NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Frank SA. Immunology and Evolution of Infectious Disease. Princeton (NJ): Princeton University Press; 2002.

Cover of Immunology and Evolution of Infectious Disease

Immunology and Evolution of Infectious Disease.

Show details

Chapter 11Classifications by Antigenicity and Phylogeny

In this chapter, I compare immunological and phylogenetic classifications of antigenic variation. Contrasts between these classifications provide insight into how natural selection shapes observed patterns of diversity. Following chapters take up other methods to infer processes of selection.

The first section describes immunological measures of antigenicity. These measures summarize the ability of specific antibodies to recognize different antigenic variants. The reactivities for various antibodies tested against different antigenic isolates form a matrix of antigenic or immunological distances between parasite variants. These distances can be used to classify antigenic variants into related clusters.

The second section notes that antigenic variants can also be classified by phylogeny. This classification scheme measures relatedness between variants by distance back in time to a common ancestor. Such distances arise from the patterns of nucleotide or amino acid differences in genomic sequences.

The third section defines possible relations between antigenic and phylogenetic classifications. Concordance commonly occurs because antigenic distance often increases with time since a common ancestor, reflecting the natural tendency for similarity by common descent. A particular pattern of discord between antigenic and phylogenetic classifications suggests hypotheses about evolutionary process. Suppose, for example, that phylogenetically divergent parasites are antigenically close at certain epitopes. This suggests as a hypothesis that selective pressure by antibodies has favored recurrent evolution of a particular antigenic variant.

The fourth section presents flaviviruses as an example of concordant antigenic and phylogenetic classifications. This example compares strains that differ by relatively long phylogenetic distances with antigenicity measured by averaging reactivity over many different epitopes. Aggregate antigenicity tends to diverge steadily over time as amino acid differences accumulate (Benjamin et al. 1984). Particular details of natural selection with regard to each amino acid substitution disappear in the averaging over many independent events.

The fifth section shows a mixture of discordance and concordance between antigenic and phylogenetic classifications for influenza A. The classifications cover a history of transfers between different host species. Antigenicity and phylogeny both separate isolates from pigs into two groups, the classical swine types and avian-like swine types more recently transferred from birds to pigs. Two bird isolates group phylogenetically with the avian-like swine types, as expected. However, antigenic measures separate the bird isolates as distinct from the relatively similar classical swine and avian-like swine groups. Perhaps host adaptation influences antigenicity of some epitopes used in this study.

The sixth section suggests that immunological pressure by antibodies drives the short-term phylogenetic divergence of influenza A. If so, then antigenic classifications over the same scale of diversity may match the phylogenetic pattern. Concordance probably depends on the percentage of amino acid substitutions explained by antibody pressure.

The seventh section considers explanations for the discordant patterns of phylogeny and antigenicity reported for HIV. Shared antigenicity over long phylogenetic distances may arise by stabilizing or convergent selection. Stabilizing selection prevents change in particular amino acids because of their essential contribution to viral fitness. Convergent selection causes recurrent evolution of the same antigenic type by repeatedly favoring that type in different times and places. Alternatively, divergent antigenicity over short phylogenetic distances can arise from intense immune pressure. Stabilizing, convergent, and diversifying selection can all occur over different temporal scales, combining to shape the relations between antigenicity and phylogeny.

The final section lists problems for future research.

11.1. Immunological Measures of Antigenicity

Immunological methods measure the reactions of antigens with antibodies or T cells. A particular test can be described by a matrix. One dimension consists of standardized immunological components such as different antibodies; the second dimension lists alternative parasite isolates or molecules to be tested for their antigenic properties. Each matrix entry contains the strength of the immunological response—a measure of antigenicity.

Immunology differentiates antigens only to the extent that the test antigens react differently with the panel of immunological agents. Thus, the measures depend on the immunological panel used for discrimination. This reiterates a key point of chapter 4, that specificity and diversity describe interactions between the host and parasite. The antigenic diversity of a parasite has meaning only in the context of the specificity of host recognition.

Different kinds of immunological panels have been used. Monoclonal antibodies (MAbs) provide a high titer of identical antibodies. Each clone responds to variation in a small region of a parasite molecule. Different clones produce antibodies with different specificities. A panel of MAbs provides a highly specific and repeatable set of determinants.

