The Dynamics of HIV-1 Adaptation in Early Infection

Human immunodeficiency virus type 1 (HIV-1) undergoes a severe population bottleneck during sexual transmission and yet adapts extremely rapidly to the earliest immune responses. The bottleneck has been inferred to typically consist of a single genome, and typically eight amino acid mutations in viral proteins spread to fixation by the end of the early chronic phase of infection in response to selection by CD8+ T cells. Stochastic simulation was used to examine the effects of the transmission bottleneck and of potential interference among spreading immune-escape mutations on the adaptive dynamics of the virus in early infection. If major viral population genetic parameters are assigned realistic values that permit rapid adaptive evolution, then a bottleneck of a single genome is not inconsistent with the observed pattern of adaptive fixations. One requirement is strong selection by CD8+ T cells that decreases over time. Such selection may reduce effective population sizes at linked loci through genetic hitchhiking. However, this effect is predicted to be minor in early infection because the transmission bottleneck reduces the effective population size to such an extent that the resulting strong selection and weak mutation cause beneficial mutations to fix sequentially and thus avoid interference. THE interaction between selection and genetic drift over multiple loci may be complex. For a single locus, the product of the effective population size (Ne) and the selection coefficient (s), Nes, adequately measures the scaled intensity of selection. However, with more than one locus, selection at one locus may reduce Ne at linked loci. This has many ramifications, not least of which is that there is no species Ne, but that Ne varies across recombinational neighborhoods of the genome (see review by Comeron et al. 2008). On the other hand, with strong selection (Nes . 1) and weak mutation (Nem ,, 1), a new beneficial mutation that survives stochastic loss may spread to fixation before the emergence of the next beneficial mutation to survive stochastic loss (Gillespie 1984; Orr 2002). These conditions, therefore, reduce linkage interactions among loci. Ne is the size of a model population that exhibits the same stochastic variation in allele frequencies as an actual population (see Charlesworth 2009 for a recent review). This is a useful quantity because it captures stochasticity due to numerous factors, including population bottlenecks and selection, allowing the model to focus on a limited set of loci and evolutionary forces of interest. Complexity is introduced by the fact that the intensity of selection is proportional to Nes, but that selection tends to reduce Ne. Selection does this in two ways: first, by generating, at least initially, additive genetic variance in fitness (without selection there is no variance in fitness), because this increases the variation in offspring number among parents (Robertson 1961; Nei and Murata 1966), and second, by reducing variation at genetically linked loci. This effect arises as a result of genetic drift because only in finite populations is there the necessary linkage disequilibrium between loci (Barton 2000). One manifestation of this linkage effect is known as hitchhiking because the frequency of an allele will increase if it is associated with a selected allele at a linked locus (Maynard Smith and Haigh 1974). This process is also known as a selective sweep (Berry et al. 1991) and is part of a more general class of processes, known as the Hill–Robertson effect (Hill and Robertson 1966; Felsenstein 1974), in which selection usually reduces Ne at linked loci (Comeron et al. 2008). It has been proposed that if adaptation is common, neutral variation may be more affected by selection at linked loci than by genetic drift, a scenario referred to as “genetic draft” (Gillespie 2000; Maynard Smith and Haigh 1974). At present, it is unclear whether neutral variation is strongly affected by linked selection in Drosophila (Andolfatto 2007; Sella et al. 2009) and in humans (Cai et al. 2009; Hernandez et al. 2011). A recent theoretical study suggests that genetic draft is an important determinant of genetic variation in Copyright © 2012 by the Genetics Society of America doi: 10.1534/genetics.111.136366 Manuscript received November 1, 2011; accepted for publication December 18, 2011 Address for correspondence: School of Molecular and Biomedical Science, Molecular Life Sciences Building, Gate 8, Victoria Drive, University of Adelaide, Adelaide, SA 5005, Australia. E-mail: [email protected] Genetics, Vol. 190, 1087–1099 March 2012 1087 human immunodeficiency virus type 1 (HIV-1) (Neher and Shraiman 2011). The interaction between loci under selection and linked loci may also be viewed as clonal interference or negative linkage disequilibrium (Comeron et al. 2008). Clonal interference refers to the reduction in the rate of fixation of a beneficial mutation caused by beneficial mutations at other loci residing on different genomes, which are therefore in