Cross-linking of a B25 Azidophenylalanine Insulin Derivative to the Carboxyl-terminal Region of the a-Subunit of the Insulin Receptor IDENTIFICATION OF A NEW INSULIN-BINDING DOMAIN IN THE INSULIN RECEPTOR*

To identify a site within the insulin receptor ectodo- main which forms a binding pocket for B25 Phe and is responsible for initiating conformational changes re- quired for high affinity binding of insulin we have used a novel photoreactive insulin, despentapeptide-(B26- B30) [B25 p-azidophenylalanine-a-carboxamide] insulin (APC insulin). This derivative has a highly photoreactive azido group incorporated into the aromatic ring of the B25 phenylalanine amide. APC insulin bound to hu- man insulin receptors overexpressed on a transfected Chinese hamster ovary cell line (P3-A) with an apparent potency of 9-fold relative to that of native insulin and stimulated lipogenesis in rat adipocytes with an average potency equal to porcine insulin. Addition of biotin to the B1 Phe amino group to form despentapeptide-(B26-B30) [Bl (6-biotinylamidocaproyl)phenylalanine B25 p-azidophenylalanine-a-carboxamide] insulin deriva- solution methods) of the tripeptide amide derivative H-Gly-Phe-Phe-(p-NH-Bot)-NH, corresponding to the sequence 23-25 of the insulin B-chain; (iii) the trypsin- catalyzed coupling of the tripeptide amide with DOPI to give despen-tapeptide-(B26-B30) [B25 Phe-(p-NH-Boc)-a-carboxamidel insulin, and (iv) the removal of the Boc group from the latter compound, and the conversion of the resulting p-NH, derivative to the final product, APC insulin. 0.5 ml of dimethyl sulfoxide, 1.2 ml of 1,4-butanediol, 0.6 ml of 0.25 M Tris acetate (pH 7.5), and 6 mg of TPCK-trypsin, after Sephadex G-50 chromatography and trifluoroacetic acid treatment, 20 mg of crude product was obtained. HPLC of this material in 2-mg batches under the conditions previously described yielded about 6 mg of highly purified product. The product upon HPLC chromatography displayed a sharp single peak.

The control of the metabolic and mitogenic functions of insulin is regulated by specific interactions between the hormone and its cell surface receptor. The insulin receptor is a cell membrane spanning glycoprotein tetramer consisting of two o~-and two P-subunits. The extracellular domain of the receptor is formed by the ligand binding &-subunit in disulfide linkage to the first 194 amino acids of the P-subunit. The intracellular portion consists of the COOH-terminal403 amino acids of the P-subunit which includes the tyrosine kinase domain.
In spite of its importance for understanding the mechanism of receptor activation, the location of the ligand-binding region of the insulin receptor remains a n unsettled issue. Given the high affinity of insulin for its plasma membrane receptor, it is obvious that multiple interactions between ligand and receptor are required to generate the necessary free energy for specific binding. Among putative residues of insulin involved in receptor binding are the highly conserved sequences A1 to A3 and B23 to B26, among others (1). Replacement of the invariant B25 Phe with Leu, as in the mutant insulin of an individual with the syndrome of hyperinsulinemia associated with mild diabetes mellitus (Insulin Chicago) ( 2 4 , along with other data indicates that the aromatic character of this residue is important (5). Moreover, in the crystal the B25 side chain is fully exposed and the conformation of molecules 1 and 2 of the insulin dimer differ at this position in that accommodation of helix movement requires the rotation of the aromatic side chain of B25 Phe by 150 degrees in molecule 2. In contrast, most of the other side chains on the monomer surface vary by no more than 14 degrees between the two molecules in the dimer.
The transmission of a hormonal signal is not a single consequence of ligand-receptor association but requires significant and perhaps sustained changes in ligand-receptor structure. Conformational changes similar to those observed in the isologous interactions of insulin in the dimer have long been considered to be an important model for the heterologous interaction of insulin with its plasma membrane receptor. Thus, the aromatic side chain at residue B25 may participate in an active way in initiating conformational changes that occur in the formation of the insulin-receptor complex (5).
