminus to uniquely label each receptor. After transfection with WTϩFLAG or WTϩMyc into 293H cells, immunoblot analysis showed that a predominately 260-kDa truncated IR dimer was secreted into the medium, which was reduced by mercaptoethanol to the 130-kDa monomeric form. The medium also contained high affinity insulin binding activity with a calculated IC 50 ϭ 0.27 nM (see Fig. 5), consistent with results presented by Brandt et al. (6).
Mutant midi-hIRs R14AϩFLAG, F64AϩFLAG, and F714GϩMyc were prepared by in vitro mutagenesis and analyzed for expression after transient transfection. Immunoblots revealed that each mutant receptor was expressed as a 260-kDa dimer in similar amounts to WT, but no insulin binding was detected (Fig. 1). These results show that alanine mutations at Arg 14 , Phe 64 ,and glycine mutation at Phe 714 that markedly   reduced insulin binding in the holo-IR (7,8) also inactivated insulin binding in the midi-hIR. However, when 293H cells were co-transfected with two complementary constructs, R14AϩFLAG and F714GϩMyc or F64AϩFLAG and F714GϩMyc, insulin binding activity was recovered in the medium. A control co-transfection performed with two noncomplementary mutants, R14AϩFLAG and F64AϩFLAG, yielded no detectable insulin binding activity, and this provided evidence that the midi-hIRs must contain complementary mutations (Fig. 1). However, we always obtained 2-6-fold less insulin binding activity than would be expected from the amount of hybrid receptor formed by random dimerization of each mutant ␣ subunit. The explanation for this discrepancy is not known, but one possibility is that mutant receptors may preferentially form homodimers due to subtle conformational changes introduced by the presence of different epitope tags at the C terminus or to the presence of the mutated residues. Thus, the amount of hybrid receptor present in the medium after co-transfection with two mutant midi-hIR may be less than expected based on the immunoblot result.
To demonstrate that the recovery of insulin binding activity was due to the formation of hybrid mutant receptors, several experiments were performed to characterize these receptors. First, when media from single mutant transfections were combined, no insulin binding activity was recovered, indicating that co-expression of the complementary mutant receptors in the same cell was necessary as would be the case in order to form the class I (␣-␣) disulfide bonds present in a hybrid dimer (data not shown). In addition, it has been shown that the hIR holo-receptor can be converted to ␣␤ half receptors by reduction of the class I disulfide bonds with low concentrations of dithiothreitol (DTT) and the half receptor retains insulin binding activity (12,13). Similarly, incubation of WT midi-hIR dimer with 2 mM DTT efficiently produced the monomeric ␣ chain that bound and chemically cross-linked 125 I(A14)insulin. However, all binding activity was lost when medium from complementary mutants was similarly treated (Fig. 2).
We also analyzed the hybrid mutant midi-hIR by performing successive immunoprecipitations with anti-FLAG and anti-Myc. In this experiment, 125 I(A14)-insulin was covalently cross-linked to R14AϩFLAG/F714GϩMyc with DSS and immunoprecipitated with anti-FLAG. The immunoprecipitated receptor was dissolved in 1% SDS, and an aliquot was re-immunoprecipitated with anti-Myc. Analysis of the samples by SDS-PAGE revealed that substantially all of the insulin-labeled receptors were immunoreactive with both antibodies, and this provides strong evidence that the active receptor is a covalent hybrid dimer (Fig. 3). However, if the 125 I(A14)-insulin-labeled R14AϩFLAG/F714GϩMyc receptor was first reduced to monomeric form by incubation with 2 mM DTT and subsequently immunoprecipitated with anti-FLAG and anti-Myc, only the R14AϩFLAG ␣ subunit was detected (Fig. 4). This result demonstrates that DSS cross-links 125 I(A14)-insulin specifically to the R14AϩFLAG ␣ subunit and therefore insulin was bound asymmetrically to the hybrid receptor. We have recently performed the same analysis with a hybrid midi-hIR in which the epitope tags were reversed, i.e. R14AϩMyc/ F714GϩFLAG, and in this case the cross-linked insulin was immunoprecipitated only with anti-Myc (data not shown). This result demonstrates that insulin cross-linking was not dependent on the epitope tag sequence.
