Proc. Nati. Acad. Sci. USAVol. 80, pp. 7390-7394, December 1983Biochemistry

Identification of a mutant human insulin predicted to contain aserine-for-phenylalanine substitution

(diabetes/HPLC/semisynthesis/insulin gene/insulin structure)

S. SHOELSON*, M. FICKOVA*, M. HANEDA*, A. NAHUM*, G. MUSSO*t, E. T. KAISER*t,A. H. RUBENSTEIN*, AND H. TAGER**Depaments of Biochemistry, Medicine and Chemistry, The University of Chicago, Chicago, IL 60637; and tLaboratory of Bioorganic Chemistry and Biochemistry,The Rockefeller University, New York, NY 10021

Communicated by Murray Rabinowvitz, August 22, 1983

ABSTRACT Using information gained from (i) the relativeHPLC retention of an abnormal insulin present in the serum ofa hyperinsulinemic diabetic patient and (ii) the loss of an Mbo IIrestriction site in one of the patient's insulin gene alleles, it wasrecently predicted that the mutant insulin contained a serine-for-phenylalanine substitution at position B24 or B25. We have nowprepared human [Ser4]insulin and [Ser5]insulin by solid-phasepeptide synthesis -and semisynthesis using an improved approachwhereby the protecting groups can be removed from the finalproduct in a single step. During reversed-phase HPLC analysis,the two semisynthetic insulins were clearly separated from normalinsulin and from each other. Analysis of the patient's immunoaf-finity-purified serum insulin by HPLC and radioimmunoassayshowed that the insulin eluted at the position of [Ser]24insulinand coeluted with the analog when the two were studied inadmixture. Additional studies showed that [SernBa]insulin and[SerB25]insulin have about 16% and 0.5% of the activity of normalinsulin, respectively, in stimulating glucose oxidation by isolatedrat adipocytes. We conclude that the patient's abnormal insulin(insulin Los Angeles) is human [SerB4]insulin and that abnormalinsulins with amino acid replacements at both positions B24 andB25 can be associated with human diabetes.

The proposal that human insulin gene mutations might resultin the biosynthesis and secretion of abnormal insulins with lowbiological activity and that these abnormal insulins might con-tribute to human diabetes has been considered for many years(1). Four years ago, a hyperinsulinemic diabetic patient (patient1) who responded normally to exogenous insulin was identified,and it was proposed that the patient secreted an abnormal in-sulin resulting from a mutation within a single insulin gene al-lele (2, 3). Chemical analysis of the insulin isolated from pan-creatic tissue obtained during an operation to drain a pseu-docyst showed that the abnormal insulin contained a leucine-for-phenylalanine substitution at position B24 or B25 (2), a find-ing confirmed by restriction mapping of the patient's leuko-cyte-derived DNA (4). As sizable amounts of pancreatic insulinfrom this patient were no longer available, and as the oppor-tunity for studying insulin isolated from pancreatic tissue is rare,a reversed-phase HPLC system for separating and identifyingimmunoaffinity-purified samples of human serum insulin wasdeveloped (5). By using this system, coupled with radioimmu-nometric detection of the hormone and with analysis of theHPLC elution patterns of semisynthetic standards of human[LeuB"]insulin and [LeuB"]insulin, the patient's abnormal in-sulin (named insulin Chicago) was identified as human [LeuB25]_insulin.

In the course of these studies, two additional patients (pa-tients 2 and 3) with clinical findings similar to those of the firstwere identified and their serum insulins were analyzed by HPLC(5). The results showed that the three patients secreted chem-ically distinct insulin variants that were clearly separated bothfrom normal insulin and from each other on reversed-phaseHPLC columns; in each case, the patient's serum insulin con-sisted of the corresponding abnormal insulin and normal hu-man insulin in a ratio of about 9:1. Restriction mapping of theleukocyte DNA derived from patient 2 further showed that oneof the patient's insulin gene alleles had undergone a mutationresulting in loss of the Mbo II restriction site (T-C-T-T-C) nor-mally attributable to the DNA sequence encoding phenylala-nine-24 and -25 of the human insulin B chain (T-T-C-T-T-C) (5);point mutations in either codon could result in substitution ofphenylalanine by leucine (TTA, TTG, or CTC), isoleucine (ATC),valine (GTC), tyrosine (TAC), cysteine (TGC), or serine (TCC).As the serum insulin derived from this patient eluted very earlyfrom reversed-phase HPLC columns (indicating an exaggeratedhydrophilic character relative to that of insulin), it was pre-dicted that the patient's abnormal insulin had undergone a serine-for-phenylalanine substitution at position B24 or B25 (5).

