A host often retains immunological memory of B and T cells stimulated by prior infections. Upon later inoculation, a host rapidly builds defense from its memory cells. Each host acquires a unique memory profile based on its infection history.
In this chapter, I discuss the immune memory profiles of the host population. The following chapter describes how the structuring of immunological memory in the host population shapes the structuring of antigenic variation in parasite populations.
The first section reviews the immune processes that govern immunological memory. I emphasize the rate at which a host can generate a secondary immune response and the rate at which immune memory decays. These rate processes determine how immunological memory imposes selective pressure on antigenic variants.
The second section discusses the different consequences of immunological memory for different kinds of parasites. For example, antibody titers tend to decay more rapidly in mucosal than in systemic locations. Thus, selective pressures on antigenic variation may differ for parasites that invade or proliferate in these different compartments. Cytopathic viruses, which kill their host cells, may be more susceptible to antibodies, whereas noncytopathic viruses may be more susceptible to CTLs that kill infected host cells. The different memory responses of antibodies and CTLs may impose different selective pressures on antigenic variation of cytopathic and noncytopathic viruses.
The third section describes the immunodominance of memory. The memory profile may differ from the pattern of immunodominance during primary infection. The immunodominance of memory affects the ease with which new parasite variants can spread. If each host has narrow memory immunodominance with protection against one or a few epitopes, then a small number of mutations can escape memory. By contrast, if hosts have broad memory profiles, then the parasites have to change simultaneously at many epitopes in order to avoid the hosts' memory responses.
The fourth section focuses on the cross-reactivity between the antigens of a primary and secondary infection. Sometimes a host first develops a memory response to a particular antigen, and then is exposed secondarily to a variant of that antigen. If the secondary variant cross-reacts with memory cells, then the host may produce a memory response to the first antigen rather than a primary response to the second antigen. This original antigenic sin can prevent the host from mounting a vigorous immune response to secondary challenge. It can also prevent a host from expanding its memory profile as it becomes infected by different antigenic variants.
The fifth section summarizes the distribution of immune profiles among hosts. This distribution determines the ability of particular antigenic variants to spread. Older hosts tend to have broader profiles because they have experienced more infections. Maternal antibodies provide short-term protection to infants, and certain antibody and T cell responses may provide temporary protection to recently infected hosts. Finally, the hosts may vary spatially in their prior exposure to different epitopes, creating a spatial mosaic in the selective pressures that favor different antigenic variants.
The final section takes up promising lines of study for future research.
9.1. Immunological Memory
Immunological memory causes different immune responses between primary and secondary exposure to an antigen. Differences include the speed, intensity, and breadth of reaction. I focus on the consequences of immunological memory for antigenic variation of parasites. Thus, I am mostly concerned with how memory affects replication and transmission of the parasite.
The state of internal memory influences whether secondary response rapidly clears or only partially reduces secondary infection (reviewed by Zinkernagel et al. 1996). The X-Y–Z model (Byers and Sercarz 1968) captures the essential features: X represents a specific, naive B or T lymphocyte clone; Y represents a partially differentiated, long-lived memory state for the specific lymphocyte; and Z represents the short-lived, fully armed effector cells that do the work of clearing infection.
Studies have supported different components of this model for some experimental systems. But many conflicting results have been obtained, and controversy continues. A recent symposium (McMichael and Doherty 2000) and many reviews summarize empirical details and opposing views (Ahmed and Gray 1996; Zinkernagel et al. 1996; Dutton et al. 1998; Ada 1999; Farber 2000; Gray 2000; Seder and Hill 2000).
In an antibody response, Y represents the long-lived memory B cell clones and Z represents the short-lived plasma B cells that secrete antibody. Ochsenbein et al. (2000) studied B cell memory against vesicular stomatitis virus (VSV) infections in mice. They found that memory cells did in fact live a relatively long time compared with antibody-secreting plasma cells. The antibody-secreting cells had a half-life of 3–10 days.
Memory cells persisted in the absence of recurrent antigenic stimulation. By contrast, the maintenance of plasma cells and circulating antibodies required continued stimulation by antigens. Circulating antibodies often protect against secondary infection by VSV, whereas memory cells alone do not.
This example highlights several questions. Are effector cells generally short-lived or long-lived? Does the maintenance of effectors require recurrent antigenic stimulation? How long do memory cells live in the absence of repeated stimulation? How long does it take for memory cells to differentiate into effector cells? Is there always a sharp distinction between memory and effector cells, or do some cell types have some memory attributes (long-lived, easily stimulated) and effector attributes (directly involved in killing)?
These issues play a crucial role in shaping the immunological structure of host populations and consequently in the evolution of antigenic variation. The various conflicting details do not provide a clear picture at present. But it is possible to discuss how particular memory processes may affect the evolution of parasite diversity. The next subsection provides one example.
Recurrent Antigenic Stimulation of Antibody Production
How does a host maintain antibody titers after an infection has apparently been cleared? Zinkernagel et al. (1996) and Ochsenbein et al. (2000) favor the need for internal storage of antigen as a source of recurrent stimulation, with perhaps repeated infection as a booster in some cases. Others studies have implicated a subset of long-lived plasma cells as a potential source of continuous antibody production without the need for recurrent stimulation by antigen (Manz et al. 1998; Slifka et al. 1998). For our purposes, we can take the following relatively safe position.
