BpaB25 insulins. Photoactivatable analogues that quantitatively cross-link, radiolabel, and activate the insulin receptor.

Benzoylphenylalanine (Bpa), a photoactivatable amino acid was incorporated into the 25 position of the insulin B-chain by a combination of chemical synthesis and enzymatic semisynthesis. BpaB25 insulin binds specifically to the insulin receptor with an affinity approximately 40% that of native insulin. Addition of biotin or epsilon-amino-hexanoic acid-iminobiotin at the epsilon-LysB29 position to form BpaB25,B29 epsilon-biotinyl (BBpa) or BpaB25,B29 epsilon-(Aha-iminobiotin) analogues had little adverse effect on receptor binding affinity and provided a convenient handle for affinity purification, gel shift assays, and blotting of cross-linked complexes with avidin. The analogues were readily 125I-iodinated with the majority of 125I being incorporated at the TyrA14 position; the monoiodo-A14 derivative was easily separated from other forms by high performance liquid chromatography. Photolysis of a complex of the insulin receptor and either BpaB25 insulin or [125I]iodo-TyrA14,BpaB25 insulin yields a covalent insulin-receptor complex. The efficiency of cross-linking with these reagents was unusually high, ranging from 60 to 100%. Furthermore, cross-linking to the insulin receptor results in kinase activation in vitro, and in intact cells insulin receptor phosphorylation and internalization were both activated. Notably, even at saturating concentrations one molecule of BpaB25 insulin covalently cross-linked each alpha 2 beta 2 receptor, demonstrating that holoreceptor activation occurs with one high affinity insulin binding site occupied and, if a second insulin binds with lower affinity it must be in a different orientation. Bpa insulin analogues form a new class of photoaffinity reagents which facilitate studies relating insulin-insulin receptor structure and function.

Benzoylphenylalanine (Bpa), a photoactivatable amino acid was incorporated into the 2 5 position of the insulin B-chain by a combination of chemical synthesis and enzymatic semisynthesis. BpaBZS insulin binds specifically to the insulin receptor with an affinity =40% that of native insulin. Addition of biotin or t-aminohexanoic acid-iminobiotin at the t-LysBZ8 position to form BpaBZs,B29'-biotinyl (BBpa) or BpaBZ6,B29'-(Aha-iminobiotin) analogues had little adverse effect on receptor binding affinity and provided a convenient handle for affinity purification, gel shift assays, and blotting of cross-linked complexes with avidin. The analogues were readily 12SI-iodinated with the majority of lzaI being incorporated at the TyrA14 position; the monoiodo-A14 derivative was easily separated from other forms by high performance liquid chromatography. Photolysis of a complex of the insulin receptor and either BpaBZS insulin or [12SI]iodo-TyrA'4,BpaB26 insulin yields a covalent insulin-receptor complex. The efficiency of cross-linking with these reagents was unusually high, ranging from 60 to 100%. Furthermore, cross-linking to the insulin receptor results in kinase activation in vitro, and in intact cells insulin receptor phosphorylation and internalization were both activated. Notably, even at saturating concentrations one molecule of BpaBZ5 insulin covalently crosslinked each aZ& receptor, demonstrating that holoreceptor activation occurs with one high affinity insulin binding site occupied and, if a second insulin binds with lower affinity it must be in a different orientation. Bpa insulin analogues form a new class of photoaffinity reagents which facilitate studies relating insulin-insulin receptor structure and function.