In this report, we describe (i) the semisynthesis of human[SerB"]insulin and [SerB25]insulin using solid-phase synthesisof the corresponding COOH-terminal B-chain octapeptides andan improved approach for amino group chemical protection, (ii)the use of these semisynthetic analogs as HPLC standards inidentifying the hydrophilic mutant human insulin called insulinLos Angeles as human [SerB"]insulin, and (iii) the biologicalactivities of both serine-substituted insulin analogs in stimu-lating glucose oxidation by isolated rat adipocytes.

MATERIALS AND METHODSSynthesis of the COOH-Terminal B-Chain Octapeptides of

Insulin Containing Serine-for-Phenylalanine Substitutions. Thepeptides a-CF3CO-Gly-Ser-Phe-Tyr-Thr-Pro-Lys-Thr and a-CF3CO-Gly-Phe-Ser-Tyr-Thr-Pro-Lys-Thr were prepared bysolid-phase methods as described (6) using 2.1 g of chloro-methylated styrene-divinylbenzene copolymer resin substi-tuted with a-t-butyloxycarbonyl-O-benzyl-L-threonine to theextent of 0.48 mmol/g. a-CF3CO-glycine was prepared es-sentially as described by others (7). Peptides cleaved from theresin with hydrofluoric acid in the presence of anisole were ex-tracted into 10% (vol/vol) acetic acid and were gel-filtered onBio-Gel P-2 using 3 M acetic acid; the respective major com-

Abbreviations: t-Boc, t-butyloxycarbonyl; DOP insulin, desoctapeptideinsulin (des-GlyB23,pheB24,pheB25,TyrB26,TrB27, ProB28, LysB29,ThrB30-insulin).

7390

The publication costs of this article were defrayed in part by page chargepayment. This article must therefore be hereby marked "advertise-ment" in accordance with 18 U.S.C. §1734 solely to indicate this fact.

Proc. Natl. Acad. Sci. USA 80 (1983) 7391

ponents were pooled and purified by reversed-phase HPLC on

a column of Zorbax ODS (0.94 X 25 cm; Dupont) using an iso-cratic mobile phase composed of 20% (vol/vol) acetonitrile pre-

pared in an aqueous buffer of 0.1 M phosphoric acid/0.02 Mtriethylamine/0.05 M NaClO4, adjusted to pH 3.0 with NaOH.The major components were pooled, desalted by gel filtrationon Bio-Gel P-2 as before, and lyophilized. The purified pep-

tides (10-16 ,umol) were dissolved in dimethylformamide (finalpeptide concentration, 3.5 mg/ml); triethylamine (10 equiv) and2-t-butyloxycarbonyl-oximino-2-phenylacetonitrile (180 equiv)were added and the solution was incubated at 220C for 3 hr. Thesolvent was then removed under reduced pressure and the res-

idue was repeatedly extracted with diethyl ether to remove sol-uble reagents and their decomposition products. Finally, therespective peptides were dissolved in 10% aqueous piperidine(final peptide concentration, 5 mg/ml) and the solutions were

incubated at 40C for 4 hr prior to removal of solvent under re-

duced pressure, gel filtration on Bio-Gel P-4 in 5 mM N-ethyl-morpholine brought to pH 8 with acetic acid, and lyophiliza-tion. The resulting peptides bearing the t-butyloxycarbonyl (t-Boc) protecting group on the £-amino function of lysine andhaving free glycine NH2 termini were obtained in 95% yieldbased on the amount of purified peptide subjected to the pro-

tection and protecting group removal procedures.Preparation of Bis-(t-Boc)Desoctapeptide Insulin. Mono-

component porcine insulin (identical to human insulin exceptfor substitution of threonine-B30 by alanine-B30, Novo Phar-maceuticals, Copenhagen) was cleaved by trypsin to remove

the COOH-terminal B-chain octapeptide and was purified as

described (6). HPLC analysis of the product using radioim-munoassay to detect small amounts of insulin-related peptides(5) revealed 1-4 parts per 10,000 contamination by des-alanine-B30-insulin and <1 part per 10,000 contamination by intactporcine insulin. Desoctapeptide insulin (des-GlyB ,PheB4,PheB2 TyrB2 Thr 27 ProB28 LysB29,ThrB3O-insulin; DOP insu-lin) (200 mg) was dissolved in 20 ml of dimethylformamide con-

taining 80 ,ul of triethylamine and 4 g of 2-(t-Boc)-oximino-2-phenylacetonitrile, and the mixture was incubated at 4°C for 6hr. Bis-(t-Boc)-DOP insulin was precipitated from the solutionin 95% yield by the addition of diethy ether (6).

