b-Sheet Models for the Ordered Filamentous Structure Formedby a Peptide That Enhances the Action of Insulin

LARRY WEAVER

Institute of Molecular Biology, Howard Hughes Medical Institute and Department of Physics, University of Oregon, Eugene, Oregon 97403

JAN STAGSTED

Receptron, Inc., 835 Maude Avenue, Mountain View, California 94043

OLAV BEHNKE†

Institute of Anatomy, University of Copenhagen, The Panum Institute, Copenhagen, Denmark

BRIAN W. MATTHEWS

Institute of Molecular Biology, Howard Hughes Medical Institute and Department of Physics, University of Oregon, Eugene, Oregon 97403

AND

LENNART OLSSON

Receptron, Inc., 835 Maude Avenue, Mountain View, California 94043

Received April 29, 1996, and in revised form October 8, 1996

Certain peptides with sequences related to part ofthe major histocompatibility complex class I anti-gen enhance the action of insulin. These peptidesalso aggregate into fibrous structures that seem tobe related to their biological activity. In the currentstudy, the 17-residue peptide with amino acid se-quence Gly-Asn-Glu-Gln-Ser-Phe-Arg-Val-Asp-Leu-Arg-Thr-Leu-Leu-Arg-Tyr-Ala is used as a represen-tative example of these bioactive molecules. As seenby electron microscopy, the peptide associates intogently twisted ribbons, 50 Å thick, in which theamount of twist decreases as the ribbons becomewider. X-ray diffraction analysis suggests that thepeptides are arranged as in an antiparallel b-sheetextending essentially endlessly along the fiber axis.The amino acid sequence of the peptide is such thatone side of the b-sheet is predominantly polar whilethe opposite side is nonpolar. This allows theb-sheetsto form multilayers with alternating hydrophobicand hydrophilic interfaces. The length of the ex-tended peptide (D54 Å) determines the thickness ofthe ribbon and the tendency of individual b-sheetsto twist accounts for the twisting of the ribbons. An

alternative model is also discussed, again based onantiparallel b-sheets, butwith adjacent sheets inter-digitated in a ‘‘side-by-side’’ fashion rather thanforming stacked layers. Comparable inactive pep-tides such as Gly-Asn-Glu-Gln-Ser- Ala-Arg-Val-Asp-Leu-Arg-Thr-Leu-Leu-Arg-Tyr-Tyr (changed aminoacids underlined) do not form ordered filamentousstructures. r 1996Academic Press

INTRODUCTION

Recently there have been a number of reports ofpeptides which aggregate in ways that may berelevant to their biological activity (Halverson et al.,1990; Goldfarb et al., 1993; Zhang et al., 1993;Perutz, 1994; Lansbury et al., 1995; Nguyen et al.,1995). The formation of amyloid deposits in Alzhei-mer’s disease (Lansbury et al., 1995) and the conver-sion of prions to insoluble, aggregated forms incertain human and animal neurodegenerative dis-eases (Nguyen et al., 1995) are cases in point.Another example of this type is a 17-residue

peptide which corresponds in sequence to part of themajor histocompatibility complex class I antigen† Deceased July 27, 1995.

JOURNAL OF STRUCTURAL BIOLOGY 117, 165–172 (1996)ARTICLE NO. 0080

165 1047-8477/96 $18.00Copyright r 1996 by Academic Press

All rights of reproduction in any form reserved.

(MHC-I)1 (Stagsted et al., 1990). This peptide, andrelated members of the class, enhance the action ofcertain hormones by inhibition of the internalizationof the cognate receptors (Stagsted et al., 1993a;Olsson et al., 1994). The bioactive peptides self-aggregate, and it appears that this aggregated formis necessary for activity (Stagsted et al., 1991). It ispossible that degradation fragments of the MHC-Imolecule may self-interact in the organism and onthe cell surface by the same mechanisms as thoseobserved for the bioactive MHC-I derived peptides.This suggests that the peptide may need to be held ina specific conformation, or to be in some multimericstate, to be active. Binding studies of the peptide tocells indicate a Kd value around 1 nM, whereas theEC50 value for biological activity is ,1 µM (Olsson etal., 1994). This suggests that a large majority of themolecules might exist as aggregates and not be freeto interact with cells. If so, the concentration ofactive peptide actually interacting with the cellscould be a minor fraction of the amount of adminis-tered peptide.In this report, we use electron microscopy and