Potential insulin contact sites on the receptor have been mapped thus far to residues within the amino terminus (exon 2) (649, the cysteine-rich region (exons 2 and 3) (9-111, or a

Insulin Receptor-binding Site 29191
more carboxyl-terminal domain that possibly includes sequences encoded by exons 6 and 7 (12)(13)(14). To explore further the organization of the insulin-binding domain of the receptor with respect specifically to the binding pocket for B25 Phe, we have used a Chinese hamster ovary cell line which overexpresses the human insulin receptor at very high levels and B25 azidophenylalanyl photoreactive insulin derivatives that quantitatively cross-link, radiolabel, and activate the insulin receptor.  (15). Phenylmethylsulfonyl fluoride, leupeptin, and pepstatin A were from Boehringer Mannheim (Meylan, France). Tricine' was from Bio-Rad. 30% Acrylamide solution was from National Diagnostics (Atlanta, GA). Lys-C endoproteinase was sequencing grade from Boehringer. Polyvinylidene difluoride membrane (Immobilon P) was from Millipore Corp. (Bedford, MA). Nylon membrane was from Amersham (Hybond N). Protein A-agarose was from Pierce. All other reagents were of the highest grade commercially available.
Synthesis of Despentapeptide-(B26-B30) B25 p-hidophenylalaninea-carboxamidel Human Insulin (APC Human Insulin)-This compound was prepared by a semisynthetic procedure involving (i) the preparation of desoctapeptide-(B23-B30) porcine (human) insulin with all functional groups unprotected (DOPI); (ii) the synthesis (by solution methods) of the tripeptide amide derivative H-Gly-Phe-Phe-(p-NH-Bot)-NH, corresponding to the sequence 23-25 of the insulin B-chain; (iii) the trypsincatalyzed coupling of the tripeptide amide with DOPI to give despentapeptide-(B26-B30) [B25 Phe-(p-NH-Boc)-a-carboxamidel insulin, and (iv) the removal of the Boc group from the latter compound, and the conversion of the resulting p-NH, derivative to the final product, APC insulin.

Desoctapeptide-(B23-B30) Porcine (Human) Insulin
(DOPI)-This compound was prepared as described previously (5) from porcine insulin, which differs only at B30 from human insulin. H.Gly-Phe-Phe-(p-NH-Boc)-NH,"t was prepared from its N"-benzylcarbonyl derivative by transfer hydrogenation over 10% palladium/ charcoal catalyst using formic acid as the hydrogen donor. The N"benzyloxycarbonyl tripeptide amide (Z-tripeptide amide) was prepared stepwise, starting with H-Phe-(p-NH-Boc)-OH (16) as follows: to a suspension of H-Phe-(p-NH-Boc)-OH (2.8 g) in DMF (30 ml), diisopropylethylamine (0.3 ml) and N-benzyloxycarbonyloxy-5-norbornene-2,3-dicarboximide (Chemalog, South Plainfield, NJ) (3.1 g) were added. After stirring for 12 h, the mixture was diluted with 0.2 M NaHCO, solution and extracted with ether. The aqueous layer, cooled to 2 "C, was acidified to pH 2.0 in HCl and the oily product extracted into ethyl acetate. The organic layer was washed with water, dried, and concentrated under reduced pressure to dryness. To a solution of the residue in acetonitrile (60 ml), N-hydroxy-5-norbornene-2,3-dicarboximide (Chemalog) (1.8 g) and N,N"dicyclohexylcarbodiimide (2.0 g) were added. After 12 h the urea by-product was filtered off and the solvent removed. The residue was dissolved in chloroform (100 ml) and ammonia gas was passed through the solution for 3 h. The precipitated Z-Phe-(p-NH-Boc)-NH, was collected, recrystallized from 95% ethanol (weight, 2.5 g; m.p. > 230 "C) and used in the following synthesis of the dipeptide derivative. To a solution of Z-Phe-(p-NH-Boc)-NH, (2.5 g) in a mixture of methanol (150 ml) and DMF (20 ml), a 10% palladiudcharcoal catalyst (1 g) was added followed by formic acid (1 ml). After 1.5 h the catalyst was filtered off, the filtrate concentrated to a small volume and diluted with 1 M Na,CO, (40 ml) and ethyl acetate (250 ml). The organic layer was washed, dried, and concentrated to give the free base as an with human insulin receptor constructs with a point mutation in the tetrabasic cleavage site ( L y~~~~-A l a ) ; PAGE, polyacrylamide gel electrophoresis; HPLC, reversed phase high performance liquid chromatography; MabIR, anti-human insulin receptor monoclonal antibody; TLCK, N"-p-tosyl-L-lysine chloromethyl ketone; MeOH, methanol; DTT, dithiothreitol.
