Insulin Receptor Activation with Transmembrane Domain Ligands

Background: The insulin receptor (IR) is a transmembrane (TM) receptor tyrosine kinase that regulates nutrient homeostasis. Results: Peptides derived from the IR-TM domain activate IR, independent of insulin binding. Conclusion: IR can be activated through a pathway that bypasses its canonical ligand-binding domain. Significance: Sites outside of canonical ligand-binding domains can be used as targets for the development of new therapeutic agents. A having the transmembrane but not related receptors, specific interac-tions the IR-TM have a competent ATP-bind-ing site and kinase activation from with which do not to that can be activated through a pathway that bypasses its canonical ligand-binding domain.

important questions about what natural ligands do, which features are sufficient for receptor activation, and whether receptor activation can be achieved using alternative methods.
Receptor oligomerization is believed to be a key step in the activation of a receptor (1), as it is believed to juxtapose cytoplasmic kinase domains and thereby facilitate their trans-phosphorylation, which stimulates catalysis and creates docking sites for downstream signaling proteins. Evidence supporting the necessity of oligomerization in receptor activation comes from studies using the following: (a) activating antibodies and lectins; (b) chimeric receptors containing the extracellular domain of one receptor linked to the intracellular domain of another; and (c) biochemically separated receptor monomers and oligomers (1). However, these studies do not distinguish whether ligands actually cause receptors to oligomerize or whether ligands stabilize preformed oligomers at the cell surface to shift equilibrium in that direction. It is also not clear whether oligomerization is sufficient for receptor activation or whether the binding of the ligand also induces additional conformational changes that facilitate phosphorylation in trans.
The insulin receptor (IR) 5 is a heterotetrameric membrane protein that is composed of two ␣ and two ␤ subunits (6,7). The ␣ subunit, which is the insulin-binding subunit, is totally extracellular and is linked to the extracellular part of the ␤ subunit as well as to the other ␣ subunit through disulfide bridges (8). The ␤ subunit is composed of an extracellular domain, a membranespanning transmembrane (TM) domain, and an intracellular kinase domain that can be activated by autophosphorylation. IR is very similar to other growth factor receptors except for the fact that its receptor protomers are covalently tethered. Thus, IR provides a special case where potential effects of ligand binding may be distinguished. Because insulin binding does not affect monomer/dimer equilibrium, the primary effect is presumably conformational. Lines of supporting evidence include the following: (a) studies using homodimeric chemical crosslinkers reveal differences in subunit contacts between ligandoccupied and -unoccupied IR (9, 10); (b) anti-receptor antibodies differentially recognize receptors depending on whether they are phosphorylated or insulin-bound (11)(12)(13)(14)(15); (c) the basal and the activated insulin receptor exhibit a differential response to sulfhydryl reagents (16); (d) changes in intrinsic fluorescence reveal real time conformational changes during discrete steps of receptor activation (insulin and ATP binding and autophosphorylation) (17); (e) analyses of x-ray crystallography structures of apoversus insulin-bound IR ectodomains show that insulin binding induces conformational changes in both insulin and the IR ectodomain (18 -20); and (f) comparisons between activation loops in nonphosphorylated versus phosphorylated IR kinase domains reveal similar conformational changes to those observed in other kinases (21,22). This raises the possibility that the ligand-induced conformational change is a primary function in other signaling systems as well. Studies with cytokine receptors support this hypothesis. For example, fluorescence energy transfer experiments indicate that in the absence of their ligands the erythropoietin (EPO) and IL-2 receptors are oligomerized and that ligand binding alters the conformation of these receptors (23,24). Moreover, crystal structures of extracellular EPO receptors reveal that they are dimeric, regardless of whether or not EPO is bound (5,25). These results indicate that in addition to dimerization, which is clearly important, ligand binding stabilizes receptor dimers at the cell surface and alters the structure of the receptor (5,26,27).