X-ray diffraction to analyze the highly organizedfilamentous structures formed by the 17-amino-acidpeptide of sequence GNEQSFRVDLRTLLRYA. Thispeptide, also referred to as [Ala]Dk-(69–85), is repre-sentative of a series of such peptides with compa-rable biological activity (Stagsted et al., 1993b). TheX-ray and electron microscopy analyses suggest thatthe peptide molecules self-interact into antiparallelb-sheets and form highly organized filamentousstructures. These filaments twist along their longitu-dinal axis with a constant length period, i.e., with aconstant length between turnover points in an indi-vidual fiber. The period was directly correlated to thewidth of the fiber.

EXPERIMENTAL PROCEDURES

Peptides

The synthesis and characterization of peptides have beendescribed previously (Constantine et al., 1993; Stagsted et al.,1993b). Briefly, peptides were assembled on a phenylacetamido-methyl resin using the t-Boc NMP/HOBt protocol of an AppliedBiosystems 430A peptide synthesizer. The peptides were purifiedto greater than 99% homogeneity by preparative high perfor-mance liquid chromatography using a Vydac C18 (2.2 3 25 cm)column and a gradient of acetonitrile in 0.1% aqueous TFA. Thedesired peptides were confirmed by sequence analysis, amino acidcomposition, and fast atom bombardment mass spectrometry.

X-Ray Diffraction

A suspension of the fibers in water (1 mM) was lightly centri-fuged to concentrate the material toward the bottom of the

centrifuge tube. The liquid containing the fibers was then drawnup into a 1-mm diameter thin-walled glass capillary tube. Theflow of the solution along the tube appeared to preferentially alignthe fibers parallel to the tube.An X-ray diffraction pattern (Fig. 1a) of this solution was taken

with graphite-monochromatized CuKa radiation using a 0.3-mmdiameter collimator and a standard precession camera.The capillary tube was then left with the end open under

reduced pressure to permit slow evaporation and concentration ofthe material. We assume that this tends to further align the fibersalong the axis of the tube. Overnight the sample was concentratedagainst the side of the tube with little if any surplus liquidapparent. A diffraction pattern of this material, again taken withmonochromatic CuKa radiation, is shown in Fig. 1b.

Electron Microscopy (EM)

The peptides were prepared in 1 mM solutions in water andstudied in the electron microscope by three different methods.Sections. Aggregated peptide was sedimented at 12 000g for

10 min at room temperature, the supernatant discarded, 2%glutaraldehyde in 0.1M sodium buffer added on top of the gel, andthe gel fixed for 48 hr. The fixed gel was then removed from thetube and postfixed for 2 hr in 1% OsO4 in cacodylate buffer. Toenhance contrast, the gel was incubated with 1% tannic acid inwater for 2 hr, rinsed briefly in water, and immersed in 0.5%uranyl acetate in water overnight. Specimens were dehydrated ingraded ethanols and finally embedded in epon. Sections werecontrasted with lead citrate.Negatively stained preparations. Solutions were applied to

Formvar/carbon-coated, glow-discharged electronmicroscope gridsand negatively stained with 1% uranyl acetate in water. Somespecimens were fixed for 5 min by floating the grids on a drop of1% glutaraldehyde in 0.1M sodium cacocylate buffer before beingnegatively stained.Shadowed preparations. Solutions were applied to pieces of

freshly split mica. The mica was then rinsed with a few drops ofdistilled water and the peptides were fixed on the mica by floatingit on glutaraldehyde, as above, for 5 min, followed by rinsing indistilled water and air drying.Shadowing with platinum was done in a Balzer freeze-etch

apparatus at a shadowing angle of 18°. After shadowing with Pt areinforcing layer of carbon was evaporated onto the specimen at90°. The resulting replicas were floated off the mica onto waterand the replicas picked up on electron microscope grids. Somespecimens were rotation shadowed at 15°.Specimens were studied in a Philips 300 electron microscope

and photographed at 10–80 000 primary magnification.

