Interactions between immunogenic peptides and MHC proteins.

Annu. Rev. Immunol. 1991. 9:527-65 Copyright © 1991 by Annual Reviews Inc. All rights reserved

ANNUAL REVIEWS

Further

Quick links to online content

INTERACTIONS BETWEEN Annu. Rev. Immunol. 1991.9:527-565. Downloaded from www.annualreviews.org by Indiana University - Purdue University Indianapolis - IUPUI on 09/20/12. For personal use only.

IMMUNOGENIC PEPTIDES AND MHC PROTEINS Jonathan B. Rothbardl and Malcolm L. Gefter2

ImmuLogic Pharmaceutical Corp., '855 California Ave., Palo Alto, California 94304, 21 Kendall Square, Cambridge, M assachusetts 02139 KEY WORDS:

MHC, peptides, biological recognition, structure/function correla tions, T-cell recognition

Abstract

The MHC class-I and class-II molecules are highly polymorphic membrane proteins, which bind and transport to the surface of cells peptide fragments of intact proteins. The peptide-MHC complexes are recognized by the antigen-specific receptor of T lymphocytes and are the basis by which the cellular immune system distinguishes self from nonself. In order to perform this function, MHC proteins simultaneously display a large spectrum of structurally divergent peptides for a sufficiently long period of time for the T cell repertoire to scan the cell effectively. Consistent with the protein's biological role, the rates of association and dissociation at physiological pH are very slow relative to other known receptor-ligand interactions. The mechanism by which the proteins do this is still poorly understood, but recent experimental results indicate that the rate determining step may be a conformational change that results in the entrapment of the peptide. A variety of binding assays have been developed that allow study of the detailed kinetics and specificity of the interaction. The optimal peptide length for binding is between 8 and 12 amino acids with the central 5-7 residues contributing the majority of the specific contacts. Determining the conformation of bound peptides has been hampered by the inherent ability of the receptor to bind manifold sequences. Consequently, strategies employing monosubstituted analogs have had only limited success. Approaches using biotinylated amino acids and other bulky substituents 527 0732-0582/9 1/0410-0527$02.00

Annu. Rev. Immunol. 1991.9:527-565. Downloaded from www.annualreviews.org by Indiana University - Purdue University Indianapolis - IUPUI on 09/20/12. For personal use only.

528

ROTHBARD & GEFTER

or multiple substitutions have generated more information. Recent experi ments demonstrating that peptides with polyalanine, polyproline, or poly glycine bind well to M HC proteins have proven that the structural require ments for binding are quite minimal. In fact, a significant factor of the selectivity for binding appears to be the avoidance of deleterious contacts, rather than the need for a large number of critical interactions. Binding experiments also have shown that several peptides can bind a large number of MHC class-I and class-II alleles. The degenerate binding indicated that the binding site of M H C proteins must have a significant number of conserved features. Solution of the crystal structures of the MHC class-I alleles A2 and Aw68 has identified a putative antigen-combining site whose overall dimensions were quite similar between the two structures. The detailed surface topology of the site varied between the two alleles due to the size and chemical properties of the side chains of the polymorphic amino acids composing the cleft. In both cases poorly resolved electron density was present in the binding site, indicating that a significant percentage of the receptors contained bound peptide ligands. Support for this hypothesis has been the identification of small molecular weight peptides isolated from denatured, purified class-I and class-II proteins. The presence of endogenous peptides in the site could explain the unusual feature that only a low percentage of purified MHC molecules can bind ligand. Recent work has shown that multiple conformational forms of MHC proteins exist and that these also are important in binding. Future experiments exploring the dynamics of binding coupled with biochemical strategies as well as the crystallographic solution of a single peptide-M HC complex will greatly increase our understanding of this unique receptor. INTRODUCTION

The MHC class-I and -II molecules are membrane glycoproteins that have evolved the remarkable capacity to bind and display on the surface of cells an extremely large number of structurally diverse peptides. Recognition of the appropriate peptide-MHC complexes by the antigen specific recep tor of T lymphocytes leads to cell proliferation and a cascade of cellular immune responses. This unique recognition mechanism is the means by which the organism originally defines self and distinguishes nonself throughout its lifetime. During embryonic development, this interaction is critical in the establishment of the T-cell repertoire in the thymus, and during the lifetime of the organism, the continual display of fragments of self-proteins is necessary for the maintenance of peripheral tolerance. Specific cellular immune responses are triggered by the display of frag-


MHC-PEPTIDE INTERACTIONS

529

ments of viral, bacterial, or parasitic proteins on either MHC class-lor class-II molecules that are recognized by T lymphocytes. Not surprisingly, due to their unique role, the MHC proteins differ from other defined cell surface receptors in several important respects. Even though the genes encoding the MHC proteins are among the most polymorphic yet defined, any single allele has the capacity to bind an extremely large number of peptides. The range of peptides is sufficiently diverse to allow the individual to develop appropriate cellular immune responses for protection against the spectrum of pathogens and yet be tolerant to a sufficient number of self proteins to avoid deleterious autoimmune syndromes. In addition, the MHC proteins must display simultaneously a large number of peptides on the cell surface for a period sufficiently long to allow the T-cell repertoire to monitor each cell. This review summarizes experimental data about how the MHC mol ecules bind such a large spectrum' of peptides. The subjects of antigen processing and presentation (reviewed in 1-3), T-cell recognition of the complex (reviewed in 4-6), and intracellular transport of proteins (reviewed in Ref. 7) are discussed only in the context of peptide binding, both due to length restrictions and because several other excellent reviews on these subjects have recently appeared. HISTORICAL BACKGROUND

Both Rosenthal (8) and Benaceraff (9) speculated that variations in the ability of different MHC gene products to interact with protein antigens accounted for the genetic differences in immune responsiveness between inbred strains of mice. As prescient as these theories were, almost ten years of investigation were necessary to understand fully how different T-cell recognition of protein antigens differs from the binding of proteins by antibodies. The antibody structure provided a foundation on which a diverse set of antigen specific receptors was postulated to exist on the surface of T cells. The principal dilemma was explaining the need of the self-MHC proteins in recognition. Did the antigen specific receptor bind the MHC proteins as well as the antigen or was a second receptor involved? At the time, a two-receptor model was more easily understood because the alternative model had a fundamental paradox: How could so many different proteins interact with MHC molecules to create unique neo antigens that exhibited features of both the protein and the MHC mol ecule? We now know that a single T-cell receptor binds a complex of fragments of protein antigens bound by MHC class-I and -II molecules on the surface of cells. The path to this understanding has been interesting, and we certainly have not reached the end of the road.