Growth factor, insulin, and cytokine receptor protomers each span the membrane only once. Bargmann et al. (28) have shown that Val 644 to Glu substitutions within the TM domain of HER2/erbB2/Neu activates this receptor tyrosine kinase in the absence of a ligand (29). We subsequently showed that chimeric IR containing the HER2/erbB2/Neu TM domain were similarly activated by the Val 3 Glu mutation and suggested that dimerization at the level of the TM domain may be sufficient to activate receptor tyrosine kinases (30,31). Subsequent studies showed that related TM domain substitutions activate the EGF, PDGF, IGF-1, and FGF receptors, presumably by similar mechanisms (30,(32)(33)(34). Based on the hypothesis that TM domain homodimerization may cause the cytoplasmic domains to juxtapose, thereby activating the receptors, we reasoned that exogenous ligands that bind to receptor TM domains may also be able to modulate receptor function. We further reasoned that these exogenous ligands may be synthetic peptides that correspond to TM domain sequences and that may interact with TM domains.

EXPERIMENTAL PROCEDURES
Peptide Synthesis-Peptide chains were assembled on Fmocamino acid Wang resin with the FastMoc protocol by using an Applied Biosystems 430A peptide synthesizer. Peptides were deprotected and cleaved from the resin using a mixture of TFA, thioanisole, phenol, ethanedithiol, and water (90:5:5:2.5:5) and purified by gel filtration and reverse phase-HPLC using a nitrile column. In certain cases, carboxyfluorescein was introduced using diisopropylcarbodiimide/1-hydroxybenzotriazole after removing the final Fmoc group. Peptide sequences were confirmed by mass spectrometry. Alternatively, the IR-TM peptide was synthesized and purified by the Tufts University Peptide Synthesis Facility.
Preparation of the Insulin Receptor-IR was purified from NIH3T3 cells expressing human IR (35). Briefly, the plasma membrane fraction was treated for 1 h in 30 mM HEPES buffer, pH 7.6, containing 2% Triton and 0.02% NaN 3 . Solubilized receptor collected from the supernatant following centrifugation at 10 5 ϫ g for 60 min was loaded onto wheat germ agglutinin (WGA)-agarose overnight. After extensive washing with HTA (30 mM HEPES, pH 7.4, 0.1% Triton X-100, and 0.02% NaN 3 ), the bound proteins were eluted with 0.3 M N-acetylglucosamine in 30 mM HEPES, pH 7.4, containing 0.1% Triton X-100. Column fractions containing insulin binding activity were pooled and stored at Ϫ80°C.
Exogenous Kinase Activity-The exogenous kinase activity of the IR was examined using poly(Glu/Tyr) (4:1) as the substrate. The TM peptides were first dissolved in trifluoroethanol (10 mM) and then diluted in HTA (10 M). The final concentration of trifluoroethanol in the reaction mixtures was 0.1%. WGApurified IR was preincubated with the TM peptides for 2 h at 21°C or overnight at 4°C. Then the mixtures were further incubated with insulin for an additional hour at 21°C and a mixture of ATP (50 M), MgCl 2 (10 mM), and MnCl 2 (8 mM) for 5 min at 21°C. To initiate substrate phosphorylation assays, [␥-32 P]ATP (5 Ci/sample) and poly(Glu/Tyr) (4:1) (0.3 mg/ml final concentration) were added. Reactions were stopped after 5 min by adding 67 mM EDTA. Aliquots of the reaction mixtures were spotted onto 3 ϫ 3-cm squares of Whatman No. 3 MM filter paper, which were washed extensively with 10% (w/v) ice-cold trichloroacetic acid (TCA) in 20 mM sodium pyrophosphate, dried, and counted.