The insulin receptor is a complex, membrane-spanning * This work was supported by National Science Foundation Grant DMB 90-04670 (to S. E. S.), National Institutes of Health Grants R01 DK36424 (to P. F. P.) and R01 DK43123 (to S. E. S.), and Joslin's Diabetes and Endocrinology Research Center Grant DK36836. Preliminary results were presented in part at the Twelfth American Peptide Symposium (Shoelson et al., 1992b). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
$Recipient of a Career Development Award from the Juvenile Diabeters Foundation International. To whom correspondence should be addressed: Joslin Diabetes Center, One Joslin Place, Boston, MA 02215. Tel.: 617-732-2528. glycoprotein present on the surfaces of all insulin-responsive tissues. Early attempts to characterize the insulin receptor relied heavily upon cross-linking radiolabeled insulins to the receptor. In fact, the =130-kDa a-subunit of the receptor was first visualized by photoaffinity cross-linking with an "'Iazidoinsulin derivative (Yip et al., 1978) and by chemically cross-linking 1251-insulin with homobifunctional reagents (Sayoun et al., 1978;Pilch and Czech, 1979). These pioneering methods were used to determine the overall a& subunit structure of the insulin holoreceptor (Yip et al., 1978(Yip et al., , 1982Czech, 1979, 1980;Massague et ul., 1980Massague et ul., , 1981. Related approaches have been used through the years to address a wide variety of questions relating insulin receptor structure-function. However, these reagents have low intrinsic efficiencies and/or specificities, i.e. the proportion of insulin-receptor complexes that are cross-linked is typically low (<2-10%; Waugh et al., 1989).' Attempts to increase cross-linking efficiencies often leads to disproportionate increases in undesirable cross-reactions. More recent attempts to develop reagents with greater efficiency and selectivity have focused primarily on derivatizing the t -or a-amino groups of LysBZ9 or PheB1, respectively, with additional modified aryl azides (Thamm et al., 1980;Ng and Yip, 1985;Yip et al., 1988;Wedekind et al., 1989;Fabry et al., 1992).
We are interested in 1) mapping residues of the insulin receptor involved directly in insulin binding, 2) delineating mechanisms of transmembrane signaling by the insulin receptor, and 3) following the intracellular itinerary of activated insulin-receptor complexes. For such studies we have developed a different approach to photoaffinity labeling the insulin receptor which relies on incorporating the photoactivatable benzophenone moiety directly into the receptor-binding surface of insulin. Unlike diazo-and azido-based photoaffinity reagents which generate highly reactive but relatively nonspecific intermediates (Chowdhry and Westheimer, 1979), the photoactivated benzophenone group reacts selectively with C-H bonds (Breslow, 1980;O'Neil et al., 1989)
Peptide Synthesis-Solid-phase synthesis of Tfa-Gly-Phe-[DL-Bpal-Tyr-Thr-Pro-Lys-Thr-OH was performed on an Applied Biosystems model 430A synthesizer using standard dicyclohexylcarbodiimide-mediated preformed symmetrical anhydride coupling protocols. Racemic Boc-Bpa and Tfa-Gly were incorporated as preformed HOBt esters; a 2-fold excess of Boc-DL-Bpa over peptide resin, and a 2.0-h coupling time was used to minimize amino acid waste. For

syntheses of H-Gly-Phe-[DL-Bpal-Tyr-Thr-Pro-Lys(biotin)-Thr-
OH, lysine was incorporated as a preformed HOBt ester of a-Boc-Lys(r-Fmoc) (Bachem). Prior to proceeding with the rest of the synthesis the c-Fmoc protecting group was removed with 30% piperidine in a 1:l mixture of toluene and dimethylformamide and the ramino group was coupled with biotin for 4 h using 4-fold excesses each of (+)biotin (Aldrich), HOBt, and BOP (Milligen/Biosearch) in A. Arrow 1 denotes peptide bond between Are'' and GlyBZ3 which is formed during enzymatic semisynthesis; arrow 2 points to PheBZ5, the position of the Bpa substitution; arrow 3 points to LysBB, the position of biotin and Aha-iminobiotin substitutions. B, p-benzoylphenylalanine (Bpa). C, biotin-lysine or biocytin. D, Iminobiotin-lysine separated by an e-aminohexanoic acid (Aha) spacer. a 1:l mixture of dimethyl sulfoxide and N-methylpyrrolindone. For the remainder of the synthesis, reactions were as described above. A related peptide with an c-aminohexanoic acid (Aha) spacer was similarly prepared. Boc-Lys(Fmoc)-Thr(Bz1)-R was reacted with piperidine to remove the K-Fmoc group and K-Z-Aha was coupled by standard protocols. The synthesis was continued to yield, following cleavage, Tfa-Gly-Phe-[DL-Bpal-Tyr-Thr-Pro-Lys(Aha)-Thr-OH.