530

ROTHBARD & GEFTER

Important initial experiments were performed by Werdelin, Benaceraff, and Goodman, who demonstrated that structurally related, as well as distinct antigens could compete for T-cell recognition by interfering with antigen presentation ( 1 0- 1 2). Several groups defined the requirements for T-cell stimulation by systematically varying the T cells and consequently their antigen specific receptor, the class-II expressing, antigen-presenting cells, and either protein or peptide antigens (reviewed in Ref. 4). These studies provided indirect evidence that the T-cell receptor interacted with both the MHC protein and the immunogen, and thus the studies eliminated the need for a two-receptor model. Heber-Katz et al refined the system still further by using a T-cell hybridoma that recognized a cytochrome C peptide in the context of two separate I-A alleles ( 1 3). By modifying the peptide sequence, she identified a particular amino acid that interacted with the MHC protein, while another, just three residues away, contacted the T-cell receptor. Although these experiments did not prove directly that peptide ligands were bound by MHC proteins or that the complex was recognized by T-cells, they established the foundation on which future experiments proving the existence of the complex were designed. As more laboratories defined T-cell determinants, the generality that linear peptide fragments and not their native conformation interacted with MHC pro teins became apparent. The demonstration that T-cell recognition could be inhibited by co incubating an excess of a peptide antagonist with the naturally immuno genic peptide provided additional indirect evidence that peptides bound to MHC class-I and -II proteins. Guillet and colleagues were able to inhibit the proliferative response of a class II-restricted hybridoma specific for a bacteriophage sequence with a staphylococcal nuclease peptide previously shown to bind the same M HC allele ( 1 4). These experiments also were the first to demonstrate that a single peptide combining site existed on class II proteins and that each complex was not the result of a unique associ ation, but that MHC molecules were a general receptor for peptides. Inhibition of class I-restricted cellular cytotoxicity also was dem onstrated prior to experimental proof of direct binding. The first antigens used were two closely related pairs of peptides, corresponding to residues of 365-380 of different strains of influenza nucleoprotein and 1 70- 1 82 of two separate H LA class-I proteins ( 1 5, 16). As more determinants were defined, unrelated peptides known to be recognized by a common restric tion element also were shown to compete. The efficiency of the competition correlated with the potency of the peptide antigens in cytotoxicity assays, demonstrating not only that the peptides must bind to the same antigen combining site, but also that their relative affinity is a factor in determining their immunogenicity. Further experimentation also revealed that peptides



53 1

unrelated not only in structure but also by the restriction element by which they were defined, could compete for recognition by cytotoxic T cells ( 1 7) . Bodmer and colleagues demonstrated that one determinant, residues 50-63 of influenza nucleoprotein, could inhibit T-cell recognition of two other nucleoprotein determinants, 1 47-1 58 and 365-380, even though each of the three peptides previously was recognized in the context of different murine class-I proteins. Consequently, the residues 50-63 must bind to all three class-I alleles, and yet T cells recognizing this peptide were present only in a single strain of mice. These authors concluded that the binding specificity of the MHC c1ass-I proteins might be significantly broader than the peptides defined by assaying the immune responses. Further aspects of degenerate binding are discussed later. EXPERIMENTS INITIALLY DEMONSTRATING BINDING MHC Class-II Molecules

Direct proof that MHC molecules bound immunogenic peptides required detergent-solubilized, affinity-purified proteins. The initial successful experi ment utilized equilibrium dialysis to demonstrate that a fluorescently labelled lysozyme peptide bound I-Ak with an apparent K d of 2-4 11M ( 1 8). The peptide (residues 46-61 of hen egg lysozyme) did not bind I-Ad . This specificity corresponded to the restriction of the T cell used to define the determinant. The formation of the peptide-Ia complex could be inhibited only by other peptides known to be recognized by T ceIls restricted by the same murine allele, which further established that binding to thc class-II protein correlated with immune responsiveness ( 1 9). Subsequently, the murine class-II peptide complex was shown to be sufficiently stable to be isolated by gel filtration, which allowed a wide diversity of peptides to be analyzed. The great number of peptides shown to bind provided a convincing argument that the formation of the peptide MHC complex was the molecular explanation for determinant selection (20, 2 1 ). These studies also provided the first evidence that not all peptides bound by MHC proteins were able to elicit a cellular immune response. A peptide from bacteriophage lambda CI repressor specifically bound I-Ed , but did not elicit I-Ed restricted T cells in vivo. The authors realized that this peptide, unlike the majority that were immunogenic, was hom ologous to the I-Ed molecule itself; we postulated that this was the basis of the nonresponsiveness. This implied that the T-cell-receptor repertoire of an individual might be critical in determining responsiveness and might also be a factor in determinant selection. The development of a method for the facile separation of peptide-MHC


532

ROTHBARD & GEFTER

complexes from excess peptide also allowed the kinetics of the interaction to be studied (20). An ovalbumin peptide bound I-N with an apparent micromolar equilibrium constant similar to that reported for the lysozyme pep tide to I_Ak. However, the rate of the formation of the complex was extraordinarily slow, and once formed, the complex was very stable. While the molecular basis for these unusual rates is still unclear, they are the most vivid evid ence that the mechanism by which MHC p rot eins bind peptide ligands is distinctly different from other defined peptide-receptor interactions. Because of the slow on and off rates, the stability of the complex is best conveyed by the rates of dissociation and the half-life of the complex and not by equilibrium constants. The unusual kinetics of binding was shown to be characteristic of HLA proteins and not simply due to detergent solubilization, by demonstrating that I-A and I-E proteins behaved similarly when embedded in planar lipid bilayers (22) and on the surface of intact T cells (23, 24). Several groups have shown that purified HLA DR binds peptides with kinetics similar but not identical to those of murine class-II proteins. Cresswell & Roche demonstrated that an iodinated peptide from influenza hemagglutinin binds D R 1 , 5, and 8 with similar rates of association. However, the rates of dissociation from each differed, so that the Kd for D R I was 1 3 nM, and approximately 25 nM for DR5 and 8. Table I shows that these equilibrium constants are significantly different from those reported for I-A and I-E. Wiley's group also has shown that the same hemagglutinin peptide, an influenza matrix peptide, and a peptide cor responding to residues 380-3915 of the circumsporozoite protein of malaria all can bind D R I (25, 26). In addition, they show that the circumsporozoite peptide bound DRwil and D Rw 1 4 and three natural variants of DR wi1 . A summary of the rates of association, dissociation, and calculated appar ent affinity constants are listed in Table 1 . MHC Class-/ Proteins

