SIV and HIV

Soudeyns et al. (1999) identified the regions of the HIV-1 envelope under strong selective pressure by analyzing the pattern of nucleotide changes in the population. They compared the rate of nonsynonymous (d_N) nucleotide replacements that cause an amino acid change versus the rate of synonymous (d_S) nucleotide replacements that do not cause an amino acid change. A high d_N/d_S ratio suggests positive natural selection favoring amino acid change; a low d_N/d_S ratio suggests negative natural selection opposing change in amino acids (Page and Holmes 1998; see chapter 15 below).

Soudeyns et al. (1999) found that regions of the envelope gene under strong positive selection corresponded to epitopes recognized by CTLs. The nonsynonymous substitutions in these epitopes typically abolished recognition by a matching CTL clone. The population of viruses accumulated diversity in the dominant epitopes over the course of infection within hosts. These results suggest that CTL attack based on specific recognition drives the rapid rate of amino acid replacements in these epitopes.

Kimata et al. (1999) studied properties of simian immunodeficiency virus (SIV) isolated from early and late stages of infection within individual hosts. The early viruses infected macrophages, replicated slowly, and the viral particles were susceptible to antibody-mediated clearance. The late viruses infected T cells, replicated more than 1,000 times faster than early viruses, and were less sensitive to antibody-mediated clearance.

Kimata et al. (1999) did not determine the viral amino acid changes that altered cell tropism of SIV. Connor et al. (1997) found that changes in the host cell coreceptors used by early and late HIV-1 correlated with changes in cell tropism, but it is not yet clear which changes are essential for the virus's tropic specificity. Connor et al. (1997) did show that the population of early viruses used a narrow range of coreceptors, whereas the late viruses were highly polymorphic for a diverse range of host coreceptors. Clearly, the virus is evolving to use various cell types.

The relative insensitivity of late SIV to antibody apparently depended on increased glycosylation of the envelope proteins (Chackerian et al. 1997). The late viruses with increased glycosylation were not recognized by antibodies that neutralized the early viruses. Viruses that escape antibody recognition gain significant advantage during the course of infection (Chackerian et al. 1997; Rudensey et al. 1998). Kimata et al. (1999) showed that, when injected into a naive host, the late SIV did not stimulate as much neutralizing antibody as did the early SIV. Additional glycosylation apparently reduces the ability of antibodies to form against the viral surface. Presumably the glycosylation also hinders the ability of the virus to initiate infection; otherwise both early and late viruses would have enhanced glycosylation.

Both the early, macrophage-tropic SIV and the late, T cell–tropic SIV used the host coreceptor CCR5 (Kimata et al. 1999). That observation contrasts with a study of early and late HIV-1 isolated from individual hosts, in which Connor et al. (1997) found that early, macrophage-tropic viruses depended primarily on the CCR5 coreceptors, whereas the population of late viruses had expanded coreceptor use to include CCR5, CCR3, CCR2b, and CXCR4.

Many recent studies focus on HIV diversification within hosts (e.g., Allen et al. 2000; Goulder et al. 2001; Saha et al. 2001).

Hepatitis C Virus

Farci et al. (2000) obtained hepatitis C virus (HCV) samples at various stages of infection within individual hosts. They sequenced the envelope genes from these samples to determine the pattern of evolution within hosts. They then compared the evolutionary pattern with the clinical outcome of infection, which follows one of three courses: clearance in about 15% of cases; chronic infection and either slowly or rapidly progressive disease in about 85% of cases; and severe, fulminant hepatitis in rare cases.

Farci et al. (2000) sampled three major periods of infection: the incubation period soon after infection; during the buildup of viremia but before significant expression of specific antibodies; and after the host's buildup of specific antibodies.

The sequence diversity within hosts identified two distinct regions of the envelope genes. The hypervariable region evolved quickly and appeared to be under positive selection from the host immune system, whereas other regions of the envelope genes had relatively little genetic variation and did not evolve rapidly under any circumstances. Thus, the following comparisons focus only on the hypervariable region.

