Insulin Receptor Autophosphorylation Occurs Asymmetrically*

The unoccupied insulin receptor is a structurally symmetric, disulfide-linked dimer, comprising two CUB halves, each with a potential insulin binding (Y subunit and a kinase active ,LY subunit. In the accompanying paper (Shoelson, J. Biol. Chem. we described the utility of a novel insulin analogue, ~-benzoylphenylalanine~~~,B29~-biotin in- sulin (BBpa insulin)’ as a probe for receptor behavior, and we determined that binding and cross-linking of one BBpa insulin molecule could fully stimulate insulin receptor autophosphorylation. Here we use this analogue to determine the symmetry of the autophosphor- ylation reaction. The CX@ half-receptor that does not covalently couple to BBpa insulin incorporates more orthophosphate than the a@ half that becomes coupled to the insulin analogue. Phosphopeptide mapping of each receptor half shows minimal differences in the phosphorylation sites or their relative contribution to the phosphate content of each half. The ki- netics of 32P incorporation into each receptor half are essentially identical over a 10-20-min time course. Phosphopeptide mapping analysis reveals that the phosphate incorporation patterns do not change between

The unoccupied insulin receptor is a structurally symmetric, disulfide-linked dimer, comprising two CUB halves, each with a potential insulin binding (Y subunit and a kinase active , LY subunit. In the accompanying paper (Shoelson, S. E., Lee, J., Lynch, C. S., Backer, J. M., and Pilch, P. F. (1993) J. Biol. Chem. 268, 4085-4091), we described the utility of a novel insulin analogue, ~-benzoylphenylalanine~~~,B29~-biotin insulin (BBpa insulin)' as a probe for receptor behavior, and we determined that binding and cross-linking of one BBpa insulin molecule could fully stimulate insulin receptor autophosphorylation. Here we use this analogue to determine the symmetry of the autophosphorylation reaction. The CX@ half-receptor that does not covalently couple to BBpa insulin incorporates 50% more orthophosphate than the a@ half that becomes coupled to the insulin analogue. Phosphopeptide mapping of each receptor half shows minimal differences in the phosphorylation sites or their relative contribution to the phosphate content of each half. The kinetics of 32P incorporation into each receptor half are essentially identical over a 10-20-min time course. Phosphopeptide mapping analysis reveals that the phosphate incorporation patterns do not change between the two a@ half-receptor forms (BBpa insulinlinked and unlinked, respectively) at different time points or concentrations of ATP ranging from 12 to 200 WM. Based on these and other data, we propose a model of insulin receptor activation whereby binding of one insulin molecule can trigger autophosphorylation in an asymmetric fashion.
The insulin receptor is a member of the ligand-activated receptor/tyrosine kinase family of transmembrane signaling proteins. Binding of a hormone/growth factor ligand to the extracellular portion of its corresponding receptor results in allosteric regulation of the receptor's intrinsic tyrosine kinase * This work was supported by National Institutes of Health Grants DK43123 (to S. E. S.), DK36424 (to P. F. P.), National Science Foundation Grant DMB 90-04670 (to S. E. S.), and a Diabetes and Endocrinology Research Center Award to the Joslin Diabetes Center (DK36836). 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.
To whom correspondence should be addressed Dept. of Biochem- activity in the cytoplasm of the cell (Ullrich and Schlessinger, 1990). For the epidermal growth factor and piatelet-derived growth factor receptors, ligand-dependent allosteric regulation of kinase activity requires the non-covalent association of two monomers formation to form functional dimers (Ullrich and Schlessinger, 1990). As discussed in the previous paper, the insulin receptor is also a functional dimer, but together with the closely related insulin-like growth factor-1 receptor, it is unusual in that the dimeric structure is maintained in the absence of ligand by a covalent disulfide linkage (the class I disulfide) (Massague and Czech, 1982;Boni-Schnetzler et al., 1986;Sweet et al., 1987). Beyond the requirement for receptor dimerization, the subsequent mechanistic steps by which ligand binding activates kinase activity is largely unknown for the insulin receptor and the other members of the receptor tyrosine kinase family.
