In this chapter, I describe the attributes of host and parasite molecules that determine immune recognition. Two terms frequently arise in discussions of recognition. Specificity measures the degree to which the immune system differentiates between different antigens. Cross-reactivity measures the extent to which different antigens appear similar to the immune system. The molecular determinants of specificity and cross-reactivity define the nature of antigenic variation and the selective processes that shape the distribution of variants in populations.
The first section discusses antibody recognition. The surfaces of parasite molecules contain many overlapping antibody-binding sites (epitopes). An antibody bound to an epitope covers about 15 amino acids on the surface of a parasite molecule. However, only about 5 of the parasite's amino acids contribute to the binding energy. A change in any of those 5 key amino acids can greatly reduce the strength of antibody binding.
The second section focuses on the paratope, the part of the antibody molecule that binds to an epitope. Antibodies have a variable region of about 50 amino acids that contains many overlapping paratopes. Each paratope has about 15 amino acids, of which about 5 contribute most of the binding energy for epitopes. Paratopes and epitopes define complementary regions of shape and charge rather than particular amino acid compositions. A single paratope can bind to unrelated epitopes, and a single epitope can bind to unrelated paratopes.
The third section introduces the different stages in the maturation of antibody specificity. Naive B cells make IgM antibodies that typically bind with low affinity to epitopes. A particular epitope stimulates division of B cells with relatively higher-affinity IgM antibodies for the epitope. As the stimulated B cell clones divide rapidly, they also mutate their antibody-binding regions at a high rate. Mutant lineages that bind with higher affinity to the target antigen divide more rapidly and outcompete weaker-binding lineages. This mutation and selection produces high-affinity antibodies, typically of type IgA or IgG.
The fourth section describes "natural" antibodies, a class of naive IgM antibodies. Each natural antibody can bind with low affinity to many different epitopes. Natural antibodies from different B cell lineages form a diverse set that binds with low affinity to almost any antigen. One in vitro study of HIV suggested that these background antibodies bind to the viruses with such low affinity that they do not interfere with infection. By contrast, in vivo inoculations with several different pathogens showed that the initial binding by natural antibodies lowered the concentrations of pathogens early in infection by one or two orders of magnitude.
The fifth section contrasts affinity and specificity. Poor binding conditions cause low-affinity binding to be highly specific because detectable bonds form only between the strongest complementary partners. By contrast, favorable binding conditions cause low-affinity binding to develop a relatively broad set of complementary partners, causing relatively low specificity. The appropriate measure of affinity varies with the particular immune process. Early stimulation of B cells appears to depend on the equilibrium binding affinity for antigens. By contrast, competition between B cell clones for producing affinity-matured antibodies appears to depend on the dynamic rates of association between B cell receptors and antigens.
The sixth section compares the cross-reactivity of an in vivo, polyclonal immune response with the cross-reactivity of a purified, monoclonal antibody. Polyclonal immune responses raise antibodies against many epitopes on the surface of an antigen. Cross-reactivity declines linearly with the number of amino acid substitutions between variant antigens because each exposed amino acid contributes only a small amount to the total binding between all antibodies and all epitopes. By contrast, a monoclonal antibody usually binds to a single epitope on the antigen surface. Cross-reactivity declines rapidly and nonlinearly with the number of amino acid substitutions in the target epitope because a small number of amino acids control most of the binding energy.
The seventh section discusses the specificity and cross-reactivity of T cell responses. Four steps determine the interaction between parasite proteins and T cells: the cellular digestion of parasite proteins, the transport of the resulting peptides to the endoplasmic reticulum, the binding of peptides to MHC molecules, and the binding of peptide-MHC complexes to the T cell receptor (TCR). Mason (1999) estimates that each TCR cross-reacts with ~105 different peptides. If a TCR reacts with a specific peptide, then the probability that it will react with a second, randomly chosen peptide is only ~10−4. Thus, the TCR can be thought of as highly cross-reactive or highly specific depending on the point of view.
The eighth section lists the ways in which hosts vary genetically in their responses to antigens. MHC alleles are highly polymorphic. The germline genes that contribute to the T cell receptor have some polymorphisms that influence recognition, but the germline B cell receptor genes do not carry any known polymorphisms.
The final section takes up promising lines of study for future research.
4.1. Antigens and Antibody Epitopes
An antigenic molecule stimulates an immune response. Each specific subset of an antigenic molecule recognized by an antibody or a T cell receptor defines an epitope. Each antigen typically has many epitopes. For example, insulin, a dimeric protein with 51 amino acids, has on its surface at least 115 antibody epitopes (Schroer et al. 1983). Nearly the entire surface of an antigen presents many overlapping domains that antibodies can discriminate as distinct epitopes (Benjamin et al. 1984).
Epitopes have approximately 15 amino acids when defined by spatial contact of antibody and epitope during binding (Benjamin and Perdue 1996). Almost all naturally occurring antibody epitopes studied so far are composed of amino acids that are discontinuous in the primary sequence but brought together in space by the folding of the protein.
The relative binding of a native and a mutant antigen to a purified (monoclonal) antibody defines one common measure of cross-reactivity. The native antigen is first used to raise the monoclonal antibody. C50mut is the concentration of the mutant antigen required to cause 50% inhibition of the reaction between the native antigen and the antibody. Similarly, C50nat is the concentration of the native antigen required to cause 50% inhibition of the reaction between the native antigen and the antibody (self-inhibition). Then the relative equilibrium binding constant for the variant antigen, C50nat/C50mut, measures cross-reactivity (Benjamin and Perdue 1996).
