Proc. Natl. Acad. Sci. USA
`Vol. 80, pp. 1137-1143, February 1983
`Review
`
`Secondary structures of proteins and peptides in amphiphilic
`environments (A Review)
`(amphiphilic surfaces/apolipoproteins/peptide hormones/peptide secondary structure/peptide toxins)
`E. T. KAISER*tt AND F. J. KE1ZDYt
`Departments of *Chemistry and tBiochemistry, The University of Chicago, Chicago, Illinois 60637; and tLaboratory of Bioorganic Chemistry and Biochemistry,
`The Rockefeller University, New York, New York 10021
`Communicated by Daniel E. Koshland, Jr., November 18, 1982
`
`Many peptides and proteins that act at lipid-water
`ABSTRACT
`interfaces assume a unique amphiphilic secondary structure which
`is induced by the anisotropy of the interface. By using synthetic
`peptides in which these inducible amphiphilic structures have been
`optimized, one can show that the amphiphilic a helix is a functional
`determinant of representative apolipoproteins, peptide toxins, and
`peptide hormones. By increasing the amphiphilicity of the struc-
`turally important regions of the molecule, one can enhance the
`biological activity of the peptide even beyond that of the naturally
`occurring polypeptide. It is proposed that rigid amphiphilic sec-
`ondary structures such as a helix, /3 sheet, or irhelix will be found
`in most medium-sizedpeptides acting at membranes and lipid-water
`interfaces.
`A major objective of our research is to achieve an understanding
`of the structural determinants of biologically active peptides
`and polypeptides sufficient to permit us to design, in a rational
`fashion, new peptides with comparable or enhanced activities.
`Much progress has been made in the ability to predict tertiary
`structure from primary sequence information, but the point at
`which a macromolecule, such as an enzyme, possessing a great
`deal of tertiary structure can be designed from first principles
`has not yet been reached. Nevertheless, we have already suc-
`ceeded in designing an enzymatic catalytic site by taking ad-
`vantage of the preexisting tertiary structure present in an en-
`zyme (1-3). In other words, we have carried out the "chemical
`mutation" of the catalytic site of an enzyme into a new reactive
`group capable of carrying out a different type of catalysis. For
`example, we have taken the hydrolytic enzyme papain and, by
`appropriate covalent modification with a flavin group, we have
`been able to obtain a semi-synthetic enzyme that oxidizes dihy-
`dronicotinamides. Ultimately, we would want to design not only
`the catalytic site of an enzyme but also a substrate binding- site.
`Before such a project can be undertaken, one has to establish
`viable mechanisms by which tertiary structures can be gener-
`ated.
`The goals of our present work are more modest, and in the
`present review we shall discuss the progress we have made in
`designing biologically active peptides and polypeptides in cases
`in which the secondary' structure rather than the tertiary struc-
`ture is the dominant factor determining the binding character-
`istics of the peptides. It is the thesis of this work that, in those
`instances in which secondary structural characteristics domi-
`nate the biological activity and physical properties of a peptide
`or polypeptide, it is now possible to achieve the rational design
`of new sequences that have little homology to the naturally oc-
`curring ones but have similar or enhanced biological activity.
`The present article focuses primarily on surface-active pep-
`tides-i.e., peptides that bind to amphiphilic surfaces such as
`phospholipid surfaces, membranes, receptors, etc. We are test-
`ing the hypothesis that peptides that bind to amphiphilic sur-
`
`faces will have important regions comprising amphiphilic sec-
`ondary structures complementary to those of the surfaces (4).
`To illustrate our approach to the design of surface-active pep-
`tides, we shall discuss the progress that has been made in pre-
`paring model peptides that mimic successfully the naturally oc-
`curring systems in the cases of apolipoproteins, peptide toxins,
`and peptide hormones.
`The biological activity of a protein-be it an enzyme, a struc-
`tural protein, or a receptor-strictly depends on its conforma-
`tion because the functions of all proteins rely on the precise spa-
`tial positioning of several functional groups with respect to each
`other. Proteins achieve a well-defined three-dimensional struc-
`ture by the juxtaposition of blocks of structural units which are
`stabilized by secondary structural forces (e.g., see refs. 5 and 6).