Because the insulin holoreceptor is composed of two structurally identical ab-heterodimers (Massague et al., 1980;U11rich et al., 1985;Ebina et al., 1985), it should have two insulinbinding sites/holoreceptor. However, most experimental evidence indicates that only one insulin molecule binds with high affinity to one insulin holoreceptor a t physiological concentrations of insulin. Three lines of evidence for this conclusion have been reported. First, insulin binding to cells and membranes was shown to exhibit negative co-operativity as determined by Scatchard analysis indicating that binding of one insulin molecule to the receptor made the binding of the second molecule more difficult (DeMeyts et al., 1973). Second and in confirmation of this hypothesis, purified ab-heterodimers prepared by mild reduction of the class I disulfides show only low affinity binding with a stoichiometry of one insulin/ a@-heterodimer, whereas purified a&-holoreceptor exhibits negative co-operativity and only one high affinity insulinbinding site (Boni-Schnetzler et al., 1987;Sweet et al., 1987). Third, double probe analysis using two different insulin analogues showed that only one analogue at a time could bind to the receptor with high affinity Shafer, 1983,1984).
In the previous paper (Shoelson et al., 1993), we have described ~-benzoylphenylalanine~~~,B29~-biotin insulin (BBpa insulin) and its utility for probing insulin receptor structure-function correlates and receptor trafficking. In particular, we confirmed that only one insulin can be cross-linked to one insulin holoreceptor by means of the gel shift assay. Under the conditions used, nearly all receptors were crosslinked and were separated from uncross-linked receptors by the addition of avidin and the gel shift. The fact that the insulin itself has no obvious intrinsic symmetry (Baker et al., 1988;Hua et al., 1991) but must interact with both ab halves for high affinity binding raises the possibility that autophosphorylation of the ( 3 subunits may be asymmetric with respect to insulin-receptor contact sites. Autophosphorylation of the insulin receptor is essential for the full activation of its protein tyrosine kinase activity toward exogenous substrates. There-fore, if autophosphorylation is an asymmetric process, receptor halves may differ in their interactions with cellular signal molecules. We show here that BBpa insulin can be used to study receptor symmetry under conditions where only those receptors covalently coupled to this ligand become phosphorylated.

EXPERIMENTAL PROCEDURES
Preparation of Insulin Receptors-The NIH-3T3 cells (line 1502) transfected with human insulin receptor cDNA were generously provided by Drs. Takashi Kadowaki and Simeon Taylor (National Institutes of Health, Bethesda, MD). Succinylavidin and insulin receptors were prepared as described in the previous paper (Shoelson et al., 1993). All experiments were performed in the presence of protease inhibitors (1 mM phenylmethylsulfonyl fluoride, 10 p~ leupeptin, 50 mTIU aprotinin, 1 p M pepstatin, 1 mM 1,lO-phenanthroline, 2.5 mM benzamidine-HCl, and 2 mM EDTA).
Gel Shift Assay-WGA-agarose-purified insulin receptors were incubated in the presence or the absence of M BBpa insulin overnight at 4 "C unless indicated otherwise and were photolyzed on ice with a 200-watt UV visible lamp (Oriel Corp.) with a 340-nm cut off filter. Receptors were then autophosphorylated in a solution typically containing 50 p~ ATP, 10 mM MgCl2, and 8 mM MnC12 in the presence of [y-32P]ATP(1 mCi/ml) at room temperature. The reactions were stopped by adding 60 mM EDTA. The preparations of the ap-heterodimeric receptor and gel shift assay were performed as previously described (Shoelson et al., 1993). Briefly for the gel shift assay, phosphorylated a&-holoreceptors or ap-heterodimeric receptors (Boni-Schnetzler et al., 1986) were incubated with a 30:l molar ratio of succinylavidin to insulin in Laemmli (1970) sample buffer containing 0.6% SDS (Pang and Shafer, 1983). In some cases, biotin was added in large excess over the concentration of succinylavidin, to block the reaction of succinylavidin and the biotinyl group of the insulin analogue (BBpa insulin). After the insulin receptor was separated on a 3-10% gradient SDS-polyacrylamide gel, it was analyzed by autoradiography. The gels were stained, destained, and dried. The insulin receptor was then visualized by autoradiography. The corresponding region was excised from the gels and the radioactivity was determined as Cerenkov counts/minute with 30% efficiency. Phosphopeptide Mapping Analysis-Phosphopeptide mapping analysis was analyzed by using an HPLC system as described in White et al. (1988). Briefly, the fixed and dried gel fragments containing the insulin receptor were washed with 20% (v/v) methanol for 12 h at 37 "C . After drying with a Savant Speed-vac, the gel fragments were rehydrated in a 2 ml of 50 mM NH4HC03, pH 8.0, containing 100 pg of TPCK-treated trypsin. After incubation for 8 h at 37 "C, an additional 100 pg of trypsin was added, and another incubation was performed for 16 h. The phosphopeptides were separated by a C18-reverse-phase column with a Beckman HPLC system. For the mobile phase, 20 mM phosphate buffer, pH 3.0, was used. Peptides were eluted by a linear gradient of acetonitrile from 5 to 25% during an 85-min interval at a flow rate of 1.1 ml/min. The fractions (1.0 min) were collected and counted as Cerenkov radiation.