Semisynthesis of Human [SerB ]Insulin and Human[SerB"]Insulin. The procedure described below was adaptedfrom Inouye et al. (8). Bis-(t-Boc)-DOP insulin (12 mg) and 12mg of either of the two protected, serine-substituted B-chainoctapeptides were dissolved in a mixture of 55 ,ul of dimethyl-formamide and 70 u1l of 1,4-butanediol. Tris buffer (45 ,ul of0.25 M Tris brought to pH 7.5 with HCl) and tosylphenylala-nine chloromethylketone-treated trypsin (1.2 mg, Worthing-ton) were added and the solution was incubated at 37°C for 12hr with frequent mixing. An additional 1.2-mg portion of tryp-sin was then added and the incubation was continued for an-

other 12 hr; the apparent pH of the incubation mixture was 6.4.At the end of the incubation, the mixture was chromatographedon a Sephadex LH-20 column (1.5 x 48 cm) equilibrated withdimethylformamide/0.5 M acetic acid, 1: 1 (vol/vol). Fractionscorresponding to the higher molecular weight components were

pooled, dried under reduced pressure, and gel filtered on Bio-Gel P-10 in 3 M acetic acid (6). To remove the protecting groupsfrom the semisynthetic products, insulin-containing fractionswere dried under reduced pressure, dissolved in anhydrous tri-fluoroacetic acid (3 ml) containing a drop of anisole, and in-cubated at 4°C for 1 hr. The solvent was then removed underreduced pressure and the products again gel-filtered on Bio-GelP-10 in 3 M acetic acid. Pooled insulin-containing fractions werelyophilized, giving [SerB24]insulin and [SerB25]insulin in theo-retical yields of 57% and 50%, respectively; HPLC analysis

showed the products to be at least 98% pure.HPLC Analysis of Human Serum Insulin. Methods have been

described for (i) the immunoaffinity purification of insulin fromhuman serum, (ii) the preparation of samples for HPLC, (iii)the HPLC elution of insulins from reversed-phase C-18 col-umns (Altex, Berkeley, CA) using an isocratic mobile phasecontaining 29.7% acetonitrile in 0.10 M phosphoric acid/0.02M triethylamine/0.05 M NaClO4 (adjusted to pH 3.0 withNaOH), and (iv) the radioimmunometric detection of fmolamounts of eluted insulin (5, 9).

Biological Assay for Insulin-Stimulated Glucose Oxidationby Isolated Rat Adipocytes. Epididymal fat pads removed from300-g Wistar rats were minced and incubated with shaking at370C for 1 hr in Krebs-Ringer/Hepes buffer, pH 7.4, contain-ing 4% (wt/vol) bovine serum albumin (which had been di-alyzed overnight against the same buffer), 3 mM glucose, andtype II collagenase (Worthington) at 3 mg/ml; 1.2 ml of thecollagenase solution was used for each gram of tissue. The re-sulting suspension was filtered through nylon mesh and the cellswere washed five times at 370C with 10-ml portions of the buff-er lacking both glucose and enzyme. A cell suspension (1% byvolume of isolated adipocytes) was prepared in the Krebs-Ringer/Hepes buffer supplemented with 4% bovine serum al-bumin/5 mM sodium pyruvate; 0.8-ml aliquots of the suspen-sion were added to plastic scintillation vials containing 0.5 tkCi(1 Ci = 37 GBq) of [1-'4C]glucose (Amersham) and insulin orinsulin analogs at concentrations ranging from 0 to 1 ,uM. Thecells were incubated at 37°C for 90 min under 95% 02/5% CO2with shaking. At the end of the incubation, individual suspen-sions were acidified with a drop of sulfuric acid and were im-mediately capped with a second vial containing filter paper sat-urated with Hyamine hydroxide. After 1 hr of shaking at 370C,scintillation fluid was added to the vial containing the adsorbedCO2 and the vial was assayed for radioactivity.