Final peptide products were cleaved from the resin, and side chain protecting groups were removed with an anhydrous mixture of 20% trimethylsilyl trifloromethanesulfonic acid, 12% thioanisole, 6% ethanedithiol, and 2% m-cresol in trifluoroacetic acid for 1 h at 4 "C (Yajima et al., 1988). Following filtration to remove the resin, peptides mixtures were extracted with ethyl acetate/hexane, and peptides were precipitated with ice-cold methyl$-butyl ether, desalted on a Bio-Gel P-2 column, and lyophilized. Isolated peptides were purified by reversed-phase HPLC (Waters' Prep 4000) on a Dynamax-300A 12-pm C8 column (41.4 X 250 mm) equipped with a matched guard column (Shoelson et al., 1992a).
sponding to L-Bpa peptide and D Bpa-peptide eluted from the HPLC.
Resolution of D-and L-Bpa-Peptides-Two major peaks corre-These were assigned by treating the Ne-deprotected peptides with leucine aminopeptidase (IUB 3.4.11.1, Worthington), which was activated by diluting 2 p1 of a stock solution (178 units/mg, 3.98 mg/ ml) in 0.1 ml of 0.2 M triethylamine acetate, 25 mM MnClz, pH 8.5, and incubating for 2 h at 37 "C. The peptides (0.1 mg) were dissolved in 50 p1 of the triethylamine acetate buffer, reactions were initiated by addition of 25 p1 of activated endopeptidase and allowed to proceed for 4 h at 37 "C, and samples were analyzed for release of free Bpa by reverse-phase HPLC (Miller and Kaiser, 1988).
To demonstrate that Bpa itself had not been iodinated and to determine which of the four insulin tyrosines had been, samples of transfected with human insulin receptor constructs and expressing lo6 receptors/cell (CHO/HIRC), were grown in suspension in modified McCoy's media (Shymko et al., 1989;Shoelson et al., 1992a).
Cells were harvested by centrifugation at 1000 revolutions/minute and solubilized at 4 "C in a solution of 0.1% Triton X-100 and 0.05 M HEPES, pH 7.6, containing 2 mM phenylmethylsulfonyl fluoride and 0.1 mg/ml aprotinin (0.1 ml cells/ml solubilizing solution). Alternatively, NIH-3T3 (1502) cells transfected with human insulin receptor cDNA (kindly supplied by Takashi Kadowaki and Simeon Taylor, National Institutes of Health, Bethesda, MD) were grown to confluence in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum and 0.6 mg/ml geneticin. Cells were lysed with 1.0 mM sodium carbonate, pH 7.4, and plasma membranes obtained by centrifugation at 20,000 X g for 1 h were solubilized with 2.0% Triton X-100 in 30 mM HEPES, pH 7.4, containing 1 mM phenylmethylsulfonyl fluoride, 10 PM 1,lO-phenanthroline, 2.5 mM benzamidine, and 2 mM EDTA. In either case, following centrifugation (130,000 X g for 1 h) to remove insoluble material the cell or membrane extracts were passed over WGA-agarose. The columns were washed extensively, and the lectin-bound proteins were eluted with 0.3 M N-acetylglucosamine in HEPES buffer (30-50 mM, pH 7.4) containing 0.1% Triton X-100, either alone (CHO/IR) or with 10% glycerol and protease inhibitors (NIH-3T3). Both were stored at -70 "C until use.
Cross-linking Time Course, Quantitation, and Specificity-WGApurified insulin receptor and [1251]iodo-TyP4-~-BpaB25 insulin (5 X lo6 cpm/ml) were incubated overnight in the dark at 4 "C. In a typical experiment, aliquots (250 pl/well, lo5 cpm) in a 24-well tissue culture plate (NUNC) were irradiated at 4 "C at a distance of 1.2 cm from the light source in a Rayonet photochemical reactor equipped with 350-nm lamps. Alternatively, for experiments shown in Figs. 4-6, samples were irradiated at 4 "C for 1 h with a 200-watt Oriel UV visible lamp equipped with a <340-nm filter. Samples (200 pl) were removed at the indicated times, and cross-linked proteins were either separated by SDS-PAGE directly or precipitated with anti-insulin receptor antibodies prior to SDS-PAGE. Cross-linked, radiolabeled proteins were identified by autoradiography of the fixed and dried gels.