Protein Assay-Protein amounts were determined by the Bio-Rad protein assay using bovine serum albumin as a standard. The insulin binding assay was performed as previously reported (O'Hare and Pilch, 1988). Polyacrylamide gel electrophoresis was performed according to Laemmli (1970).
Materials-The CIB-reverse-phase HPLC column and Bradford protein assay solution were obtained from Bio-Rad. The reagents for cell culture were purchased from GIBCO. WGA-agarose was from E. Y. Laboratories. [y-3ZP]ATP was from Amersham Corp. Acetonitrile for HPLC phosphopeptide mapping analysis was purchased from Fisher. TPCK-treated trypsin was from Worthington Enzyme.

RESULTS
In the accompanying article, we have shown that one molecule of BBpa insulin could be cross-linked to an a&-heterotetrameric receptor, and this complex was fully capable of autophosphorylation. While it was not clear whether one or both /3 subunits is undergoing phosphorylation under these conditions, the covalent linkage of BBpa insulin allowed us to address this question. There are three possible ways that autophosphorylation might occur that can readily be distin-guished by a gel shift assay (Fig. 1). The first possibility is that only the p subunit on the receptor half to which insulin is cross-linked becomes autophosphorylated (Fig. 1A). The second way is that autophosphorylation and covalent coupling of insulin occur exclusively on different ab-heterodimers (Fig.  1B). The last way is that autophosphorylation may occur on both p subunits, regardless of where insulin is bound and cross-linked (Fig. IC). These models would be differentiated by the 32P autoradiogram patterns of the ap-heterodimeric receptors in the gel shift assay as depicted in the idealized gels of Fig. 1.
The gel shift assay of the photocoupled and autophosphorylated insulin receptors was performed as shown in Fig. 2. Notably, all of the 32P-labeled azP2-receptors were shifted to the higher molecular weight form in the presence of succinylavidin (Fig. 2, lane 4 ) . This is consistent with the fact that only those receptors covalently linked to the BBpa insulin could be autophosphorylated. In other words, the photocoupling efficiency of the BBpa insulin was 100% in terms of phosphorylation activity, although the total efficiency of cross-linking of BBpa insulin is normally about 70% as reported in the companion paper (Shoelson et al., 1993). Following mild reduction, the phosphorylated ab-heterodimeric receptors migrated as two distinct forms in the presence of succinylavidin, shifted and non-shifted (Fig. 2, lane 8). These data agree with model C of Fig. 1 where autophosphorylation occurs on both p subunits of the insulin receptor. Biotin could inhibit the effects of succinylavidin without the loss of radioactivity (Fig. 2, lanes 5 and 9). Although the autophosphorylation occurred on both ap-heterodimeric receptors, the degree of phosphorylation was not equal. The shifted a/3 halfreceptors, those covalently linked to BBpa insulin, accounted for 40% of the radioactive phosphate and the non-shifted cupreceptors contained the other 60% (Fig. 2, lane 8). There was no loss of phosphate upon reduction of holoreceptors into a@heterodimers, and the 60:40 ratio of non-shifted to shifted species was observed under all experimental conditions (see below) over dozens of experiments.
To determine if there were any effects of UV irradiation on autophosphorylation activity and the autophosphorylation patterns, the conditions for photolysis were examined. There was no difference in the amount of 32P incorporation nor the amount of shifting whether photolysis was performed before or after autophosphorylation (data not shown). Without photolysis, no receptors were shifted in the presence of succinylavidin. We confirmed that autophosphorylation of the insulin holoreceptor is an intramolecular mechanism (Shia et al., 1983;White et al., 1984) under the assay conditions employed for photocoupling and autophosphorylation of BBpa-linked receptor. Also, the relative amount of 32P that was shifted by succinylavidin remained around 40%, with no significant difference over the range of insulin receptor concentration examined (data not shown).