RESULTS

Approaches to the semisynthesis of insulin analogs bearing aminoacid substitutions in the COOH-terminal region of the insulinB chain most often rely on trypsin-catalyzed peptide bond for-mation between the COOH-terminal arginyl residue of desoc-tapeptide insulin and the NH2-terminal glycyl residue of achemically synthesized octapeptide (10). Although synthesis ofthe protected octapeptide can be accomplished by solution-phasetechniques (8, 10, 11), our procedures have relied on rapid andautomated solid-phase methods (6). Retention of a protectinggroup on lysine-B29 during cleavage of the peptide from thesolid support, however, requires careful peptide design whensolid-phase methods are used. Our earlier procedure, identi-fied as Approach I in Fig. 1, was based on incorporation of ly-sine into the peptide as its e-CF3CO derivative. Although en-zyme-catalyzed coupling of the synthetic octapeptide to bis-t-Boc-DOP insulin could then proceed readily, removal of pro-tecting groups from the disulfide bond-containing product re-quired two steps, one of which involved incubation in aqueouspiperidine at pH 12 (6). Our improved method, identified asApproach II in Fig. 1, incorporates glycine-B23 as its CF3COderivative and results in the synthesis of the a-CF3CO octa-peptide with a free lysyl E-amino function. Subsequent pro-tection of the E-amino function with the t-Boc group prior totreatment with piperidine yields a t-Boc-lysine-B29-protectedpeptide with a free NH2-terminal glycyl residue; enzyme-cat-alyzed peptide bond formation can then proceed as before. Al-though this approach requires an additional step, all reactionsproceed with excellent yield. More important, the improvedapproach obviates the need for incubating the semisynthetic

Biochemistry: Shoelson et al.

7392 Biochemistry: Shoelson et al.

Approach I

c-TFA-Octapept ide

c ac,-bis-t-BOC-DOP-insul in

+ trypsin

ct,,a'-biS-t-BOC-E:-TFA-insul in

piperidine -

a,a'-bi s-t -BOC- insul1i n

trifluoroacetic acid

insulin

Approach II

a-TFA-Octapept ide

t -BOC-OPA

c.-TFA-e -t-BOC-octapept ide

piper idi ne

£ -t-BOC-octapept ide

xa,a'-bis-t-BOC-DOP insulin

+ trypsin

aa ',-tris-t-BOC-insulin

trifluoroacetic acid

insulin

FIG. 1. Approaches to thetrypsin-catalyzed semisynthesisof insulin analogs bearing aminoacid replacements in the COOH-terminal octapeptide of the in-sulin B chain. In both cases, theprotected COOH-terminal octa-peptides are synthesized by solid-phase methods. Approach I in-corporates lysine-B29 as its e-CF3CO derivative (6). ApproachII incorporates glycine-B23 as itsa-CF3CO derivative and permitsintroduction ofa t-Boc protectinggroup at lysine-B29. Approach Irequires incubation of the prod-uct insulin in aqueous piperi-dine (a condition potentially sup-porting disulfide-bond inter-change), whereas Approach IIdoes not. The compound identi-fied as t-Boc-OPA is 2-(t-Boc)-ox-imino-2-phenylacetonitile. TFArepresents CF3CO.

product at increased pH, a condition under which undesirableside reactions involving disulfide bond interchange might oc-cur. Final protecting group removal occurs with trifluoroaceticacid in a single step.