RESULTS

X-Ray Diffraction

The diffraction pattern of the suspended material(Fig. 1a) was quite diffuse, but showed sharp arcscorresponding to a repeat distance of 4.8 Å. On drying,the pattern became much more distinct (Fig. 1b),presumably due in part to the higher concentrationof the sample and in part to a better orientation ofthe fibers along the capillary tube. The spacings ofthe observed pattern are summarized in Table 1. Thearc corresponding to a d-spacing of 4.8 Å becomesmuch more obvious and is the dominant feature inthe meridional direction. There is also a weaker, butsharp reflection at 2.4 Å (i.e., 4.8 Å/2). In theequatorial direction, a strong arc appears at a spac-

1 Abbreviations used: BFU, basic filamentous unit; CD, circulardichroism; EM, electronmicroscopy;MHC,major histocompatibil-ity complex.

166 WEAVER ET AL.

ing corresponding to d 5 11.3 Å, with weaker reflec-tions at d-spacings of 23.1 and 7.8 Å. This combina-tion suggests an overall repeat distance of 23.1 Åwith second and third order reflections. It might benoted that although the sample in Fig. 1b wasconcentrated by evaporation, it was not rigorouslydried and dehydrated. Therefore, we assume thatthe sample retains a limited amount of water.

Electron Microscopy

In the electron microscope, the active peptidepresents itself primarily as filaments (Figs. 2a and2e) or ribbon-like aggregates of filaments (Figs.2c–2e).

The most striking structures are the ribbons.These are essentially straight (Figs. 2d and 2e) andcan curve slightly over their flat side (Figs. 2b and2d). The longer ribbons are twisted regularly alongtheir length (Figs. 2a and 2c). The period of twist(from crossover to crossover) may vary from 1500 to3000 Å in different ribbons, but is constant in anindividual ribbon. The ribbons are somewhat pliable,as seen when they adapt to underlying structures(Fig. 2e), but they may break when bent over theedge (not illustrated). Some ribbons seemed to frayat their ends (Fig. 2e), suggesting a basic filamen-tous unit (BFU) of the polymer. The BFU measures,40 Å (Figs. 2d and 2e) and may show a faintlydemarcated midline groove, indicating that it con-sists of two still thinner filaments (arrow, Fig. 2e).The BFUs may aggregate laterally forming ribbons(Figs. 2d and 2e) with a lateral periodicity of ,70 Åand a thickness of ,50 Å (Figs. 2b, 2d, and 2e).Ribbons vary in width depending upon the numberof participating BFUs.The ribbon-like structures are not seen in electron

micrographs of inactive peptides (data not shown).

DISCUSSION

Both the X-ray diffraction patterns and the elec-tron microscopy images show clear evidence of well-

FIG. 1. (a) X-ray diffraction pattern of a suspension of the peptide in a capillary tube (horizontal). Experimental details are in the text.The arc with d-spacing of 4.8 Å is indicated by the arrowhead. (b) X-ray diffraction of the sample as in a, but air-dried. The reflection on themeridian with d 5 4.8 Å is marked with an arrowhead. The strong reflection on the equator has d 5 11.6 Å. Other reflections are listed inTable 1.