Thus, the consistent result of quantitatively asymmetric receptor autophosphorylation could be explained if different tyrosine residues were reactive on each /3 subunit. Insulin receptor autophosphorylation sites can be separated into two clusters of tyrosine residues: three at positions 1146, 1150, and 1151 (numbering of Ullrich et al., 1985), and the other two within the COOH terminus a t residues 1316 and 1322 (White et al., 1988;Tavare and Denton, 1988;Tornquist et al., 1988). The 60:40 ratio of non-shifted to shifted receptor halves could be explained by the predominant use of one site exclusively on one receptor half. Thus, we determined the phosphorylation sites of each receptor half following photocoupling of BBpa insulin, autophosphorylation, and separa-

SDS-PAGE AUTORADIO -GRAPHY
AVIDIN tion of shifted from non-shifted ab-receptor heterodimers on SDS-polyacrylamide gel electrophoresis. Fig. 3A shows a representative phosphopeptide mapping profile by HPLC. The maps of shifted and non-shifted receptor were qualitatively very similar with the only difference being that pY1, the triphosphorylated peptide, migrated more slowly when derived from the shifted @receptor heterodimer in two-thirds of the maps. The significance of this is unknown. However, as expected from the results of Fig. 2, the radioactivity of trypsin-digested peptide from the shifted, BBpa insulin crosslinked ap-heterodimeric receptor was 50% lower than that from the non-shifted half-receptor. When the radioactivity of the each fraction was expressed as a percentage of total radioactivity/receptor half (Fig. 3B), phosphopeptideprofiles from the two a@ half-receptor forms were identical, meaning that all the phosphorylation sites were used in the same relative proportion on both the shifted and non-shifted apheterodimeric receptors. Fig. 3C shows data from five different experiments confirming that each phosphotyrosine peptide contributes to autophosphorylation asymmetry to the same extent.
A comparison of HPLC profiles from the insulin holoreceptor stimulated by native insulin with that stimulated by BBpa  C insulin revealed no differences in the phosphorylation sites or in the distribution of radioactivity in each peptide (data not shown). We were surprised that only quantitative asymmetry was seen between insulin-coupled and -uncoupled receptor halves, and we surmised that time of autophosphorylation and/or ATP concentrations might influence these results.
We therefore examined peptide mapping profiles for shifted and non-shifted receptors as a function of these parameters. The kinetics of incorporation of 32P into the a&-receptor and into each a@-heterodimeric half-receptor were essentially identical, and we observed no significant difference in the relative amount of 32P that was shifted by succinylavidin a t any time point (data not shown). We performed phosphopeptide mapping analysis by HPLC after 1, 3, 10, and 20 min of autophosphorylation, and the data showed no significant difference in the phosphorylation sites nor in the relative amount of 32P incorporation in each site (Fig. 4).
Finally, we investigated autophosphorylation patterns a t 12.5, 50, and 200 mM ATP. At all concentrations, the degree of shift in the a@ half-receptor was maintained at 40% (data not shown). Phosphopeptide mapping revealed that there was no difference in phosphorylation sites between shifted and  1 -5 ) or the dithiothreitol-reduced crlj half-receptor (/nnc>s 6-IO) were incubated in the ahsence (Inncps I-9,6. 7, and 1 0 ) or presence (lanes 4, .5,H, and 9 ) of succinylavidin. To compete with the hiotinyl group of HHpa insulin, exogenous hiotin was added (lanes 9 , 5 . 7. and 9). The samples were separated in 51 :{-IOr; gradient gel and visualized by autoradiography.
non-shifted receptor forms, although the proportion of each phosphopeptide was different at the various concentration of A'rP (Fig. 5) as has previously been shown (Kohanski and Schenker, 1991). IIISCIJSSION We wish to determine the mechanism by which insulin binding to the insulin receptor a subunit can transduce a signal to /-I subunit(s) result.ing in autophosphorylat.ion and activation of the receptor's tyrosine kinase activity toward exogenous substrates. Insulin binding occurs toward the amino terminus of the (r subunit with the first 400 amino acids being necessary but not sufficient for this binding (Kjeldsen et al., 1991;DeMeyts et al.. 1990;Yip et al., 1990;Schumacher ct al., 1991;Zhang and Roth, 1991). A conformational change occurs upon ligand binding that alters the interface between each (r and also, between each / j subunit Kohanski, 1988: Waugh andPerlman rt al., 1989: Baron et al., 1990. Aut.ophosphory1ation then occurs, most probahly by a "trans" mechanism where one n[f half-receptor initially phosphorylates the other (Treadway e t al., 1991; Frattali ct 01.. 1992). Hybrid recept.ors composed of a kinase inactive tu[$ heterodimer and a COOH terminus truncated cui$-heterodimer were shown to undergo insulin-stimulated autophosphorylation that could only occur by a trans mechanism, although a certain amount of autophosphorylation was attributed to a cis mechanism. Interehngly, this hybrid receptor, composed of a kinase inactive half and a kinase active half, was unable to mediate insulin-stimulated exogenous protein tyrosine kinase activity consistent with a dominant negative effect for receptor possessing one normal and one kinase defective half (Frattali et al., 1992). Other studies have been performed consistent with a cis mechanism of autophosphorylation.