Direct binding of peptides to purified class-I molecules has been more difficult to demonstrate than for class-II proteins. In one study the per centage of class-I that could bind radiolabeled peptide was so small that identification of the complex required an affinity column after removal of excess peptide by gel filtration (27). The binding appeared to correlate with immune responsiveness because significantly higher binding was seen using the class-I allele through which the peptide was originally defined. In an attempt to observe a greater amount of the class-I peptide complex, Levy et al developed a solid phase binding assay by attaching peptides to the bottom of a 96-well plate (28). Specific binding was seen when the plates were incubated with radiolabeled class-I proteins. A much higher


Table 1

Thermodynamics and kinetics of peptide binding to MHC class I I-proteins

A- Rate constants for peptide class-II association and dissociation Peptide OVA323-339

Cyt c 8 8- 1 04 HEL 46-6 1 Mat 1 7-31

Class I I

Association (M -' s-')

I_Ad

1 .9

1 .6

1 .3

3.1

I_E k I-Ak DRI

Conditions

Reference

1 0-5

pH 7.2, 37°C

20

1 0-5

pH 5.6, 37°C

Dissociation (s- ')

0.2

1 .7

1 00

1 .2

4xl O-ss-'

6.3

x X x x

1 0-4

1 0- 3

pH 4.6, 37°C (intermediate formation) (intermed ... . .. final complex)

70

0.87

detergent only

87

5.5

5 mM phosphatidylserine

10

2.5

x

x

1 0-6

1 0 -6

37°C

T. S. Jardetzky (pers. commun.)

HA307-3 1 9

bio-HA307-3 1 9 •

DRI

1 20

1 .6 x 1 0-6

DR5

1 \0

2.7

DR8

1 60

DRI

1 .3

4.4

x

x

37°C

25

37°C

24*

1 0-6 10-6

(slow)

Cell surface binding measured using fluoresceinated streptavidin.

� (")

�

tT1 'tI

B-Binding affinities of pep tides for class II-proteins Peptide

Class I I

Quantity (unit)

Value

Reference

S

NBD-HEL46 -61

I-A" I_A k I-Ad

Kd (/lM)

2

18

Z

Kd ({lM)

1 -8

62

HEL46-61 analogs OVA323-339 RNase 20mers

I-A, E k .d

Kd (/lM)

2

20

IC,o (IlM)

20-1250

99

MAT 1 7-3 1

DRI

Kd (/lM)

0.25

T. S. Jardetzky (pers. commun.)

MAT1 7-31

DRI

1 /2 max. binding t o cells (/lM)

0.04

31

HA307-3 1 9

DRI

IC,o ({lM)

HA307-3 1 9

DRI

Kd (/lM)

0.5 0.0 1 3

DR5

0.024

DR8

0.028

tT1

..., tT1

:>:I :> (")

gZ '"

T. S. Jardetzky (pers. commun.) 25

Vl w w


534

ROTHBARD & GEFTER

percentage of the added class-I proteins bound the peptide in this con figuration than in solution phase assays. Up to 40% bound in the plate assay, whereas less than I % was able to bind peptide in solution. This discrepancy and the large amount of degenerate binding found using this assay (discussed later) implied that the interaction might be different in this assay than when the peptide-MHC complex is formed in solution. Collectively these experiments using purified MHC proteins have been crucial in establishing that these proteins are receptors for peptide ligands. However, to determine whether the methods of purification affect the ability of the class-II molecule to bind peptide, the results with isolated M HC proteins must be compared with the binding observed on cell surfaces. PEPTIDE BINDING TO MHC PROTEINS ON OR WITHIN CELLS

Peptides were shown to be the actual form of proteins recognized by T cells by adding the peptides either to antigen-presenting cells in pro liferation assays or to cellular targets in cytotoxicity assays (4, 29). The success of these assays demonstrated that M HC-peptide complexes were being formed when the ligand was added exogenously. Several new methods have been developed that are sufficiently sensitive to detect pep tide-MHC complexes on the surface of cells. The first group to identify bound ligand on intact T cells used insulin modified with a photo reactive probe (30). A B-cell line was incubated with the modified, radiolabeled peptide for various times between I min and I hr: The cells were photo lyzed, washed, disrupted, and separated into subcellular fractions by dis continuous sucrose gradient centrifugation. As assayed by SDS-PAGE, M HC-peptide complexes were formed within 1 min after exposure to peptide and were concentrated in fractions composed of the plasma mem brane and endosomes. No complexes were apparent in the fractions con taining Iysosomes or the Golgi. Ceppellini et al (3 1 ) also reported relatively rapid binding to cIass-II molecules on or within cells. This group assayed the ability of a radio labeled peptide from influenza matrix protein to bind to class-II proteins, without the use of a cross-linking agent, by incubating the peptide with human B-cell lines, then washing and counting the cells. A significant amount of the binding could be inhibited by coincubation with anti-DR monoclonal antibodies. The radioactive signal was apparent within 30 min, and a maximum signal was seen at 1 hr. However, the authors could not distinguish binding from uptake of the radiolabeled peptide because they did not isolate peptide-M HC complexes.