Site-directed mutagenesis has been used to create epitopes that vary by only a single amino acid. This allows measurement of relative binding caused by an amino acid substitution. Studies differ considerably in the methods used to identify the amino acid sites defining an epitope, the choice of sites to mutate, the amino acids used for substitution, and the calculation of changes in equilibrium binding constants or the free-energy of binding. Benjamin and Perdue (1996) discuss these general issues and summarize analyses of epitopes on four proteins.
Five tentative conclusions about amino acid substitutions suffice for this review. First, approximately 5 of the 15 amino acids in each epitope strongly influence binding. Certain substitutions at each of these strong sites can reduce the relative binding constant by two or three orders of magnitude. These strong sites may contribute about one-half of the total free-energy of the reaction (Dougan et al. 1998).
Second, the other 10 or so amino acids in contact with the antibody may each influence the binding constant by up to one order of magnitude. Some sites may have no detectable effect.
Third, the consequences of mutation at a particular site depend, not surprisingly, on the original amino acid and the amino acid used for substitution.
Fourth, theoretical predictions about the free-energy consequences of substitutions based on physical structure and charge can sometimes be highly misleading. This problem often occurs when the binding location between the antibody and a particular amino acid is highly accessible to solvent, a factor that theoretical calculations have had difficulty incorporating accurately.
Fifth, antibodies raised against a particular epitope might not bind optimally to that epitope—the antibodies sometimes bind more strongly to mutated epitopes. In addition, antibodies with low affinity for an antigen can have higher affinity for related antigens (van Regenmortel 1998).
4.2. Antibody Paratopes
An antibody contains a variety of binding sites. Each antibody binding site defines a paratope, composed of the particular amino acids of that antibody that physically bind to a specific epitope. Approximately 50 variable amino acids make up the potential binding area of an antibody (van Regenmortel 1998). Typically, only about 15 of these 50 amino acids physically contact a particular epitope. These 15 or so contact residues define the structural paratope. Only 5 or so of these amino acids dominate in terms of binding energy. However, in both epitope and paratope, substitutions both in and away from the binding site can change the spatial conformation of the binding region and affect the binding reaction (Wedemayer et al. 1997; van Regenmortel et al. 1998; Lavoie et al. 1999).
The antibody's 50 or so variable amino acids in its binding region define many overlapping groups of 15 amino acids. Thus, an antibody has a large number of potential paratopes. A paratope does not define a single complementary epitope; rather it presents certain molecular characteristics that bind antigenic sites with varying affinity. This leads to four aspects of antibody-antigen specificity.
First, an antibody can have two completely independent binding sites (paratopes) for unrelated epitopes (Richards et al. 1975). Bhattacharjee and Glaudemans (1978) showed that two purified mouse antibodies (M384 and M870) each bind methyl α D-galactopyranoside and phosphorylcholine at two different sites in the antigen-binding region of the antibody.
Second, an antibody presumably has many overlapping paratopes that can potentially bind to a variety of related or unrelated epitopes. I did not, however, find any studies that defined for a particular antibody the paratope map relative to a set of variable epitopes. The potential distribution of paratopes may change as a B cell clone matures in response to challenge by a matching antigen—I take this up in the next section (4.3), on Antibody Affinity Maturation.
Third, a single paratope can bind two unrelated epitopes (mimotopes, Pinilla et al. 1999; Gras-Masse et al. 1999). Kramer et al. (1997) scanned a library of synthetically generated peptides for competition in binding to a monoclonal (purified) antibody. X-ray diffraction of three competing peptides showed that they all bound to the same site on the antibody (Keitel et al. 1997).
Fourth, a particular epitope can be recognized by two different paratopes with no sequence similarity. For example, Lescar et al. (1995) used x-ray diffraction to study the physical contact between guinea fowl lysozyme and two different antibodies. The two antibodies contact the same 12 amino acids of the antigen. However, the antibodies have different paratopes with no identical amino acids in the region that binds the antigen. The two antibodies also have different patterns of cross-reactivity with other antigens.
Experimental studies of specificity frequently compare pairwise affinities between an epitope and various paratopes or between a paratope and various epitopes. In these pairwise measures, one first raises antibody to a monomorphic (nonvarying) antigenic molecule and then isolates a single epitope-paratope binding—in other words, one raises a monoclonal antibody that binds to a single antigenic site. Variations in affinity are then measured for different epitopes holding the paratope constant or for different paratopes holding the epitope constant.
Alternatively, one can challenge a host with a polymorphic population of antigens. One controlled approach varies the antigens only in a small region that defines a few epitopes (Gras-Masse et al. 1999). If exact replicas of each epitope occur rarely, then antibodies will be selected according to their binding affinity for the aggregate set of varying epitopes (mixotopes) to which they match. This method may be a good approach for finding antibodies with high cross-reactivity to antigenic variants of a particular epitope.
4.3. Antibody Affinity Maturation
The host's naive B cells make antibodies of the immunoglobulin M class (IgM). An antibody is a secreted form of a receptor that occurs on the surfaces of B cells. Each B cell clone makes IgM with different binding characteristics—that is, the variable binding regions of the IgMs differ. The host has a large repertoire of naive B cells that produce a diverse array of IgM specificities.
An antigen on first exposure to a host will often bind rather weakly to several of the naive IgM. Those B cell clones with relatively high-affinity IgM for the antigen divide rapidly and come to dominate the antibody response to the antigen.