Polyclonal antibody serum contains the diverse antibody specificities raised by a host against a particular challenge. The host may be challenged with a peptide, with a whole molecule, or with an entire parasite. Polyclonal sera from different hosts form a panel that can be used to test novel antigens. The response of each polyclonal serum aggregates reactions against many antigenic epitopes. Thus, polyclonal measures tend to be broader measures of total differences between antigens when compared with monoclonal measures, but it is harder to know exactly what differences the polyclonal technique measures.

T cell immunity has generally not been used to form an immune panel for discrimination of antigenic diversity. T cell recognition depends on the processing of peptides, their binding to MHC molecules, and the affinity of T cell receptors for peptide-MHC complexes. Until recently, it has been difficult to control these steps in a repeatable and measurable way. A new method, tetramer binding (Altman et al. 1996; Doherty and Christensen 2000), may allow some progress in this area. This method first creates peptide-MHC complexes, then measures the percentage of a host's T cells that bind those peptide-MHC complexes. A test of parasite variability could take the following form: challenge a host with a particular parasite type, then compare the host's T cell response against peptide-MHC complexes with peptides derived from different antigenic variants.

Immunological tests can be conducted with intact parasites, whole molecules, or molecular fragments such as peptides. Each choice has its benefits and limitations. Peptides provide small, controllable, and easily studied variants. However, antibodies normally respond to exposed, three-dimensional conformations rather than naked, sometimes linearized peptides. Whole molecules maintain three-dimensional structure and provide a broader aggregate measure of variation over the entire molecule. The shape of the molecule may, however, be altered when combined into a whole parasite, and many parts of the surface of the naked molecule may be inaccessible when in the intact parasite. Whole parasites provide the most realistic aggregate measure of differentiation. Assays may be technically difficult with the whole parasite, and results from such assays do not focus on specific variant epitopes (Nyambi et al. 2000a).

A completed immunological test fills the matrix of reaction strengths for each immunological agent and parasite isolate. The matrix can be used to classify the parasite isolates into related groups according to the degree of similarity in their immunological reactivity. The classification provides a basis to type new isolates according to immunological properties. The matrix may also be used to identify the major determinants of antigenic differences, which can be helpful in the design of vaccines against antigenically variable parasites.

11.2. Phylogeny

The amount of change between two antigens relative to a common ancestor provides another way to classify antigenic variants. One typically reconstructs the phylogenetic relationships of evolutionary descent by analyzing the patterns of change in the nucleotide or amino acid sequences that encode antigenic molecules (Page and Holmes 1998; Rodrigo and Learn 2000).

Allelic variants of a gene can usually be arranged into a phylogenetic pattern of evolutionary descent—a gene tree. That phylogeny by itself simply describes the lineal history of antigenic variants without regard to the processes that shaped the pattern of descent. The phylogenetic history provides a necessary context for interpreting evolutionary process (Hughes 1999).

11.3. Hypothetical Relations between Immunology and Phylogeny

Suppose that we have constructed an immunological classification of four parasite isolates, P1, P2, P3, and P4. The four parasites group into two clusters, shown in figure 11.1.

Figure 11.1. Hypothetical relationship of four parasite isolates based on immunological reactions.

Figure 11.1

Hypothetical relationship of four parasite isolates based on immunological reactions. The clustering shows that the pair P1 and P3 reacts in a similar way to immunological agents, the pair P2 and P4 reacts in a similar way, and the two pairs differ in (more...)

Figure 11.2 shows a phylogenetic classification with the same groupings as the immunological pattern in figure 11.1. If a phylogenetic analysis provides the same classification, then immunological distance increases with phylogenetic distance. The parasites may, for example, accumulate genetic differences randomly throughout their genomes. Parasites that diverged from a more distant common ancestor have more genetic differences both inside and outside the tested antigenic regions, with no concentration of differences in the antigenic sites. Alternatively, natural selection on the antigenic sites may be driving apart the clusters. Then both antigenic and nonantigenic sites provide the same phylogenetic pattern, clustering P1/P3 versus P2/P4, but the differences between the clusters would likely be concentrated disproportionately in the antigenic sites.

Figure 11.2. Hypothetical phylogenetic relationship of four parasite isolates based on nucleotide or amino acid sequences.

Figure 11.2

Hypothetical phylogenetic relationship of four parasite isolates based on nucleotide or amino acid sequences. The same clustering occurs as in fig. 11.1. The white lineages have the antigenic properties of the P1/P3 immunological grouping, and the black (more...)