Those hosts that eventually cleared the virus had similar or higher rates of viral diversification before antibodies appeared than did those patients that developed chronic infection. By contrast, after antibodies appeared, chronic infection was correlated with significantly higher viral diversity and rates of evolution than occurred when the infection was eventually cleared. It appears that hosts who cleared the infection could contain viral diversity and eventually eliminate all variants, whereas those that progressed to chronic infection could not control viral diversification. The rare and highly virulent fulminant pattern had low viral diversity and rates of evolution. This lack of diversity suggests either that the fulminant form may be associated with a single viral lineage that has a strong virulence determinant or that some hosts failed to mount an effective immune response.

Generality of Within-Host Evolution of Antigens

HIV and HCV share several characters that make them particularly likely to evolve within hosts. They are RNA viruses, which have relatively high mutation rates, relatively simple genomes, simple life cycles, potentially high replication rates, and potentially high population sizes within hosts. HIV and HCV also typically develop persistent infections with long residence times in each host. If the mutation rate per nucleotide per replication is 10⁻⁵ and the population of viruses is on the order of 10¹⁰ within a host, then there are 10⁵ point mutations at every site in every generation. For every pair of sites, there will usually be at least one virus that carries mutations at both sites. Thus, there is a tremendous influx of mutational variation.

Other RNA viruses such as influenza also have high mutation rates and potentially large populations within hosts, but the hosts typically clear infections within two weeks. Some within-host evolution very likely occurs, but it does not play a significant role in the infection dynamics within hosts.

DNA-based pathogens produce much less mutational variation per replication. But large population sizes, long infection times, and hypermutation of epitopes could still lead to significant evolution within hosts. At present, the persistent RNA infections have been studied most intensively because of their obvious potential for rapid evolutionary change. As more data accumulate, it will be interesting to compare the extent and the rate of within-host evolutionary change in various pathogens.

Stochastic Switching versus Ordered Parasitemias

In Trypanosoma brucei, lineages switch stochastically between variants. Turner and Barry (1989) measured the switch probability per cell per generation for changes between particular antigenic types. Each entry in table 7.1 shows log₁₀ of the probability that a cell expressing a particular variant, designated by a number in the left column, switches to another variant designated by a number in the column headings.

Table 7.1

Log₁₀ of probability of switching between antigenic variants in Trypanosoma brucei.

The different rows in table 7.1 summarize data from five separate experiments. Overall, it appears that each type can potentially switch to several other types, with the probability of any transition typically on the order of 10⁻⁴ to 10⁻². Trypanosoma brucei stores and uses many different antigenic variants, perhaps hundreds (Vickerman 1989; Barry 1997). Thus, the limited sample in table 7.1 does not provide a comprehensive analysis of switch probabilities between all types.

Switches between types within a cellular lineage occur stochastically. But the sequence of variants that dominate sequential waves of parasitemia tends to follow a repeatable order (Gray 1965; Barry 1986). For example, figure 7.1 shows the date at which different variants first appeared in Trypanosoma vivax infections of rabbits. Some separation occurs between variants that arise early versus late.

Figure 7.1

The sequence of appearance for nineteen antigenic variants of Trypanosoma vivax in rabbits. Each column shows a different antigenic variant. The rows are the day since inoculation at which a variant was first detected during an infection. The days of (more...)

Temporal separation in the rise of different antigenic variants allows trypanosomes to continue an infection for a longer period of time (Vickerman 1989). If all variants rose in abundance early in the infection, they would all stimulate specific immune responses and be cleared, ending the infection. If the rise in different variants can be spread over time, then the infection can be prolonged.

The puzzle is how stochastic changes in the surface antigens of individual parasites can lead to an ordered temporal pattern at the level of the population of parasites within the host (Agur et al. 1989; Frank 1999; Turner 1999). Four hypotheses have been developed, none of which has empirical support at present. I briefly describe each idea.