The dividing B cell clones undergo affinity maturation for particular epitopes. During this process, elevated mutation rates occur in the DNA that encodes the antibody binding region. This hypermutation in dividing B cell lineages creates a diversity of binding affinities. Those B cells with relatively higher binding affinities are stimulated to divide more rapidly than B cells with lower affinities. This process of mutation and selection creates high-affinity antibodies for the antigen.
The B cells that win the competition and produce affinity matured antibodies switch from producing IgM to immunoglobulin G (IgG). This class switch occurs by a change in the nonvariable region of the antibody that is distinct from the variable binding region.
Wedemayer et al. (1997) studied the changes in the variable antibody binding region during affinity maturation to a particular epitope. The matured antibody had an affinity for the epitope 30,000 times higher than the original, naive antibody. This increased affinity resulted from nine amino acid substitutions during affinity maturation. Wedemayer et al. (1997) found that the naive antibody significantly changed its shape during binding with the epitope. By contrast, the mature antibody had a well-defined binding region that provided a lock-and-key fit to the epitope. Wedemayer et al. (1997) speculated that naive antibodies may have more flexible binding regions in order to recognize a wide diversity of antigens, whereas matured antibodies may develop relatively rigid and highly specific binding sites.
Wedemayer et al.'s (1997) study suggests that naive IgM may bind a broader array of antigens at lower affinity, whereas matured IgG may bind a narrower array of antigens at higher affinity. Most analyses of epitope binding focus on IgG antibodies that have been refined by affinity maturation. Recently, attention has turned to the binding characteristics and different types within the IgM class, including the natural antibodies.
4.4. Natural Antibodies—Low-Affinity Binding to Diverse Antigens
Some purified antibodies bind to a wide array of self- and nonself-antigens. These polyreactive antibodies are sometimes referred to as natural or background antibodies because they occur at low abundance independently of antigen stimulation (Avrameas 1991). Natural antibodies are typically of the IgM class and have few mutations relative to the germline genotype, suggesting that natural antibodies usually have not gone through hypermutation and affinity maturation to particular antigens (Harindranath et al. 1993).
Chen et al. (1998) sampled the antibody repertoire of adult and newborn humans. They tested B cells for ability to bind insulin and β-galactosidase. Among adults, 21% of B cells bound insulin, 28% bound β-galactosidase, and 11% bound both antigens. Among newborns, 49% bound insulin, 54% bound β-galactosidase, and 33% bound both antigens. They concluded that low-affinity background reactivity commonly occurs in antibodies. Not surprisingly, newborns have a higher percentage of polyreactive antibodies than adults because adults have been exposed to many challenges and have a higher percentage of specific IgG antibodies.
Llorente et al. (1999) studied the natural antibody repertoire against the gp120 antigen of HIV-1. Among uninfected human blood donors, gp120 bound 2–5% of peripheral B cells. Llorente et al. (1999) analyzed in more detail the full repertoire of a single uninfected donor. None of the IgG isolates bound gp120, whereas 86% of the IgM clones bound the HIV-1 antigen. The IgM binding affinities were low, about an order of magnitude lower than a specific IgG antibody for gp120 that has been through the affinity maturation process. The low-affinity IgM antibodies did not inhibit in vitro infection by HIV-1. The authors suggested that these polyreactive antibodies do not provide protection against infection in vivo.
Ochsenbein et al. (1999) tested the role of natural antibodies in immunity against infection. They compared the ability of antibody-free and antibody-competent mice to resist infection against various viruses and the bacterium Listerium monocytogenes. In early infection kinetics, the pathogens were detected in concentrations one to two orders of magnitude lower in antibody-competent mice. Natural IgM but not IgG were found against most of the pathogens tested. By contrast with Llorente et al.'s (1999) conclusions, Ochsenbein et al. (1999) suggest that natural antibodies can help to contain infection during the early stages of invasion.
4.5. Affinity versus Specificity
The consequences of antigenic variation depend on how the host's immune system recognizes and reacts to variants. For example, if host immunity reacts in the same way to two parasite genotypes, then the host immune response does not exert differential effects of natural selection on those variants.
The ability of host immunity to discriminate between antigenic variants can be measured in different ways. For the sake of discussion, I focus on antibody-antigen binding. The same issues apply to any binding reaction.
An antibody's equilibrium affinity for different antigens can be compared by the relative inhibition tests described above in section 4.1, Antigens and Antibody Epitopes. Measures of relative inhibition can be easily translated into the free-energy difference in binding between an antibody and two different antigens (Benjamin and Perdue 1996).
Dynamic rather than equilibrium aspects of affinity drive certain processes in host immunity. For example, B cells compete for antigen to stimulate clonal expansion and enhanced expression of the associated antibodies. Several authors have argued that different processes influence the selection and maturation of antibodies during different phases of the immune response (reviewed by Lavoie et al. 1999; Rao 1999).
Rao (1999) summarizes the argument as follows. The early stimulation of B cells in response to initial exposure to an antigen depends on relative equilibrium binding affinities of the B cell receptors and associated antibodies. Those B cells that receive a threshold level of stimulation increase secretion of antibodies. Typically, a variety of B cells receive threshold stimulation. Thus, the early immune response tends to produce diverse antibodies that recognize various epitopes.