`Because secondary structures can be assumed only by peptides
`larger than a certain minimal size, proteins with the usual struc-
`tural organization have to be at least 50-60 amino acids long.
`Below this limit, additional structural restraints are required to
`ensure the maintenance of a unique tertiary structure. Thus,
`the smaller proteins achieve rigidity by the presence of nu-
`merous intrachain disulfide bonds, and cyclic oligopeptides owe
`their unique conformation to the small peptide ring. Without
`these intrachain covalent bonds, medium-size peptides in gen-
`eral will not assume a unique, rigid, stereospecific conforma-
`tion.
`Yet, many oligopeptides serve as biological agents of exqui-
`site specificity, even though they exist in solution in a multitude
`of ill-defined conformer states. In order to be able to express
`their specific function, these peptides must be induced to as-
`sume a special conformation. For these peptides, the inducer
`is usually a protein-an enzyme, a receptor, an antibody, etc.
`The induction is usually a result of stereospecific interactions
`between the ligand and the protein. Such interactions are lim-
`ited by the number of groups that can be accommodated at a
`given "active site," usually not more than four or five groups.
`Thus, protein-ligand interactions are ideally suited to induce a
`specific conformation for small peptides acting as ligands.
`It is most likely that most peptide hormones containing less
`than 10 amino acids achieve their active conformation this way.
`However, a large number of biologically active peptides have
`10-50 amino acids and no intrachain linkages. How do these
`peptides maintain their active conformation? Do they undergo
`protein-induced conformational transitions? If not, is only a small
`fraction of the peptide active at any time? Because many of these
`peptides act on cell surfaces, they are in a very amphiphilic en-
`vironment at the locus of their activity. We thought that in the
`case of many of these peptides this amphiphilic environment
`might induce the active conformation.
`
`Abbreviations: apo, A-I, apolipoprotein A-I; HDL, high density lipo-
`protein.
`
`1137
`
`Downloaded by guest on March 22, 2020
`
`Novo Nordisk A/S Ex. 2007, P. 1
`Mylan Institutional v. Novo Nordisk
`IPR2020-00324
`
`

`

`1138
`
`Review: Kaiser and Kezdy
`
`APOLIPOPROTEIN
`The first macromolecule to draw our attention in the course of
`our development of models for surface-active peptides was apo-
`lipoprotein A-I (apo A-I), the principal polypeptide constituent
`of high density lipoproteins (HDL) (7-9). From circular di-
`chroism spectra of apo A-I solutions that suggested that the
`polypeptide was quite a-helical and from the amino acid se-
`quence, it was proposed by Segrest et al. (10) a number of years
`ago that there are "amphipathic"-or, to conform to modern
`usage, amphiphilic-helical regions in the peptide chain. Sub-
`sequently, Fitch suggested (11) that through approximately two-
`thirds of the apo A-I molecule not only were there such am-
`phiphilic a-helical regions but that these might be repeating units
`approximately 22 amino acids in length punctuated by helix
`breakers such as glycine or proline and connected by short pep-
`tide segments (hinges). The repeating pattern that Fitch found
`was not perfect; indeed, there were several cases in which hy-
`drophobic residues appeared in otherwise hydrophilic regions
`of the helix or in which the converse was true. Nevertheless,
`Fitch's analysis of the repeating units appeared to us (4, 7, 11)
`to provide a very attractive hypothesis that could account for the
`ready ability of apo A-I to bind to the surface of the HDL par-
`ticle.
`In particular, if one considers the structural organization of
`the HDL particle with the phospholipids in a surface mono-
`layer-the phospholipid head groups protruding into the aqueous
`solution and the fatty acyl chains oriented toward the interior
`of the particle-it could be readily visualized that the amphi-
`philic a-helical regions of apo A-I might lie on the surface of the
`HDL particle with the axes of the helices approximately tan-
`gential to the particle surface. This would allow the hydrophobic
`sides of the helices to penetrate into the space between the
`phospholipid head groups, thereby making contact with the hy-
`drophobic chains of the phospholipid. The hydrophilic sides of
`the helices would be oriented toward the aqueous solution in
`the same direction as phospholipid head groups.