oil. Asolution of this product in DMF (20 ml) was mixed with Z-Phe-OH, 1-hydroxybenzotriazole, and N,N"dicyclohexylcarbodiimide (10 mmol of each) in DMF (30 ml) (activated for 30 min at room temperature before adding the free base). After 24 h the mixture was diluted with 0.1 M cold KHCO, and the precipitated Z-Phe-Phe-(p-NH-Bot)-NH, was filtered, dried, crystallized from 95% ethanol (weight, 3.0 g; m.p. > 250 "C), and used in the following synthesis of the tripeptide. Z-Phe-Phe-(p-NH-Boc)-NH, (3 g) was deprotected at the NH, terminus by transfer hydrogenation as described above. The resulting free base was dissolved in DMF (20 ml) and mixed with an activating mixture of Z-Gly-OH, I-hydroxybenzotriazole, and N,N"dicyclohexylcarbodiimide (10 mmol of each) in DMF (30 ml). The reaction mixture was processed as described above in the synthesis of the dipeptide and the resulting Z-Gly-Phe-Phe-(p-NH-Bot)-NH, was purified by recrystalization from 95% ethanol; weight 2.3 g; m.p. > 250 "C.

C33H39O7N5
Calculated: C 64.2 H 6.36 N 11.3 Found: C 64.6 H 6.35 N 11.6 This material (0.5 g) was dissolved by warming in a mixture of methanol (50 ml) and DMF (15 ml). After cooling to room temperature, 10% palladium/charcoal catalyst (1 g) was added followed by formic acid (1 ml). After 3 h the catalyst was filtered off, the filtrate concentrated to a small volume and mixed with anhydrous ether (100 ml). After cooling, the formic acid salt of the tripeptide H-Gly-Phe-Phe-(p-NH-Boc)-NH, was precipitated, collected, dried (0.3 g), and used in the trypsin-catalyzed coupling with DOPI. The pH was adjusted to 7.0 with 50% acetic acid. After the addition of TPCK-trypsin (6 mg), incubation for 5 h at 37 "C, and dilution with 1 M acetic acid (4 ml), the insulin derivative was separated on a precalibrated column of Sephadex G-50 (2.5 x 60 cm) in 1 M acetic acid, collected, and lyophilized. The residue was treated with trifluoroacetic acid, 10% anisole (10 ml; 30 min a t room temperature), and the deblocked insulin analogue was precipitated with anhydrous ether and isolated by centrifugation; weight, 20 mg. This product was purified by HPLC (batchwise) on a Vydac 218 TP column (0.45 x 25 cm) a t a flow rate of 0.5 mumin with a 10-50% linear gradient: A = 0.1% trifluoroacetic acid; B = 0.1% trifluoroacetic acid in acetonitrile. The analogue, eluted at 48.17 min as a sharp peak, was recovered by concentration and lyophilization of the effluent; weight, 8 mg.

Despentapeptide-(B26-B30) [B25 p-Azidophenylalanine-a-carboxa-
midel Insulin (APC Human Insulin)-This compound was prepared from the analogue described above using the method of Garbay-Jaurequiberry et al. (18). All operations were done under reduced, indirect light to minimize destruction of the azido group. Briefly, an icecold solution of despentapeptide-(B26-B30) [B25 Phe-(p-NH,)-a-carbox-amide1 insulin (8 mg) in 0.1 N HCl (0.8 ml) was treated with aqueous 1 M sodium nitrite solution (17.5 1 1 1 ) for 15 min. To this solution (4 "C) a n aqueous 1 M sodium azide solution (17.5 111) was added. After 15 min stirring, the mixture was diluted with a saturated picric acid solution (1 ml). The precipitated picrate of the azido insulin analogue was isolated by centrifugation and converted to the hydrochloride as described previously (19). HPLC, as described above, led to the isolation of the insulin analogue in a highly purified form; weight, 6 mg. Upon HPLC rechromatography this product exhibited a single sharp peak with a retention time of 44.91 min. Amino acid analysis after acid hydrolysis was in agreement with the theoretically expected values (data not shown).

Receptor-binding Site
L B l (6-Biotinylamidocaproyl)Phenylalanine] Porcine Insulin-This compound was prepared essentially according to the method of Hofmann et al. (20). Briefly N"'A1)jV4B29'-bis-(t-butoxycarbonyl) porcine insulin (21) (62 mg), 6-biotinylamidocaproate N-hydroxylsuccinimide ester (Sigma) (80 mg), and imidazole (28 mg) were dissolved in dimethyl sulfoxide (4 ml). After 24 h, the mixture was loaded into a Sephadex G-50 column (2.5 x 60 cm) equilibrated and eluted with 0.1 M NH,HCO, solution. The effluent under the main peak was collected and lyophilized. The dry material was treated with trifluoroacetic acid, 10% anisole (10 ml; 30 min) and the deblocked biotinylated insulin precipitated from the solution with the addition of anhydrous ether; weight, 45 mg. The crude product was purified by HPLC. Batches of about 3 mg of protein each were chromatographed under the conditions described above to yield 25 mg of highly purified product.