The ratio of plasma to memory cells likely rises with recurrent antigenic stimulation. A higher concentration of plasma cells and antibodies provides greater protection and more rapid clearance. The benefit for maintaining plasma cells depends on how rapidly the infection develops within the host. Slow infections may allow memory cells to differentiate into an antibody response sufficiently rapidly to contain the infection. Fast infections may spread so quickly that memory cells cannot differentiate antibody-secreting plasma cells fast enough to contain the infection, but memory cells may aid in eventual clearance.
The immunological structure of host populations as it affects parasite transmission depends on plasma:memory ratios, which in turn may be affected by recurrent stimulation by internally stored antigen or extrinsic reinfection. Plasma:memory ratios more strongly influence parasites that grow relatively quickly within hosts.
Helper T Cell Memory
Memory B cells have higher densities of MHC class II molecules on their surfaces than naive B cells (Janeway et al. 1999). Presumably this allows antigens taken up by the B cell receptor to stimulate more strongly helper T cells, which in turn signal the memory B cells to differentiate into antibody-secreting plasma cells.
The CD4+ helper T cells themselves appear to differentiate into a memory form after sufficient initial stimulation (Janeway et al. 1999). Memory CD4+ helper T cells provide stronger stimulation to B cells than do naive CD4+ cells (Scherle and Gerhard 1986; Croft and Swain 1992; Swain 1994; Marshall et al. 1999).
The speed of an antibody response may be enhanced by CD4+ memory cells. This raises some interesting questions concerning the selective pressures that influence antigenic variation in parasites. Suppose, for example, that during initial exposure a host produces a dominant immune response to a parasite's B cell epitope, b, and to a CD4+ T cell epitope, t. Thus, we can write the initial parasite genotype as b/t.
In a host with memory against the b/t parasite, how well would antigenically variant parasites succeed with genotypes b′/t, b/t′, or b′/t′, where the primes denote variants? For example, how much advantage does the host gain by CD4+ memory against a parasite with an altered B cell epitope, or from the parasite's point of view, what is the fitness of the parasite genotype b′/t relative to b′/t′ in a host previously exposed to b/t? If the difference in fitness is sufficiently large, then the selective intensity on the epitope t may be strong. This would be interesting to know because most attention currently focuses on the obviously strong selective pressure for changes in the epitope b.
Zinkernagel et al. (1996) summarize limited evidence suggesting that helper T cell memory does not play an important role in shaping antigenic variants of parasites. Helper T cells cross-react between influenza strains (Hurwitz et al. 1984; Mills et al. 1986; Scherle and Gerhard 1986) and between vesicular stomatitis virus (VSV) strains (Gupta et al. 1986; Burkhart et al. 1994). This cross-reactivity does not protect hosts against secondary infection, but it can accelerate antibody response and reduce the time until clearance (Scherle and Gerhard 1986; Marshall et al. 1999).
In influenza infections, the dominant epitopes of helper T cells focus on hemagglutinin, a major surface molecule of influenza. The T cell epitopes are very near the B cell epitopes that dominate protective immunity (Wilson and Cox 1990; Thomas et al. 1998). It may be that amino acid changes in hemagglutinin between antigenically variant strains are sometimes selected by memory helper T cells. However, for amino acid replacements in hemagglutinin, it is difficult to separate the potential role of memory helper T cells from the obviously strong effects of antibody memory.
The level of memory helper T cells can be measured by the time required for naive B cells to switch from initial IgM secretion to later IgG secretion. When assessed by this functional response, helper T cell memory appears to be short-lived for influenza (Liang et al. 1994) and VSV (Roost et al. 1990). In mice infected with VSV, memory T cell help that accelerated the IgM to IgG switch lasted only fourteen to twenty-one days. Other assays find that memory helper T cells remain for several months after initial infection (Gupta et al. 1986); it appears that eventually the number of memory cells drops below a threshold or the memory cells lose a complementary signal. It will be interesting to learn whether limited functional helper T cell memory applies generally to all vertebrates or varies for different hosts or host-parasite combinations.
CD4+ cells have other functions in addition to stimulating B cells. For example, CD4+ cells influence CTL response and the response of other effectors such as macrophages. The limited evidence available does not demonstrate a strong role for CD4+ memory with regard to these effector-stimulating functions (Stevenson and Doherty 1998); however, the potentially diverse memory effects for these cells must be considered (Whitmire et al. 2000).
Important attributes of memory include the speed and intensity of response to antigen and the time decay of these quantitative responses (Seder and Hill 2000). CTL memory has been measured in various ways, for different hosts and different kinds of parasites (Zinkernagel et al. 1996; Dutton et al. 1998; Stevenson and Doherty 1998). Preliminary data suggest that patterns of immunodominance in the primary response do not necessarily carry through to the memory pool (Belz et al. 2000; Rickinson et al. 2000; Seaman et al. 2000). In some cases, it seems that T cell clones increased to high abundance in the primary response suffer greater reductions as the cellular populations are regulated in the memory phase (Rickinson et al. 2000).