TABLE IX-Ray Spectra Observed fromAir-Dried Peptide

Meridional reflections Equatorial reflections

d spacing (Å) Comment d spacing (Å) Comment

5.4 Sharp, weak 23.1 Diffuse, medium4.8 Sharp, strong 11.6 Diffuse, strong3.8 Diffuse, weak 7.8 Diffuse, weak3.2 Diffuse, weak2.4 Sharp, weak2.3 Diffuse, weak2.0 Diffuse, weak

167FILAMENTOUS STRUCTURES FORMED BY MHC-I OLIGOPEPTIDES

ordered structure. The peptide is therefore not disor-dered. Rather, it forms some type of repetitive orquasirepetitive aggregate.The most obvious feature of the X-ray diffraction

patterns (Figs. 1a and 1b) is the sharp arc correspond-ing to a repeat distance of d 5 4.8 Å. Because the arcis sharp it must come from a motif that is repeatedmany times, and must therefore extend, at mini-mum, over several thousand Angstrom units. Thetypical width of the ribbons seen in the electronmicrographs is 200–1000 Å, whereas the length ismuch greater (Figs. 2a and 2c). In addition to thereflection at d 5 4.8 Å, the meridional arcs atd-spacings of 5.4 and 2.4 Å are also sharp, whereasall of the equatorial reflections are diffuse (Fig. 1,Table 1). This indicates that the ribbon-like fibersmust be aligned along the capillary tube, as wouldalso be expected from the behavior of other rod-like

structures such as tobacco masaic virus, which tendsto orient along a capillary tube (Holmes et al., 1975).The strong reflection at 4.8 Å is strongly sugges-

tive of antiparallel b-sheet, in which the expecteddistance between hydrogen-bonded polypeptidestrands is close to 4.7 Å (Arnott et al., 1967; Fraserand MacRae, 1973; Rees and Sternberg, 1973). Theindividual polypeptide chains must be aligned per-pendicular to the fiber axis. Also, in order to explainboth the length of the fibers and the sharpness of the4.8 Å reflection, many polypeptide chains must bearranged in succession to form an essentially infiniteantiparallel b-sheet.If the peptide is aligned in an extended conforma-

tion, as in a b-sheet, the overall length of the strandis expected to be 16 3 3.35 Å 5 54 Å (Rees andSternberg, 1973) (Figs. 3a and 3b). As can be seen inFig. 3a, the amino acids on one side of the sheet are

FIG. 2. Electron micrographs of biologically active peptides. All bars are 0.1 µm. (a) Negatively stained section of gel at lowmagnification showing randomly oriented, rather straight ribbons, some of which are twisted (arrows). Magnification, 356 000. (b)Negatively stained section of gel showing the thickness of the ribbons seen in transverse sections (arrows). Magnification, 3136 000. (c)Negatively stained preparation of filaments showing regularly twisted ribbon. Magnification, 3136 000. (d) Preparation, as in a, of ribbonsshowing filamentous substructure (arrow) and association with globular polymers (arrowheads). Magnification, 3136 000. (e) Ribbon bentover the flat side (arrowhead) and fraying at the end into subfilaments (arrow). Magnification, 3136 000.

168 WEAVER ET AL.

predominantly hydrophobic, whereas they are pre-dominantly hydrophilic on the other side. In particu-lar, all the charged residues are segregated on oneside of the sheet. If successive strands are arrangedto form an antiparallel b-sheet (Fig. 3b) the aminoacid side chains on one side of the sheet are predomi-nantly nonpolar (in Fig. 3a these are the side chainsbetween strands 1 and 2, 3 and 4, etc.). On the otherside of the sheet the side chains are almost entirelypolar (i.e., those between strands 2 and 3, etc.). Thisis consistent with the individual b-sheets assem-bling to form alternate hydrophobic and hydrophiliclayers, as illustrated in Fig. 4a. It seems likely that

pairs of sheets would tend to form back-to-backdimers with their hydrophobic surfaces away fromthe solvent. The association of the polar surfaceswould be expected to be weak but might be promotedby electrostatic and other polar interactions. Figure3b suggests that there are positive and negativecharges that would tend to favor association ofadjacent layers, but there are also some positive–positive and negative–negative charge pairs as well.In the latter case, however, the charges are arrangedsuch that small changes in the conformation of theindividual side chains might lead to charge networksthat would, overall, be favorable.