Despentapeptidee-(B26-B30) [Bl (6-Biotiny1umidocaproyl)phenylalanine, B25 p-Azidophenylalanine-a-carboxamidel Insulin (Bio-APC
Human Insulin)-This compound was prepared as described previously for the corresponding non-biotinylated analogue. Approximately 7 mg of the respective B25 Phe-(p-NH,) analogue were processed and after HPLC, about 200 pg of highly purified azido derivative were obtained. This product, on HPLC rechromatography exhibited a sharp single peak (retention time 46.83 min). Quattro mass spectrometric (VG) analysis gave a molecular weight of 5600 for the synthetic compound. The calculated weight is 5597.
Biological Evaluation-The ability of the two photoreactive insulin analogues to stimulate, relative to porcine insulin, the conversion of [3-3H]glucose into an organic extractable form by isolated rat adipocytes (lipogenesis) was determined as described previously (22). For APC insulin the potency was equal to that of natural porcine insulin (data not shown). The Bio-APC insulin displayed a potency about 83% relative to porcine insulin (data not shown).
Cell Line and Culture-The stably transfected Chinese hamster ovary cell line (P3-A) expresses human insulin receptors with a point mutation in the tetrabasic proreceptor cleavage site (Ly~'~~-Ala) which does not affect its processing or binding characteristics, as described elsewhere (23). These cells express very high levels of insulin receptor protein (20-50 millionkell) and exhibit normal insulin-dependent autophosphorylation of the p-subunit of the receptor. Accordingly, this cell line was considered to be suitable for studying ligand binding.
Insulin Binding Assay-Established clonal cell lines were plated at the indicated density numbertdwell in 12or 24-well dishes. After 24 h of incubation, cells were washed twice with Hank's balanced salt solution supplemented with 20 m M Hepes (pH 7.5) (buffer A) and incubated in buffer A with 10 mg/ml bovine serum albumin containing 10-30 fmol of '261-insulin and various concentrations of unlabeled bovine insulin for 16 h at 4 "C. The cells were washed three times with ice-cold buffer A and then solubilized in 250 pl of 0.2 M NaOH for 30 min at 37 "C for measurement of radioactivity. Specific binding was determined by subtracting the amount of 1261-insulin bound in the presence of excess M) unlabeled insulin.
Insulin Receptor Purification-Cells (-5 x 10') were solubilized in 5 ml of 1% Triton X-100, 50 m~ Hepes (pH 7.8), 150 m~ NaCl containing 1 m~ phenylmethylsulfonyl fluoride, 5 pg/ml pepstatin A, and 5 pg/ml leupeptin for 30 min at 4 "C. Cell lysates were centrifuged at 14,000 rpm for 20 min at 4 "C, and the supernatant fraction was then incubated with 1 ml of wheat germ agglutinin-agarose for 2 h at 4 "C. The resin was then transferred into 20-ml columns (Bio-Rad) and washed with 30 ml of 50 m~ Hepes (pH 7.8), 500 m~ NaCl, 0.1% Triton X-100. The adsorbed glycoproteins were eluted with 0.3 M N-acetylglucosamine in the above buffer. 1-ml fractions were collected and protein concentration was measured by the Bradford assay (Bio-Rad). The protein peaks were stored at -80 "C. Binding was measured with polyethylene glycol 6000 precipitation according to the method of Hedo (24).
Preparation of z251-APC Insulin-Radioiodination ofAPC insulin was conducted in subdued light using a described method (25). One mg of APC insulin was dissolved in 0.002 N HCl and 1.6 nmol of APC insulin was incubated in 0.2 M NaHPO, (pH 7.0), combined with 1 mCi of NalZ6I for 2 min, then added with 1 nmol of chloramine T. The mixture was vortexed, allowed to react for 2 min. Another 1 nmol of chloramine T was added and the reaction was terminated by addition of acetyltyrosine. To separate free ['261]iodide from the peptide, iodination mixtures were diluted with 50 1.11 of 0.01 N HC1 and applied on a NAP-10 column (Pharmacia Biotech Inc.). The '261-APC insulin was eluted with 0.01 N HCl containing 20 pg/ml bovine serum albumin and stored at -80 "C until used.