competition (Gerrish and Lenski 1998). This is equivalent to negative linkage disequilibrium between beneficial alleles because each allele is linked to a deleterious or neutral alternative allele at the other loci. Such interactions may be interpreted in terms of selection reducing Ne at linked loci (Keightley and Otto 2006) and clearly have important consequences for the efficiency of natural selection and the evolutionary maintenance of recombination (Felsenstein 1988; Kondrashov 1993; Keightley and Otto 2006). The nature of the interaction between selection and drift in HIV-1 has been controversial. On one hand, a very large viral census population size within a patient of 107–108 infected host cells (Chun et al. 1997) and a high mutation rate of 1025 per nucleotide per generation (Sanjuan et al. 2010) suggest that every possible point mutation occurs numerous times each viral generation (Coffin 1995). Together with evidence of strong selection by the immune system (Williamson 2003), this would suggest highly deterministic evolution (Coffin 1995; Overbaugh and Bangham 2001). On the other hand, the within-patient Ne during chronic infection has been routinely estimated to be only 103, suggesting that stochastic genetic drift is a powerful force in HIV-1 evolution (Leigh Brown 1997; Nijhuis et al. 1998; Rodrigo et al. 1999; Drummond et al. 2002; Seo et al. 2002; Achaz et al. 2004; Shriner et al. 2004b). In addition, variation among patients in the rate and pattern of the evolution of HIV-1 resistance to antiviral drugs has been attributed to the effects of genetic drift (Leigh Brown and Richman 1997; Nijhuis et al. 1998; Frost et al. 2000). Here, recent estimates of killing rates of infected cells by the immune system and patterns of fixation by viral mutants that escape immune recognition in the earliest stages of HIV1 infection are used to investigate the dynamics of the virus’s adaptation to the earliest immune responses. The early stages of infection by HIV-1 are considered an opportune time to control viral replication (Haase 2010; McMichael et al. 2010). With the transmission of the virus from one host to another, and especially with sexual transmission, the viral population goes through a severe population bottleneck, drastically reducing genetic variation (Mcmichael et al. 2010). Furthermore, events in early infection determine the viral population size in clinically asymptomatic chronic infection, which is proportional to the rates of disease progression and viral transmission (Ho 1996; Mellors et al. 1996; Quinn et al. 2000; Fideli et al. 2001). Consequently, recent studies have attempted to characterize, in unprecedented molecular detail, the earliest immune responses to the virus and the virus’s adaptive responses to this selection (Asquith et al. 2006; Goonetilleke et al. 2009; Fischer et al. 2010). Sexual transmission of HIV-1 is thought to typically involve a single viral genome (Keele et al. 2008; Abrahams et al. 2009; Salazar-Gonzalez et al. 2009; Fischer et al. 2010; Novitsky et al. 2011). This is followed by a rapid expansion of the virus population to a peak at 21–28 days postinfection (p.i.), known as peak viremia, and then an initially rapid decline in numbers, reaching a moderately stable level 1–2 orders of magnitude below the peak, known as the viral set point (De Loes et al. 1995; Ho 1996; Kinloch-McMichael et al. 2010). The earliest effective HIV-1-specific immune response detected is that of CD8+ T cells, first observed just prior to peak viremia (McMichael et al. 2010). Prior to this point there is no evidence of changes to amino acid frequencies in viral proteins due to immune selection. After this point, amino acid mutations that provide escape from immune detection (escape mutations) are observed to spread to fixation. Typically, eight such adaptive fixations occur throughout the early chronic phase of infection (McMichael et al. 2010), giving approximately one fixation every 22 days on average. Although immune responses and viral adaptation to these in early HIV-1 infection have been described in some detail, the dynamics of both are poorly understood. In particular, a transmission bottleneck of a single genome seems difficult to reconcile with extremely rapid adaptation. In addition, the targeting by CD8+ T cells of several different epitopes simultaneously may interfere with the adaptive response at each epitope. The present study investigates how the transmission bottleneck and potential clonal interference affect the adaptive dynamics of HIV-1 in early infection. Stochastic simulations showed that, given realistic values for important population genetic parameters that permit rapid adaptive evolution, a transmission bottleneck of a single genome is not inconsistent with the observed rapid adaptation to the earliest immune responses. However, selection under these conditions is predicted to result in only minor reductions in Ne at linked loci because the transmission bottleneck reduces Ne to such an extent that beneficial mutations spread to fixation sequentially rather than simultaneously.