535

A nonradioactive binding assay utilizing biotinylated peptides also has been developed (23, 24). A variety of T cell determinants have been bio tinylated, and incubated with either murine monocytes or human B-cell lines. The cells were washed, treated with fluorescent streptavidin, and analyzed by flow cytometry. A fluorescent signal between 5 and 25 times that observed in the absence of the peptide was seen, but it was only approximately I % of that seen when the cells were stained with a flu orescent anti-DR antibody. In the case of B-cell lines, binding of the biotinylated peptide required cell surface expression of human class-II proteins and was inhibited by anti-HLA DR monoclonal antibodies as well as by a variety of unbiotinylated peptides. In contrast with the other two studies using radio labeled pep tides binding to cells, the rates of associ ation and dissociation of the biotinylated peptide were similar to those reported for purified MHC class-II proteins. Even though binding could be seen after 30 min, the fluorescent signal increased over the 1 6-hr period. A possible explanation for these differences in apparent rates is that detec tion of the biotinylated compounds requires binding of avidin, which when fluorescent will be significantly quenched i f inside a cell. Consequently the major part of the signal will arise from peptide-MHC complexes expressed on the cell surface, whereas the signals apparent in assays using a radio active peptide could arise from populations that are incorporated in intra cellular vesicles as well as from those molecules associated with MHC proteins. That the only method to inhibit the fluorescent signal from the biotinylated analogs was to perform the assay at low temperatures sup ports this hypothesis. The fluorescent signal was not inhibited by reagents that interfered with cytoskeletal elements, such as colchicine, cytochalasin B, or sodium azide. Compounds that affected the pH of lysosomes, such as ammonium chloride or chloroquine also had no effect on the ability of the peptide to bind. These results are consistent with the peptide simply binding to a small subfraction, approximately I % , of the MHC proteins on the cell surface. Possible explanations for the low percentage of the class-II proteins binding peptide are discussed later. Cell surface binding assays have complemented the results of assays using purified proteins and, because that they do not require purification of the individual MHC allele, have allowed a much larger number of alleles to be screened. ALLELE SPECIFIC OR DEGENERATE BINDING

MHC Class-II Proteins

The initial binding experiments using purified I-A and I-E proteins demon strated that several peptides bound to the Ia allele by which the T cell


536

ROTHBARD & GEFTER

was restricted, and originally such experiments were used to define the determinant. For example, the lysozyme peptide corresponding to residues 46-6 1 bound I-N, but not I-Ad ( 1 8); an immunogenic peptide of oval bumin bound I _Ad , but not I-Ed , I -Ak, or 1- Ek (20). Only when such binding exhibited the expected restriction could one be confident that the binding was specific. A more extensive study analyzing the capacity of I I known determinants to bind Ad , Ak , Ed , or Ek provided further evidence that the binding correlated with the M HC restriction pattern associated with the immune responses to these peptides (32). Of the I I peptides, 8 bound best to the Ia molecule that was the restriction element through which they were defined, I bound to both I-E alleles, another bound three of the four la proteins, and I peptide bound them all. The data were quite consistent with the notion that the capacity of an M HC protein to bind a particular peptide was the predominant factor in determinant selection. However, when the same peptides were screened for their ability to inhibit a single radiolabeled peptide in the assay, greater degeneracy of binding became apparent: Each peptide examined bound to more than a single MHC allele. In most cases (J I of the 1 2) the binding to the "appropriate" MHC protein was strongest. In general, degenerate binding was more common between the two I-E alleles and the two I-A alleles than was binding to both an I-A and I-E protein (33). Binding of peptides to human class-II proteins appears to be significantly more degenerate than binding to murine molecules. The first evidence of extensive degenerate binding was an analysis of human T-cell responses to peptides from the circumsporozoite protein of malaria (34). Sinigaglia et al identified a peptide that not only could be recognized by human T cells restricted by seven separate alleles, but also was capable of inducing strong proliferative responses in several strains of inbred mice. Two other groups recently have reported peptides in tetanus toxin that were recog nized by a variety of T cells with a broad range of DR restriction (35, 36). Lanzavecchia et al identified two peptides that were recognized by all human donors who had been primed with the toxin, regardless of their MHC alleles. By using deletion mutants, one peptide was shown to contain three separate epitopes; however, this was not the case for the second peptide. A similar pattern of responsiveness was observed to a set of analogues that were either shortened or monosubstituted. This pattern indicated that this single sequence could bind a wide spectrum of DR alleles. Ho et a l found two other regions in the same protein, which also could be presented to their respective T-cell clones by antigen-presenting cells of many HLA class-II specificities. Do these data contradict the murine studies? We do not believe so. As



537

mentioned earlier, binding to several alleles of I-A or I-E was reported in many of the original papers but was de-emphasized relative to the cor relation of strong binding to the allele, by which the original immune response was identified. Most likely, had more alleles been examined, greater degeneracy would have been observed. Studying degenerate bind ing extensively required the purification of a large number of alleles, however, and this was both expensive and labor intensive. Cell surface binding assays have been very useful because they cir cumvent these problems. Busch et al have demonstrated that a biotinylated analogue of a defined determinant from influenza hemagglutinin could bind class-II proteins on each of 22 homozygous B-cell lines (24). Though not identical for each of the cell lines, the signal was shown to be specific for HLA-DR in each case by using cells transfected with DR molecules and by inhibition with anti-DR, but not with anti-DQ or anti-HLA class I antibodies. This degenerate binding was characteristic of each of three other known T-cell determinants tested. Importantly, their capacity to bind the various alleles was quantitatively different in each case. Partial confirmation of this broad degeneracy in binding to HLA DR proteins has been the demonstration that the HA peptide can bind purified DR l , 5, and 8 (25). The differences between these results and the initial experiments using purified I-A molecules are not due to the different experimental protocols; recent experiments, reported by Grey and Unanue, have con firmed the allele-specific binding that used biotinylated forms of the oval bumin and lysozyme peptide assayed on murine splenocytes (H. McDevitt, personal communication). If generally true, why do the human proteins appear less selective than murine class-If molecules? One potentially significant aspect is that the majority of the binding studies in murine systems have analyzed I-A alleles, whereas in the human, only DR proteins have been analyzed. I-A should be compared with HLA-DQ, due to their structural similarities, and not to DR proteins which, like J-E molecules, have a conserved lX chain. Half of the proposed antigen binding site is composed of residues from the lX chain. Consequently, I-E and DR proteins will share more structural features than DQ or I-A alleles, and greater degeneracy in binding would be expected. I-A and DQ alleles exhibit several significant variations in primary structure. In particular, the I-A alleles can be divided into two separate groups as distinguished by residues 65-67 of the f3 chain. In I-A b, d, and q, these residues are proline, glutamic acid, and isoleucine. In the sequences of I-A f, k, s, and u, two of the corresponding residues are deleted; the third is a t yrosine. The crystallographic structures of these alleles need to be solved in order to determine the exact ramifications these differences have for the tertiary structure; however, this is the most