By contrast, dynamic association rates of reaction rather than equilibrium binding constants may determine the next phase of antibody response. Rao's (1999) lab compared antibodies that had developed in response to two related antigens. These antibodies were isolated from the later stages of the immune response and had therefore been through affinity maturation. They found no detectable difference in the equilibrium binding affinities of an antibody to the antigen to which it was raised versus the other antigenic variant. By contrast, the on-rates of antigen binding did differ.
Apparently, those B cell receptors with higher rates of antigen acquisition outcompete B cell receptors with lower rates of acquisition. This makes sense because affinity maturation occurs when the B cell clones are highly prone to apoptosis (suicide) unless they receive positive stimulation. Thus, the selection process during affinity maturation tends to optimize antigen acquisition rates rather than equilibrium binding constants.
Other studies have also analyzed the maturation of antibody binding properties during the course of an immune response (reviewed by Lavoie et al. 1999). Those studies also found differences in how the affinity constants and rates of association and dissociation changed over time. The appropriate type of affinity and measure of immune recognition depend on the dynamic processes of the immune response. I take this up in more detail in chapter 6.
Specificity defines another dimension of immune recognition. Specificity is the degree to which an immune response discriminates between antigenic variants. A simple approach measures the relative binding affinities of purified antibodies or T cell receptors for different antigens. Discrimination depends on the range of parasite variants bound, on the binding affinity, and on the stringency of the conditions under which one conducts the assay.
Figure 4.1 shows that relatively low-affinity binding can often provide greater specificity when measured at intermediate stringency. This occurs because low-affinity receptors bind fewer kinds of antigens as conditions limit the assay's sensitivity for low-affinity binding. Thus, the relative specificity of different antibodies or T cells depends on both affinity and conditions of measurement.
Figure 4.1
Affinity versus specificity of host immunity. The two triangles show the range of antigens bound by a particular antibody or T cell. A narrow range implies high discrimination between parasite antigens and high specificity. Detection of binding depends (more...)
4.6. Cross-Reaction of Polyclonal Antibodies to Divergent Antigens
I have discussed the cross-reactivity of a particular antibody to antigenic variants and the extent to which a particular antigenic variant reacts with different antibodies. These issues focus on the affinity and specificity of particular binding reactions when one perturbs either the antibody or the antigen. For example, affinity decreases in a highly nonlinear way with amino acid substitutions in either the antibody or the antigen. Substitutions in just a few key amino acids can reduce the equilibrium binding constants by several orders of magnitude.
Those studies of affinity and specificity were typically conducted with purified (monoclonal) antibodies of a single type. By contrast, the immune response to an antigen often raises many different antibodies to the various exposed epitopes on the antigen. The initial polyclonal response may narrow over time as the various B cell clones receive positive or negative signals for expansion and the development of memory. I consider those dynamics in a later section. Here, I am concerned with the nature of cross-reactivity of the polyclonal immune response to a whole antigen as compared with the cross-reactivity of a monoclonal antibody to the antigen.
Benjamin et al. (1984) emphasized that the strength of cross-reactivity between antigens decreases linearly with the number of amino acid substitutions when measured by the polyclonal sera of the full immune response (see update in Prager 1993). The linear relationship explains 80% or more of the total variation in cross-reactivity between phylogenetically diverged variants of proteins such as myoglobin. Other major studies have focused on lysozyme c and cytochrome c.
The linear relationship between polyclonal cross-reactivity and amino acid substitutions arises because the surface of a protein antigen appears to present a nearly continuous and overlapping set of epitopes. Each exposed amino acid probably contributes only a small amount to the total binding between all antibodies and all epitopes.
4.7. T Cell Epitopes
Antibodies bind directly to free pathogens in the blood, lymph, or mucosal surfaces. Antibodies cannot get at intracellular pathogens. To fight intracellular infections, the host uses cell-mediated defenses spearheaded by the cytotoxic T lymphocytes (CTL), also referred to as CD8+ T cells. The entire pathway leading to a CTL response against a pathogen has several components that act as regulatory checks and balances. But the bottom line often amounts to CTLs killing host cells that contain specific epitopes of intracellular pathogens.
I start with a brief outline of specific recognition and then expand on the key issues. First, host cells cut up the proteins of an intracellular pathogen. Second, transporter of antigen presentation (TAP) moves peptides from the cytosol to the endoplasmic reticulum. Third, some of these cut-up peptides bind to the host cell's major histocompatibility complex (MHC) class I molecules. Fourth, peptide-MHC complexes move to the cell surface, exposing the bound peptide to the outside. Fifth, the T cell receptors (TCR) on free T cells can bind to certain peptide-MHC combinations. Sixth, if the TCRs of some CTLs bind pathogen peptide-MHC complexes on the surface of a cell, and some supporting signals prevail, then the CTLs kill the host cell.
Several variations on T cell immunity occur. But this scenario captures the essential features of specific recognition: cutting pathogen peptides, transporting peptides to the endoplasmic reticulum, binding of pathogen peptides to class I MHC, presentation on cell surfaces, and binding of specific TCRs to the peptide-MHC complex (Deng et al. 1997; Davis et al. 1998; Germain and Štefanová 1999).
The last part of this section addresses the relation between specificity and cross-reactivity for the TCR. On the one hand, each TCR probably cross-reacts widely with different epitopes. On the other hand, T cell responses appear to be highly specific—variant epitopes often avoid the T cell response generated against the initial challenge. The resolution may follow from numerical considerations. Each TCR reacts with many different epitopes, perhaps as many as 105 different peptides. However, if a TCR reacts with one epitope, then the probability that it reacts with another, randomly chosen epitope may be as low as 10−4 because of the large number of possible epitopes. In terms of the total number of epitopes bound, TCRs appear widely cross-reactive, but in terms of the frequency of epitopes bound, TCRs appear highly specific.