Cellular Studies-Chinese hamster ovary (CHO) cells expressing either wild type (CHO-IR WT ) or mutated human IR (CHO-IR PDGFR/TM , -IR K1006A , or -IR Y3F ) were grown to confluence in 6-well dishes containing DMEM supplemented with 10% fetal bovine serum (FBS) and 0.45 mg/ml G415. Prior to experiments, the cells were quiesced overnight in DMEM containing 0.2% BSA. To initiate experiments, the cells were incubated with the peptides for 30 min and then with insulin (10 Ϫ10 M) for 5 min, all at 37°C. The peptides were originally dissolved in DMSO (10 mM) and subsequently diluted with DMEM (the final concentration of DMSO in the culture medium was 0.5%). To terminate the assays, cells were washed with ice-cold PBS containing a protease and phosphatase inhibitors (1 mM PMSF, 1 M leupeptin, 0.3 M aprotinin, 1 M pepstatin, 25 mM benzamidine HCl, 100 mM NaF, 2 mM sodium vanadate, 30 mM sodium pyrophosphate, 1 mM ammonium molybdate, 2 mM EGTA). Cells were subsequently lysed with 1% Triton X-100 in 30 mM HEPES buffer containing the same protease and phosphatase inhibitors. Activation of insulin signaling was determined by Western blot. Proteins were immunoprecipitated from cell lysates for 2 h at 4°C using immobilized anti-IR and anti-IRS-1 antibodies (protein A-Sepharose beads, Amersham Biosciences) (36). After extensive washing, proteins were eluted with Laemmli sample buffer, separated by SDS-PAGE, and transferred to Immobilon PVDF membranes (Millipore). Blots were probed with the indicated antibodies, and proteins were detected by chemiluminescence (Pierce). The activations of the MAP kinase and AKT were determined in the total cell lysates by using phosphospecific MAP kinase (Promega) and AKT (Ser-473; Cell Signaling) antibodies.
Confocal Microscopy-CHO-IR cells were treated with 1 M fluorescein-labeled IR-TM peptide as described above, washed extensively, and fixed with 4% formaldehyde in PBS. Cells were then incubated with IR antibody (4 g/ml) and a rhodaminelabeled secondary antibody examined using a Zeiss confocal microscope.
Materials-CHO cells that express wild type or mutant IR (CHO-IR WT , CHO-IR PDGFR/TM , -IR K1006A , and -IR Y3F ) have been described previously (30,37). NIH3T3 cells that express wild type or mutant IR were generously provided by Dr. Simeon Taylor (38). WGA-agarose was purchased from EY Labs. The anti-phosphotyrosine (4G10) and anti-IR antibodies that were used for confocal studies were from Millipore; the anti-activated MAPK antibody was from Promega, and the rhodaminelabeled donkey anti-rabbit IgG was from Jackson ImmunoResearch. All other reagents were from Sigma.

RESULTS
In Vitro Activation of IR by the IR-TM Peptide-Although the TM domains of insulin, growth factor, and cytokine receptors appear to adopt similar hydrophobic ␣-helical structures, the sequences of these membrane-spanning domains vary dramatically (Table 1). Even otherwise highly homologous receptors differ significantly within their TM domains. For example, insulin and IGF-1 receptors share 56% identity overall, but their TM domains are only 21% identical. We decided to test whether exogenous TM sequences altered receptor activity and, if so, whether this was sequence-specific.
Peptides corresponding to various TM domain sequences (Table 1) were incubated with solubilized IR, and the effects on tyrosine kinase activity were assessed. In the absence of insulin, the insulin receptor TM (IR-TM) peptide stimulated a 3-4-fold increase in the IR-catalyzed phosphorylation of the artificial substrate, poly(Glu/Tyr) (Fig. 1). Stimulation was concentration-dependent and showed a logarithmic dose-response profile (ED 50 ϳ1 M). Under identical conditions, insulin stimulated a maximal 6-fold increase in substrate phosphorylation, with an ED 50 value of ϳ0.02 M (Fig. 1). When submaximal insulin and IR-TM were combined, their effects were additive. However, IR-TM had no additional effects when added to maximal insulin (e.g. 10 Ϫ7 M). This suggests that the two peptides might activate the IR by similar mechanisms.
To begin to assess a mechanism for IR-TM-mediated activation of IR, we asked whether IR-TM competed for or otherwise altered insulin binding. Binding of 125 I-labeled insulin was not affected by IR-TM (data not shown), suggesting that IR-TM and insulin bind at distinct nonoverlapping sites. Because the IR-TM peptide stimulates IR kinase activity in the absence of insulin, activation is not through the modulation of insulin binding. Instead, the IR-TM peptide appears to be an insulinmimetic as opposed to an insulin sensitizer.