A correspondence generally occurs between phylogenetic distance and the differences measured on particular characters, reflecting the natural tendency for similarity by common descent. Sometimes a particular force disrupts this natural concordance.

Figure 11.3 shows a discordant pattern between phylogenetic and immunological distance. In this case, broad similarity over the nucleotide or amino acid sequence phylogenetically groups P1 with P2 and P3 with P4. The immunological test, which focuses on only a narrow subset of the total sequence, highlights antigenic divergence within closely related phylogenetic pairs. The pattern shows recurrent evolution of an antigenic type.

Figure 11.3. Alternative phylogenetic pattern that clusters P1 with P2 and P3 with P4.

Figure 11.3

Alternative phylogenetic pattern that clusters P1 with P2 and P3 with P4. The white lineages share the P1/P3 immunological grouping and the black lineages share the P2/P4 immunological grouping shown in fig. 11.1. The gray lineages show that the immunological (more...)

Many processes can generate the discordant pattern of figure 11.3. Suppose, for example, that only two variants can occur at a particular epitope because of conformational constraints on the function of the parasite molecule. If an epidemic begins with a parasite in state one, then host immunity will eventually favor the spread of state two. Conversely, an initial epidemic beginning with state two leads eventually to replacement by state one. Pairs of closely related lineages will often be of opposite state.

Functional hypotheses can often be tested by comparison of the predicted and observed phylogenetic patterns. For example, the functional constraint that an epitope can exist only in two alternative, antigenically distinct states predicts a discordant pattern between phylogenetic and immunological classifications. Alternatively, an observed discordance between phylogenetic and immunological classifications may lead to a functional or process-oriented hypothesis. That hypothesis can be tested by using other methods to infer function or process—for example, whether an observed epitope is indeed constrained to two alternative states by structural and functional attributes.

11.4. Immunology Matches Phylogeny over Long Genetic Distances

The flaviviruses illustrate broad correspondence between immunological and phylogenetic distances. This group includes well-known pathogens such as yellow fever, dengue fever, and West Nile virus. Kuno et al. (1998) built a phylogeny from the nucleotide sequences of seventy-two viral strains. These viruses span a diverse group, with nucleotide sequence identities of 69% or higher within the fourteen phylogenetic clades and lower percentages of identities between clades.

The flavivirus clades identified by molecular phylogeny correspond closely to the antigenic classification by Calisher et al. (1989) based on reactions to polyclonal antisera, although the two methods do disagree on the classification of a few strains. Two factors probably contribute to the close match between antigenic classification and molecular phylogeny. First, distinct clades have fairly large nucleotide sequence differences. Thus, both phylogenetic and antigenic groupings measure broad-scale divergence. Second, the antigenic analysis used polyclonal antisera, so that each test agent averaged broadly over many antigenic sites.

11.5. Immunology-Phylogeny Mismatch with Radiations into New Hosts

Influenza A isolates show a mixture of concordant and discordant relations between antigenic distance and phylogeny. Figure 11.4 illustrates the phylogenetic pattern for eleven isolates and an immunological matrix of reactivities to monoclonal antibodies (Brown et al. 1997). The phylogenetic analysis separated three clusters: classical swine types on the left, avian types in the middle, and avian-like isolates obtained from swine on the right.

Figure 11.4. Phylogeny of influenza isolates based on nucleotide sequences from the HA1 region of the hemagglutinin (HA) gene.

Figure 11.4

Phylogeny of influenza isolates based on nucleotide sequences from the HA1 region of the hemagglutinin (HA) gene. The isolates were obtained from swine (sw), turkey (ty), and duck (dk). The avian isolates were closer to the swine isolates on the right (more...)

The immunological reactivities divide the swine and avian-like swine into distinct clusters, matching the phylogenetic classification. The avian isolates are immunologically relatively distant from the other clusters and from each other, creating dissonance between phylogeny and antigenicity. It may be that the avian isolates have differentiated more strongly at the sites recognized by some of the monoclonal antibodies than they have when averaged over the entire sequenced region. It would be interesting to know more about the sites to which the different monoclonal antibodies bind. Perhaps some of those sites are influenced by selective pressures for attachment to host cells or for avoidance of host defense that differ between birds and pigs.