Autophosphorylation of Cross-linked Receptors-Mixtures of insulin receptor and BBpa insulin, incubated overnight in the dark at 4 "C, were photolyzed at 350 nm for 1 h at 4 "C. Receptors were phosphorylated with 50 p~ ATP, 10 mM MgCI2, and 8 mM MnCI2. For identification of phosphorylated proteins by autoradiography [y-32P]ATP (1 rCi/pl) was added. Reactions were terminated at the times shown by the addition of 60 mM EDTA. a@-Heterodimers were prepared by treatment with 2.5 rather than 1.25 mM dithiothreitol (Boni-Schnetzler et al., 1986) because the class I disulfides of BBpa insulin-cross-linked receptors were more resistant to reduction. For gel shift assays, phosphorylated receptors were incubated with a 30fold excess of succinylavidin over BBpa insulin in Laemmli sample buffer containing 0.6% SDS (Pang and Shafer, 1983). In some cases excess free biotin was added to block the insulin-avidin interaction.
Proteins were separated by SDS-PAGE on 3-10% gradient gels. For 3ZP-labeled receptors, gels were fixed, dried, and subjected to autoradiography. Radiolabeled bands were excised and 32P incorporation was determined by Cerenkov counting. For Western blots, proteins were transferred to PVDF membranes (Immobilon-P, Millipore) which were blocked, in turn, with 30 mM HEPES, 5% milk, 0.01% NaN3, pH 7.4. Receptors were identified by sequential incubation for 16 h at 4 "C with R1064 anti-insulin receptor antisera (diluted 1:20 in phosphate-buffered saline containing 2% bovine serum albumin and 0.05% Tween-20) and IzsI-protein A. Alternatively, the biotin moiety of receptors cross-linked with BBpa insulin was detected by incubation of the PVDF membranes with 1251-streptavidin.
of CHO/HIRC cells in 6-well dishes were incubated at 4 "C for 3 h in Internalization of Cross-linked Receptors-Confluent monolayers F-12 media containing 50 mM HEPES, 0.1% bovine serum albumin, and ['251]iodo-TyP4-BpaB25 insulin. The cells were irradiated at 350 nm for 30 min at 4 'C in the photochemical reactor 5 cm from the light source. Cells were washed with phosphate-buffered saline at 4 'C, and internalization was initiated by rapid warming to 37 "C. At the indicated times cells were chilled by aspirating media and adding ice-cold phosphate-buffered saline. Internalized uersus cell-surface pools of receptor were differentiated by subsequent treatment of the cells with 0.5 mg/ml trypsin (Worthington) in phosphate-buffered saline at 4 "C. After terminating proteolysis with soybean trypsin inhibitor, cells were solubilized and cross-linked receptors were immunoprecipitated and separated by SDS-PAGE (Backer et al., 1991).

Synthesis and
Radiolabeling of Bpa Irzsulins-The synthetic scheme for Bpa described by Kauer et al. (1986) proceeds smoothly and results in an excellent yield of pure product. Bpa-containing octapeptides corresponding to residues B23-B30 of insulin were readily synthesized using standard coupling cycles for solid-phase peptide synthesis. Coupling of biotin at the e-position of lysine was complete, and biotin was stable to the subsequent repetitive couplings and cleavage with either trifluoromethanesulfonic acid or trimethylsilyl trifluoromethanesulfonate. Separations of D-and L-isomers of the a-trifluoroacetyl-or ebiotinyl-[D~]-Bpa peptides were easily accomplished using reversed-phase HPLC. Treatment of the Ne-deprotected peptides with leucine aminopeptidase, an L-amino acid-specific endopeptidase, liberates free Bpa from only one of each peptide pair to assign D-and L-isomers of Bpa within peptide enantiomers (Miller and Kaiser, 1988).
Trypsin-catalyzed semisyntheses of full-length Bpa insulin analogues were straightforward using Bpa-substituted octapeptides. Synthetic yields (60-70% based on insulin/DOI ratios) were similar to those observed previously (Shoelson et al., 1983a(Shoelson et al., , 1992aNakagawa and Tager, 1986). The analogues required no special handling procedures and were stable at -20 "C. Under mild 1251-iodination conditions, the analogues behaved like native insulin, with a predominance of incorporation occurring at the TyrAI4 position. Neither the iodination conditions nor radioactivity per se appeared to affect the analogues adversely.