Activation of IR in Cells-Next, we determined whether the IR-TM peptide could activate insulin signaling responses in cultured cells. We first examined the subcellular localization of IR-TM peptide by using confocal microscopy. FITC-labeled IR-TM exhibited a peripheral pattern of fluorescence in CHO-IR cells, consistent with incorporation of the peptide into the plasma membrane ( Fig. 2A, left panel). As expected, immunofluorescence staining with a rhodamine-labeled antibody confirmed that IR was also located in the plasma membrane ( Fig. 2A, middle panel). Overlays showed that IR and IR-TM were co-localized in the plasma membrane ( Fig. 2A, right  panel).
We next asked whether the peptide had related bioeffects in cells. Thirty min after addition of the IR-TM peptide, cells were lysed, and IR phosphorylation was determined by Western blotting (Fig. 2B). The IR-TM peptide stimulated a dose-dependent increase in IR tyrosine phosphorylation by 2.3-fold at a concentration as low as 1.25 M and increased it to 13.6-fold at 10 M (the limit of peptide solubility), compared with the control (Fig. 2B, lane 1). When CHO-IR cells were incubated with the IR-TM peptide prior to stimulation with submaximal insulin, the effects appeared to be additive (Fig. 2B, lane 6 versus  lane 7). Tyrosine phosphorylation IRS-1 increased in parallel with IR phosphorylation (Fig. 2B), indicating that the IR-TM peptide activated IR in cells as does insulin. At 10 M, the highest dose of the IR-TM peptide used, the level of IR or IRS-1 phosphorylation was roughly equivalent to that seen with a submaximal 10 Ϫ9 M dose of insulin. Concomitant with its stimulation of proximal signaling, the IR-TM peptide stimulated AKT and MAPK, both of which are more distal end points downstream in the insulin action cascade (Fig. 2B).
Specificity of TM Domain Sequence-The specificity of IR-TM-mediated receptor activation was studied using two approaches. We first tested whether alternative peptide  sequences activated IR. Peptide sequences corresponding to the TM domains of other receptor tyrosine kinases failed to activate IR in in vitro kinases assays (Fig. 3A). Sequences studied included those corresponding to the TM domains of the IGF-1, PDGF, HER2/erbB2/Neu, and TrkA receptors. Similar studies conducted with CHO-IR cells yielded identical results (Fig. 3B). IR-TM activated both substrate phosphorylation and IR autophosphorylation, whereas TM domain peptides from heterologous receptors did not.
As a second approach to address specificity, we used a chimeric IR in which the TM domain had been substituted with the corresponding domain from a heterologous receptor. The sequence of IR PDGFR/TM is identical to WT IR, with the exception of the TM domain, which has the sequence of the PDGF receptor (30,31). Although IR-TM consistently activates WT IR, IR-TM was unable to activate IR PDGFR/TM . In contrast, insulin activates IR PDGFR/TM normally (30,31).
As additional controls, we compared the ability of either insulin or IR-TM to activate two kinase-defective receptors. The single amino acid substitution in the ATP-binding site of IR K1006A renders it incapable of catalyzing phosphoryl transfers, whereas the three tyrosines in the activation loop of IR Y3F were substituted with phenylalanine (39,40). Neither insulin nor IR-TM was capable of activating either receptor (Fig. 4). Our findings suggested that IR-TM might interact with IR through its TM domain.
Activation of "Unresponsive" IR-Numerous mutations have been identified in the IR of patients with extreme insulin resistance, including some that block insulin binding (41,42). The S323L substitution within the insulin-binding site of IR abolishes the insulin binding without affecting surface expression or intrinsic kinase activity (38). We have found that IR S323L , expressed in NIH3T3 cells, is activated by IR-TM but not insulin (Fig. 5). In fact, IR-TM activates IR S323L as potently as it activates WT IR. These observations are consistent with the notion that IR-TM bypasses the insulin-binding site of IR and After washing the cells and lysing them, the IR and IRS-1 in the lysates were immunoprecipitated (IP) with specific antibodies and resolved by SDS-PAGE. The proteins were transferred onto PVDF membranes and subjected to Western blot analysis using an anti-phosphotyrosine antibody or antibodies specific for the IR or IRS-1. The proteins were visualized with chemiluminescence. Activated MAPK or ATK in total cell lysates was detected by an anti-phosphospecific MAPK or an anti-phospho-specific AKT antibody (Ser-473), respectively. Numbers represent the folds of specific phosphorylations (phosphorylation/amount of the protein). IB, immunoblot.