First, the antigenic variants may differ in growth rate. Those that divide more quickly could dominate the early phases of infection, and those that divide more slowly could increase and be cleared later in the infection (Seed 1978). Computer studies and mathematical models show that variable growth rates alone can not easily explain wide separation in the times of appearance of different variants (Kosinski 1980; Agur et al. 1989). Only with a very large spread in growth rates would the slowest variant be able to avoid an immune response long enough to develop an extended duration of total infection. Aslam and Turner (1992) measured the growth rates of different variants and found little difference between the variants.

Second, parasite cells may temporarily express both the old and new antigens in the transition period after a molecular switch in antigenic type (Agur et al. 1989). The double expressors could experience varying immune pressure depending on the time for complete antigenic replacement or aspects of cross-reactivity. This would favor some transitions to occur more easily than others, leading to temporal separation in the order of appearance for different antigenic variants. This model is rather complex and has gained little empirical or popular support, as discussed in several papers (Barry and Turner 1991, 1992; Agur 1992; Muñoz-Jordán et al. 1996; Borst et al. 1997).

Third, the switch probabilities between antigenic variants may be structured in a way to provide sequential dominance and extended infection (Frank 1999). If the transition probabilities from each variant to the other variants are chosen randomly, then an extended sequence of expression cannot develop because the transition pathways are too highly connected. The first antigenic types would generate several variants that develop a second parasitemia. Those second-order variants would generate nearly all other variants in a random switch matrix.

The variants may arise in an extended sequence if the parasite structures the transition probabilities into separate sets of variants, with only rare transitions between sets. The first set of variants switches to a limited second set of variants, the second set connects to a limited third set, and so on. Longer infections enhance the probability of transmission to other hosts. Thus, natural selection favors the parasites to structure their switch probabilities in a hierarchical way in order to extend the length of infection.

Turner (1999) proposed a fourth explanation for high switch rates and ordered expression of variants. The parasite faces a trade-off between two requirements. On the one hand, competition between parasite genotypes favors high rates of switching and stochastic expression of multiple variants early in an infection. On the other hand, lower effective rates of switching later in an infection express variants sequentially and extend the total length of infection.

Many Trypanosoma brucei infections in the field probably begin with infection by multiple parasite genotypes transmitted by a single tsetse fly vector (MacLeod et al. 1999). This creates competition between the multiple genotypes. According to Turner (1999), competition intensifies the selective pressure on parasites to express many variants—variation allows escape from specific immunity by prior infections and helps to avoid cross-reactivity between variants expressed by different genotypes. These factors favor high rates of stochastic switching.

The effective rate of switching drops as the infection progresses because the host develops immunity to many variants. Effective switches occur when they produce novel variants, and the rate at which novel variants arise declines over the course of infection. Those novel variants, when they do occur, can produce new waves of parasitemia, promoting parasite transmission.

Turner's idea brings out many interesting issues, particularly the role of competition between genotypes within a host. But his verbal model is not fully specified. For example, delayed expression of some variants and extended infection depend on the connectivity of transition pathways between variants, an issue he does not discuss. The problem calls for mathematical analysis coupled with empirical study.

Role of Prior Exposure

Hosts that have recovered from an infectious parasite that switches antigenic type may retain immune memory for many antigenic variants. Successful reinfection would require a parasite to express a variant for which the host lacks specific memory. Antigenic variants expressed from an archival library can help a parasite to overcome immune memory of previously infected hosts.

The role of antigenic variation in avoiding immune memory from prior infections depends on several factors. How many variants stimulate memory during a typical infection? What percentage of infected hosts recover and survive? What is the rate of death among surviving hosts (population memory decay) relative to the rate at which naive, newborn hosts enter the population? What percentage of hosts suffer a primary infection? What percentage become reinfected?

Again, these interacting quantitative factors can be combined into a mathematical model. A model would suggest what conditions must be met for archival antigenic variation to be an effective strategy to avoid host immune memory.

Predator-Prey Feedback with Specific Immune Cells

Consider two variants, x and y, each variant attacked by specific immunity, I_x and I_y. In the simplest case of persistence, each matching pair fluctuates independently. Thus, as x increases, I_x rises and causes x to decline. A decline in x lowers stimulation and causes I_x to fall, which allows x to rise, and so on. A similar cycle happens with the predatory immune type, I_y, preying on the antigenic type, y.