Intracellular Peptide Processing and Transport
MHC class I molecules typically bind peptides with 8–10 amino acids. Any short peptide within any pathogen protein is a candidate for MHC binding. Peptides have polarity, with carboxyl (C-terminal) and amino (N-terminal) ends. Many processes cut proteins into shorter peptides, but the proteasomes seem to be particularly important for generating peptides of the right length for MHC presentation (Rock and Goldberg 1999; York et al. 1999). Proteasome digestion creates a nonrandom population of peptides relative to the potential set defined by the amino acid sequence of whole proteins. Digestion appears to be particularly specific for the C-terminal cut, less so for the N-terminal cut (Niedermann et al. 1999).
The MHC class I molecules have biases for certain amino acids at the C-terminal site of peptides (Rammensee et al. 1995). Niedermann et al. (1999) have shown a correlation between the preferred C-terminal cuts of proteasome digestion and the favored C-terminal amino acids bound by MHC class I molecules. Thus, the proteasomes appear to be preferentially generating peptides that can be presented by MHC.
A peptide that binds MHC with relatively high affinity may not be generated in sufficient quantity to be a dominant epitope for immune recognition. In vitro studies of proteasome digestion provide the easiest way to quantify peptide generation. Although in vivo results may differ, the preliminary data from in vitro studies provide interesting hints. For example, mice were immunized with chicken ovalbumin, and the CTL response was studied by in vitro reactions with peptides presented by the class I MHC molecule Kb. The peptide Ova257SIINFEKL264 dominated the CTL response (Chen et al. 1994; Niedermann et al. 1995), where the capital letters define the peptide sequence based on the single-letter amino acid code and the subscripts give the location of the peptide within the primary sequence of the protein. A secondary, weaker response to Ova55KVVRFDKL62 can be generated under some conditions.
Dick et al. (1994) found that proteasome digestion of full-length ovalbumin proteins yielded Ova257SIINFEKL264 but not Ova55KVVRFDKL62. Niedermann et al. (1995, 1996) studied in detail the proteasome digestion patterns of synthetic peptides with 22 or 44 amino acids containing the two potential epitopes. They found major cleavage sites at the two ends of the dominant Ova257SIINFEKL264 epitope and relatively little cleavage within the epitope. By contrast, a dominant cleavage site occurred between amino acids 58 and 59 of Ova55KVVRFDKL62, yielding only traces of the intact epitope. Several other studies reviewed by Niedermann et al. (1999) support the hypothesis that proteasomal digestion frequently reduces the number of copies of potential epitopes sufficiently to prevent a strong CTL response.
It may eventually be possible to predict the probabilities of proteasomal cleavage sites (Niedermann et al. 1999). However, many factors influence the concentration of different peptides available for MHC binding. For example, sequences flanking antigenic peptides influence cleavage (Yewdell and Bennink 1999). Interferon-γ changes the distribution of proteases affecting antigenic peptide production, perhaps enhancing peptides with MHC binding motifs (York et al. 1999). TAP transports different peptides at different rates from the cytosol to the endoplasmic reticulum, where MHC class I binding of peptides occurs (Yewdell and Bennink 1999).
I have focused on the MHC class I molecules, which are present in many cell types. By contrast, MHC class II molecules operate in specialized antigen-presenting cells. Those cells take in antigen from the outside environment, process the proteins into peptides, bind peptides to MHC class II molecules, and present the peptide-MHC complexes on the cell surface. These peptide–class II complexes bind to a subset of CD4+ T cells with specific, matching T cell receptors. The CD4+ cells are often called helper T cells because they provide stimulation to CTLs or to B cells and antibody production.
Different cellular locations and proteases occur in the MHC class I and class II pathways. Nakagawa and Rudensky (1999) and Villadangos et al. (1999) review the proteases involved in class II peptide processing. Kropshofer et al. (1999) review transport and loading of peptides onto class II molecules.
Intracellular Production and Exogenous Uptake of Antigens
The timing and quantity of production for different antigens in infected cells has received relatively little attention (Schubert et al. 2000), but certainly affects the influx of peptides available for MHC binding. In addition, exogenous antigens may be taken up by antigen-presenting cells and carried to lymphoid tissue for presentation to T cells (Schumacher 1999; Sigal et al. 1999). Intracellular production and exogenous uptake of antigens most likely influence the distribution of epitopes presented to T cells.
Peptide-MHC Binding
The class I MHC molecules bind peptides of 8–10 amino acids. For peptides with 9 amino acids (nonamers), the 20 different amino acids that can occur at each site combine to make 209 = 512 × 109 different peptide sequences. The human genome has three loci with class I molecules that present to CTLs. These loci are highly polymorphic; thus, each diploid individual typically carries six different class I alleles for CTL presentation. Clearly, if the molecules encoded by these six alleles are to bind and present a reasonable fraction of parasite peptides, then each molecule must bind to a large diversity of peptides.
Class I binding is indeed highly degenerate with regard to peptide sequence. Yewdell and Bennink (1999) estimate that each molecule binds approximately 1/200 of the possible peptide sequences, or on the order of roughly 107 different nonamers. An individual with six different alleles binds approximately 6/200 = 3% of candidate peptides. Here, binding means with sufficient affinity to stimulate a CTL response.