`Although this picture appeared attractive, the question arose
`as to how such a proposal might be tested. As the application of
`physical chemical methodology to the study of protein-lipid in-
`teractions progresses, it should be possible eventually to de-
`termine directly the structural characteristics of apo A-I bound
`at the surface of the HDL particle. However, at the present
`time, solution of this structural problem by physical methods
`is a most formidable undertaking. Our approach to testing the
`hypothesis that it is theiaduction of amphiphilic a-helical seg-
`ments in the apo A-I molecule that allows it to bind effectively
`to the phospliolipid surface present in the HDL particle has
`been an organic chemical one. Specifically, wehave undertaken
`to test the structural basis of peptide binding to the phaspho-
`lipid surface by designing and preparing peptides that are pre-
`dicted to have the desired secondary structural characteristics
`(4). In taking this approach, one route would be to prepare seg-
`ments corresponding to regions of the naturally occurring apo
`A-I molecule. Although for a polypeptide for which tertiary
`structure is crucial to biological and physical properties it would
`be unsatisfactory to prepare relatively short regions of the mol-
`ecule and hope that a reasonable simulation of the properties
`of the intact molecule might be achieved, such an approach does
`not appear comparably unattractive for peptide systems in which
`tertiary structure is of little importance and the secondary
`structural characteristics dominate the basic properties. Never-
`theless, we have not considered that the preparation of peptides
`corresponding to the native sequences of the naturally occur-
`ring system is an attractive route to testing the importance of
`the secondary structural characteristics that we have been ex-
`amining (4). There are two reasons for this which can be ap-
`
`Proc. Natl. Acad. Sci. USA 80 (1983)
`preciated from a consideration of our studies on peptide models
`for apo A-I.
`First, we had prepared a number of peptide segments cor-
`responding to the naturally occurring sequences, but we found
`that the shorter segments were not effective models for apo A-
`I. For larger segments, such as a 44-amino acid peptide chosen
`from a region that appears to have a considerable potential to
`form amphiphilic a helices, we have observed a moderately ef-
`fective simulation of some of the properties of the apo A-I mol-
`ecule, but a peptide corresponding to a prototypic 22-amino acid
`amphiphilic a-helical region of apo A-I was a poor model for the
`intact polypeptide (12). Thus, the preparation of the natural
`segments of apo A-I does not give a particularly attractive entry
`into models for the polypeptide. However, beyond the lack of
`success of such peptides in simulating the properties of apo A-
`I, it must be pointed out that, even if the smaller peptides had
`been found to possess properties similar to those of the intact
`polypeptide, one could not be sure that the reason the small
`peptides were active was that their amphiphilic a-helical struc-
`ture was important. The alternative proposal could be made that
`a specific amino acid sequence corresponding to'a kind of "ac-
`tive site" might have been prepared fortuitously by choosing the
`right peptide segment of the intact molecule for synthesis.
`For the reasons just outlined, in our design of models we have
`chosen to take a different route. If a secondary structural char-
`acteristic such as an amphiphilic a helix is indeed crucial to the
`biological and physical properties of a peptide, then it should
`be possible to construct a new sequence with minimal homology
`to the natural sequence but having a similar secondary structure
`(4). For example, we have designed, synthesized, and charac-
`terized a docosapeptide, peptide 1, which, in its structural char-
`acteristics and fundamental properties, epitomizes the apo A-I
`molecule (7). Peptide 1,
`
`5
`10
`Pro-Lys-Leu-Glu-Glu-Leu-Lys-Glu-Lys-Leu-Lys-
`15
`20
`Glu-Leu-Leu-Glu-Lys-Leu-Lys-Glu-Lys-Leu-Ala,
`consists principally of three types of residues: glutamate, as the
`negatively charged hydrophilic residues; lysine, as the posi-
`tively charged residues; and leucine, as the neutral hydrophobic
`residues. These amino acid residues were chosen because in each
`of their categories they have very high helix-forming potential
`(13). Additionally, because we have a great deal of experience
`in preparing peptides by the solid-phase method (14, 15) using
`alanine as the COOH-terminal residue, this residue was chosen
`as the COOH terminus of the model peptide. Because proline
`and glycine residues function as helix breakers in the naturally
`occurring sequence, we decided to place a proline residue at the
`NH2 ternminus of the model peptide, although this was probably
`not essential to its design. As can be-seen from the helical pro-
`jection (16) of peptide1 (Fig. 1), approximately one-third of the
`amphiphilic a helix surface is hydrophilic, and the rest is hy-
`drophobic.