Resource Competition

It could be that the two cycles progress independently, with coexistence of the antigenic types. Or there can be various forms of coupling between the cycles. For example, the parasite types x and y may compete for a host resource, R, such as host cells to infect or the uptake of a limiting nutrient (Smith and Holt 1996).

Direct competition between the parasite variants creates indirect interactions between the specific immune types. Consider what happens if y increases. Figure 7.2a shows that a rise in y stimulates I_y and reduces the available host resources, R. A drop in R depresses x, which in turn lowers stimulation to I_x. Overall, if we ignore all feedbacks, an increase in y enhances I_y, and depresses x and I_x.

Figure 7.2

Pathways of interaction between a host resource, R, two populations of parasites with different antigenic types, x and y, and two populations of specific immune cells, I_x specialized for x, and I_y specialized for y. (a) The various direct effects of an (more...)

Feedbacks occur, and their consequences must be followed. Continuing with the example, figure 7.2b shows that a rise in I_y grazes y to a lower value, allowing host resources, R, to recover, which enhances x and stimulates I_x. Figure 7.2c shows that several other feedbacks exist even in this highly oversimplified network of interactions—to trace all consequences is beyond normal intuition. Analysis requires mathematical models (Nowak and May 2000).

For this particular example, it turns out that resource competition by itself typically reduces the potential for coexistence of antigenic variants compared with the case in which no competition occurs. If I_y drives y to extinction in the absence of competition, then additional competition for resources will usually not save y. Rather, the competition from x further decreases y's chances for survival.

Several studies suggest that resource competition between parasites may sometimes influence the within-host dynamics of infection. In persistent malaria infections, competition between Plasmodium for susceptible erythrocytes apparently plays an important role (Gravenor et al. 1995; Hetzel and Anderson 1996; Gravenor and Lloyd 1998). Wodarz et al. (1999) proposed that the spread of human T cell leukemia viruses (HTLV-1) between host cells is limited by availability of susceptible, uninfected T cells.

Turner et al. (1996) inferred density-dependent effects on the growth rate of the blood-borne parasite Trypanosoma brucei. Although Turner et al. (1996) did not demonstrate parasite competition for host resources, it seems likely that some resource limitation arises because of the very high parasite densities that occur at peak parasitemia, on the order of 10⁶ to 10⁷ parasites per ml of blood. These studies did not directly discuss antigenic variation, but they suggest that resource competition may be important.

Resource Specialization

Antigenic variants may specialize on different host resources. For example, parasite surface molecules may influence tropism for host cell type or efficiency in uptake of different host nutrients. To the extent that antigenic variants do differ in their use of host resources, coexistence becomes easier to maintain by reducing the direct competition between the variants.

Variation in tissue tropism appears to be associated with antigenically variable surface molecules in Neisseria gonorrhoeae (Gray-Owen et al. 1997; Virji et al. 1999). In Neisseria, variable cell tropism may be important in sequentially colonizing different tissues as invasion and spread develop, with little direct competition between the antigenic variants.

HIV provides a potential case of competition and resource specialization between variants. Connor et al. (1997) found changes in coreceptor use by early and late HIV-1 correlated with changes in cell tropism. The population of early viruses used a narrow range of coreceptors, whereas the late viruses were highly polymorphic for a diverse array of host coreceptors. As the population of viruses builds and depresses the abundance of commonly infected cell types, diversification to different cell tropisms reduces competition.

Spatial Segregation

Variable resource concentration can also favor antigenic diversity. Consider a contrast between two antigenic variants. The first has a surface antigen that provides superior entry into host cells, but this variant is cleared at a higher rate. The second variant has a lower rate of entry into host cells, but is cleared at a lower rate. The first type interferes with the second when both are common.

This infection-clearance trade-off can promote coexistence by spatial segregation. For example, host compartments with low resource levels cannot sustain the first type—limited host cells reduce the production rate below the high clearance rate. The second, weaker competitor dominates. By contrast, in compartments with high resource levels, the stronger type dominates by outcompeting the weaker type.