PhotoaffEnzty Labeling of the Insulin Receptor"P3-A cells (10' cells) were incubated with 1251-APC insulin (5 x los cpm) in 3 ml of buffer A with 1% bovine serum albumin overnight in the dark at 4 "C. In a typical experiment, the medium was removed and P3-A cells in 10-cm culture dishes were irradiated at 4 "C at a distance After overnight incubation at 4 "C, the immunoprecipitates were sedimented and washed three times with buffer B. Lys-C Endoproteinuse Digestion-The agarose bound metabolically labeled insulin receptor was washed once with digestion buffer (1% Triton X-100, 10% acetonitrile, 100 LTLM Tris-HC1, pH 8.0). Digestion was then carried out with 20 pdml Lys-C endoproteinase for the periods indicated at 37 "C in water bath incubator. The reaction was stopped by the addition of 0.15 m~ TLCK and the digested fragment was eluted with (-300 p1) buffer B until the radioactivity of eluate was close to background values.
Streptauidin Complexing-900 pl of eluate containing the digest was incubated with 20 pl of streptavidin-agarose (Pierce) for 16 h. After washing of the agarose four times with 1 ml of buffer B, the agarose was heated for 5 min at 95 "C in 40 pl of 1.5 x Laemmli's sample buffer with or without 100 m~ dithiothreitol just prior to PAGE analysis with glycine (26) or Tricine (27) buffers.
Protein Tkansfer-Peptide transfer from polyacrylamide gels to membranes (Immobilon-P Millipore, Hybond-N; Amersham) was performed in 25 m M Tris base, 192 m~ glycine, 20% MeOH (pH 8.2) at 4 "C for 16 h at 14 V, 65-125 mA. The blot was analyzed by autoradiography on Kodak X-AR film at -80 "C. The area of the membrane corresponding to the appropriate band was cut out from the blot and divided into small pieces (3 x 2 mm). The radiolabeled fragments were eluted with 88% formic acid at room temperature by sonicating for 15 s x 2 at a setting of 40-60 watts at 20 kHz (SonicatorTM Cell Disruptor, model W185F, Heat Systems-Ultrasonics, Inc.).
Amino Acid Sequence AnalysisSequencing was performed on an Applied Biosystems 470A gas-phase sequencer on samples dissolved in 50 p1 of 88% formic acid. The anilinothiazolinone derivatives obtained at the completion of each sequence cycle were extracted in n-butyl chloride and 400-pl aliquots counted after dilution into 20 ml of scintillator (Optifluor 0, Packard) in a liquid scintillation spectrometer.

RESULTS
Both APC insulin and Bio-APC insulin ( Fig. 1) competed efficiently with [1251]iodo-insulin for binding to wheat germ agglutinin-purified insulin receptor preparations (Fig. 2). The results show that the ED,, values, determined as the concentrations of analogues causing half-maximal inhibition of 12,1insulin binding, were 1.8 x 10"' M for APC insulin and 4.8 x 10"l M for Bio-APC insulin, as compared to 1.7 x lo-' M for native bovine insulin. Therefore, the relative receptor binding potency for APC insulin was 9.4 and for Bio-APC insulin was 34.9 times higher than that of native insulin. Furthermore, in the stimulation of lipogenesis in rat adipocytes, both derivatives showed potencies similar to porcine insulin.
Insulin binding on intact P3-A cells is shown in Fig. 3. In the experiments performed on whole cells, biphasic displacement curves typical of insulin binding to placental insulin receptor preparations were obtained. P3-A cells bound insulin with high affinity and the ED,, was 1. 254 nm. We examined the effect of this short wavelength UV light on both insulin binding and 12'I-insulin. Using solubilized wheat germ agglutinin-purified insulin receptors from P3-A cells, insulin binding activity decreased progressively to about 11% of control values over a 30-min W exposure with a halfmaximal effect after 10 min (data not shown). UV exposure of "'I-insulin tracer lead to slightly increased trichloroacetic acid nonprecipitable radioactivity after only 30 s exposure and reduced trichloroacetic acid precipitable radioactivity to 46% of control at 30 min (data not shown). This deiodinating effect of UV light was most rapid during the initial 5 min of exposure and was also half-maximal a t 10 min. These results suggested that the optimal conditions of photolysis for avoiding the destructive effects of this shortwave UV light would be exposures of less than 60 s.