Competitive displacement assays were performed to compare the binding affinity of the hybrid mutant receptors with that of WT. Fig. 5 shows representative competitive assays performed with WT, R14ϩFLAG/F714AϩMyc, and F64AϩFLAG/ F714GϩMyc. There were no significant differences in the calculated values for affinity binding constants with WT ϭ 0.27 nM Ϯ 0.38, R14AϩFLAG/F714GϩMyc ϭ 0.13 nM Ϯ 0.09, and F64AϩFLAG/F714GϩMyc ϭ 0.38 nM Ϯ 0.37. In addition, Fig. 5 shows that curvilinear Scatchard plots were also obtained for both WT and hybrid receptors. These results demonstrate that the hybrid mutant receptors bound insulin with complex kinetic properties similar to WT midi-hIR.

DISCUSSION
Utilizing alanine-scanning mutagenesis, Whittaker et al. (7,8) have reported that mutations at Arg 14 , Phe 64 , and Phe 714 in the holo-IR resulted in a marked loss of insulin binding activity.
Here we have shown that secreted truncated IR dimer constructs (midi-hIRs) with an alanine or glycine mutation at Arg 14 , Phe 64 , or Phe 714 were efficiently expressed and correctly folded but also exhibited no detectable insulin binding activity. However, by co-expressing two midi-hIRs with complementary mutations, R14AϩFLAG/F714GϩMyc or F64AϩFLAG/ F714GϩMyc, a hybrid receptor was obtained that bound insulin with high affinity similar to wild type midi-hIR and yielded a curvilinear Scatchard plot.
Several control experiments were performed to ensure that the insulin binding activity was attributable to the formation of a hybrid receptor. Thus, we showed that treatment with 2 mM DTT reduced the ␣-␣ disulfide bond in the dimer to yield inactive midi-hIR monomer. Similarly, insulin binding activity was not recovered when two preformed mutant midi-hIRs were incubated in vitro, and the co-expression of two noncomplementary mutants, i.e. R14AϩFLAG and F64AϩFLAG, did not yield active hybrid receptor.
We have also shown that 125 I(A14)-insulin is bound asymmetrically to the hybrid midi-hIRs containing R14AϩFLAG/ F714GϩMyc or F64AϩFLAG/F714GϩMyc. After cross-linking with DSS and reduction, the labeled A-chain was selectively immunoprecipitated with anti-FLAG. Because DSS can crosslink with the primary amine at Gly A1 , this implies that the insulin N-terminal A-chain binds to the C-terminal region of the ␣-subunit. This result is consistent with the recent finding reported by Wan et al. (14), who used photo-affinity-labeled insulin analogue to identify that Val A3 interacts with the C-terminal region of the IR ␣ chain.
The recovery of high affinity insulin binding activity in the hybrid mutant midi-hIR suggests that insulin can bind contralaterally to the two ␣ subunits in the holo-IR and this property may be important for IR signaling. It is now clear that the signaling mechanisms for a number of hormones, including growth hormone and the epidermal growth factors, involve ligand-mediated receptor dimerization (15)(16)(17). In the case of insulin, the conundrum has been that the IR already exists as a covalently linked dimer. Nonetheless, several models have been proposed in which a single insulin molecule is shown to bind both ␣ subunits, and this binding induces a conformational change in the IR to activate the intrinsic tyrosine kinase (18 -21). Our results suggest that by constructing a hybrid holo-IR with a complementary mutation in each ␣ subunit using sequences analogous to those described here, the validity of a cross-linking mechanism can be tested by characterizing hybrid holo-IR insulin binding and tyrosine kinase activities.