Other Factors

There are, of course, many other factors that influence the abundance of antigenic variants and immune cells. The immunogenicity of the antigenic types may differ, varying the rate of parasite killing and the stimulatory signals to the immune cells. Each parasite carries many different antigenic determinants, raising once again issues of immunodominance, which must be understood in terms of populations of parasite variants and populations of immune cells with matching specificities.

Each factor can have complex dynamical consequences. Mathematical studies show that even rather simple interactions often lead to fluctuating abundances because of the nonlinear processes inherent in population dynamics. Thus, fluctuating abundances of antigenic variants and matching immune specificities may often occur in persistent infections (Nowak and May 2000).

1. Mutational distance required for escape

How many amino acid substitutions are needed for new variants to escape immunity against the original epitope? Does escape usually arise from a single substitution, or are multiple substitutions often required? If laboratory mice can be used as a model, it would be interesting to infect replicates of a common host genotype by a cloned pathogen genotype. One could then study the relative effect of genotype and stochastic factors on the number of substitutions in escape variants and the genetic pattern of diversification in escape. I discuss relevant preliminary studies in later chapters on experimental evolution.

2. Transmissibility of escape variants

Epitopes often occur in key surface molecules used for attachment or in important enzymes such as replication polymerases. Escape variants gain by avoiding specific immunity but may impose costs by lowering other components of parasite fitness. I mentioned earlier that SIV isolated late in infections had increased glycosylation of surface antigens that reduced susceptibility to antibodies. The glycosylation also reduced the degree to which viruses stimulated an antibody response when injected into new hosts. It would be interesting to know if glycosylation reduces transmissibility or some other component of viral fitness.

Escape within a host does not necessarily reduce transmissibility or other components of fitness. Goulder et al. (2001) studied human mothers with the MHC allele B27 infected by HIV-1. B27 recognizes an immunodominant and highly conserved CTL epitope in the viral Gag protein. Escape mutants at this epitope persist and enhance progression to disease. Mothers can transmit this escape variant to their offspring, who then target a subdominant B27 epitope and fail to contain the infection. These escape variants remain stable and do not revert to the original type when passaged in cell culture. It would be interesting to study similar issues with SIV, in which it would be possible to do experiments in vivo and test whether there is a fitness cost associated with CTL escape.

3. Dynamics of infection in parasites that switch between archival variants

Antigenic switching from archival libraries generates interesting dynamics within the host. Typically, the first variants increase rapidly, causing a high density of parasites within the host. Specific immunity then rises against those initial variants, causing a decline in the parasite population within the host. The initial, dense parasitemia generates variants by occasional switching. The variants rise in abundance during or after the decline of the first parasite burst.

These dynamics point to several questions. What is the basic timing for the initial growth of the parasite population, the rise in specific immune cells, and the decline in the initial parasitemia? What are the densities and the diversity of antigenic variants during the initial parasitemia? What are the timings and the shapes of the growth curves for the populations of antigenic variants?

At what parasite density do the variants begin to stimulate a specific immune response? That stimulatory threshold sets the pace at which the host can raise a new wave of immunity to combat the second parasite wave. How many variants rise in the second wave? What is the timing and pattern of new variants generated by parasites in the second wave? How do the coupled dynamics of specific immune cell populations and matching parasite variants together determine the total length of infection and the fluctuating density of parasites available for transmission? What determines the order in which parasite variants rise in successive parasitemias?

4. Immune dynamics in different body compartments

Different parasite surface molecules may cause infection of different body compartments. The surface molecules that affect tissue tropism may also be strong antigenic determinants. I mentioned that diversifying tissue tropisms during the course of an infection can diversify antigenic variation within the host. In addition, the dynamics of host immunity and the ability of immune effectors to attack parasites likely vary among body compartments. Thus, variants with certain tropisms may sequester themselves in refuges from immune pressure. These protected sites may provide a source of chronic infection or generate relapses after apparent clearance of the initial infection.