A few general conclusions arise from this work: secondary CTL responses are typically faster and more intense than primary response, and the strength of the secondary response can decay over time. More important, the relations between CTL response and clearance depend strongly on the kinds of parasites.
9.2. Kinds of Parasites
"Every infection is a race" (Mims et al. 1993). The parasites race against immune effectors, which may eventually kill parasites faster than they are born. Each kind of parasite has its particular site of infection, pattern of spread between tissues, and rate of increase. Immunological memory therefore influences the host-parasite race in a different way for each kind of parasite.
In this section, I highlight some of the interactions between immunological memory and parasite attributes. I discuss memory-parasite interactions with regard to the type of immune cell involved, the kinetics of parasite spread, and the kinetics of immune effector response.
There are four main classes of immune cells that can be enhanced by primary infection to provide greater protection against later infections: plasma B cells, memory B cells, effector T cells, and memory T cells (Ahmed and Gray 1996).
The plasma cells secrete antibody. These effector B cells usually produce mature immunoglobulins such as IgG in systemic sites and IgA on mucosal surfaces. Circulating IgG often remains at significant titers throughout life. IgG can sometimes prevent infection by binding to inoculum before the parasites replicate in the host. IgG causes most cases of long-lived protective immunity against pathogens such as measles, yellow fever, polio, mumps, smallpox, and many other viruses and bacteria (Plotkin and Orenstein 1999; Knipe and Howley 2001).
IgA is often raised to high titers on mucosal surfaces in response to infection. IgA antibodies provide effective protection against pathogens that initially invade mucosal sites, such as influenza through the nasal mucosa, rotaviruses and many bacterial pathogens via the intestinal mucosa, and gonorrhea via the urethral epithelium (Mims 1987; Ada 1999). However, IgA titers decline relatively rapidly after infection, lasting on the order of months rather than years, as is often the case for IgG.
Memory B cells proliferate and differentiate into plasma cells upon secondary infection. If the pathogen is not immediately cleared by existing antibodies and the pathogen's initial replication is relatively slow, then the memory B cells may have time to differentiate into plasma cells and clear the pathogen before widespread infection develops. Differentiation of memory B cells into plasma cells depends on stimulation by CD4+ helper T cells (Ochsenbein et al. 2000).
Once widespread infection becomes established, memory B cells can help to produce a more specific, rapid, and intense antibody response. However, the relative roles of antibodies and T cells in clearing established infection vary depending on the attributes of the pathogen (Mims 1987; Janeway et al. 1999). For example, Zinkernagel et al. (1996) emphasized the distinction between cytopathic (cell-killing) and noncytopathic viruses. They also distinguished between viruses exposed to antibody in the blood or mucosa and those hidden from antibody in peripheral tissue.
Antibodies play a key role in clearing cytopathic viruses on mucosa or circulating in the blood. CTLs may be relatively ineffective against cytopathic viruses when the rate at which viruses infect, replicate, and kill a cell is greater than the rate at which CTLs kill viruses in infected cells. The dynamics of this race could be analyzed by mathematical models that compare the viruses' birth and death rates in light of the killing action mediated by antibodies and effector T cells.
For viruses that circulate in systemic infections, memory IgG antibodies may often protect against infection. By contrast, for mucosal infections such as those by rotaviruses and many bacterial pathogens, memory IgA antibodies often decline below protection level, but memory B cells can play an important role in defense by differentiating IgA-secreting plasma cells (Ahmed and Gray 1996).
Effector CTLs dominate clearance of noncytopathic intracellular pathogens and infected peripheral tissue that limits access to antibody (Zinkernagel et al. 1996). Effector T cells typically have a short half-life when not stimulated by antigen. Thus clearance before significant infection develops can occur by various scenarios. First, recent stimulation by antigen can boost effector T cell density to protective levels. Stimulation can occur by persistent antigen maintained in the host or by recurrent infection. Second, slowly spreading infections may allow differentiation of effector T cells from memory T cells in time to control initial spread of the pathogen. Third, memory antibody may clear the pathogen before the initial infection becomes established.
Lack of symptoms during secondary infection may result from rapid clearance of the parasite or from control of the infection that still allows some parasite replication and transmission. It is important to distinguish between clearance and controlled infection when studying the population dynamics and evolution of the parasite.
In summary, parasite attributes determine the type of host memory that impedes secondary infection. I mentioned the site of initial invasion (e.g., mucosal versus systemic), the site of widespread infection (e.g., epithelial, systemic, peripheral), and the lifetime of infected cells for intracellular parasites (e.g., cytopathic versus noncytopathic). Other parasite factors can tip the balance between clearance and widespread infection of a secondarily inoculated host. For example, the number of parasites in the inoculum frequently influences whether an infection is cleared quickly or spreads widely.
These various parasite attributes and the rate parameters that govern parasite birth and death within hosts must be measured against the kinetics of immunological memory and the response to secondary infection. The quantitative outcome influences the selective pressure imposed on various parasite epitopes by host memory. Such selective pressure, in turn, shapes the distribution of antigenic variation in parasite populations.