11.6. Short-Term Phylogenetic Diversification Driven by Immunological Selection

Influenza A variants with substitutions in the hemagglutinin molecule rise to high frequency in epidemics, causing a selective sweep through the influenza population. Isolates obtained in a particular year tend to trace their ancestry back to a common progenitor lineage just a few years into the past (Bush et al. 1999). Thus, the temporal sequence of the population is dominated by lineal replacements rather than bifurcating divergence.

Host antibodies mainly attack hemagglutinin. Immune selective pressure on hemagglutinin appears to drive the lineal replacements—put another way, immunological pressure drives change in the population-wide pattern of phylogenetic descent. Thus, the phylogenetic pattern of change may match the immunological pattern of change. Concordance probably depends on the percentage of amino acid substitutions explained by antibody pressure and the degree to which the antibody panel used for classification measures aggregate divergence.

11.7. Discordant Patterns of Phylogeny and Antigenicity Created by Within-Host Immune Pressure

Nucleotide sequences of the HIV-1 envelope divide this virus into eleven lineages within the M (major) group, into an O (outlier) group, and into an N (new) group (Robertson et al. 1999; The phylogenetic distance between isolates does not predict well the strength of shared immunological response (Vogel et al. 1994; Nyambi et al. 1996, 2000b; Weber et al. 1996; Zolla-Pazner et al. 1999).

Vaccines must stimulate an immune response against most viral genotypes in order to provide sufficient protection. A candidate vaccine might, for example, include isolates from each of the common phylogenetic lineages. This provides good coverage of diverse pathogens when antigenicity corresponds to phylogeny. However, phylogeny does not predict antigenicity for HIV. In order to choose HIV variants for a vaccine, one needs to divide viral diversity into antigenic types that together cover most of the antigenic range.

A few studies have developed antigenic classifications for HIV-1 (Vogel et al. 1994; Nyambi et al. 1996; Zolla-Pazner et al. 1999). Nyambi et al. (2000b) analyzed binding by twenty-eight monoclonal antibodies to twenty-six intact viral particles. The viral genotypes were sampled among eight different clades of the M group.

Nyambi et al. (2000b) grouped viral isolates according to their similarity in binding to the antibodies. Such grouping defines antigenic similarities of epitopes between the viral samples. The twenty-six viral isolates formed three immunological clusters. The immunotypes did not correspond to phylogenetic lineages (genotypes), geographic origin, or tropism for different host-cell receptors (CXCR4 versus CCR5). Thus, diverse genotypes share common epitopes, and similar genotypes can be differentiated by antibody binding, causing a mismatch between phylogeny and antigenicity. Nyambi et al. (2000a) emphasized that many antibodies that bind do not neutralize. Further studies must determine if the observed antibody binding can influence viral fitness in vivo.

Joint studies of phylogeny and antigenicity in HIV call attention to stabilizing, convergent, and diversifying natural selection.

First, shared antigenicity over long phylogenetic distances may be caused by stabilizing selection. Under stabilizing selection, a mutation that changes an epitope has opposing effects. The mutation allows escape from immune recognition but also reduces some functional aspect of the epitope. External viral epitopes sometimes affect binding or entry to host cells. Strong stabilizing selection of epitopes leads to conservation of amino acid composition over all phylogenetic scales of divergence.

In some cases, stabilizing selection may allow certain amino acid replacements that preserve geometric structure and charge. For example, Nyambi et al. (2000a) argue that parts of the genetically variable V3 loop of the HIV envelope have highly conserved structure and antigenicity. Binding affinity to monoclonal antibodies may be a better measure of antigenic conservation than amino acid sequence.

Second, shared antigenicity over long phylogenetic distances may be caused by convergent selection. Suppose a small set of alternative structures for a parasite epitope retain similar function. This functional set constrains the range of acceptable escape mutants. Immune pressure favors change of epitopes within the acceptable set, eventually returning to the original epitope. Phylogenetic pattern will reveal short-term changes and occasional long-term similarity. T cell pressure based on MHC binding may be particularly likely to create such patterns. A parasite that can escape from a particular host's MHC array will be favored. The next host will likely have a different MHC pattern, perhaps favoring a return to the epitopes lost in the previous host. Testing for this pattern requires detailed data over different temporal scales.

Nyambi et al. (2000b) found that the tip of the V3 loop was conserved antigenically between genotypes, whereas other parts of the V3 loop separated isolates into different antigenic groups. The genetic variants of the V3 loop may fall into relatively few conformational, antigenic types. The range of types may be constrained by stabilizing selection, causing short-term phylogenetic fluctuations between types but occasional convergence to past types within phylogenetic lines of descent.