When t,he expressed cytosolic portion of t.he / j suhunit monomer was tested, autophosphorylation occurred at the same sites as in the nat.ive receptor with one group concluding that this was a cis phenomenon occurring in monomers (Herrera et al., 1988;Villalba et al., 1989) and another concluding that oligomerization was required (Cohb et al., 1989). Another approach to this issue used trypsin-activated and -truncated recept.ors which contained the entire \ j subunit and part of the cr suhunit near its COOH terrninrrs and showed a cis autophosphorylation mechanism was ocnlrring (Shoelson ( s t al., 1991). The kinetics of nutol)hf)sy)horyl~ltic,rl l'or the trurlcated receptors, however, do not match that of the insulin holoreceptor. 'rhus, a11 o f the constructs descrit)cd i n this paragraph exhibit some almormal t)iochemic;ll properties a s cornpared t o native insulin holoreceptor.
For this reason, we have used RHpa insulin t o s t d y autophosphorylation within intact insulin holoreceptors kvith t h e following criteria in mind: 1 J the andogwe must hind t f J the insulin receptor without changing the receptor's structural and functional characteristics. 2 ) I t must cross-link t o insulin receptors with high specificity a n d efliciency so the receptorligand complex is amenatlle t o convcntional biochemical techniques. 3 ) 'The insulin nnnlogue must have A reporter group which would a l l o w one t o easily follow the covalently crosslinked insulin-insulin receptor complex. H H p a insulin does in fact meet all these criteria. A s r e p (~r t~d in the previous paper, the60-100"; cross-linkingefficiencyoftheanalo~~~e t o insulin receptors is exceptional with 5 ' ; ( l r less tleing achieved by previously descrihed techniques ISh~)elsc~n V I d . , 1 9 9 3 ) . Also. the hiotin-avidin interaction allowed the tlt4red w s y determination o f ligand-linked rvceptor ant1 rtwptor half'. A p p r opriate controls revealed that the 1)iotin-avitlin inter;lction w a s stoichiometric, and the cross-linking conditions wew tvit h o w effect on receptor st ructure m d f'unction.
We show for the first time that inslllin-ind1~cc.d ilutophosphorylation is asvmmet ric with 40'; o f t h e r~~c e~) t o r~i~~c o r~) o rated phosphate being found o n the same half' t o which Hl3pt1 insulin is photocoupled (Fig. 2). However, somewh;tt unexpectedly, the specific tyrosine residues phosphorylated are nearly identical on each receptor half as determined hy phosphopeptide mapping analysis using Hl'l,C (Fig. 3 (Figs. 4 and 5 ) . We had originally postulated that i f insulin binding was asymmetric. this might induce asymmetry i n the kinase domains such that different tyrosine residues might I)e phosphorylated on the half-receptor covalently coupled t o HRpa insulin as compared t o the other half. h u t this d o e s not appear to be the case. 'To explain our results, we propose the working nlodels shown in Fig. 6. The insulin-hinding site is dtq)ic.ted ;IS being between the trlf halves 1)ec;luse as previously d i s c w w l . high affinity insulin binding o f one insulin molecule requires hot h halves ofthe receptor, i.v. specific contact o f insulin with hoth trd-heterodimers. We propose that Iigi111d rcceptor cont;lc.t will consist of a predominant interaction with o n e receptor half (the right sidv ' I d in Fig. 6 ) ;mtl R less conlplett. c.ont;lc.t with the other C h f t ) half receptor since a second insulin can bind to the complex with the l o w affinity char;lc~tc~risti(. f J f insulin binding to an individual crd-insulin recc.ptrrr half (Roni-Schnetz1erc.t al., 1985: Sweet nt ol., 1985). The rcwntly described three-dimensional structure for the ext racellular domain of the human growth hormone receptor (human growth hormone-t)inding protein) Idr Yos VI nl.. lW2t provides a general precedent for the participation ( J I two receptor subunits in hinding one molecule ol'hormone. 1,ikr the insrllin receptor, the human growth hormone-t)intling protein is composed of'two symmetrical s u h n i t s t h a t hind only one ligand molecule with high affinity. alheit with non-cooperative k inetic interactions (runningham V I d . . 1 9 9 1 ). For the insulin receptor, ligand binding is proposed t o trigger autophosphor-

FIG. 3. Phosphopeptide mapping