FIGURE 3. Ability of peptides corresponding to the TM sequences of other receptors to activate the IR in vitro and on cells.
A, peptides corresponding to the TM domains of the IR, IGF-1, PDGF, erbB2, and TrkA receptors were incubated at 10 M with the WGA-purified IR overnight at 4°C. The tyrosine kinase activity of the IR was then measured as described in the legend of Fig.  1 in the presence or absence of 10 Ϫ8 M insulin. *, p Ͻ 0.05. B, CHO-IR cells were grown as described previously and treated with the TM peptides of various receptors at 5 M for 30 min at 37°C. The cells were washed and lysed, and the phosphorylation status of the IR was analyzed as described in Fig. 2. Numbers represent the folds of specific phosphorylations (phosphorylation/amount of the protein). IP, immunoprecipitation; IB, immunoblot.
activates IR by interacting at an alternative site, possibly the TM domain.

DISCUSSION
We have found that a synthetic peptide corresponding to the TM domain sequence of IR possesses the ability to both activate purified IR and IR in cultured cells. The effects of submaximal doses of IR-TM and insulin are additive (Figs. 1 and 2). It appears that the IR-TM peptide influences the TM domain of the IR, as substituting the native TM domain abrogates IR-TM effects without altering insulin's ability to activate the receptor (Fig. 4). Moreover, as shown in Fig. 5, a mutation in the IR ␣ subunits that abolishes insulin binding does not prevent the receptor from being activated by the IR-TM peptide. These data clearly demonstrate that the sites of interaction (signal initiation) are distinct between the two ligands, insulin and IR-TM peptide, but that they have indistinguishable net effects on the receptor kinase and downstream signaling. This suggests that insulin and IR-TM activate IR through similar mechanisms.
Conformational changes are key steps in the activation of the IR, including those caused by insulin binding to the IR ectodomain (18,19,43) and ATP binding to the intracellular IR kinase domain (21,22). Although the nature of the conformational changes that are induced by IR-TM are unclear, studies on other membrane proteins provide some general insights into the structural changes that TM domains can undergo. For example, the dimerization of TM domains in glycophorin A has been well characterized by various biochemical and mutational studies (44,45). The structure has also been resolved by an NMR technique (46). These studies show that the glycophorin A TM domain is constitutively dimerized as two parallel ␣-helices held at a constant distance by a network of hydrogen bonding between the amino acids facing each other on the inner dimer surface. Because of the high degree of hydrogen bonding between the two helices, the dimerized TM domains cannot even be dissociated during SDS-PAGE, a technique that has been used to analyze the effects of TM domain dimerization in the insulin and EPO receptors. When the TM domain of the IR was replaced with the TM domain of glycophorin A, the ␤ subunits of the chimeric receptor remained dimerized in conditions that reduced the wild type receptor ␤ subunits and rendered them entirely monomeric (47). This indicates that the chimeric receptor can exist as a heterotetrameric (functional dimeric) receptor even in the absence of disulfide bonds between the two ␣ subunits (48). When the chimeric receptor was tested in vitro and in cells, it responded to insulin less well than wild type IR, which suggests that IR activation requires the separation of the two TM domains. This notion was supported by the structure of IR, as determined using cryoelectron microscopy. This analysis showed that the distance between the two TM domains was increased by insulin binding to the ␣ subunits (49 -51). By contrast, the TM of the EPO receptor appears to behave differently during the activation of this receptor. Comparison of the crystal structures of the liganded and unliganded EPO receptors suggests that the binding of the ligand initiates lateral (translational) movements that juxtapose the cytoplasmic kinase domains and promote trans-phosphorylation (Fig. 6A). Replacement of the EPO receptor TM domain with that of glycophorin A revealed that this dimerization event was independent of the ligand used and that this replacement of the TM domain did not impair the activation of the receptor by the EPO ligand (52). Thus, the TM domain replacement experiments with two different receptor types reveal completely different outcomes and suggest that the activation of some receptors requires that their TM domains be dimerized, although other receptors require the separation of their TM domains.