`First we asked whether peptide1 forms an ahelixin solution.
`The ultraviolet circular dichroism spectra over a range of pep-
`tide concentrations showed that there was considerable a-he-
`lical character for it. The measured []222 nm was dependent on
`the peptide concentration. Quantitative analysis of the concen-
`tration dependence of the mean residue ellipticity gave results
`that could be interpreted in terms of a monomer-tetramer equi-
`librium for the peptide. At low peptide concentration we es-
`nm that the peptide was approximately
`timated from the[O]0
`30% a-helical, whereas at high concentration it was calculated
`to be about 50% helical (7), a value in reasonable accord with
`
`Downloaded by guest on March 22, 2020
`
`Novo Nordisk A/S Ex. 2007, P. 2
`Mylan Institutional v. Novo Nordisk
`IPR2020-00324
`
`

`

`Review: Kaiser and Kezdy
`
`19 Glu
`Pro
`
`5 Glu
`16 Lys
`
`9 Lys
`20 Lys
`2 Lys
`13 Leu
`
`Leu3
`Leu 21 Leu
`
`10
`
`Peptide 1 represented to show amphiphilic a-helical seg-
`FIG. 1.
`ment on an Edmundson helical wheel.
`
`that estimated for apo A-I. The interpretation that the doco-
`sapeptide is involved in a monomer-tetramer equilibrium was
`confirmed by a molecular weight measurement: at high peptide
`concentration the molecular weight corresponded to that of the
`tetramer. For apo A-I there is a similar aggregation of the pep-
`tide chains, resulting in the formation of polymeric forms up to
`an octamer (17). Thus, it is only the helical conformation of pep-
`tide 1 that is conducive to self-association by forming tetrameric
`peptide micelles.
`Another method of physical characterization of the apo A-I
`molecule has been through monolayer studies at the air-water
`interface. When the docosapeptide model was compared to apo
`A-I in behavior at the air-water interface, considerable simi-
`larity between these peptide systems was seen. In particular,
`the collapse pressure measured for the monolayer formed by
`either was approximately 21 dynes/cm (17).
`A very important characteristic of apo A-I is its binding to
`phospholipid. It was the consideration of the way in which this
`molecule might bind to a phospholipid surface that initiated our
`modeling studies. Because we wished to make as quantitative
`a comparison as possible between the binding of our docosa-
`peptide model and of apo A-I to phospholipids, we had to de-
`velop an appropriate experimental approach. We prepared
`unilamellar vesicles from egg lecithin according to the Korn and
`Batzri injection procedure (18), and we measured the binding
`of the peptides to the vesicles by using either a rapid ultrafil-
`tration method or rapid gel filtration through a crosslinked
`Sepharose column (7). The peptide concentrations generally were
`measured with reagents such as fluorescamine or o-phthalal-
`dehyde which detect free amino groups. When the binding of
`either apo A-I or peptide I to the egg lecithin vesicles was mea-
`sured, we observed saturation behavior, and the dissociation
`constants measured for the peptides were nearly identical. Spe-
`cifically, Kd was approximately 10-6 M for the binding of apo
`A-I to egg lecithin vesicles and 2 x 10-6 M for the binding of
`peptide 1. Next, we measured the relative binding to choles-
`terol-containing vesicles in which the phospholipid/cholesterol
`ratio was 4: 1. In this case, the binding of the apo A-I molecule
`to the vesicles was improved somewhat from that seen with pure
`egg lecithin, giving a Kd approximately one-third as large. In
`contrast, peptide 1 bound about 50% less tightly to the mixed
`vesicles than it did to pure egg lecithin (8).