Solid-phase synthesis and purification of the octapeptides a-CF3CO-Gly-Ser-Phe-Tyr-Thr-Pro-Lys-Thr and a-CF3CO-Gly-Phe-Ser-Tyr-Thr-Pro-Lys-Thr yielded homogeneous productswith the expected amino acid compositions. Protection of theoctapeptide lysyl E-amino groups and removal of protection fromtheir glycyl a-amino groups resulted in derivatives yielding dansylglycine after treatment with dansyl chloride, acid hydrolysis,and thin-layer chromatography (12); the absence of E-dansyllysine in the acid hydrolysate showed the effectiveness of thesequential protection and protecting group removal proceduresoutlined in Approach II. Incubation of each of these peptidesand bis-(t-Boc)-DOP insulin with trypsin in aqueous-organic so-lution then gave the tris-(t-Boc) semisynthetic insulin products.Fig. 2 a and b show the reversed-phase HPLC elution profilesobtained for aliquots of the reaction mixtures in which pro-tecting groups were removed corresponding to the serine-B25and -B24 octapeptides, respectively. In each case, only residualoctapeptide, residual DOP insulin, and the semisynthetic in-sulin were observed after 24 hr of incubation. Coupling yieldswere 74% and 70%, respectively, for the serine-B24 and -B25insulin analogs. The HPLC-derived time course for the trypsin-catalyzed semisynthesis is shown (Fig. 2 Inset). Notably, thereaction was nearly complete after 2 hr of incubation. The ad-dition of a second aliquot of trypsin after 12 hr did not increasethe yield of product, a result suggesting that thermodynamicequilibrium had indeed been reached and that the extent ofpeptide bond formation had not been limited by enzyme in-activation. Fig. 2c shows the isocratic reversed-phase HPLCseparation of normal human insulin from eight natural andsemisynthetic insulin analogs; all analogs shown have amino acidsubstitutions in the COOH-terminal octapeptide of the insulinB chain. Notably, human [Ser5"]insulin and human [SerB"]in-sulin are clearly separated from each other, from normal human

insulin, and from other analogs bearing COOH-terminal B-chainamino acid replacements.

As the abnormal human insulin from patient 2 had a greatlyexaggerated hydrophilic character, and as the correspondinghuman insulin gene had apparently undergone mutation at anucleotide site corresponding to insulin residue B24 or B25, wepredicted that the insulin variant was either human [SerB24]in-sulin or human [SerB5I]insulin (5). Our synthesis of both insulinanalogs as standards for comparison with the natural mutant in-sulin and our ability to detect fmol amounts of insulin by ra-dioimmunoassay after HPLC separation permitted a direct testof our prediction. As shown in Fig. 3b, only about 4% of thepatient's serum insulin eluted at the position taken by normalhuman insulin. The majority of the serum insulin eluted at aposition identical to that of the serine-B24 human insulin an-alog. Chromatography of a mixture of the patient's serum in-sulin and the semisynthetic serine-B24 analog resulted in a sin-gle peak of immunoreactive insulin with neither asymmetry norbroadening, as shown in Fig. 3c. Furthermore, the small amountof normal insulin in the patient's serum remained well resolvedeven in the presence of the synthetic analog. On the basis ofthe cochromatography of the patient's abnormal insulin and the[SerB']insulin standard, the abnormal insulin called insulin LosAngeles was identified as human [SerB24]insulin.To understand the functional consequences of substituting

serine for phenylalanine at positions 24 and 25 of the insulin Bchain, we examined the biological activities of both semisyn-thetic insulin analogs. Fig. 4 compares the activities of normalinsulin, human [SerB5]insulin and human [SerB25]insulin instimulating [1-'4C]glucose oxidation by isolated rat adipocytes.Notably, both analogs stimulate glucose oxidation to the samemaximal extent as insulin, and all three peptides show paralleldose-response relationships. Nevertheless, both analogs showdecreased biological potency relative to that of normal insulin:based on data derived from three experiments in which the ac-tivities of the analogs were studied in triplicate and the con-centrations of the peptides producing half-maximal biologi-

Proc. Natl. Acad. Sci. USA 80 (1983)