538

ROTHBARD & GEFTER

dramatic difference found in any of the class-l or class-II H-2 or HLA sequences. In molecular models of the class-II molecule (37), these residues compose part of a potential pocket in the binding site (38). Interestingly, many of the reported examples of clear specificity in binding have com pared binding to an allele of one of these groups with binding to a member of the other set; for example, the ovalbumin peptide binds I_Ad , but not I-A k. Whether these examples of specific interactions are the exceptions or the rule has significant practical ramifications. If degenerate binding is commonplace among DR proteins, then subunit or peptide reagents might be broadly useful as components for vaccines in the outbred human popu lation. However, it makes the development of allele specific antagonists for partial immune suppression in autoimmune therapy much more diffi cult. The experiments demonstrating degenerate binding do not contradict the idea that the peptide-MHC interactions are fundamental to deter minant selection, because no peptide bound all alleles equally well. Vari ations in the rates of association or dissociation clearly will have an effect on determining a peptide's relative immunodominance. However, these results do emphasize that the capacity of peptide to bind is necessary but not sufficient to elicit an immune response. Aspects of the intracellular location of the protein from which the peptide is derived and the way the peptide is generated within the cell are important factors in deter mining the relevance of a peptide in the generation of the cellular immune response . . MHC Class-! Proteins

Degenerate binding has also been seen in the case of peptides binding to MHC class-I proteins. The initial indication came from experiments previously described, by Bodmer et aI, where a peptide defined with K k restricted T cells was able to inhibit T-cell recognition of two other peptides restricted by Kd and D b ( 1 7). Several purified M HC proteins also have been shown to bind to a common peptide attached to a plastic surface. In the initial report Bouillot and colleagues demonstrated that not only HLA A2 but also B37 and B27 could bind to an influenza matrix peptide (residues 57-68) (28). This work was extended by Frelinger et ai, who demonstrated that degenerate binding of class-I proteins was extremely common when peptides were attached to a solid phase (39). Approximately 90% of 1 02 peptides tested bound HLA A2, B27, and B8 equally well. Thc binding was detected directly with radiolabeled class-J proteins or indirectly using monoclonal antibodies specific for each class-I allele. Chen et al concurrently showed that 1 5 of 64 peptides analyzed from viral antigens, HLA A, B, and C proteins, and clatherin light chains bound HLA A2. 1 , Aw68. 1 , Aw69, B44, and B5. The pattern of binding of each



539

peptide to these MHC proteins was similar, arguing that the variations in the binding sites had only minimal effects on their specificity (40). In contrast, using the ability of peptides to inhibit class I-restricted cytotoxicity, Carreno et al have demonstrated that HLA-B37 and HLA A2. 1 bind largely nonoveriapping sets of peptides, a fact consistent with the specificities ofT cells restricted by these class-I proteins (4 1 ). A possible explanation for these conflicting results is that the details of the complex detected in solution might be quite distinct from that identified using solid phase assays. Significant conformational changes could be involved in peptide binding, and a plastic surface might limit the conversion between forms. If an intermediate form of the complex was less selective than the final complex, greater degeneracy might be more apparent using the solid phase assay than using one in solution. Little can be concluded in this case until more is understood of how MHC complexes are formed (see below) and about the active i.e. binding form of MHC class-I and -II proteins that bind peptide. Peptides can not only bind to multiple alleles of either class-lor class II proteins but, as recent experiments have demonstrated, can bind to both classes of MHC molecules also. The initial demonstration was the generation of class II-restricted murine T cells against peptides previously shown to be recognized by either murine or human class I-restricted T cells (42). Hickling et al subsequently demonstrated directly that four of five peptides that were known to be recognized by class I-restricted T cells, also bound to several DR alleles (43). Collectively these experiments provide support for the similarities in the M HC class-I and class-II proteins of both mouse and human. Even though the vast majority of variation in sequence between alleles is located in residues that compose the antigen combining site, many aspects of the site, including its dimensions and some of its chemical features, must be conserved. STRATEGIES OF IDENTIFICATION OF CRITICAL RESIDUES

The unusual combination of specific and degenerate recognition of peptide sequences by M HC proteins has made the identification of the critical factors involved much more difficult than with other peptide receptors. An integral part of proving that T-cell determinants corresponded to linear fragments of proteins was defining their length requirements. A pattern of responsiveness was often seen in both T-cell proliferation and cytotoxicity assays (reviewed in 44-46). The minimal peptide can be defined either as the shortest peptide that generated the maximal response and was therefore the most potent, or as the shortest peptide that stimulates any response. The former definition usually requires pcptides of 1 1 or 1 2 residues in length,


540

ROTHBARD & GEFTER

while the shortest peptides that can generate any responses often need only have 7 or 8 amino acids. Two examples of pentapeptides being recognized by T cells have been described (47, 48); both peptides were bound by H-2 Ld. Although a pentapeptide to date has not been shown to bind directly to M HC class-II proteins, a core of five to seven critical residues has been shown to be characteristic of most T-cell determinants (46, 49). This has been determined not only by the use of truncated peptides, but also by analyzing the effects of substitutions at each position in the determinant. For example, in an analysis of an ovalbumin peptide, Grey and colleagues used peptides truncated at either end to define the core peptide cor responding to residues 327 to 333 (50). An analysis of the effects of modifying these residues demonstrated that the majority of substitutions within this region had a much greater effect on binding than did modi fications of residues outside of this central core. Similar results have been observed in detailed analyses of peptides known to bind HLA DR (26, 5 1-53), HLA A2 (54), H-2Kd ( 1 6, 55), I-A (5 1 , 56, 5 7), and I-E (58-61 ). In all cases a diverse set of amino acids could be substituted for the residues flanking the core, but few substitutions could be made within the cen tral residues without dramatically affecting both T-cell recognition and binding. T-cell determinants exhibit a clear length requirement, but the import ance of individual side chains appears to be restricted to the central amino acids of the peptide. This could be explained (a) if the main chain atoms and not the side chains of the flanking residucs makc important contacts with the MHC protein, (b) if the flanking residues contribute to the ability of the peptide to fold into an appropriate conformation for optimal binding, or (c) if the flanking residues orient the carboxyl and amino groups into critical positions in the binding site. These possible explanations are not mutually exclusive. For example, Allen et al and Rothbard et al have shown that acetylating and amidating several peptides dramatically improved their antigenicity and immunogenicity in the cases where the peptides can be tested in mice (52, 62). Even though these modifications were shown to help helix formation, and though acetylation or amidation alone did not result in similar effects, neither group of investigators could detect any difference in the rate at which the modified peptides bound, nor in their affinity to MHC class-II proteins. Consequently, the molecular basis for their increased potency is not understood. The position of the terminal charges of a peptide clearly can have dramatic effects on binding. Two examples are the variations in the ability of shorter analogues of the cytomegaloviral peptide to bind H-2Ld and the enhanced ability of an influenza nucleoprotein peptide to bind H-2Kd (46, 63). In the former case, as the peptide was decreased in length from the optimal span of nine