The specificity of MHC binding influences which parasite epitopes dominate an immune response. Current understanding of MHC binding is rather crude. Prediction of which parasite sequences would bind strongly to MHC molecules might help in vaccine design and in understanding the different patterns in immune response between different individuals. Given this widespread interest, the field is moving rapidly.
The three human class I loci that present to CTLs have 614 currently known alleles (http://www.anthonynolan.com/HIG/index.html). Many of the MHC molecules have been characterized by a specific binding motif—the amino acid sequence pattern to which they typically bind (Marsh et al. 2000).
Buus (1999) reviews the different methods to estimate binding motif and alternative techniques for prediction of binding. Details vary for the different alleles, but often an MHC class I molecule has two anchor positions near the ends of the peptide among the 9 or so amino acids of the peptide. Each anchor position has a favored amino acid or sometimes a limited set of alternatives. However, prediction based on anchor positions is only moderately successful; about 30% of peptides carrying the predicted motif actually bind, and sequences lacking anchor residues can bind. More complex statistical approaches have improved prediction above 70%.
Class II molecules also bind a region of about 10 amino acids. However, by contrast to class I molecules, the class II molecules have open-ended binding grooves, allowing class II molecules to bind peptides of varying lengths in which differing numbers of amino acids hang out of each end of the groove (Marsh et al. 2000). These varying peptide lengths have made it difficult to establish binding motifs; thus relatively less is known about class II binding.
Class II molecules appear to be less specific (more degenerate) in their binding compared with class I molecules (Marsh et al. 2000). A few detailed studies of class II binding have been developed (e.g., Latek and Unanue 1999). It may be that class II's relatively less specific binding has to do with its role in stimulating helper T cells that regulate the immune response rather than in directly killing parasites, but there is little evidence for this at present.
The class I and class II molecules bind only to peptides. Recent work has shown that the CD1 MHC system presents lipids and glycolipids to T cells, providing an opportunity for T cell response to nonprotein antigens (Porcelli and Modlin 1999; Prigozy et al. 2001). No doubt this system plays some role in immunity, but its relative importance is not clear at present.
T Cell Receptor Binding to Peptide-MHC Complexes
The immune system can generate highly specific memory responses against particular antigens. For example, first infection by a measles virus typically leads to symptomatic infection and eventual clearance. Second infection rapidly induces specific antibody and T cell responses based on a pool of memory cells from prior infection. This type of observed memory response naturally led to the belief that TCR recognition is highly specific for particular epitopes. However, recent work demonstrated that the TCR binds in a highly degenerate way, each TCR binding on the order of 104–107 different epitopes (Mason 1999). In addition, limited data suggest that a single peptide stimulates several different T cell clones (Maryanski et al. 1997; Mason 1999). However, different studies and different methods have given variable estimates for the number of clones stimulated by a single peptide (Yewdell and Bennink 1999).
How can the TCR binding be so degenerate, yet the immune response be so specific? Most of the details are not understood at present, but some reasonable hypotheses are beginning to take shape.
I first review relevant aspects of T cell binding. The TCR on CTLs (CD8+ T cells) binds to peptide-MHC class I complexes presented on the surface of most cell types. The TCR on the helper (CD4+) T cells binds to peptide-MHC class II complexes presented on the surface of specialized antigen-presenting cells. Binding and signal strength are quantitative factors, but again I use binding qualitatively to mean sufficient signal strength to stimulate a T cell response against an epitope.
Now consider some magnitudes with regard to the problem of recognition (Mason 1999). Define T as the number of T cell clones with different TCRs in the naive T cell repertoire, T(p) as the number of T cell clones that respond to the same peptide, N as the number of possible peptides to be recognized, and P(t) as the number of different peptide-MHC complexes recognized by a particular TCR. Thus,
because both sides describe the total number of recognition specificities over all T cell clones. Immune response probably depends more on the frequencies of T cell clones that respond to particular peptides, F(p) = T(p)/T, rather than total numbers of clones in the body, so it is also useful to write
which shows that the frequency of T cell clones responding to a particular peptide equals the frequency of presented peptides stimulating a particular T cell clone. If we assume that, in the naive T cell repertoire, each clone with a unique TCR has about the same number of cells, then F(p) is also the frequency of individual T cells that respond to a particular peptide. Equation (4.2) can be rewritten to emphasize cross-reactivity of the TCR as
which is the probability that if a T cell receptor binds one epitope, then it also binds to a second, randomly chosen epitope.
The number of possible specificities, N, for peptides with n amino acids, is 20n×S, where S is the fraction of peptides that can be presented by MHC alleles. Considering nonamers with n = 9, and setting S ≈ 10−2 as discussed above for MHC binding, we have N ≈ 209 ×10−2 = 5 ×109.
The lower bound for F(p), the frequency of T cell clones that respond to a peptide, occurs when every T cell is unique and each peptide stimulates only a single T cell. Mice have about 108 T cells, so the minimum value of F(p) is 10−8, and thus, from equation (4.3), the minimum value for the number of peptides bound by a TCR is P(t) ≈ 50. This is certainly a gross underestimate, because each TCR clone has more than one cell on average, and each peptide likely stimulates more than one clone. Nonetheless, this extreme case shows that the magnitude of the recognition problem demands some degeneracy in TCR binding in mice.