Proc. Natl. Acad. Sci. USA 80 (1983) 7393

jl0 4 8CD tionTime, hr

.00

CD)0-oCU

8)~~~~~~~~

0 -

0 ~~~~~3060Elution time, min

FIG. 2. Analysis of the formation and separation of semisyntheticinsulin analogs by reversed-phase HPLC. Aliquots (0.1-0.3 AD) of thereaction mixtures resulting from the trypsin-catalyzed coupling of bis-(t-Boc)-DOP insulin to each of the two serine-substituted protected oc-tapeptides were removed after selected periods, the protecting groupswere removed from the peptides by incubation in 25 41. of trifluoro-acetic acid for 15 min at 220C, the solvent was removed under reducedpressure, and the products were applied to the HPLC column. (a and b)Isocratic HPLC analysis of the mixtures representing the semisyn-thesis of human [Ser525]insulin and human [SerB24]insulin, respec-tively. The acetonitrile concentration was 29.2% (vol/vol); absorbancewas measured at 214 nm. The positions of insulin without protectinggroups and DOP insulin are shown; the octapeptide from which pro-tecting groups had been removed eluted in the void volume of the col-umn. (Inset) Time course for formation of [SerB2S]insulin as determinedby integration of absorbances obtained during HPLC analysis; the cal-culation of yield was based on the amount of bis-(t-Boc)-DOP insulinused and its conversion to [SerB25]insulin. (c) Isocratic HPLC elutionprofile of a mixture of nine semisynthetic and natural insulin analogs.The acetonitrile concentration was 29.7% and absorbance was deter-mined at 214 nm. The upper abscissa scale corresponds to a and b; thelower scale corresponds to c. The insulins in c are as follows: 1, human[SerB2S]insulin; 2, human [SerB24]insulin; 3, porcine [Ala 24]insulin; 4,human [LeuB26]insulin; 5, porcine [LeuB25]insulin; 6, normal humaninsulin; 7, human [LeuB24]insulin; 8, normal porcine insulin; 9, porcine[LeuB24]insulin.

cal activity were analyzed, human [SerB2S]insulin and human[Ser"5]insulin exhibit only 16% and 0.5% of the potency of thenatural hormone, respectively. These values do not differ greatlyfrom those reported for the biological activities of [LeuB24]in-sulin and [LeuB5I]insulin [about 20% and 2% of the activity ofnatural insulin, respectively (6, 13, 14)].

Because (i) the diabetic patient expressing [Ser5 ]insulin hadtotal fasting insulin levels corresponding to about 90 micro-units/ml of serum (15), (ii) the ratio of normal to abnormal in-sulin in the patient's serum was about 0.04: 1 (Fig. 3b), (iii) ourradioimmunoassay detects [SerB24]insulin and normal insulinequivalently (data not shown), and (iv) [SerB24]insulin has only16% of the relative activity of normal human insulin (Fig. 4),we can calculate the biologically effective concentration of in-sulin occurring in the patient's serum in the fasting state: about

5- SerB25 SerB24 Normal

T- a

x

I> I

CY

06-

cc

0

o 1a)

C0

E

0 80 160Fraction

FIG. 3. Analysis of serum insulin by reversed-phase HPLC andidentification of the human insulin variant. (a) Reversed-phase sepa-ration of semisynthetic standards of human [SerB25linsulin and hu-man [SerB]insulin from normal human insulin; both serine-substi-tuted analogs elute before the natural hormone, but separation of thetwo semisynthetic insulins is complete. (b) HPLC elution profile for im-munoaffinity chromatography-purified serum insulin from the dia-betic patient; the insulin was derived from a fasting sample of the pa-tient's serum. The column effluent was collected in fractions and aliquotsof each were subjected to radioimmunoassay for detection of insulin.Small amounts of normal insulin were detected, but the majority oftheserum insulin eluted at the position of the human [SerB2]insulin stan-dard. (c) Radioimmunometric HPLC elution profile obtained from theanalysis of a mixture of equivalent amounts of the patient's serum-de-rived insulin and the [SerB24]insulin standard.

4 microunits/ml of insulin activity arise from the normal hor-mone, whereas about 13 microunits/ml of insulin activity arisefrom the insulin variant. The sum (17 microunits/ml) seems toplace the normal weight patient within the normal fasting range

100

co

E

E

-0

75

50

25

of-11 -10 -9 -8 -7 -6 -5

log molar peptide concentration

FIG. 4. Comparison of the potencies of human [SerB24]insulin (i),

human [SerB26]insulin (A), and normal insulin (e) in stimulating [1-14C]glucose oxidation by isolated rat adipocytes. Mean data from a sin-gle experiment in which the activities of the three peptides were de-termined in triplicate at various hormone concentrations are shown.Porcine insulin was used as the control. Basal and maximally stimu-lated glucose oxidation corresponded to conversion of 0.15% and 1.63%of the labeled glucose to 14CO2, respectively; data are plotted as per-centage of maximal insulin-stimulated glucose oxidation. This exper-iment is representative of others carried out on three separate occa-sions.

UII

Biochemistry: Shoelson et al.