54 1

residues to five amino acids, significant changes in the ability of the ana logues to bind the restriction element were observed. Deletion of either the first residue, tyrosine 168, or the last, leucine 1 76, destroyed the capacity of the peptide to be recognized by the T cell. That neither could compete with the parent sequence indicates that both residues make important contacts with the restriction element. However, a shorter peptide, PHFMPTN, lacking both of these residues, was 103 times more potent than PHFMPTNL. The heptapeptides could not be competed by a 103 molar excess of YPHFMPTN, although both octapeptides contained the complete heptapeptide sequence. These results are best explained by the deleterious effects that the a amino and carboxyl charges have on formation of the peptide complex. Their position in the site would dramatically change as the peptide length was modifie d. Additional evidence that the terminal charges of short peptides can have significant effects on binding was the discovery that a deletion of a single amino acid in a nucleoprotein cytotoxic determinant resulted in a peptide that was three orders of mag nitude more potent. Further experimentation made clear that the effects of the deletion were two-fold: (a) a threonine was replaced by glycine at a critical position, and (b) the peptide was shortened, which obviously modified the position of the terminal carboxylate. Two other groups have analyzed the capacity of the flanking sequences to affect binding. Sette et al (64) tried to correlate the capacity to bind to MHC proteins with the predicted conformation of tetrapeptide sequences linked at both ends of their core peptide. Their results indicated that the flanking regions had significant effects on binding, but a simple correlation with the predicted conformation was not possible. This is not surprising since the secondary structure cannot confidently be predicted from the primary sequence. A more thorough study using the same strategy has been done by Anderson et al (65). However, few generalizations could be made except that the effect of tetrapeptide extensions on a central peptide core of ten amino acids was not consistent with the peptide adopting a single regular conformation throughout its entire length. As useful as the strategy of using analogue peptides containing mono substitutions has been in identifying the chemical and physical require ments of a particular residue in binding, it has not been generally useful for identifying which amino acids make important contacts with the MHC proteins. In retrospect, this is due to the inherent ability of this particular class of receptors to bind an extraordinarily wide diversity of peptides. The initial study was done by Allen et al (66) who used a set of hom ologous peptides substituted with alanine for each residue in turn. The results from both binding and T-cell proliferation assays led this group to postulate that the peptide bound in a helical conformation. However, a


542

ROTHBARD & GEFTER

more extensive study attempting to correlate the inability of mono substituted analogues to stimulate a T-cell hybridoma with their capacity to bind the restriction element, was unable to define the orientation of an ovalbumin determinant when it was part of an I-A-peptide complex (50). A regular conformation could not be identified primarily because the majority of peptides containing single substitutions were still able to bind the MHC protein. A complementary strategy was to exchange multiple residues between determinants and to transfer the T-cell specificity between peptides. Unfor tunately, to investigate exhaustively the exchange of all groups of amino acids between two determinants requires the synthesis of a great number of hybrid peptides. For example, if two twelve-amino-acid peptides are aligned colinearly, there are 212 possible ways the residues can be combined. If alternate alignments are considered, then the possibilities are further increased. To make the approach practical, this method requires an align ment of the two peptides. The success of this strategy is based on the following assumptions: (a) the two peptides adopt a similar conformation when bound, (b) both peptides bind in the same location in the binding site and contact many common MHC residues, (c) substitution of amino acids from one sequence into the second will not prevent the hybrid peptide from adopting the appropriate conformation for binding, and (d) substitution of a sufficient number of amino acids results in the retention of the surface necessary for binding to either the MHC protein or the T-ceIl receptor. The first successful use of this strategy was by Guillet et al who, by mutating a single amino acid in a determinant from a phage tail protein, were able to convert it into a sequence that could be recognized by a staphylococcal nuclease specific ceIl (67). Rothbard et al used this approach more broadly in an attempt to determine whether the structural similarities they identified in peptide determinants had merit ( 51). Two peptides previously shown to be recog nized by DRI -restricted T cells were aligned based on homologous residues at relative positions at 1 , 4, 5, and 8, which would form a structurally similar surface if the two pep tides adopted a helical conformation. Based on this alignment, residues from the one peptide were sequentially sub stituted in the sequence of the second peptide. The T-cell specificity was transferred between determinants by exchanging six amino acids. However, an equivalent proliferative response required approximately 1 00 times as much peptide as the natural sequence. Even though these experiments supported the model, the experimental design had two short comings: (a) the only experimental readout was the proliferation of a T cell clone; MHC binding was not examined; and (b) not all possible hybrid peptides were synthesized, and the proposed amino acids necessary for T-