Third, distinct antigenicity between phylogenetically close isolates implies very rapid diversifying selection. Escape mutants within hosts follow this pattern. For example, Nyambi et al. (1997) studied the temporal pattern of escape mutants and antibody profiles over the course of a single infection of a chimpanzee by simian immunodeficiency virus (SIV), a near relative of HIV. They analyzed eleven consecutive serum (antibody) samples and eight SIV isolates taken at about four-month intervals over a total of forty-one months. They tested the eighty-eight pairwise reactions between serum antibodies and viral isolates. The data showed viral escape mutants emerging at intervals of about fifteen months, each escape followed approximately eight months later by new antibody responses that matched the escape variants. Several studies of HIV escape mutants have been published (see references in Beaumont et al. 2001).

Combinations of diversifying, stabilizing, and convergent selection may determine the relationship between HIV phylogeny and antigenicity (Holmes et al. 1992). Diversifying selection within hosts favors escape variants that avoid antibodies or T cells. Diversifying selection between hosts favors mutants that avoid MHC recognition or immunological memory. Stabilizing selection constrains the range of variants. Convergent selection causes recurrence of previous antigenic types in response to diversifying selection and the stabilizing constraints that limit the range of alternative forms. Together, these factors group HIV isolates into a limited number of immunological types. The immunological classification does not match the phylogenetic classification.

Holmes et al. (1992) observed both diversifying and convergent selection within a single host infected by HIV. They sequenced the V3 loop of the viral envelope from eighty-nine isolates collected over a seven-year period. The isolates evolved over time through a series of replacements, with different sequences dominating in frequency at different times. Two divergent lineages formed about three years after infection. Most subsequent isolates mapped to one of these two major lineages. The same sequence of 6 amino acids at the tip of the V3 loop evolved convergently in the two lineages.

In summary, phylogeny provides the historical context in which to interpret immunological patterns. Hypotheses about natural selection can be tested by mapping the sequence of immunological changes onto the lineal history of descent.

11.8. Problems for Future Research

1. Adaptation to different hosts

Relations between antigenicity and phylogeny suggest hypotheses about how natural selection shapes antigenic variation. Consider, for example, the data for influenza A in pigs and birds (fig. 11.4). Antigenicity groups isolates according to current host species, whereas phylogeny groups isolates according to the history of transfers between species.

Adaptation to different hosts may shape antigenicity. This could occur by adaptation of viral surfaces to host receptors associated with attachment. Or the nature of immune pressure might differ significantly between birds and pigs. Such hypotheses, suggested by statistical patterns of association between phylogeny and antigenicity, must be tested by molecular studies.

2. Optimal sampling for evolutionary inference

Most antigenic and phylogenetic data were collected for reasons other than analyzing relations between antigenic and phylogenetic classifications. Little thought has been given to the sampling schemes that maximize information about evolutionary process. Ideal studies require analysis of the interactions between evolutionary process, methods of measurement, and statistical inference.

Different sampling schemes may be needed to study different kinds of natural selection. For example, stabilizing selection impedes change. To detect relatively slow antigenic change, one should probably sample over relatively long phylogenetic distances. The average divergence of genomes over long distances sets a standard against which one can detect reduced antigenic change at sites constrained by stabilizing selection.

By contrast, diversifying selection accelerates change by favoring antigenic types that differ from the currently prevalent forms. To detect relatively rapid change, one should probably sample over relatively short phylogenetic distances. This sets a low level of background change against which rapid, diversifying change can be detected.

3. Antigenic selection may shape phylogeny

The degree of match or discord between antigenic and phylogenetic classifications may depend on the demographic consequences of selection. If selection on a few closely linked epitopes determines the success or failure of a parasite lineage, then phylogeny may follow antigenicity. By contrast, selection may strongly influence patterns of antigenic change without absolutely determining success or failure of lineages. In this case, antigenicity does not constrain phylogeny.

Mathematical models would clarify the various relations that may arise between antigenic and phylogenetic classifications. Those relations depend on the time scales of differentiation, the epitopes used for antigenic classification, and the antibodies used to discriminate between variant epitopes.

Copyright © 2002, Steven A Frank.
Bookshelf ID: NBK2378