`Thus, although peptide 1 is still a reasonable model for apo
`A-I in its binding to the mixed vesicles, it does not behave as
`well compared to apo A-I as it does in the binding to pure egg
`lecithin. In examining the proposed helical regions (17) of apo
`
`Proc. Natl. Acad. Sci. USA 80 (1983)
`
`1139
`
`A-I, we noted that, in some instances, polar hydrophilic resi-
`dues were located in predominantly hydrophobic regions of the
`helices. In particular, there were two arginine residues (Arg-116
`and Arg-123) located in what otherwise would be hydrophobic
`regions of apo A-I (7, 8). We wondered if the 3-OH group of
`cholesterol might have a deleterious interaction with the hy-
`drophobic portion of the amphiphilic a helix of 1 that is in-
`serted into the vesicles. In that event, the rather favorable in-
`teraction of apo A-I with the mixed lecithin/cholesterol vesicles
`might be due to the presence of polar arginine residues in the
`predominantly hydrophobic regions of the respective amphi-
`philic a helices. Therefore, we undertook to incorporate an ar-
`ginine residue into the hydrophobic region of an amphiphilic a
`helix like 1 in order to assess the role of the 3-OH function of
`cholesterol in peptide-cholesterol interaction.
`The newly designed docosapeptide, peptide 2,
`10
`5
`Pro-Lys-Leu-Glu-Glu-Leu-Lys-Glu-Lys-Arg-Lys-
`20
`15
`Glu-Leu-Leu-Glu-Lys-Leu-Lys-Glu-Lys-Leu-Ala
`contained an arginine residue at position 10 in what would be
`otherwise a completely hydrophobic region on the amphiphilic
`a helix (see helical projection of peptide 1 in Fig. 1). Binding
`of this peptide to cholesterol-containing vesicles was appreci-
`ably tighter than to the pure egg lecithin vesicles. This behavior
`was strongly reminiscent of that of apo A-I itself (19). Although
`the differences in the binding constants for the interaction of
`peptide 2 with the cholesterol-containing vesicles and with the
`pure egg lecithin vesicles are small, they nevertheless are com-
`parable to the difference seen for apo A-I. This clearly dem-
`onstrates that, through the approach that we have undertaken,
`it is possible to fine-tune the binding of amphiphilic peptides
`to lipid or phospholipid surfaces. Because we use an experi-
`mental system that allows us to determine directly thermody-
`namic quantities such as dissociation constants, we are able to
`assess the interactions of peptides with lipids or phospholipids
`quantitatively.
`
`PEPTIDE TOXIN
`In considering the modeling of increasingly complicated am-
`phiphilic secondary structures, we thought that it would be
`worthwhile at the next stage to try to design apeptide containing
`not only an amphiphilic secondary structural feature but also an
`active center. In surveying possible candidate peptides for this
`type of modeling our attention was drawn to the bee venom toxin,
`melittin. This 26-amino acid peptide is an activator of phos-
`pholipase A2 and is able to lyse erythrocytes. Consideration of
`the amino acid sequence of the peptide and of various data in
`the literature concerning binding to phospholipids and other
`physical characteristics (20), led us to the conclusion that the
`NH2-terminal 20 amino acids of the peptide chain might be
`forming a rather hydrophobic amphiphilic a helix (21). Accord-
`ing to this picture, the proline residue present in this region might
`cause a kink in the helix. Alternatively, melittin might have a
`structure in which the proline would be flanked on either side
`by two shorter helical segments. In addition to the amphiphilic
`a helix, according to our analysis, there appears to be a hexa-
`peptide active site region at the COOH terminus containing a
`cluster of positive charges. Indeed, melittin-(1-20), which lacks
`this hexapeptide portion, does not lyse erythrocytes but ap-
`parently is quite capable of binding to them (22).
`To test our structural hypothesis for melittin we designed
`
`Downloaded by guest on March 22, 2020
`
`Novo Nordisk A/S Ex. 2007, P. 3
`Mylan Institutional v. Novo Nordisk
`IPR2020-00324
`
`

`

`1140
`
`Review: Kaiser and Kezdy
`
`peptide 3,
`
`5
`H2N-Leu-Leu-Gln-Ser-Leu-Leu-Ser-Leu-Leu-
`10
`15
`Gln-Ser-Leu-Leu-Ser-Leu-Leu-Leu-Gln-Trp-
`20
`25
`Leu-Lys-Arg-Lys-Arg-Gln-Gln-CONH2
`which was made as nonhomologous as possible to the native
`peptide in the NH2-terminal 20 amino acids (21). Because our
`picture of melittin suggested that the proline residue probably
`was notan important factor in the lytic activity of the peptide,
`in our structural modeling we decided to replace this residue
`with serine. The hydrophobic side of the helix in peptide 3 (Fig.