Photolabeling the Insulin Receptor with Bpa Imulins-A
time course for photolytic cross-linking between lzsI-L-BpaR'' insulin and the insulin receptor is illustrated in Fig. 3A. Under these conditions the reaction was essentially complete within 30-60 min (Fig. 3B). A single cross-linked species was observed with an apparent molecular mass of 135 kDa. In an additional experiment cross-linked protein in a WGA preparation was either separated directly by SDS-PAGE or precipitated first with a monoclonal anti-insulin receptor antibody (18-44, Prigent et al., 1990). The same cross-linked protein observed in the absence of antibody (e.g. Fig. 3 A ) was precipitated by the antibody (data not shown).

Cross-linked (Biotinylated) Receptor-Avidin Binding and the
Gel Shift Assay-The ability of succinylavidin to bind BBpa insulin-cross-linked receptor was tested by Western blotting an SDS-PAGE transfer membrane with I2'I-streptavidin (Fig.  4). Cross-linked holoreceptor and (YP half-receptors were equally recognized by l2'1-streptavidin (lanes 1 versus 5).
Succinylavidin in the applied sample completely blocked "' 1streptavidin recognition (lanes 3 and 7 ) . When free biotin was added to the applied sample in a 30-fold molar excess over succinylavidin, "'1-streptavidin recognition of the blotted TyP4-BpaHZs insulin (IO5 cpm) and WGA-purified insulin receptor, incubated together a t 4 "C for 16 h, were photolyzed for the indicated times as described under "Experimental Procedures." Reaction mixtures were diluted with sample buffer and separated directly by SDS-PAGE. A, an autoradiogram of a representative experiment shows cross-linked insulin receptor a-subunit at -135 kDa. The diffuse band identified at the gel front represents free lzsI-BpaRZS insulin. B, fragments from both bands of the gel were excised and counted directly. Percentage covalent cross-link in this case refers to the percentage of total added 'zsI-BpaR2s insulin which formed a covalent adduct. This was calculated as ((cprn 135 kDa band)/(cpm 135 kDa band + cpm lower band)) X 100%. proteins was restored (lanes 4 and 8). Therefore, in the presence of succinylavidin alone all cross-linked (biotinylated) receptors were associated with succinylavidin and remained so throughout SDS-PAGE separation and protein transfer (lanes 3 and 7 ) .
Based on results indicating that succinylavidin remained associated with the cross-linked, biotinylated receptor during SDS-PAGE (and previous results of Pang and Schaffer, 1983), an insulin receptor gel shift assay was developed. Under nonreducing conditions cross-linked versus uncross-linked receptor forms were not separated by SDS-PAGE (Fig. 5, lane 1 shifted form of tu& holoreceptor was ohserved, and the amount of shifted t r / j half-receptor was always hall' that of the holoreceptor (gel fragments were excised and counted: 78.4 2 0.7!';, v~rsu.s :38.2 -C l.lrA, n = 5). Therefore, at IO-' M RRpa insulin each holoreceptor w a s cross-linked hy one molecule of insulin.
ALidin Binding l o Non-dcvnturcd Bpn Insulin ('ross-1inl:c.d Rcrcptors-Whereas the gel shift assays clearly show that cross-linked receptors are recognized by succinylavidin in the presence of O.6"h SDS, in the atlsence of denaturants succinylavidin does not hind to HHpa insulin or the HHpa insulin cross-linked receptor ((lata not shown). Assuming that this was caused by steric hindrance related to native protein structure, we prepared an additional analogue with an (aminohexanoic acid spacer hetween Lvs""' and the hiotin moiety of HHpa insulin. With the increased length of the flexihle connector, avidin recognition occurred in the ahsence of protein denaturants:' Iminohiotin (Hofmann r.1 al., 1980;Fudem-Goldin and Orr, 1990) was used in place of hint in t o facilitate separation of complexes for hinding site mapping studies. Autokinnsc Aclivily oj ('ross-1inf:c.d Hcwptors-To determine whether insulin receptor arltophosphorylation was act.ivated following photolabeling with HHpa insulin, photolyzed complexes were reacted with [y-:T]A'rI', separated hv SDS-PAGE, and identified hv autoradiography (Fig. 6). Under nonreducing conditions in the absence of succinylavidin, the molecular sizes of uncross-linked and cross-linked holoreceptors were virtually indistinguishable (lnnc 2 vcrsus 3 ) . However, the presence of succinylavidin in the sample applied to SDS-PAGE resulted in a shift in apparent molecular size of the majority of arltophosphorylatecl holoreceptor (lnnc 6 ) . Therefore, covalently cross-linked receptors were antophosphorvlated, and the cross-link does not interfere with kinase activation. Similar results were ot)t;lined following autophosphorylationofrross-linked receptors in tells (data not shown).