FIG. 3. (a) Schematic illustration showing the sequence of peptide [Ala85]Dk-(69–85). The one-letter code GNEQSFRVDLRTLLRYAcorresponds to Gly-Asn-Glu-Gln-Ser-Phe-Arg-Val-Asp-Leu-Arg-Thr-Leu-Leu-Arg-Tyr-Ala. In an extended b-sheet conformation, alternateside chains are directed to opposite sites of the strand, as shown. Nonpolar side chains (Ala, Leu, Phe, Val) are shown as solid circles,partially nonpolar side chains (Thr, Tyr) as solid squares, and polar side chains (Arg, Asn, Asp, Gln, Glu, Ser) as arrowheads; charges arealso shown. All the charged side chains are on one side of the strand, and most of the hydrophobic side chains are on the other. (b) Peptidestrands arranged in an antiparallel b-sheet. The side chains between strands 1 and 2 extend toward the viewer while those betweenstrands 2 and 3 extend away from the viewer. (c) Alternative arrangement of the peptide chains in an antiparallel b-sheet. Thisarrangement maintains the alignment between strands 1, 3, 5, . . ., and between strands 2, 4, 6, . . ., and optimizes the nonpolar contacts onone side of the sheet (the side chains between strands 1 and 2, 3 and 4, etc.) and the polar contacts on the other side (the side chains betweenstrands 2 and 3, 4 and 5, etc.). The figure also suggests how the ends of adjacent b-sheets might interdigitate. The arrangement shown isjust one of many others that might be considered.

169FILAMENTOUS STRUCTURES FORMED BY MHC-I OLIGOPEPTIDES

What would be the expected separation betweenthe alternate hydrophobic and hydrophilic layers?In known polypeptides that form b-sheet structuresthe intersheet distance can vary substantially,ranging, for example, from 4.5 Å for [Ala-Gly]n to15.2 Å for dehydrated [Lys(HCl)]n (Fraser andMacRae,1973). Because the composition of amino acids in thepresent peptide seems to be representative of proteinstructures in general, it may be appropriate to use,as a guide, the b-sheet sandwich structures that arecommonly observed in a number of different globularproteins (e.g., see Branden and Tooze, 1991; Kuhn etal., 1994). Inspection of such sandwiches shows thatthe separation between the polypeptide backbone onopposite sides of the sandwich is about 11.0–11.5 Å.In order to be explicit we will assume that the actualvalue is 11.5 Å and will take this as the expectedseparation between a pair of b-sheets with theirhydrophobic surfaces in contact. Presumably theseparation between the sheets with their polar sur-faces in contact will be comparable, or perhapsslightly larger, to allow for some water of hydrationtrapped between the polar side chains. We willassume that the value is the same, i.e., 11.5 Å. Theoverall repeat distance is thus 11.5 Å 1 11.5 Å 5 23Å, which agrees well with the lowest-order equato-rial reflection (d 5 23.1 Å; Table 1). Furthermore,the intensity of the equatorial reflection at d 5 11.6Å provides direct support for the assumption that

the successive b-sheets are stacked with a separa-tion of 11.5 Å. The electron micrographs do not showany evidence for structures with spacings of 23.1 Å.Possibly it is not apparent in the electron micro-graphs in part because the more obvious peri-odicity in this direction is 11.6 Å rather than 23 Å(the 11.6 Å reflection is strong while the 23 Åreflection is weaker; see also Fig. 4a), and alsobecause a spacing of 11.6 Å (or 23 Å) is too small toresolve in the electron micrographs.A partial ‘‘alanine scan’’ of the peptide has been

reported (Stagsted et al., 1993b). Of the residuessubstituted with alanine the three that have thegreatest effect on aggregation (and also reduce bio-logical activity) are Phe 6 = Ala, Leu 10 = Ala, andLeu 14=Ala (numbering as in Fig. 3a). According tothe model (Figs. 3b and 4a), these three bulkyresidues contribute to the hydrophobic interfacebetween adjacent b-sheets and it would be expectedthat their substitution with alanine would weakenthis association.The model for the association of the peptides (Fig.