9.3. Immunodominance of Memory
A host's immunological memory profile depends on three factors. First, which parasite variants have infected that host in the past? Second, to which epitopes did the host respond?—the immunodominance of primary response. Third, to which of the primary epitopes has the host retained memory?—the immunodominance of the memory profile.
The immunological profile of each host and the variation of profiles between hosts influence the selective pressures imposed on parasite antigens. For the profile of each host, consider as a simple measure of immunodominance the number of epitopes to which a host retains protective antibody. If a host retains protection against n epitopes, then a variant parasite strain must differ in at least n sites to avoid all memory. If the mutation rate per site is μ, then the probability is μn that a progeny of the original strain is an escape variant with all of the n necessary differences.
Several laboratory experiments of influenza have studied the origin of escape variants when neutralizing antibody pressure is imposed against viral epitopes (Yewdell et al. 1979, 1986; Lambkin et al. 1994). For antibodies against only a single epitope, escape variants arise often because only a single mutation is needed. The mutation rate of influenza is on the order of μ = 10−5 per nucleotide per generation. Thus, a moderate-size population of viruses likely has at least a few escape mutants. By contrast, antibody selection against two or more epitopes rarely yields escape mutants, because the probability of multiple mutations, μn, becomes small relative to the effective size of the population.
These laboratory experiments show that a broader antibody response against multiple epitopes impedes the origin of new variants. By contrast, a more focused immunodominant response allows the rapid evolution of escape variants.
Similarly, persistent viral infections within hosts respond differently to narrow versus broad CTL pressure (Wodarz and Nowak 2000). A highly immunodominant CTL response allows rapid evolution of escape mutants and continuing change within hosts. By contrast, a broad CTL response against multiple epitopes impedes the origin of escape variants and leads to relatively slow evolution of viruses within a host.
To determine the selective pressures imposed on parasite populations, the immunodominance of each host's memory profile must be placed in the context of variation in memory profiles between hosts. Suppose, for example, that a parasite has two distinct antigenic sites. A parasite with genotype A/B at the two sites sweeps through the population, infecting all hosts. One-half of the host population maintains memory against both antigens, one-quarter has immunodominant memory against A only, and one-quarter has immunodominant memory against B only.
Now consider how this distribution of memory profiles influences the success of antigenic variants. A mutation at a single site, for example B, yields an altered parasite, A/B′. This mutant can attack the quarter of the host population with memory only against B. As the parasite spreads, a second mutation to A′/B′ allows attack of the remaining hosts.
This example shows that strongly immunodominant host profiles limited to one or a few sites allow parasite mutants with few changes to succeed. Once the variant parasite begins to spread between susceptible hosts, additional mutations allow attack against hosts with different immunodominant profiles or against hosts that developed broader immunity against multiple antigenic sites.
Influenza evolution may proceed by this sort of sequential accumulation of variation, with new epidemic strains differing from the previous epidemic strain at several sites (Natali et al. 1981; Underwood 1984; Wang et al. 1986; Wilson and Cox 1990; Lambkin and Dimmock 1996; Cleveland et al. 1997; Nakajima et al. 2000). Surveys of human populations and laboratory studies of mice and rabbits support this hypothesis by showing that individuals often have narrowly focused antibody responses and that individuals vary in the antigenic sites to which they develop antibodies.
In the laboratory, studies show that individual mice infected with human influenza often produce antibody responses focused on a limited number of antigenic sites—probably just one or two sites (Staudt and Gerhard 1983; Underwood 1984; Thomas et al. 1998). Individual mice differed in the antigenic sites to which they raised antibodies. Individual variation in antibody response probably occurs because stochastic recombinational and mutational processes generate antibody specificity (Staudt and Gerhard 1983).
Surveys of human populations find that individuals previously exposed to influenza vary in antibody memory profiles (Natali et al. 1981; Wang et al. 1986; Nakajima et al. 2000). Wang et al. (1986) studied immune memory profiles of individuals when measured for three nonoverlapping sites of the hemagglutinin surface glycoprotein. For samples collected from the early years of the Hong Kong influenza subtype epidemics (1969 and 1971), 33% of individuals had antibodies to all three sites, 50% had antibodies for two sites, and 17% had antibodies for only one site. Approximately equal numbers of individuals lacked antibody to any particular site, suggesting that each site was equally likely to stimulate an antibody response. Most individuals sampled in 1978 had antibodies for all three sites. It appears that after several years of repeated exposure to various strains of the Hong Kong subtype, individuals had acquired a wider repertoire of antibodies.
Human children tend to have particularly narrowly focused antibody profiles against influenza (Natali et al. 1981, 1998; Nakajima et al. 2000). This may occur either because of children's relatively smaller number of exposures or because of their narrower response per infection.