7394 Biochemistry: Shoelson et al.

(4-22 microunits/ml). Additional studies will be required todelineate the causes of diabetes in this patient and in otherswho secrete abnormal products of the human insulin gene.

DISCUSSIONOur prediction that an abnormal human insulin identified inthe serum of a diabetic patient had serine rather than phenylal-anine at position B24 or B25 (5) prompted us to synthesize andstudy both human [SerBM]insulin and human [SerB"]insulin.The clear resolution of normal human insulin and the two serine-substituted insulin analogs by our reversed-phase HPLC sys-tem and the cochromatography of the patient insulin with hu-man [SerB24]insulin identify the patient's insulin variant as hav-ing a serine-for-phenylalanine replacement at position B24. Theseinvestigations have resulted in the structural identification ofan abnormal insulin gene product associated with human dia-betes: the previous example was the assignment of insulin Chi-cago as human [LeuBu]insulin (5); this study shows that insulinLos Angeles is human [SerB24]insulin. By using standard no-menclature, these two abnormal insulins can now be desig-nated human insulin B25 (Phe -* Leu) and human insulin B24(Phe -- Ser), respectively. It is thus clear that insulin gene mu-tations resulting in amino acid substitutions at positions B24and B25 can be associated with human diabetes, and that suchsubstitutions can result in replacement of phenylalanine by bothhydrophilic and hydrophobic amino acid residues.

As previously noted (2, 6), mutations resulting in insulins withamino acid substitutions in the evolution-invariant tetrapeptidesequence B23 to B26 can provide important clues with regardto both insulin structure-function relationships and the roles ofparticular amino acid side chains in insulin action. Althoughsubstitution of either phenylalanyl residue within this se-quence by leucine (a hydrophobic residue) or by serine (a hy-drophilic residue) results in analogs having lower than normalactivity (refs. 6, 13, and 14; this report), the loss in activity isnot equivalent for the B24 and B25 positional isomers: replacingphenylalanine-B25 (a residue with a side chain on the surfaceof the insulin monomer) with either amino acid yields analogswith only about 1% of normal potency, whereas replacingphenylalanine-B24 (a residue with an inwardly directed sidechain; see refs. 13 and 16) with leucine or serine yields analogswith 10-20% of normal potency. As the relative hydrophobic-ities of the side-chain replacements play little role in deter-mining the activities of the final products, the loss of a phenyl-alanyl side chain appears to be the specific cause of decreasedactivity in each case. For analogs with replacements at positionB25, decreased activity likely results from the loss of a specificside-chain contact with the receptor (perhaps involving iW-or-bital interactions between hormone and receptor aromatic ringsystems); for analogs with replacements at position B24, de-creased activity is probably due to conformational changes inthe insulin monomer resulting from altered side-chain bulk,rather than from altered side-chain functionality (6, 8, 11, 17).HPLC separations of human [SerB"]insulin, human [LeuB4]-insulin, and porcine [LeuB2]insulin from their respective B25-substituted counterparts (Fig. 2c) are consistent with theseconformational changes and show the induction of greater hy-drophobic contact area among invariant residues of insulin whenphenylalanine-B24 is replaced by either a hydrophilic or a hy-drophobic residue.

Although quantitative measures of side-chain hydrophobic-ity havebeen used to analyze peptide retention by reversed-phase HPLC columns, none of four such measures [based onfree energy changes for aqueous/organic phase transfer (18, 19),fragmental hydrophobicity constants (20-22), or experimental

HPLC data (23)] successfully predicted the elution pattern ofinsulin analogs shown in Fig. 2c (data not shown). In fact, ourprediction that insulin Los Angeles arose from a serine-for-phenylalanine substitution (a prediction based on HPLC anal-ysis) required (i) knowledge of the site of the proposed sub-stitution (position B24 or B25, as determined by loss of an MboII restriction site in the patient's abnormal insulin gene) and (ii)analysis of the probabilities of different amino acid substitu-tions (assessed by considering amino acid replacements that mighthave arisen via single nucleotide changes at the correspondinggene loci). Assignment of structure to abnormal insulins iso-lated from human serum thus requires a multidisciplinary ap-proach involving chemical, physical, and molecular biologicalmethods. Our synthesis of human insulin B24 (Phe -- Ser) per-mits the continuing study of its behavior at the insulin receptor,its stimulation of cellular processes, and its metabolic clearanceusing both in vivo and in vitro systems.