4a) is in essence like the structure of silk (Arnott etal., 1967). The sheet-like nature of the model imme-diately explains the ribbons seen in the electronmicrographs. Also, the thickness of the ribbon deter-mined microscopically (,50 Å) corresponds to thelength of the extended peptide (54 Å) (Figs. 3a and3b).

FIG. 4. (a) Schematic illustrations showing how the individual [Ala85]Dk-(69–85) peptides might associate to form the ribbon-likestructures seen by electronmicroscopy. The horizontal lines correspond to single peptides; the direction within each b-sheet is vertical, as isthe fiber axis. The polar and nonpolar contacts between successive layers of the b-sheet are indicated. (b) An alternative model for theribbon-like structures seen by electron microscopy. Here two antiparallel b-sheets, each shown in Fig. 3c, are stacked ‘‘back-to-back’’ withtheir nonpolar surfaces away from solvent. The fiber axis is vertical and the individual peptides (shown as straight lines 54 Å long) are atright angles, with their directions indicated by the arrowheads. The distance of 11.5 Å corresponds to the separation between the backboneatoms. Including side chains, the expected thickness of the sheet would be about 23 Å.

170 WEAVER ET AL.

It is well known that b-sheets in globular proteinsare usually twisted rather than flat (Chothia, 1973).For a single b-sheet, or for a pair of b-sheets stackedtogether as in a b-sheet sandwich, the twist anglecan be substantial, e.g., 10° or more per polypeptidechain.Asmore andmore sheets are stacked together,however, the amount of twisting will necessarily bereduced. This could explain why the twisting of thewider sheets is less than that of the narrower ones(data not shown).Thus, the model shown in Fig. 4a is strongly

supported by the X-ray data and is consistent withthe primary electron microscopy observations. Themodel does not, however, explain the lateral periodic-ity of 70 Å seen within the ribbons (Fig. 2e). Thedistance of 70 Å would correspond to three succes-sively stacked b-sheet sandwiches (i.e., 70 Å/23.1Å 5 3.0). This set of three b-sheet sandwiches wouldalso have to correspond to the basic filamentous unitseen in the electron micrograph (Fig. 2). It is notobvious, however, why the association of three, ratherthan some other number, of b-sheet sandwicheswould form a folding unit.In order to try to explain the 70 Å lateral periodic-

ity within the ribbons we have explored another typeof association, as shown in Fig. 3c. As is required bythe X-ray observations, it assumes that the peptidesare associated in an antiparallel b-sheet. In order tomaximize the contact between the hydrophobic sidechains on the same side of the sheet (i.e., those sidechains shown between strand 1 and strand 2 in thefigure), strand 1 is displaced six amino acids relativeto strand 2. If strand 3 is assumed to remain inregister with strand 1, as in Fig. 3c, the positive andnegative charges on the opposite side of the sheetgroup together in pairs. This includes not onlyArg 15of strand 2 and Asp 9 of strand 3, but also Arg 7 ofstrand 3 and the carboxy terminus of strand 2. Onecan also envisage that the ends of the strands of onesheet can interdigitate with the ends of the strandsof an adjacent sheet as shown in Fig. 3c. In this wayGlu 3 of strand 2 can interact favorably with theamino terminus of strand 3 and vice versa. Theside-by-side repeat distance is 20 residues 3 3.35Å 5 67 Å. Overall, the arrangement shown in Fig. 3cmaximizes the possible electrostatic interactions be-tween adjacent strands (i.e., six favorable interac-tions and no unfavorable ones). This can be com-pared to the alignment in Fig. 3b, which includesfour favorable electrostatic pairs (including the aminoand carboxy termini) and three unfavorable ones.(Many other alignments might also be considered.)Presumably the sheets shown in Fig. 3c would

associate in pairs with their hydrophobic faces to-gether in the manner shown in Fig. 4b. In the figurethe sheets are aligned to maximize the overlap