These observations on mice and humans support the hypothesis that individuals have narrowly focused antibody memory and that individuals vary in the antigenic sites to which they respond. This combination of individual focus and population variability creates a heterogeneous pattern of selection on parasites. After a widespread epidemic by a single parasite type, the parasite must acquire several new mutations before it can again spread widely through the population. Stepwise changes can occur by first changing at one site and attacking a subset of the population with a dominant response against that site. The new mutant strain can then accumulate a second change that provides access both to hosts with a dominant antibody response to the second mutant site and to hosts with antibodies against both the first and second mutant sites. Additional mutations allow attack against broader sets of immunological profiles.
This description certainly oversimplifies the actual process. However, the immunodominance of individual hosts for particular epitopes and the population variability of immune profiles can create important selective pressures on parasites.
9.4. Cross-Reactivity and Interference
A host's secondary response to an antigen depends on immunological memory to that antigen. Typically, memory leads to a faster and more vigorous secondary response. Suppose, however, that a host first develops a memory response to a particular antigen, and then is exposed secondarily to a variant of that antigen. If the secondary variant cross-reacts with memory cells, then the host may produce a memory response to the first antigen rather than a primary response to the second antigen. A memory response to the first antigen rather than a primary response to the variant is called original antigenic sin. I reviewed aspects of this phenomenon in chapter 6.
A memory response based on previously encountered, cross-reactive antigens has three consequences for the immunological structure of host populations. First, cross-reaction may aid protection or clearance against secondary challenge. This occurs if the cross-reactive memory effectors have sufficient affinity for the variant antigen (Kaverin et al. 2000; Roden et al. 2000; Sonrier et al. 2000; Stalhammar-Carlemalm et al. 2000).
Second, cross-reaction may interfere with the secondary response. This occurs when cross-reactive memory effectors do a poor job of clearing secondary challenge but respond sufficiently to repress a new, primary response against the variant antigen (Good et al. 1993; Klenerman and Zinkernagel 1998; Nara and Garrity 1998; Ferguson et al. 1999).
Third, the host may fail to develop an increasingly broad memory profile over the course of repeated exposures to different variants. This occurs when a new variant stimulates cross-reactive memory rather than a specific primary response, preventing memory particular for the new variant (Fazekas de St. Groth and Webster 1966a, 1966b; Smith et al. 1999).
9.5. Distribution of Immune Profiles among Hosts
The distribution of immune profiles influences selective pressures on antigenic diversity. Several factors shape the distribution of immunity. I have already mentioned the immunodominance of individual immune profiles and the tendency for the pattern of immunodominance to vary among individuals. I also discussed how cross-reactivity can affect clearance of secondary challenge and the development of memory over a host's lifetime. In this section, I add a few more factors that affect the distribution of immune profiles.
Age Structure of Hosts
An individual becomes exposed over time to an increasingly diverse array of parasite genotypes. Thus, older individuals typically have a broader memory profile than do younger individuals. Age-related patterns have been measured by serological surveys, which describe the presence or absence of circulating antibodies to a particular strain of parasite or to a particular antigen. Many surveys have been published for a wide variety of parasites and hosts (Anderson and May 1991, pp. 49–54).
Here are just a few example pathogens for which broader immunological profiles have been reported in older hosts compared with younger hosts: influenza (Dowdle 1999), Plasmodium (Gupta and Day 1994; Barragan et al. 1998), human T cell leukemia virus type 1 (HTLV-1) (Larsen et al. 2000), and hepatitis A, B, and C viruses (Chapman et al. 2000).
The best data on age effects come from studies of the influenza A virus. Most neutralizing antibodies against influenza bind to hemagglutinin, the virus's dominant surface molecule (Wilson and Cox 1990). Three major subtypes of hemagglutinin have circulated in human populations since about 1890, labeled H1, H2, and H3. Antibodies to one subtype cross-react relatively little with the other subtypes. Significant variation occurs within each subtype. Although antibodies to a particular variant do not always protect against infection by other variants of the same subtype, the antibodies to variants of a subtype do often cross-react to some extent.
These patterns of cross-reaction allow one to measure immunological profiles of individuals with regard to previous exposure to each of the three subtypes. By measuring individuals of different ages, a picture emerges of the past history of exposure and immunity to the different subtypes.
Figure 9.1a shows the percentage of individuals with antibodies to H1 born in different years. H1 is the subtype that caused the famous 1918 pandemic that killed tens of millions of people (Oxford 2000). Note that antibodies against H1 occur in 80–90% of individuals who were less than twenty years old during the pandemic years, suggesting widespread distribution of the disease. The drop in the seropositive level for individuals born before 1900 may be explained by the typically lower percentage of adults than children infected by influenza epidemics (Nguyen-Van-Tam 1998). There may also be some decay in immune memory among older individuals. The large drop in seroprevalence after 1922 suggests that H1 declined in frequency after the pandemic. Perhaps because of widespread immunity to H1, variants of this subtype had difficulty spreading between hosts.
Figure 9.1b shows a similar picture for the H3 subtype associated with a pandemic in 1890. Cohorts born in the years before the pandemic had very high seroprevalence, suggesting widespread infection. Seroprevalence declined sharply in those born just after the pandemic, implying that H3 had nearly disappeared from circulation. Figure 9.1b also shows data on H2 seroprevalence and a possible pandemic in 1900, but those data are more difficult to interpret.