Based on our findings and those of others, we propose that the TM domains of IR in the nonactivated basal state are constitutively dimerized and that when insulin binds the ␣ subunits, the TM domains dissociate from one another (Fig. 6B). , or wild type receptors were grown as described previously and treated with 5 M IR-TM peptide for 30 min at 37°C. After the cells were washed and lysed, the IR was immunoprecipitated (IP) and its phosphorylation state was determined by Western blotting analysis using antibodies specific for phosphotyrosine and the IR. IB, immunoblot. FIGURE 5. IR-TM peptide can activate a mutant IR that lacks insulin binding activity. NIH3T3 cells that express either the wild type receptor or an IR mutant that lacks insulin binding activity (S323L) were grown to confluence in 6-well plates as described previously, treated with 5 M IR-TM peptide for 30 min at 37°C, and then stimulated with 10 Ϫ9 M insulin for 15 min. After the cells were washed and lysed, the IR was immunoprecipitated and its phosphorylation was assessed by Western blot analysis using antibodies against phosphotyrosine and the IR. The bands were quantitated with densitometry, and the fold of specific phosphorylation (phosphorylation of the IR/amount of the IR) was calculated. The results are shown with the average of three independent experiments. *, p Ͻ 0.05; **, p Ͻ 0.01.
This transition in turn brings the two ␤ subunits of IR into closer proximity (53) and results in IR autophosphorylation and the activation of the kinase domains of the ␤ subunits through a trans-phosphorylation mechanism (54). We further propose that the IR-TM peptide directly intercalates the interfaces of the two helices of the TM domains, thereby disrupting their interaction and resulting in the same dissociation of the TM domains that is induced by the binding of insulin to the receptor (Fig. 6C). This causes the ␤ subunits to transit from their basal state to an activated state, which promotes tyrosine kinase domain activation in the absence of insulin binding. However, at present, we cannot exclude possible alternative mechanisms. For example, instead of the two ␤ subunits of IR staying apart in the basal state (Fig. 6B), they may be interlocked with each other, an interaction that is then disrupted by insulin and IR-TM peptide binding to the IR; this in turn activates the IR kinases (43,55). It is also possible that the protomers may rotate on ligand or IR-TM peptide binding such that the intracellular kinase domains are similarly aligned with little energetic expenditure. This is the mechanism proposed for the activation of the growth hormone (GH) receptor (Fig. 6D) (56,57); comparison of the free and ligand-bound forms of the GH receptor in terms of the extracellular domain structures reveals that ligand binding causes one receptor chain to rotate toward the other receptor chain, causing the two C-terminal extracellular domains of the receptor to bind together (58). This rotation of extracellular domains has also been proposed to induce the rotation of the TM domain, which in turn causes the GH receptor to become fully activated. An elegant study by Waters and co-workers (58,59) supports this notion. They showed that forcing the GH receptor TM domain to rotate by inserting Ala into the ␣-helix of the TM domain can activate the GH receptor in the absence of GH binding (58,59). Interestingly, the NMR structure of the monomeric IR-TM domain in detergent micelles shows an ␣-helix structure with a kink that is induced by Pro-961 (60). This resembles the proposed model for the GH receptor TM domain (Fig. 6D) and suggests that IR may be activated by a mechanism similar to the one activating the GH receptor. However, the molecular mechanism by which IR is activated will not be fully understood until high resolution images of the holo-IR structure become available.
IR has been targeted for many years by researchers seeking small molecule insulin mimetics, which has had some success. Insulin mimetics identified to date include peptide dimers obtained by phage display screening that can bind the canonical insulin-binding sites (61, 62), a peptide from the MHC I antigen that binds to the extracellular domain of the IR and regulates the endocytosis of the receptor (63,64), and small organic compounds that target the intracellular kinase domain of the receptor (65,66). Our findings open up the possibility of targeting the IR-TM domain with novel therapeutic agents. However, identification of such agents would require considerable effort, including the screening of chemical libraries for compounds that activate the insulin holoreceptor. Such efforts would be complicated by the hydrophobic nature of the membrane. More importantly, our findings show that sites outside the canonical ligand-receptor-binding site can serve as potential targets in drug discovery research that aims to identify activators of receptor signaling pathways.