between the hydrophobic regions (cf. Fig. 3c). Thiscan be considered as an alternative model for theribbons seen in the electron micrographs. In supportof this, the individual b-sheet strands are at rightangles to the fiber axis, consistent with the promi-nent X-ray reflection with d spacing of 4.8 Å. Also,the side-by-side periodicity of the sheets (67 Å) couldexplain the 70 Å spacing seen in the electron micro-graphs (Figs. 2e and 2f). On the other hand, it isunclear whether the rather tenuous overlapping ofthe ends of the b-sheet strands would be sufficient tomaintain an extended ribbon-like structure. Theexpected thickness of the model shown in Fig. 4b isabout 11.5 Å from backbone-to-backbone, or roughlytwice this (,23 Å) if the side chains are included.This is substantially less than the thickness of theribbon-like structures seen by electron microscopy(,50 Å). Also, because the distances of 11.5 and23 Å are not repeated in the structure (Fig. 4b), thismodel does not explain the reflections with spacingsof 11.6 and 23.1 Å seen in the X-ray patterns.In summary, the model shown in Fig. 4a is consis-

tent with all of the available X-ray data and most ofthe electron microscopy data with the exception ofthe 70 Å lateral periodicity seen in the ribbons (Fig.2e). We do not mean to suggest that this model isunique. One alternative model shown in Fig. 4b wasdeveloped in an attempt to explain the 70 Å feature,but does not account for some of the principalfeatures in the X-ray pattern. One might also con-sider b-sheet models based on the structure of pec-tate lyase C (Yoder and Jurnak, 1995). In this casethe polypeptide chain follows a long spiral withsuccessive strands separated by 4.86 Å. We do notfavor such models, however, in part because therelative amplitudes of the circular dichroism spectrado not correspond (Stagsted et al., 1991; Sieber et al.,1995).The important result is that both models suggest

that the biologically active peptide associates in astructure that is based on an antiparallel b-sheet.Spontaneous assembly of oligopeptides into b-sheet-like assemblies has also been observed in othercontexts (Halverson et al., 1990; Zhang et al., 1993;Perutz, 1994; Lansbury et al., 1995; Nguyen et al.,1995).The observation that the filamentous form of the

peptide is associated with biological activity sug-gests that the association of the peptide into themultimeric structure confers some property that isabsent in the monomeric state. Perhaps the forma-tion of the multimer greatly enhances the potency ofthe peptide by holding it in a biologically activeconformation. One can also imagine that the mul-timer might contain a receptor binding surface thatdoes not exist on an isolated peptide. Knowledge of

171FILAMENTOUS STRUCTURES FORMED BY MHC-I OLIGOPEPTIDES

the conformation of the peptide in the aggregatedstate may provide insight into the possible interac-tion with its target at the cell surface and, at thesame time, may suggest how the peptide can bedispersed in an active form from the aggregates andthereby be made more bioactive.

We thank Dr. Avram Goldstein for valuable suggestions anddiscussions throughout the work and Dr. Susumu Ohno for hisencouragement to pursue the observations described in thispaper.

REFERENCES

Arnott, S., Dover, S. D., and Elliott, A. (1967) Structure ofb-poly-L-alanine: Refined atomic coordinates for an anti-parallel b-pleated sheet, J. Mol. Biol. 30, 201–208.

Branden, C., and Tooze, J. (1991) Introduction to Protein Struc-ture, Garland, NewYork.

Chothia, C. (1973) Conformation of twisted b-pleated sheets inproteins, J. Mol. Biol. 75, 295–302.

Constantine, K. L., Mapelli, C., Meyers, C. A., Friedrichs, M. S.,Krystek, S., and Mueller, L. (1993) Micelle-bound conforma-tional preferences of a peptide derived from a murine majorhistocompatibility complex class I molecule, J. Biol. Chem. 268,22 830–22 837.

Fraser, R. D. B., andMacRae, T. P. (1973) Conformation of FibrousProteins and Related Synthetic Polypeptides, Academic Press,NewYork.