The time course of photolytic cross-linking between 1251-APC insulin and insulin receptors on intact P3-A cells is illustrated in Fig. 4. Under these conditions the reaction was essentially complete within 20 s. A single cross-linked species was observed with an apparent molecular mass of 135 kDa on SDS-PAGE. On quantitative analysis the efficiency of cross-linking of '2'I-APC insulin to the insulin receptor was roughly 70% (calculated as the incorporation of radioactivity into insulin receptor a-subunitkotal radioactivity applied onto the SDS-PAGE).
To determine the site of cross-linking we needed to be able to detect the radioactivity of '251-photoreactive insulin even after proteolysis. We therefore chose Lys-C endoproteinase for digestions since there are no Lys residues in the insulin derivatives we used. The specificity of this proteinase is reported to be high (30). Digestion of APC insulin-insulin receptor complexes with Lys-C endoproteinase gave the results shown in Fig. 5 upon analysis by SDS-PAGE (8 and 15% gels) in the presence or absence of reducing agent.
To determine if the 5.5-kDa band contained a part of the insulin receptor, its relative position on the SDS-PAGE was compared with those arising from the photoprobes alone. In the composition of the reaction mixture Triton X-100 did not affect photolysis and migration of the bands in Tricine-SDS-PAGE. '251-Bio-APC insulin migrated a t 4 kDa under nonreducing conditions and a t 3 kDa under reducing conditions, representing the insulin A-and B-chains comigrating (Fig. 6, lanes 2-51. After photolysis, 12jI-Bio-APC insulin still migrated a t -4 kDa under nonreducing conditions (Fig. 6, lanes 6 and 7). However, an extra band at 4 kDa, in addition to the 3-kDa bands, was observed under reducing conditions, suggesting the formation of a B-chain dimer by photolysis. The 5.5-kDa band after digestion of '"1-Bio-APC insulin and insulin-receptor complex with Lys-C endoproteinase clearly differed in size from those bands formed only from '2sI-Bio-APC insulin. These findings suggested that the 5.5-kDa band could contain a peptide derived from the insulin receptor, and we therefore examined the amino acid sequence of this 5.5-kDa fragment.
In order to obtain sufficient material for microsequencing,

P3-A cells in five 10-cm culture dishes (1-2 x lo7 cells/dish)
were metabolically labeled with various individual SH-labeled amino acids and successively photoaffinity labeled with ' "1-Bio-APC insulin (see "Experimental Procedure"). The photoaffinity labeled receptor was digested with Lys-C endoproteinase for 24 h. The resulting photoaffinity labeled fragments were isolated by incubation with streptavidin-agarose and then analyzed on SDS-PAGE. The protein bands in the SDS gels were transferred to polyvinylidene difluoride or nylon membranes. After autoradiography, the radioactive 5.5-kDa bands were cut out and eluted with 88% formic acid. The eluate was directly applied to an amino acid sequencer. In a typical experiment 40% of the total counts (usually 5-6 million counts of lZ5I-Bio-APC insulin) was bound to the receptor. 80% of the 12sI-Bio-APC insulin and insulin-receptor complex was isolated after immunoprecipitation with MabIR 83-14. After digestion, 80% of radioactivity became attached to streptavidin-agarose; -10% of total radioactivity initially applied to the SDS gel was found in 5.5-kDa fragment. Protein transfer yielded about 50% recovery and typically 20,00040,000 cpm of y radioactivity was subjected to an amino acid sequence analysis. In experiments using [3H]Phe or [3H]Val for labeling the insulin receptors, the radioactive residues found on amino acid sequence analysis of the 5.5-kDa fragment obtained under reducing conditions corresponded uniquely with the sequence of peptide Thr704 to Lys718, as predicted by a computer analysis of potential receptor peptides generated by Lys-C endoproteinase digestion (Fig. 7). The extra peaks at positions 14 and 19 were seen even under reducing conditions and represented [12sI]iodine atA14 andA19 (tyrosines) in the probe; as shown in Fig. 8 (DTT(+)) Tricine-SDS-PAGE gave a broad smear of radioactivity. In separate experiments this was found to be due to a mixture of reduced insulin A-and B-chains (data not shown). Accordingly, the observed ' ' ' 1 radioactivity a t these positions probably originated from small amounts of co-migrating insulin A-chain, but not from the B-chain since the NH, terminus of the B-chain was blocked by biotin. In labeling experiments using ["HlLeu or [3H]Tyr, amino acid sequence analysis of the -7-kDa fragment obtained under nonreducing conditions (Fig. 8, D m ( -) ) , also yielded radioactive tritium peaks at appropriate positions corresponding to the sequence of peptide Thr704-Lys71R (Fig. 7). To determine whether this fragment extended to the most COOHterminal residue of the a-subunit (Ser731), [3HlGly-or L3H1Arglabeled insulin receptors were also digested and analyzed. However, no 3H radioactive peaks corresponding to these amino -100 fmol of '*'I-Bio-APC insulin in 1 ml of buffer B or 0.1 M phosphate buffer with 1% bovine serum albumin was spread in a 60-mm Petri dish and photolyzed as described under "Experimental Procedures." The mixture was then incubated with streptavidin-agarose and the bound material analyzed directly by 10% Tricine-SDS-PAGE (lanes 2-9). 