`2) was made up of leucine residues which we chose for their
`high helix-forming potential, hydrophobicity, and electrical
`neutrality. Although glutamine seemed to be the optimal choice
`for the neutral hydrophilic residues, some serine residues were
`included to-increase the hydrophilicity of the model, permitting
`us to match the amphiphilicity of the native peptide. In our model
`peptide 3, a tryptophan was retained at position 19 for studies
`of intrinsic fluorescence, and the COOH-terminal hexapeptide
`portion of melittin was also preserved.
`Circular'dichroism measurements on solutions of the model
`peptide at pH 7.0 suggested(concentration dependency of the
`mean residue ellipticity at 222 nm) that peptide 3 might be ag-
`gregating. By sedimentation equilibrium centrifugation pep-
`tide 3 appeared to be tetrameric at a concentration of 2.5 X 10-
`M (21). Melittin similarly forms tetramers (20, 23). From the
`circular dichroism data, the helix content of peptide 3 was cal-
`culated to be 69% for the tetramer and 35% for the monomer
`(21). In the case of melittin the corresponding values are 48%
`and 18%, respectively.
`Both peptide 3 and melittin form stable monolayers at the
`air-water interface. Their surface pressure-area curves show
`discontinuities at 45 and22 dynes/cm, respectively, indicating
`collapse of the monolayers. The higher collapse pressure and
`also the larger limiting area per molecule of peptide 3 compared
`to melittin indicated that peptide 3 is able to form a longer am-
`phiphilic segment than melittin.
`The hemolytic activity of the model peptideas measured by
`a 30-min-incubation assay was found to be appreciably greater
`than that of melittin (21). Peptide.3 has a higher surface affinity
`than native melittin, consistent with a more extended helical
`structure for the model peptide, and this is important for cell
`lysis. We have examined the kinetics of the lysis of human
`erythrocytes by melittin and model peptide 3 (24). A complete
`19Trp
`Leu
`1
`12 Leu
`Leu8
`Leul155Leu
`i~~6Leu
`Ser4
`\
`9 Leu
`
`Se
`
`l3Leu
`Ser'7
`6Leu
`Set141
`Gln3 107iLeu
`Lysl2
`Gin
`Arg-Lys-Ar-g-Gln-Gln-NH2
`22
`23
`24
`25
`26
`FIG. 2.
`Peptide 3, represented to show amphiphilic a-helical seg-
`ment on an Edmundson helical wheel.
`
`Proc. Natl. Acad. Sci. USA 80 (1983)
`description of the results is beyond the scope of this article, but
`briefly our findings show that melittin or the model peptide binds
`rapidly to the outer surface of the erythrocyte membrane, and
`the surface-bound peptides produce transient openings through
`which hemoglobin molecules can escape. At the same time,
`melittin loses its ability to cause rapid lysis, presumably by
`translocation through the bilayer. In a substantially slower pro-
`cess, internalized melittin produces transient membrane open-
`ings in a steady state. Comparison of the results with peptide
`3 to those with melittin show that, on a molar basis, the synthetic
`analogue produces a fast process that is similar to that caused by
`melittin but is more efficient than the latter in the slow phase.
`The results wethave obtained indicate that the functional units
`sufficient for the activity of melittin-like cytotoxic peptides are
`a 20-amino acid amphiphilic a helix with a hydrophobic/hydro-
`philic ratio that is >1 and a short segment with a high concen-
`tration of positive charges.
`
`HORMONE
`In view of our findings with the model peptides which simulate
`the activity and properties of apo A-I-and melittin and the lim-
`ited possibilities of well-defined secondary structures for pep-
`tides,. we explored the possibility that amphiphilic secondary
`structures could be of importance in hormones as well (21, 25).