543

cell recognition could have been fortuitously identified. This is possible because the amino acids that were not exchanged were quite homologous. These criticisms were addressed in a study using cytotoxic T cells (55). In this case residues from an influenza nucleoprotein peptide were substituted for corresponding amino acids in an H LA CW3 peptide also seen by Kd_ restricted cells. This series of experiments was more exhaustive, because (a) a number of different alignments of the two peptides were used as the basis for exchange of amino acids; (b) the ability of the hybrid molecules to bind the restriction element was examined indirectly by competitively interfering with T-cell recognition; and (c) once a hybrid peptide was recognized successfully by the T-cell clone, the importance of each sub stituted amino acid was tested by reverting them individually to the original residue of the other determinant. Only one of the four possible alignments of the two peptides tested resulted in hybrid peptides that could stimulate the clone or compete for binding of the natural sequence consistently throughout the set. As with the class-II system, substitution of six amino acids resulted in partial recognition; however, the presence of a seventh improved the potency of the peptide. Five of the six residues were required for recognition, and the spacing of the six essential amino acids was consistent with the peptide adopting a helical conformation when bound. Substitution of less than five of these amino acids in the sequence did not interfere with the peptide's capacity to compete with the natural sequence for recognition but did eliminate all recognition by the T cell, proving that these modifications affected the interaction with the T-cell receptor of the clone but not with the restriction element. A third group has also successfully transferred a T-cell specificity from one sequence to another structurally distinct peptide (68). Using mono substitutions, four positions were shown to be critical for T-cell recognition of an RNase peptide. When these four residues plus an additional amino acid were inserted into a peptide from influenza hemagglutinin, Lorenz et al were able to stimulate an RNase specific hybridoma. In addition, the peptide was able to elicit T cells that recognized the natural sequence when injected into mice. Collectively, these experiments argue that there is a common location in the binding site that many pep tides share. However, MHC-peptide contacts can only be inferred from these experiments because of the lack of specific chemical and physical requirements to interact with the MHC proteins. The degree of this lack of specificity has been emphasized by two recent experiments. An analysis of the ability of a set of monosubstituted analogues of an influenza hemagglutinin (residues 307-3 1 9) to bind HLA D R l revealed that the majority of residues in the peptide could tolerate a diversity of amino acids without affecting binding (52 and unpublished results). The


544

ROTHBARD & GEFTER

single exception was position 309 which appeared to require a bulky hydrocarbon sidechain (either tyrosine, phenylalanine, leucine, isoleucine, valine, or methionine). To test whether this was the single most important contact in the formation of the complex, thrcc pcptides were assayed for binding to purified DR 1 , (a) a 1 3 amino acid containing tyrosine at 309, a lysine at 316, and alanine at all other positions, (b) another with only the lysine at 3 1 6 in polyalanine backbone, and (c) a third with aspartic acid at 309, a lysine at 316, and alanine at all other positions (69). Sur prisingly, the polyalanine peptide with only the tyrosine and lysine bound D R I equally as well as the natural hemagglutinin sequence. The peptide with alanine rather than tyrosine at 309 bound approximately an order of magnitude less well, and the analogue with aspartic acid at 309 bound very poorly. Because the readout of these experiments was the capacity of the analogues to inhibit the binding of a radiolabeled peptide, a tyrosine at 309 might be essential for inhibition of a fast forming intermediate documented by Sadegh-Nasseri and McConnell (70), but not for the for mation of a stable complex. This was shown not to be the case because the polyalanine peptide with tyrosine also was able to form a complex sufficiently stable to be isolated by both gel filtration and SDS-PAGE in the absence of reductants. Independently, Maryanski and colleagues have recently shown that a polyproline backbone can also form a stable peptide-MHC class-I complex (7 1 ). Murine cytotoxic T cells specific for a fragment of human HLA class-I antigens were assayed for their ability to recognize closely related sequences from different MHC alleles and monosubstituted analogues. Based on the ability of the analogues to compete with the natural sequence for recognition, three discontinuous amino acids (a tyrosine separated by six amino acids from a threonine, leucine pair) were identified that were important for binding H-2 Kd . To determine whether these residues alone could allow a peptide to bind, they attempted to use polyglycine and polyproline peptides as spacers between the critical residues (Table 2). A pentaproline spacer was an excellent inhibitor. A peptide containing six prolines also inhibited, as did those containing either four prolines or five glycines. Maryanski and colleagues also demonstrated that in a second K d determinant, influenza nucleoprotein, 1 47- 1 58, tyrosine also makes a fundamental contact. These experiments have demonstrated that the majority of the amino acid side chains are not essential for the formation of either MHC class-I or class-II peptide complexes. However, in both experimental systems the polyalanine or polyproline containing peptides still needed to be about ten amino acids long, consistent with a minimal length requirement for a peptide backbone. Other than the tyrosine side chain there appears to be

MHC-PEPTIDE INTERACTIONS Table 2

545

Comparison of competitor analogs containing the tyr, thr, leu

motif and either proline or glycine spacers. (Taken from reference 71 with permission.) Competitor efficienc y


relation to peptide (AYP,TLA)

Peptide

Sequence

A24 170-182

RYLEN GKETLQRA ---------L -- N-------L --- T

0.29

Kdjl70-182

1.03

(±O.38)

194A

AYPPTALA

0.002

(±0.0008)

0.039

(±0.02)

(±0.13)

194B

AYPPPPTLA

194C

AYPPPPPTLA

1 9 4E

AYPPPPPPTLA

0.083

204

AAPPPPPTLA

0.002

(±0.001)

G6

YPPPPPTLA

0.002

(±0.0009) (±0.007)

(1.0) (±0.03)

G9

AYGGGGTLA

0.008

GIO

AYGGGGGTLA

0.1 I

(±0.06)

Gil

AYGGGGGGTLA

0.03

(±0.017)

Competition experiments were performed with CTL clone CW3/70 1.1 and antigenic peptide CW3

170-182 (O.05-D.l mM).

relatively little need for specific contacts. These experiments imply that a

significant amount of the selectivity for binding by MHC proteins might be the avoidance of deleterious interactions between side chains of the peptide and the MHC proteins rather than highly selective interactions characteristic o/peptide hormone receptors. The deleterious contacts between receptor and ligand recently have been

used to identify important interactions between MHC class-II proteins and peptides. Realizing that MHC proteins exhibited considerable tol erance for monosubstitutions, Rothbard et al have sequentially substituted a lysine modified with long chain biotin at each position in several T-cell determinants (52, 72, 73). When assayed for binding with either purified or class-II proteins on cell surfaces, greater variations in binding were apparent than when analogues containing monosubstitutions with natural amino acids were screened. The pattern of fluorescence observed when a set of biotinylated analogues of a hemagglutinin peptide were used is shown in Figure lao Strong fluorescence was present when lysine-LCB either was placed at the amino terminus or was substituted for proline-307, lysine308, valine-31O, asparagine-313, lysine-3l6, or alanine-318. In contrast, no detectable fluorescence was observed with peptides containing lysine-LCB at residues 311 and 312, while substitution at 309, 314, 315, or 317 resulted in reduced fluorescence. A variety of controls allowed the authors to conclude that the differences in fluorescence arise from the varying capacity

546

ROTHBARD & GEFTER

s.}

• MAJA (OR1)

10

III u C III U WI III


...