2 pg of insulin receptors photoaffinity labeled with '2'1-Bio-APC insulin were proteolytically digested and analyzed (lane 1 ). Lanes 2, 4, 6, and 8, were ""I-Bio-APC insulin alone in buffer with 0.1% Triton X-100 or in buffer with 0.0058 Triton X -100 (lanes 3, 5, 7, and 9). A 5.5-kDa cross-linked fragment is evident in lane 1. This band is distinct from the 4-kDa band observed in lanes 2, 3, 6, 7, 8, or 9. See text for details. Dm acid residues were observed, indicating that Lys-C endoproteinase cleavage a t Lys718 was efficient. We assume that in the experiment with [3H]Arg, we were not able to detect a tritium peak a t residue 14, the penultimate Arg residue in the Thr704-Lys7lS peptide, due to loss of the small Arg-Lys dipeptide from the Polybrene carrier in the sequencer. DISCUSSION Our results demonstrate that active insulin derivatives which possess a highly photoreactive group in the para position of the aromatic ring of B25 phenylalanine become cross-linked to the COOH-terminal region of the a-subunit of the insulin receptor within a peptide that extends from Thr704-Lys718.
The despentapeptide insulin molecule, which has been adopted in the present experiments as the basis of a photoaffinity probe, has unusual features. In contrast to the insulin derivatives used in most previous receptor cross-linking studies, which have utilized photoactivatable groups attached via linkers to residues of insulin that are not directly involved in insulin binding, the APC insulins incorporate a relatively small azido group on the para position of the phenylalanyl side chain of B25, a residue predicted to participate directly in receptor contacts (5). Crystallographic and NMR studies have revealed that the residual COOH-terminal domain of the B-chain in despentapeptide insulin assumes a conformation quite different from that seen in either 2-zinc or 4-zinc insulin (31,32). Nonetheless, despentapeptide insulin exhibits biological potency comparable to that of native insulin (5). In contrast, a "mini proinsulin," (des-(B3O) insulin), a single chain molecule with a peptide linkage between residues LysRZ9 and Glfl', shows only about 0.1% or less of the normal affinity for the insulin receptor even though it exhibits no major changes in structure relative to crystalline porcine insulin (33,34). These observations underscore the need for flexibility in the COOHterminal region of the B-chain for receptor interaction. Replacement of B25 phenylalanine by serine, leucine, or homophenylalanine leads to a 100-fold decrease in receptor affinity (5), but this decrease in affinity is partially reversed (by up to 40-fold) when COOH-terminal residues B26 to B30 are deleted. These findings indicate the importance of aromaticity at B25 and support the existence of a specific aromatic binding pocket for this residue in the a-subunit. In our design of a receptor probe we speculated that a possible decrease in affinity due to the modification of B25 Phe with the azido group would be reversed by deletion of the COOH-terminal5 residues and amidation of B25. Tager had earlier shown that such a derivative having B25 tyrosine amide had %fold increased receptor affinity (5). The high affinity binding ofAPC insulin achieved in our experiments is consistent with his findings.
Although we have not identified the specific amino acid residue(s) in this peptide which are covalently linked to the B25 Phe, it is in a relatively hydrophobic region that is predicted to form an a-helix. From the amino acid sequence of the insulin receptor Phe705 and T y r 7 0 8 together with the adjacent Phe701 could form an aromatic cluster in an a-helical configuration. These findings, however, are in contrast to previous suggestions based on molecular graphics simulations that PheR9 of the insulin receptor interacts with Phe2s of the insulin molecule (7). The earlier conclusions were based on the modeling and observations of decreased insulin binding in studies of mutated receptors with substitutions for Phe residues 88 or 89. However,

Insulin Receptor-binding Site
. OOj direct evidence that B25 contacts PheSg was not obtained. On the other hand, their assumptions regarding the importance of aromatic-aromatic interactions in B25 Phe binding lend support to our findings. Such interactions are common in globular proteins and probably are very important for stabilization of protein tertiary structure (35,361. Furthermore, aromatic side chains often form hydrophobic pockets that preferentially bind aromatic substrates (37). Accordingly, the aromatic residues in the COOH-terminal region of the a-subunit may form a binding pocket necessary for high affinity binding of insulin.