We thank Dr. Ken Inouye for providing us with human [LeuB24]in-sulin and human [LeuB5]insulin and Arlene Timosciek and PamelaCadenhead for expert assistance in preparing the manuscript. This workwas supported by National Institutes of Health Grants AM 18347, AM13941, and AM 20595.

1. Kimmel, J. R. & Pollack, H. G. (1967) Diabetes 16, 687-694.2. Tager, H., Given, B., Baldwin, D., Mako, M., Markese, J., Rub-

enstein, A., Olefsky, J., Kobayashi, M., Kolterman, 0. & Poucher,R. (1979) Nature (London) 281, 121-125.

3. Given, B. D., Mako, M. E., Tager, H. S., Baldwin, D., Markese,J., Rubenstein, A. H., Olefsky, J., Kobayashi, M., Kolterman, 0.& Poucher, R. (1980) N. Engl. J. Med. 302, 129-135.

4. Kwok, S. C. M., Chan, S. J., Rubenstein, A. H., Poucher, R. &Steiner, D. F. (1981) Biochem. Biophys. Res. Commun. 98, 844-849.

5. Shoelson, S., Haneda, M., Blix, P., Nanjo, K., Sanke, T., In-ouye, K., Steiner, D., Rubenstein, A. & Tager, H. (1983) Nature(London) 302, 540-543.

6. Tager, H., Thomas, N., Assoian, R., Rubenstein, A., Saekow, M.,Olefsky, J. & Kaiser, E. T. (1980) Proc. Nati. Acad. Sci. USA 77,3181-3185.

7. Weygand, F. & Csendes, E. (1952) Angew. Chem. 64, 136.8. Inouye, K., Watanabe, K., Tochino, Y., Kobayashi, M. & Shi-

geta, Y. (1981) Biopolymers 2, 1845-1858.9. Robbins, D. C., Blix, P. M., Rubenstein, A. H., Kanazawa, Y.,

Kosaka, K. & Tager, H. S. (1981) Nature (London) 291, 679-681.10. Inouye, K., Watanabe, K., Morihana, K., Tochino, Y., Kanaya,

T., Emura, J. & Sakakibara, S. (1979)J. Am. Chem. Soc. 101, 751-752.

11. Wollmer, A., Strasburger, W, Glatler, V., Dodson, G. G., McCall,M., Danho, W., Brandenburg, D., Gattner, H.-G. & Rittel, W.(1981) Hoppe-Seyler's Z. Physiol. Chem. 362, 581-592.

12. Gray, W. R. (1972) Methods Enzymol. 25, 121-138.13. Keefer, L. M., Piron, M.-A., DeMeyts, P., Gattner, H.-G., Dia-

conescu, C., Saunders, D. & Brandenburg, D. (1981) Biochem.Biophys. Res. Commun. 100, 1229-1236.

14. Kobayashi, M.,Ohgaku, S., Iwasak, M., Maegawa, H., Sigeta, Y.& Inouye, K. (1982) Biochem. J. 206, 597-603.

15. Haneda, M., Bergenstal, R., Freidenberg, G., Wishner, W., Blix,P., Provow, S., Polonsky, K. & Rubenstein, A. (1982) Diabetes 31,Suppl. 2, 4 (abstr.).

16. DeMeyts, P., Van Obberghen, E., Roth, J., Wollmer, A. & Bran-denburg, D. (1978) Nature (London) 273, 504-509.

17. Assoian, R. K., Thomas, N. E., Kaiser, E. T. & Tager, H. S. (1982)Proc. Nati. Acad. Sci. USA 79, 5147-5151.

18. Nozaki, Y. & Tanford, C. (1971)J. Biol. Chem. 246, 2211-2217.19. Bigelow, C. C. & Channon, M. (1976) Handb. Biochem. Mol. Biol.

1, 209.20. Rekker, R. F. (1977) The Hydrophobic Fragmental Constant (El-

sevier, New York), p. 301.21. Molnar, I. & Horvath, C. (1977)J. Chromatogr. 142, 623-640.22. Pliska, V., Schmidt, M. & Fanchere, J.-L. (1981)J. Chromatogr.

216, 79-92.23. Meek, J. L. (1980) Proc. Nati. Acad. Sri. USA 77, 1632-1636.

Proc. Natl. Acad. Sci. USA 80 (1983)