Figure 9.2 illustrates the estimated mortality rate associated with influenza infection in two severe (pandemic) years. The H2 pandemic of 1957 caused relatively high mortality among older people compared with the H3 pandemic of 1968–69. Older people often suffer higher mortality from influenza than do younger people (Nguyen-Van-Tam 1998), so the pattern in 1957 appears to be typical. The contained mortality among older individuals in 1968–69 may have been caused partly by immunological memory to the H3 pandemic of 1890 and consequent protection against this subtype.
The age structure of immunity profiles has probably influenced the waxing and waning of the various influenza A subtypes over the past 110 years. Influenza causes uniquely widespread and rapid epidemics; thus the details of age-related immune profiles and antigenic variation likely differ in other pathogens. Malaria is perhaps the only other disease for which existing data suggest interesting hypotheses.
In areas with endemic Plasmodium falciparum infection, hosts often pass through three stages of immunity (Gupta and Day 1994; Barragan et al. 1998; Mohan and Stevenson 1998). Maternal antibodies provide significant protection for newborns up to six months of age. After maternal antibodies fade, high infection rates with severe disease frequently occur until the age of two to three years. Acquired immunity develops gradually over the following years, with significant reduction in the severity of symptoms. However, even healthy adults often have subclinical infections. Maintenance of protection requires repeated exposure. Individuals who depart and live in malaria-free areas for many months become significantly more susceptible upon return (Neva 1977; Cohen and Lambert 1982).
The slow buildup of immunity partly depends on the high antigenic variation of Plasmodium falciparum (Marsh and Howard 1986; Forsyth et al. 1989; Iqbal et al. 1993; Gupta and Day 1994). An individual apparently requires exposure to several of the locally common variants before acquiring a sufficiently broad immunological profile to protect against disease (Barragan et al. 1998).
Transmission of Maternal Antibodies
The rate at which susceptible hosts enter the population plays an important role in the dynamics of parasite strains. Newborns, memory decay, and migration provide the main sources of new susceptible hosts.
The susceptibility of newborns is complicated by maternal transmission of antibodies (Zinkernagel et al. 1996). Offspring of mice and humans obtain IgA antibodies in milk and IgG antibodies through the placenta (Janeway et al. 1999, pp. 326–327). IgA protects the gut epithelium and mucosal surfaces. The newborn inherits circulating IgG titers in the blood that match the mother's antibody levels. The infant receives the particular antibody specificities generated by the mother's history of exposure to particular antigens. Thus, the infant has a temporary memory profile that matches its mother's.
Maternal antibodies have a half-life of 3–6 months (Nokes et al. 1986; Anderson and May 1991, pp. 49–54). Infection of a baby early in life may be cleared by maternal antibody, thereby failing to stimulate an immune response and generate long-lasting memory (Albrecht et al. 1977).
Other vertebrates also transmit maternal antibodies to newborns (Zinkernagel et al. 1996). For example, bovines produce highly concentrated antibodies in the first milk (colostrum), which must be absorbed via the calf's gut during the first twenty-four hours after birth (Porter 1972). In this first day, the calf does not digest the immunoglobulins and is able to take up most antibody classes by absorption through the gut epithelium. Birds transmit maternal antibodies through the egg (Paul 1993).
Short-Term Protection from Recent Infection
IgA antibodies on epithelia can prevent initial infection by pathogens (Mims 1987, p. 251). For example, IgA may prevent attachment of Vibrio cholerae to the intestinal epithelium, gonococcus to the urethral epithelium, or chlamydia to the conjunctiva. IgA titers on epithelia often decay quickly after infection. Thus, protection against infection by IgA typically lasts for a few months or less.
Most vaccines protect by elevating the level of circulating antibodies and perhaps also memory B cells. The need for occasional vaccine boosters to maintain protection against some pathogens suggests that antibody titers or the pool of memory B cells decline in those cases. When long-term protection requires no boost, it may be that a lower threshold of antibodies or memory B cells protects against infection or that some regulatory mechanism of immunity holds titers higher.
In human influenza, T cells stimulated during infection provide some protection against later infection (McMichael et al. 1983b). But that protection wanes over a three-to-five-year period (McMichael et al. 1983a). A study of chickens also showed T cell–mediated control of secondary infection (Seo and Webster 2001). In that case, the secondary infection happened within 70 days of the primary challenge. The time decay of protection was not studied.
Measurements of memory decay have been difficult partly because laboratory mice provide a poor model for long-term processes of immunity (Stevenson and Doherty 1998). Laboratory mice typically live up to two years. It is difficult to separate decay of immunity from aging when immune memory in a mouse declines over many months.
Spatial Structure of Hosts
I discussed above how immune memory profiles may be stratified by age. Memory profiles may also be stratified by spatial location of hosts. At present, few spatial data exist. Thus, I confine my comments to a few conceptual issues.
To begin, consider the temporal pattern of measles epidemics prior to widespread vaccination (Anderson and May 1991, chapter 6). Data from England and Wales in 1948–1968 show a regular cycle of epidemic peaks every two years. The cycle may be explained by the threshold density of susceptible individuals required for an infection to spread. Just after an epidemic, most individuals retain memory that protects them from reinfection. The parasite declines because each infected individual transmits the infection to an average of less than one new susceptible host.