Mason (1999) suggests that a more realistic description follows if one accepts the experimental estimate by Butz and Bevan (1998) that, in the naive repertoire, three different viral epitopes each stimulated a frequency F(p) ≈ 10−4 of mouse CTL cells. In a mouse with 107–108 naive CTLs, this gives an estimate of 103–104 CTLs potentially responding to each epitope. Using F(p) = 10−4 in equation (4.3) gives the number of peptides bound by a single TCR as P(t) ≈ 5 × 105, a value that is in line with other estimates obtained by various methods (Mason 1999).
The estimated frequency F(p) ≈ 10−4 based on Butz and Bevan (1998) refers to the frequency of individual T cells responding to a peptide. It was not clear from that study how many different T cell clones were involved. It is challenging to estimate the number of different clones from the naive repertoire that respond to a particular peptide, although recent technical breakthroughs may soon provide more data (Yewdell and Bennink 1999). Among the better studies available, Maryanski et al. (1996) estimated that an epitope from the human class I molecule HLA-CW3 stimulated fifteen to thirty different T cell clones in a mouse. The response in this case may have been limited because the human MHC molecule is similar to mouse MHC molecules, causing the tested peptide to be seen as similar to a self-peptide of the mouse. In another study by the same research group, the clonal diversity of CTLs responding to a Plasmodium berghei peptide was much higher than against HLA-CW3, but the methods did not permit a comparable estimate for the number of clones (Jaulin et al. 1992).
Humans have about 1011 T cells compared with about 108 T cells in mice. If the frequency, F(p), of T cells responding to a peptide is about the same in humans as in mice, then, from equation (4.3), the cross-reactivity of each TCR receptor, P(t), is about the same in humans as in mice. The value estimated above is each TCR binding P(t) ≈ 5 × 105 different peptides.
How can such high cross-reactivity be reconciled with observed specificity? First, the probability of any particular T cell cross-reacting with two different epitopes remains low. If a T cell reacts with one epitope, the probability that it reacts with a second, randomly chosen epitope is P(t)/N ≈ 10−4.
Second, the observed specificity has to do with the number of different T cell clones that actually expand in response to an epitope. The number of expanding clones is certainly lower than the potential set of clones that bind sufficiently strongly to stimulate a response (Yewdell and Bennink 1999). Competition between clones for the epitope and for other stimulatory signals limits clonal expansion. I return to this topic in chapter 6. Here I simply note that the broad and highly cross-reactive repertoire of the naive T cells may be important for fighting primary exposure, much as the natural antibodies provide background protection against first infection. The secondary or memory response may be much narrower because it is limited to those binding clones that received additional stimulatory signals during primary infection.
Summary of T Cell Epitopes
Yewdell and Bennink (1999) calculate an overall probability of 1/2,000 for a peptide of a foreign antigen to stimulate a dominant CTL response. By their calculation only ~1/5 of potential epitopes survive proteolytic digestion and transport to the endoplasmic reticulum for loading onto MHC molecules; of these, only ~1/200 bind MHC molecules above the threshold affinity required for immunogenicity; finally, limitations in the TCR repertoire for binding peptide-MHC complexes cause ~1/2 of presented peptides to stimulate a response. This is only a very rough approximation based on the limited data available.
MHC presentation and TCR binding are just the first steps in a T cell response. Typically, several TCRs may receive sufficient stimulation, but only a subset continue to develop strong clonal expansion. I discuss the factors that influence which clones do and do not expand in chapter 6.
4.8. Every Host Differs
The epitopes that stimulate an immune response depend on an interaction between the host and parasite. Different hosts vary in several attributes of immune recognition; thus the dominant epitopes will change from one host to the next even for an unvarying parasite.
The MHC class I and II molecules are the most strikingly polymorphic of all human loci. The three main class I loci for presenting peptides, designated A, B, and C, currently have 175, 349, and 90 alleles described, respectively. The class II molecules have separate designations for individual components of each molecule. The highly polymorphic components tend to be in the β1 chains that contact bound peptides (Marsh et al. 2000). The β1 chains for the DR, DQ, and DP class II molecules currently have 246, 44, and 86 alleles described, respectively. The IMGT/HLA on-line database lists recent allelic counts, as described in Robinson et al. (2000).
The MHC molecules shape the TCR repertoire. As T cells mature in the thymus, they bind to MHC molecules presenting self-antigens. Those TCRs that bind too strongly cause the associated T cells to die. Those TCRs that bind too weakly fail to provide sufficiently strong reinforcing signals, again causing the associated T cells to die. Fewer than 1% of T cells pass these checks to survive (Marsh et al. 2000). Thus, the naive TCR repertoire is strongly influenced by the particular MHC alleles of each individual. The individual naive repertoires lead to different TCR clones being stimulated in different individuals when challenged by the same epitope (Maryanski et al. 1996, 1999). Because helper T cells influence antibody response and other aspects of immune regulation, the variable TCR repertoire may have additional consequences beyond CTL variability.
Proteolysis of antigens and transport of peptides determine the peptides available for MHC binding. Strong challenge by a particular parasite could lead to selection favoring or disfavoring specific patterns of proteolysis. However, I am not aware of any evidence for proteasome polymorphism. The peptide transporter, TAP, is polymorphic: the two subunits TAP1 and TAP2 have six and four sequences listed in the IMGT/HLA database (Robinson et al. 2000). So far, no functional differences among alleles have been found (Marsh et al. 2000).