0 :J

u: c:

I'll (II

0 METTE (DR1) FJ RJ 2.2.5 (no DR, DQ)

8

� 6 ;r.

,I,

4

� 2

'"

0

N

fzL

Il

. Ii"h Jh

m

h

M

7,l

h

3 07 308 3 0 9 3 1 0 3 1 1 3 1 2 3 1 3 3 1 4 3 1 5 3 1 6 3 1 7 3 1 8 3 1 9

Position of Biotinylation

b.)

T·Cell

MHC Figure J

Binding o f analogues of HA307- 3 1 9, biotinylated a t each position, t o D R I

homozygous, EBV-transformed B cells.

(a)

Each peptide (50 J.lM) was incubated with

transformed B cells homozygous for D R I D w l (MAJA, closed bars, and M ETTE, open bars), or with DR, DQ-negative mutant B cells (RJ 2.2.5, cross-hatched bars). (b) Model of

the conformation adopted by the peptide when bound to D R 1 , based on the data shown in panel (a). Residues 309 to 3 1 7 are folded into a helix, permitting residues 3 1 0, 3 1 3, and 3 1 6, which are not sensitive to biotinylation, to point away from the antigen combining site of HLA DR \ .

o f the class-II molecule to bind the analogues, reflecting the differential effect of biotinylation at each position on the affinity of the interaction. If this interpretation is correct, then the assay provides a quantitative mea sure of the involvement of each amino acid of the peptide in the formation



547

of the complex: The lower the signal, the more important the amino acid is. Therefore residues tyrosine-309, lysine-3 1 1 , glutamine-3 l 2, and, to a lesser extent, threonine-3 1 4, leucine-3 1 5, and leucine-3 1 7 contributed to the formation of the complex. The peptide might bind to D R l in a variety of ways. However, the distinct variations in fluorescence observed when the different analogues were assayed implied that the number of conformations and orientations of the bound peptide was limited. In all possible modes of binding, residues 3 1 1 and 3 1 2 form critical contacts with the restriction element, because biotinylation at these positions eliminated the fluorescent signal at the peptide concentration used. A further constraint on the possible con formations of the bound peptide was that the fluorescent profile peaked at every third residue (3 1 0, 3 1 3, and 3 1 6) within the central portion of the peptide, with significantly less fluorescence in between. This periodicity suggests that the central core adopted a regular conformation consistent with a helix. However, the results were not consistent with the peptide being helical over its entire length because analogues containing lysine LCB at 308 and 3 1 8 resulted in a strong signal. The ability to tolerate substitution with biotinylated lysine at both ends of the peptide might be explained by an increased accessibility of the termini of the helix or by deviations from an ideal a-helix. A model based on this interpretation of the pattern of fluorescence is shown in Figure l b. It consists of a helical core (residues 309-3 1 7) with the two amino acids at each end of the peptide exhibiting greater conformational freedom and not being modeled as part of the repeating structure. The same group has also shown that the hemagglutinin peptide does not bind all alleles equivalently (72). When assayed on different DR4 alleles, the set of peptides generated similar but not identical patterns of fluorescence. Interestingly, the differences in the observed fluorescence could be rational ized by rotations of the peptide in the binding site. To determine whether other peptides bound D R l , four other peptides were examined using the same strategy. As shown in Figure 2, the resulting fluorescent profiles are not identical. There are significant variations; but, on further inspection, interesting similarities can be identified. For each set of peptides, a central core of five residues can be identified that exhibits a common pattern of relative fluorescent signals that are generalized as low, high, low, low, high, as seen in the original hemagglutinin pattern. In all cases with the exception of myoglobin 1 1 0- 12 1 , there are peaks of fluorescence at relative positions i, i + 3, and i + 6 (corresponding to 3 1 0, 3 1 3, and 3 1 6 in the hemagglutinin sequence). In the proposed model (Figure I b) thcsc residues correspond to the first turn of the helix. The analogues containing substitutions in the residues composing the second turn of the helix generated fluorescent

548

ROTHBARD & GEFTER

120

MYO 68

TUB

100 Q)

(.) s: Q) (.)


III Q)

... o ::J

u::

s: til Q) :2:

80 60

r=

-

20 o

'

Antibody responses to non-immunogenic synthetic peptides induced by co-immunization with immunogenic peptides.

Degenerate binding of immunogenic peptides to HLA-DR proteins on B cell surfaces.

Interactions between model proteins and deoxyribonucleic acids.

MHC-NP: predicting peptides naturally processed by the MHC.

Uneven tissue distribution of minor histocompatibility proteins versus peptides is caused by MHC expression.

Getting some "backbone": how MHC binds peptides.

Directional interactions and cooperativity between mechanosensitive membrane proteins.

Functional interactions between microRNAs and RNA binding proteins.

Binding interactions between long noncoding RNA HOTAIR and PRC2 proteins.

Molecular interactions between ribosomal proteins. Evidence for specificity of interaction between isolated proteins.

Strategy for analysing the co-operativity of intramolecular interactions in peptides and proteins.

Usefulness of ELISA Methods for Assessing LPS Interactions with Proteins and Peptides.

Peptides and proteins as drugs.

Human antimicrobial peptides and proteins.

proteins.

Donor major histocompatibility complex (MHC) peptides are presented by recipient MHC molecules during graft rejection.

GAS2-like proteins mediate communication between microtubules and actin through interactions with end-binding proteins.

Modulation of T cell responses with MHC-derived peptides.

Atheroprotective Vaccination with MHC-II Restricted Peptides from ApoB-100.

Binding of truncated peptides to the MHC molecule IA (d).

Observations of interactions between myofibrillar proteins by chemical cross-linking.

Major Histocompatibility Complex (MHC) Class I and MHC Class II Proteins: Conformational Plasticity in Antigen Presentation.

Properties of MHC class I presented peptides that enhance immunogenicity.

Outer membrane proteins preferentially load MHC class II peptides: implications for a Chlamydia trachomatis T cell vaccine.