The next epidemic must wait until the population recruits enough newborns who are too young to have been infected in the last epidemic. An epidemic then follows, leaving most of the population protected until the next cycle of recruitment and spread of infection. Probably all parasite populations wax and wane to some extent as protective memory spreads with infection and the pool of susceptibles rebuilds by recruitment or by decay of immune memory.
These temporal fluctuations may also be coupled to spatial processes (Rohani et al. 1999; Earn et al. 2000). Imagine the spatial landscape of a population as a checkerboard of distinct patches. Epidemics may rise and fall synchronously in all patches, or epidemics may occur asynchronously over space. Suppose, for example, that half of the patches, labeled P1, have epidemics in odd years, whereas the other half of the patches, labeled P2, have epidemics in even years. One can visualize this dynamic landscape by imagining a peak in each patch rising during an epidemic and falling back to the ground between epidemics. Over an asynchronous landscape, some peaks are rising and others are falling at any time.
Measles virus effectively has only one antigenic type—a host's first infection and recovery provides lifelong protection. The spatiotemporal landscape of measles spread follows the waxing and waning of the numbers of infected individuals, driven by immunological memory, recruitment of newborns, and migration between patches.
Now imagine a parasite with distinct antigenic variants, for which memory to one variant does not provide any cross-protection against the other variants. The variants behave in effect as completely distinct parasites. In a patch, the waxing and waning of one variant may be synchronized with or uncoupled from the dynamics of the other variants. If the variants change asynchronously within patches, then the spatiotemporal landscape is covered by multiple surfaces of rising and falling peaks, the surfaces moving independently of each other.
I have discussed infection landscapes in a rather abstract way. But there is nothing out of the ordinary about hosts spread over space and infected over time by different antigenic variants of a parasite. The difficulty is to identify what general consequences arise from the interaction between antigenic variation and spatial processes. The landscape I have described so far has strains of antigenic variants that do not interact or interfere with each other. Thus, each strain changes independently of other strains, and no interaction occurs between space and antigenic variation.
Now consider antigenic variants for which some pairs of variants cause cross-reactive memory. It is not so easy to imagine the spatiotemporal landscape because the spread of each variant has differing quantitative effects on the dynamics of other variants. One simple analogy with age structure hints at the sort of processes that may occur.
In influenza, it may be that children have immunodominant memory focused on only one or a few antigenic sites. In the simplified example I discussed above, at first a virus strain with two sites, A/B, spreads. Some children develop immunodominant memory against A; other children develop immunodominant memory against B. Adults may have memory for both A and B. Mutant viruses can spread through the entire population in two steps. First, a mutant A′/B can attack children with memory against A. Then a second mutant, A′/B′ can attack everyone. The key is that different classes of hosts provide a pathway of connectivity by which single mutations of the virus can eventually spread through the entire population.
Connectivity may also occur over space. Figure 9.3 shows four patch types that have previously been infected by A/B, A′/B, A/B′, and A′/B′, respectively. This spatial distribution of immunological memory creates a stepwise pathway of connectivity for a parasite. For example, if a parasite A′/B′ first invades patch 1, then it can by a single mutation change into A/B′ and attack patch 2. Single mutational steps take that parasite to patch 3 and then on to patch 4.
This simplified description of spatial movement highlights two points. First, patches 1 and 3 fluctuate between A/B and A′/B′, whereas patches 2 and 4 fluctuate between A′/B and A/B′. These patch identities occur because immunological memory to both antigens imposes a barrier to any variant except the type with changes at both sites. Gupta et al. (1996, 1998) have emphasized the emergence of strain structure caused by this type of immunologically imposed selection. Second, my description extends Gupta et al.'s model to spatially structured host populations. With spatial structure, alternating regions of the host population can be dominated by the different pairwise sets of parasite strains. I will return to these issues in the next chapter, which focuses on the population structure of antigenically variable parasites.
Mathematical models could be developed to explore the interactions between antigenic variation and spatiotemporal dynamics. However, almost no data exist to compare with the models, and so there has been relatively little work along these lines. Some spatial data exist for influenza, but the scale of sampling and the measurement of cross-reactivity probably need to be enhanced before much can be concluded. This will not be easy to do in the short term. But eventually methods will improve for typing strains, and more data will become available.
9.6. Problems for Future Research
The processes that govern immunological memory within hosts and the distribution of immune profiles among hosts remain poorly understood. Throughout this chapter, I have emphasized important topics for further work. Rather than repeat all of those issues, I list here some hypotheses about kinetics that deserve study both empirically and mathematically. Many patterns of antigenic variation turn on these rate processes that drive evolutionary dynamics.
(a) A higher rate of recruitment into the host population reduces the selective pressure on antigenic variation
Consider two species, a long-lived species with an average life span of L years, and a short-lived species with an average life span of S years. If L = 70, then newborns replace approximately 1/70th of the population per year. By contrast, if S = 7, then newborns replace approximately 1/7th of the population each year. Immunological memory decays faster at the population level in short-lived than in long-lived species, perhaps reducing the relative fitness advantage of antigenic variants in short-lived compared with long-lived hosts.