`We examined firstthe question of which hormones might con-
`tain amphiphilic helical regions-and focused our attention on
`peptides that are at least 20 amino acids in length and do not
`contain multiple disulfide bridges. We concluded that human
`/3-endorphin (Fig. 3), a 31-amino acid peptide with' potent opi-
`ate activities, is an excellent candidate for the structural ap-
`proach used for the design of models for apo A-I and melittin
`(25). Our analysis of the amino acid sequence of f3-endorphin
`revealed a potential amphiphilic a or7rhelix in the COOH-ter-
`minal region, residues 13-29. In the a-helical arrangement,
`residues 13-29 would form an amphiphific structure in which
`the hydrophobic domain would cover half of the surface of the
`helix and would twist around the length of the helix with the
`hydrophilic residues being either neutral or- basic. In the ii-he-
`lical form, these residues would form a similar amphiphilic struc-
`ture except that the hydrophobic domain would run straight along
`the length of the helix. The importance of lipid interactions in
`the stabilization of helical structure on the opiate receptor has
`been suggested for P-endorphin. Also, the possibility exists that
`an amphiphilic helical structure in j3-endorphin could stabilize
`the molecule against proteolytic degradation, either through in-
`tramolecular hydrophobic interaction with the enkephalin re-
`gion or by intermolecular interactions.
`Unlike residues 13-29, residues 6-12 of human,fendorphin
`have little propensity for formation of secondary structure and
`are not hydrophobic. We have suggested that residues 6-12 serve
`as a spacinglink between the specific enkephalin sequence (res-
`idues 1-5) and the amphiphilic helical region (25). We have now
`designed, synthesized, and tested three /3endorphin models,
`peptides 4, 5, and 6 (ref. 26; unpublished data).
`Peptide 4 ishomologous to ,endorphin in residues 1-19 and
`nonhomologous in residues 20-31 (Fig. 3). Residues 13-31 of
`peptide 4 could form an. amphiphilic a helix which would be
`similar to that postulated for,8-endorphin in this region, except
`that the hydrophobic domain would extend to the COOH ter-
`minus and lay straight along the length of the helical structure
`(Fig. 4 Left). Peptide 4 exhibited behavior similar in many re-
`spects to that of /3-endorphin, including strong binding to both
`8- and A-opiate receptors, high activities in opiate assays on guinea
`pig ileum and rat vas deferens preparations, and considerable
`resistance to proteolytic degradation. The most notable differ-
`
`Downloaded by guest on March 22, 2020
`
`Novo Nordisk A/S Ex. 2007, P. 4
`Mylan Institutional v. Novo Nordisk
`IPR2020-00324
`
`

`

`Review: Kaiser and Ke'zdy
`
`Proc. Natl. Acad. Sci. USA 80 (1983)
`
`1141
`
`,f3Endorphin
`
`10
`5
`H2N-Tyr-Gly-Gly-Phe-Met-Thr-Ser-Glu-Lys-Ser-Gln-Thr-Pro-
`
`Peptide 4
`
`H2N-Tyr-Gly-Gly-Phe-Met-Thr-Ser-Glu-Lys-Ser-Gln-Thr-Pro-
`
`Peptide 5
`
`H2N-Tyr-Gly-Gly-Phe-Met-Ser-Gly-Ser-Gly-Ser-Gly-Ser-Pro-
`
`Peptide 6
`
`H2N-Tyr-Gly-Gly-Phe-Met-Thr-Ser-Glu-Lys-Ser-Gln-Thr-Pro-
`
`15
`20
`25
`30
`Leu-Val-Thr-Leu-Phe-Lys-Asn-Ala-Ile-Ile-Lys-Asn-Ala-Tyr-Lys-Lys-Gly-Glu-OH
`
`Leu-Val-Thr-Leu-Phe-Lys-Gln-Leu-Leu-Lys-Gln-Leu-Gln-Lys-Leu-Leu-Gln-Lys-OH
`
`Leu-Leu-Gln-Leu-Trp-Gln-Lys-Leu-Leu-Lys-Gln-Leu-Gln-Lys-Leu-Leu-Gln-Lys-OH
`
`Leu-Leu-Lys-Leu-Leu-Gln-Lys-Leub-Leu-Leu-Gln-Lys-Leu-Phe-Lys-Gln-Lys-Gln-OH
`FIG. 3. Amino acid sequences of 13-endorphin and peptides 4, 5, and 6.