B25 Phe has long been recognized an important residue in insulin binding. Aromaticity at this point is also conserved in the insulin-like growth factors. Residues 23-26 of the COOHterminal region of the B-chain are considered to be part of the active site of the hormone. A point mutation a t B25 resulting in a leucine for phenylalanine substitution has been found in a diabetic patient (4) and is associated with greatly reduced binding affinity. This region also plays a crucial role in dimer formation within the zinc-insulin hexamer. In a predicted model for insulin and insulin-receptor interaction the p-aromatic ring of B25 Phe participates in an active way in initiating conforma- tional changes that occur in the insulin-receptor complex (5).
Thus the filling of a receptor binding pocket by the p-aromatic side chain of B25 may well facilitate or initiate additional conformational changes that probably occur both in the COOH-terminal B-chain domain of insulin and in the receptor site (5,381. Since the insulin receptor was photoaffinity labeled a t a saturating concentration of Bio-APC insulin in order to determine the amino acid sequence at the cross-linked site(s), there is a possibility that the binding site we have detected might be a low affinity site. However, Shoelson et al. (38) reported that each holoreceptor (a,pz) was cross-linked by one molecule of insulin even at saturating concentrations M) using a somewhat similar photoreactive insulin analogue (BpaRZ5 insulin) which binds to the receptor with high affinity, and that this is sufficient to activate a reaction cascade starting with autophosphorylation. The cross-linking site in the receptor for this insulin derivative has not yet been reported. It has also been observed that each receptor half-molecule (alpI) binds insulin with low affinity (39). These findings and other data suggest the possibility that to achieve the high affinity binding state one insulin molecule must interact with sites in both a-subunits within each holo-receptor (40). After a high affinity binding state has been achieved, a lower affinity site might be created to bind a second insulin molecule. Insulin binding to the lower affinity site may occur in a different orientation, however, since the previously mentioned Bpa insulin, which has a photoreactive benzoylphenylalanine residue a t B25, only detected a single high affinity site (38). We thus believe that the site we have identified is involved in the formation of the high affinity binding state.
The insulin binding region identified in this study is also of interest because this occurs just adjacent to the sequence encoded by exon 11. Alternative splicing of a single gene tran-

Insulin
Receptor-binding Site 29197 script yields two insulin receptor mRNA species and two receptor isoforms, HIR-A and HIR-B, which differ by the 12amino acid sequence encoded by exon 11. This additional sequence is inserted at residue 716 of HIR-A (exon 11-1 to yield HIR-B (exon 11+) (41,42,53). The two isoforms possess distinct functional properties and are expressed in a tissuespecific fashion (43-48). In addition to the pathophysiological significance of changes in HIR A B expression in the skeletal muscle associated with development of non-insulin-dependent diabetes mellitus (49-51), we have postulated and demonstrated that the absence of this 12-amino acid segment encoded by exon 11 affects the folding andor conformation of the proreceptor so as to confer decreased sensitivity to insulin.' Thus cleavage at the a,P-subunit junction (residues 720-723 in the A isoform) is mandatory to allow the insulin binding site to achieve its normal high affinity binding conformation. In addition to the receptor region which we have identified here, other regions are also potentially important for high affinity binding. Specifically, the NH,-terminal domain (residues 1-68), as well as the cysteine-rich region (residues 205-310) have both been proposed as candidates for binding regions by several groups (6)(7)(8)(9)(10)(11). A more COOH-terminal domain possibly including the sequence (residues 450-601) encoded by exons 6 and 7 has also been proposed on the basis of insulin and insulin-like growth factor 1 receptor chimeras (12). With respect to insulin binding, our data also are compatible with the results obtained in studies using anti-insulin receptor antibodies that inhibit insulin binding. Shaefer et al. (52) and Gustafson et al. (11) have described inhibitory anti-insulin antibodies that recognize epitopes located within the COOH terminus of the a-subunit. Among naturally occurring mutations reported so far that impair insulin binding, no mutation has yet been found within this COOH-terminal region of the a-subunit. On the other hand, all COOH-terminal deletions of the ectodomain that extend into the &-subunit have greatly reduced or absent insulin binding (52). All these observations might be reconciled if the NH,-and COOH-terminal segments of the a-subunit are in close proximity in the native receptor.