Experiments presented in this report show that insulin can be substituted with Bpa and that L-BpaBZ5 insulin binds to the insulin receptor with near-normal potency and specificity. Irradiation of the L-BpaBL5 insulin-receptor complex results in covalent bond formation between the two proteins. The photolabeling reaction is notable for its unique efficiency, which yields 60-100% of the adduct. Bpa insulin is stable as a reagent to usual laboratory conditions and to 'z51-iodination. Furthermore, cross-linked receptors are activated kinases, either in solubilized preparations or in intact cells, and activation leads to (apparently) normal internalization.
Crosslinking with '251-monoiodinated L-BpaBZs insulin facilitates identification of photolabeled holoreceptors, a-subunits, and fragments derived from them following proteolytic or chemical fragmentati~n.~ The ability of the analogue to cross-link with unusually high efficiency concomitant with activating the receptor provides a new tool for testing a broad range of questions relating the structure and function of insulin and the insulin receptor. In this study we determined the number of insulin molecules that were cross-linked to activated receptors. Notably, even at saturating BBpa insulin concentrations M ) one insulin cross-linked each a&-holoreceptor. When the crosslinked holoreceptors were separated into aP half-receptors and analyzed quantitatively, half of the ap half-receptors were cross-linked. Therefore, we have concluded that one insulin binds to the insulin holoreceptor with high affinity in an orientation appropriate for high efficiency cross-linking.
Whereas this study provides no evidence for or against binding of a second insulin molecule to the insulin-receptor complex, previous findings indicate that a second insulin binds with lower affinity (DeMeyts et al., 1976;Sweet et al., 1987;Boni-Schnetzler et al., 1987;Yip and Jack, 1992). Crosslinking with Bpa insulin demonstrates, however, that the second insulin must bind to the receptor in a different orientation than the first. If two insulins were bound in similar orientations then at saturating concentrations two insulins would cross-link per a&-holoreceptor, which clearly does not occur.
Homobifunctional cross-linking, crystallographic, and NMR studies of insulin structure indicate that the COOH terminus of the B-chain separates from the body of insulin (the core structure composed of A-and B-chain helices) during or prior to receptor binding (Brems etal., 1991;Cutfield et al., 1981;Derewenda et al., 1991;Hua et al., , 1992Nakagawa and Tager, 1989). This is particularly pertihent to the current study because the B25 position is at the center of the region thought to undergo a structural transition. Thus, we can speculate as to how structural flexibility in insulin's COOH terminus might translate into asymmetry of binding and @-subunit autophosphorylation (see accompanying paper by Lee et al.). High affinity binding (the first insulin) must be accompanied by multiple contacts between insulin and the receptor, including residues of the COOH-terminal B-chain and the helical core. Separation of the B-chain COOH terminus from the helical core may also expose additional residues which would be buried in the well known crystal structures of insulin (e.g. IleA2 and ValA3). That only one BBpa insulin cross-links the receptor suggests that the COOHterminal B-chain of the first molecule might occupy a shared cleft between the two binding sites located at the interface between two a-subunits. Once the first insulin has bound, the COOH-terminal B-chain of the second would be unable to S. E. Shoelson, unpublished observations. occupy the cleft and would therefore be unable to cross-link (BBpa insulin) or provide additional binding energy for generation of a second high affinity complex. Such a model for asymmetric binding (Lee et al., 1993) is consistent with previous studies of insulin-receptor binding and in keeping with recent results with the occupied growth hormone-binding protein (receptor ectodomain) crystal structure (de Vos et al., 1992). Future studies to map the insulin binding pocket of the receptor and structural studies of the complex will be needed to sort out how insulin-activated receptor signaling occurs in a more detailed fashion.