Somatic mutation and recombination between various germline loci generate the DNA that encodes the TCR (Janeway et al. 1999). These generative mechanisms allow each individual to produce a huge variety of TCR binding specificities. The intensity of direct selection on germline polymorphisms may be rather weak because the somatic mechanisms of variation and selection shield the germline from the selective processes imposed by diverse antigens. However, the germline alleles do set the initial conditions on which somatic processes build, so it is certainly possible that germline polymorphisms influence individual tendencies to react to particular antigens.
The limited data available show some germline polymorphisms for the TCR (e.g., Reyburn et al. 1993; Hauser 1995; Moffatt et al. 1997; Moody et al. 1998; Sim et al. 1998; see also http://imgt.cines.fr). One interesting study found an interaction between a human germline polymorphism in a subunit of the TCR (VA8.1) and an MHC class II polymorphism (HLA-DRB1) (Moffatt et al. 1997). The authors analyzed two variants of the VA8.1 allele and the six most common HLA-DRB1 alleles. Individuals with enhanced allergic response to a dust mite antigen tended to have one of the two VA8.1 variants combined with the HLA-DRB1*1501 allele.
Moffatt et al. (1997) measured allergic response by the titer of IgE antibodies, which are known to be directly involved in stimulating allergic symptoms (Janeway et al. 1999). Most likely, the TCR and MHC class II polymorphisms influence IgE via helper T cells, because TCR binding to antigens presented by MHC class II stimulate helper T cells, and such T cell response is typically required for antibody stimulation. Thus, specific recognition by the TCR and MHC can affect specific recognition by antibodies.
The B cell receptor (BCR) is generated by a process of somatic recombination and mutation similar to the process that generates TCRs. Antibodies are secreted forms of the BCR. I did not find any comparable reports of germline variation in the alleles that make up the components of the BCR. Hauser (1995) suggested that somatic hypermutation (affinity maturation) of the BCR protects the germline from direct selection. The TCR has limited somatic mutation after the initial genetic recombinations, perhaps exposing germline TCRs to more intense selective pressures than BCRs. However, the difference may simply reflect less study of the BCR germline genes.
Finally, a complex network of quantitative or threshold signals regulates many aspects of the immune response. Quantitative aspects of immune regulation probably also vary among individuals. Those variable regulatory controls may influence different hosts' specific responses to antigens, because response often depends on a cascade of quantitative signals triggered by an antigen. Thus, it seems likely that variable patterns of specific recognition between individuals will be influenced by quantitative variability in regulatory control.
4.9. Problems for Future Research
1. Recognition by naive versus affinity-matured antibodies
Wedemayer et al.'s (1997) study suggests that naive IgM and affinity-matured IgA/IgG antibodies differ in the nature of the antibody-epitope bond. IgM antibodies may have rather flexible binding regions that significantly change shape when attaching to an epitope. By contrast, the affinity-matured antibodies seem to have relatively rigid, well-defined binding regions that provide a highly specific lock-and-key fit to the epitope. I suspect that an epitope must change more drastically to escape from flexible IgM binding than from the more rigid IgA/IgG binding. Thus, naive and affinity-matured antibodies may impose different selective pressures on the molecular changes between antigenic variants.
Affinity-matured antibodies in a memory response probably play the dominant role in blocking repeat infection by most pathogens (Janeway et al. 1999; Plotkin and Orenstein 1999). But some parasites may be more strongly limited by naive antibodies. For example, IgM antibodies seem to play the dominant role in fighting a sequence of different antigenic variants of the spirochete Borrelia hermsii (Barbour and Bundoc 2001). This parasite switches its antigenic surface molecules during a single infection. The parasite achieves this by occasionally copying into a single expression site different genes for antigenic variants stored within the genome. In this case, the major selective pressure differentiating antigenic variants most likely concerns cross-reaction with previously expressed variants during the same infection cycle.
In a parasite that rapidly switches antigenic variants, a particular epitope may need relatively more changes in amino acid composition to escape cross-reaction with naive IgM antibodies. By contrast, other pathogens may require relatively fewer amino acid changes to escape affinity-matured IgA/IgG antibodies.
It would be valuable to have more data on the degree to which IgM or affinity-matured antibodies dominate against different parasites. The IgM response may dominate only against parasites that present to the host a rapid sequence of antigenic variants. In those cases, escape from IgM rather than affinity-matured antibodies may determine the molecular changes needed for antigenic variants to escape cross-reactivity.
2. Binding stringency affects the affinity-specificity relationship
Figure 4.1 shows how specificity and cross-reactivity change with interactions between binding affinity and binding stringency. The degree of cross-reactivity determines how much molecular change parasites require to escape immune pressure directed against related antigens. Thus, the combined effects of binding affinity and stringency influence cross-reactivity, which in turn shapes molecular aspects of antigenic variation.
Controlled experiments in vitro could apply monoclonal antibodies with different affinities to cultured parasites under different binding stringencies. The type of molecular change required to escape immune pressure should vary in response to interactions between stringency and affinity.
3. Equilibrium versus kinetic aspects of affinity
Different stages of the immune response probably depend on different aspects of binding affinity to antigens. For example, clearance of antigens by antibodies may depend on the equilibrium affinity of antibody-epitope bonds, whereas the relative stimulation of different B cell lineages may depend on kinetic rates of association with antigens. Different kinetic consequences probably follow from different molecular attributes of binding between immune effectors and antigens. Complete understanding of antigenic variation requires one to trace the chain: types of molecular variation → aspects of binding kinetics → control of the immune response.