If an antigenic variant has a fitness cost relative to the wild type, then a greater offsetting fitness benefit occurs in the species with longer life span and fewer naive hosts. Thus, a high-cost variant could gain an advantage in long-lived but not in short-lived hosts. Antigenic variation may therefore occur more often in long-lived hosts than in short-lived hosts.
(b) Highly infectious parasites spread more widely in host populations and induce a higher percentage of hosts to have immunological memory against particular antigens
Highly infectious parasites therefore face more severe selective pressure for antigenic change. Here "highly infectious" means a higher basic reproductive number, R0, that is, a higher number of secondary infections caused by an infected host in a naive population. The density of immunological memory in the host population against parasites with different R0s should be studied mathematically to refine this idea.
(c) Rapidly transmitting parasites face less immunological pressure for change within hosts
If most parasite transmission occurs before the onset of strong, specific immunity, then relatively little pressure for antigenic change occurs within each host. But rapidly transmitting parasites may induce a greater density of immunological memory in the host population as noted in the previous item.
(d) Spatially homogeneous populations develop a higher and more uniform density of immunological memory than spatially heterogeneous populations
All hosts have the same high exposure rate to parasites in a well-mixed, spatially homogeneous population. By contrast, spatially heterogeneous populations may maintain temporarily isolated refuges in which hosts have low exposure. Those refuges could provide a source of hosts with limited immune memory, reducing the intensity of selection favoring antigenic variation.
The dynamics are complex because isolated host populations may have less prior exposure and immune memory but also may be less accessible to invasion by parasites and less able to transmit parasites back into the bulk of the host population. The net effect depends on the spatial connectivity of patches, rates of parasite transmission, and rates at which immune memory builds up and decays.
(e) Heterogeneity in immune memory profiles between age classes or spatial locations favors the stepwise spread of new antigenic variants
In this chapter, I discussed how new variants often need to change in several epitopes in order to spread through a host population with a high density of prior exposure. Heterogeneity enhances the chance of multiple antigenic changes by providing a sequence of susceptible host classes separated by the need for only a single antigenic change.
2. Memory decay
(a) Faster time decay of immunological memory may reduce the selective pressure on antigenic variation
Fast decay could potentially reduce the density of immunological memory across the host population. However, the faster the immunological memory decays, the more rapidly the parasites may reinfect hosts. The net effect on memory depends on the balance between these forces.
(b) The effect of immunological memory decay depends on the kind of parasite
The site of initial invasion determines which immune effectors can potentially block first infection. Epithelial invasion interacts mostly with IgA, a memory class that tends to decline relatively rapidly. By contrast, systemic invasion interacts mostly with IgG, a memory class with a relatively long half-life.
Clearance of extracellular parasites depends mostly on antibodies. If infection spreads primarily to epithelial tissue, IgA plays a key role, whereas IgG dominates against many systemic infections. Antibody memory can increase the rate of clearance.
Once intracellular infection becomes established, the key immune effectors depend on kinetics. Antibodies dominate against intracellular parasites that rapidly kill host cells. In these fast cytopathic parasites, reproduction within cells occurs sufficiently quickly that CTLs cannot reduce the spread of the parasite. Antibody memory, if it does not prevent infection, can increase the clearance rate of fast cytopathic parasites.
Slow cytopathic parasites or noncytopathic parasites that persistently infect host cells must be cleared by killing infected cells. Most killing of infected cells seems to be done by effector CTLs. Thus, the rate of memory decay in CTLs may govern protection against noncytopathic infections that are not blocked during initial invasion by antibodies.
3. Heterogeneity of immune profiles among hosts
The dynamics of immune profiles can be complex because the spread of parasite variants affects the immune structure of the hosts, and the immune structure of the hosts determines the selective pressures on different classes of antigenic variants. Some preliminary theoretical work along these lines has appeared (Gupta et al. 1996; Andreasen et al. 1997; Gupta et al. 1998; Lin et al. 1999). Similar processes of reciprocal coevolution arise when hosts and parasites have multiple genetic determinants that influence the outcome of an attack (Frank 1993, 1994).
Measles apparently can vary its dominant surface antigen, hemagglutinin, and limited variation does occur (Griffin 2001). So it is an interesting puzzle why antigenic variants do not to spread. Perhaps the very high R0 of measles causes the common strain to spread so widely in the host population that no differences occur between hosts in immune memory profiles. Thus, there is no single-step mutational change that allows a variant to spread to some hosts. The only "nearby" susceptible class arises from the influx of naive newborns.
Influenza A may not evolve as quickly and vary antigenically as much in birds as it does in humans (Webster et al. 1992). This suggestion is based on very limited evidence and must be confirmed by further study. If the pattern holds, then many plausible explanations exist. For example, the process of host invasion and spread during an infection probably differs in birds and humans, which may influence the role of immunity in clearance and in subsequent protection. Another possibility is that relatively short-lived species such as many birds have a larger class of naive hosts than comparably long-lived humans, reducing the relative pressure for antigenic variation in birds.