`ences between peptide 4 and P-endorphin included the ability
`quence of /3-endorphin. [There were three additional sequence
`of peptide 4 to self-associate at low concentration, the slowness
`homologies (Ser-10, Leu-14, and Leu-17) which are coinciden-
`of its action in the two assays, and its considerable nonspecific
`tal to the design of peptide 5.] In the region residues 6-12, an
`binding to rat brain homogenates.
`alternating Ser-Gly sequence was used to mimic the hydrophilic
`Peptide 5 was designed as a complete structural model of 3-
`spacer region proposed for the corresponding /3-endorphin res-
`endorphin. Only the [Met5]enkephalin region at the NH2 ter-
`idues. This sequence is likely to form a flexible hydrophilic pep-
`minus, the specific opiate receptor recognition site, and Pro-13
`tide chain with little tendency for secondary structure forma-
`as a helix-breaking residue were retained from the natural se-
`ton. For peptide 5 (Fig. 4 Center), residues 14-31 were selected
`
`G
`
`Gln
`
`Gin
`
`Lys
`
`Lys
`
`Gln
`
`Lys
`
`Lys
`
`Gin
`
`Gin
`
`Gln
`
`Lys
`
`Gin G
`
`Gln
`
`Lys
`
`Lys
`
`Gln
`
`Lys
`
`Thr
`
`Lys
`
`Gin
`
`Lys
`
`Gin
`
`Lys
`
`Gln
`
`Gln
`Lys
`Residues 13 through 31 represented on an a-helical net (26). (Left) Peptide 4. (Center) Peptide 5. (Right) Peptide 6.
`
`Lys
`
`FIG. 4.
`
`Downloaded by guest on March 22, 2020
`
`Novo Nordisk A/S Ex. 2007, P. 5
`Mylan Institutional v. Novo Nordisk
`IPR2020-00324
`
`

`

`1142
`
`Review: Kaiser and Ke'zdy
`to permit formation of an amphiphilic a helix of a type similar
`to that designed in the case of peptide 4. Peptide 5 has an amino
`acid sequence identical to that of peptide 4 for residue 21-31
`and substantial similarity in residues 13-21. In 6-receptor bind-
`ing, peptide 5 is slightly less potent than &-endorphin but, un-
`like peptide 4, it is 60 times more potent (in Tris buffer at 250C)
`in its affinity for pi-receptors. Peptides 4 and 5 differ the most
`at residues 6-12, in the proposed hydrophilic spacer region. Thus,
`in view of the somewhat contrasting results seen for peptides
`4 and 5 in their affinity for pL-receptors, it seems likely that this
`region is important in determining this affinity. Because pep-
`tides 4 and S have affinities for 6-receptors similar to the affinity
`of 13-endorphin, these receptors are presumably less sensitive
`to the nature of the residues linking the enkephalin sequence
`to the proposed helical regions of these substrates.
`In the guinea pig ileum assay, the opiate activities of peptides
`4 and 5 are similar to those of P3-endorphin, peptide 4 having
`slightly greater potency and peptide S having somewhat less po-
`tency. Although this assay is not sensitive to the nature of a-en-
`dorphin analogues provided that the enkephalin region is not
`altered, the activity of peptide 5 in particular still provides strong
`support for our structural hypothesis because this peptide has
`only four residues homologous to f3-endorphin in its sequence
`from residues 6-31.
`In the rat vas deferens assay, opiate activity is thought to de-
`pend on a class of opiate receptors, called E-receptors, which
`appear to be very specific for P-endorphin and show only low
`sensitivity toward [Met5]enkephalin. The high activities of pep-
`tides 4 and 5 that we have observed in this assay, compared to
`those of truncated f3-endorphin (1-21) and shorter NH2-ter-
`minal fragments of j3-endorphin, demonstrate that all of the ma-
`jor structural features of (3-endorphin necessary for this activity
`have been retained in the design of both model peptides. In this
`assay, peptide 4 is more active than peptide 5. It seems rea-
`sonable to conclude, therefore, that the hydrophilic spacer re-
`gion proposed for re