Goldfarb, L. G., Brown, P., Matti, H., Ghiso, J., Frangione, B., andGajdusek, D. C. (1993) Synthetic peptides corresponding to adifferent mutated regions of the amyloid gene in familialCreuzfeldt–Jacob disease show enhanced in vitro formation ofmorphologically different amyloid fibrils, Proc. Natl. Acad. Sci.USA 90, 4451–4454.

Halverson, K., Froser, P. E., Kirschner, D. A., and Lansbury, P. T.,Jr. (1990) Molecular determinants of amyloid depositions inAlzheimer’s disease: Conformational studies of synthetic b-pro-tein fragments, Biochemistry 29, 2639–2644.

Holmes, K. C., Stubbs, G. J., Mandelkow, E., and Gallwitz, U.(1975) Structure of tobacco mosaic virus at 6.7 Å resolution.Nature 254, 192–196.

Kuhn, P., Tarentino, A. L., Plummer, T. H., Jr., and Van Roey, P.(1994) Crystal structure of peptide-N4-(N-acetyl-b-D-glucosami-nyl)asparagine amidase F at 2.2Å resolution, Biochemistry 33,11699–11706.

Lansbury, P. T., Jr., Costa, P. R., Griffiths, J. M., Simon, E. J.,Auger, M., Halverson, K. J., Kocisko, D. A., Hendsch, Z. S.,Ashburn, T. T., Spencer, R. G. S., Tidor, B., and Griffin, R. G.(1995) Structural model for the b-amyloid fibril based oninterstrand alignment of an antiparallel-sheet comprising aC-terminal peptide,Nature Struct. Biol. 2, 990–998.

Nguyen, J. T., Inouye, H., Baldwin, M.A., Fletterick, R. J., Cohen,F. E., Pruisner, S. B., and Kirschner, D. A. (1995) X-raydiffraction of scrapie prion rods and PrP peptides, J. Mol. Biol.252, 412–422.

Olsson, L., Goldstein, A., and Stagsted, J. (1994) Regulation ofreceptor internalization by the major histocompatibility class Imolecule, Proc. Natl. Acad. Sci. USA 91, 9086–9090.

Perutz, M. (1994) Polar zippers: Their role in human disease, Prot.Sci. 3, 1629–1637.

Rees, A. R., and Sternberg, M. J. E. (1973) From Cells to Atoms.An Illustrated Introduction to Molecular Biology, BlackwellSci., Oxford.

Sieber, V., Jurnak, F., and Moe, G. R. (1995) Circular dichroism ofthe parallel b helical proteins pectate lyase C and E. Proteins23, 32–37.

Stagsted, J., Reaven, G. M., Hansen, T., Goldstein, A., and Olsson,L. (1990) Regulation of insulin receptor functions by a peptidederived from a major histocompatibility complex class I anti-gen, Cell 62, 297–307.

Stagsted, J., Baase, W. A., Goldstein, A., and Olsson, L. (1991) Apreformed, ordered structure of a 25-residue peptide from amajor histocompatibility complex class I antigen is required toaffect insulin receptor function, J. Biol. Chem. 266, 12844–12847.

Stagsted, J., Olsson, L., Holman, G. D., Cushman, S. W., andSatoh, S. (1993a) Inhibition of internalization of glucose trans-porters and IGF-II receptors. Mechanism of action of MHC classI-derived peptides which augment the insulin response in ratadipose cells, J. Biol. Chem. 268, 22809–22813.

Stagsted, J., Mapelli, C., Meyers, C., Matthews, B. W., Anfinsen,C. B., Goldstein, A., and Olsson, L. (1993b) Amino acid residuesessential for biological activity of a peptide derived from amajorhistocompatibility complex class I antigen, Proc. Natl. Acad.Sci. USA 90, 7686–7690.

Yoder, M. D., and Jurnak, F. (1995) The parallel b helix and othercoiled folds. FASEB J. 9, 335–342.

Zhang, S., Holmes, T., Lockshin, C., and Rich, A. (1993) Spontane-ous assembly of a self-complementary oligopeptide to form astable macroscopic membrane, Proc. Natl. Acad. Sci. USA 90,3334–3338.

172 WEAVER ET AL.