analysis of shifted and non-shifted a@ half-insulin receptor. A , insulin receptor was phosphorylated with 50 p~ [r-32P]ATP for 3 min and separated as described under "Experimental Procedures." The gel fragments corresponding to the shifted and non-shifted receptor halves were excised, rehydrated, and digested extensively with TPCK-treated trypsin prior to separation of phosphopeptides by a CI8 reverse-phase HPLC as described under "Experimental Procedures.'' The p y l peak is a tri-phosphorylated peptide, pY4 and pY5 are bisphosphopeptide forms from the "tri-tyrosine" residues. pY2 and pY3 are the peptides from the COOH terminus region. B , the amount of 32P radioactivity in each peak from the shifted (closed) and non-shifted (hatched) cup half-receptors were added up and set to 100% for each half. The contribution of each peak to the total for each half was then calculated and these results are an average f S.E. of five HPLC phosphopeptide maps and are presented in B. The percentage shifted and non-shifted for each phosphopeptide is shown in C where the radioactivity in each phosphopeptide was set to 100%. Conditions were as exactly as described in Fig. 3 from HPLC profiles, the proportion of each peptide was calculated as the percentage of the total counts/minute in each a0 half-receptor, and were compared between the shifted (closed) and non-shifted (open) receptor species. Panel A shows the peptides from the tri-tyrosine region and panel B shows the peptides from the COOH terminus region. ylation in trans to the subunit to which insulin has become covalently linked as shown for step 1 in model A. This initial reaction would immediately be followed by phosphorylation of the second receptor half by the first, autophosphorylated half (step 2 ) which would occur to a lesser extent due to conformational differences of an unknown nature. We propose the trans mechanism of autophosphorylation for the insulin receptor based on the predominance (90% or more) of the trans pathway for receptors with one kinase active half (Boni-Schnetzler et al., 1988;Treadway et al., 1991;Frattali et al., 1992). Similar experiments also support a trans phosphoprylation mechanism for epidermal growth factor and platelet-derived growth factor receptors (King et al., 1988;Honneger et al., 1990;Kelly et al., 1991). In these cited examples, there can be no second step because the phosphorylated receptor half is kinase inactive, but this may also may be the case for the holoreceptor as shown in models B and C. These models depict transphosphorylations, but the site of phosphate incorporation is exclusively on the opposite side of the ligand photocoupling site (model B ) or exclusively on the same side (model C). These differences could arise in linked, and autophosphorylated with 12.5,50, and 200 p~ ATP for 3 min. The reaction was stopped by adding 60 mM EDTA. The peptides from the shifted and non-shifted half-receptors were prepared as described under "Experimental Procedures." From HPLC profiles, the proportion of each peptide was calculated as the percentage of the total counts/minute of each @receptor form and compared between the shifted (closed symbols) and non-shifted (open symbols). Panel A shows the peptides from the tri-tyrosine region, and panel B shows the peptides from the COOH terminus region. at least two possible ways. The first is that there may be sufficient flexibility in the covalently coupled insulin molecule such that it can interact with and trigger autophosphorylation via either receptor half. The carboxyl terminus of the insulin B chain is known to be flexible and moves upon binding of insulin to its receptor (Hua et at., 1991). The second possibility is that the site of covalent linkage of the insulin derivative may be very close to a-a contact sites with 60% of receptors becoming coupled to half, and the rest, the other half, thus leading to the observed distribution of phosphate following autophosphorylation.

10-
Model A can hypothetically be experimentally distinguished from models B and C because in the former case, all ligandbound receptors would be phosphorylated on both half receptors whereas this would not be the case for the latter models. However, the biotin group on BBpa insulin is not accessible to avidin when bound to undenatured receptor (data not shown). We are currently further exploiting the type of technology presented here for BBpa insulin, and we are synthesizing additional insulin derivatives in order to gain a further understanding of insulin receptor structure-function correlates regarding receptor autophosphorylation and their physiological consequences.