Syst. Biol. 44(1)

NOTICE: This ms is the penultimate version, and differs in minor ways from the published version.--Editor

The AIDS Pandemic Is New, But Is HIV New?

David P. Mindell,1 Jeffrey W. Shultz,2 and Paul W. Ewald3

1Department of Biology and Museum of Zoology,
University of Michigan, Ann Arbor, Michigan 48109-1079, USA E-mail:

2 Department of Entomology, University of Maryland,
College Park, Maryland 20742, USA

3 Department of Biology, Amherst College, Amherst, Massachusetts 01002, USA

Keywords: AIDS, HIV phylogeny, direction of infection, retrovirus evolution

The AIDS pandemic is a new problem for humans, but it is unclear whether the human immunodeficiency virus (HIV) giving rise to AIDS is also new to humans. Either (1) HIV has recently infected humans, in which case we have a new virus and a new disease, or (2) HIV infected humans long ago (being mild and/or restricted in range until recently), in which case we have an old virus and a new disease. There are precedents for each scenario among known viruses causing diseases (see Shope and Evans, 1993). The new virus and old virus scenarios have profoundly different implications for understanding the mechanisms of HIV propagation and the etiology of AIDS, for combating AIDS, and, potentially, for efforts to prevent future epidemics. The terms "new" and "old" are ambiguous beyond denoting relative age; so, for purposes of this article we consider a new virus one that has infected its host species within the past 50 years or so.

The first view, that HIV has only recently contacted humans, entails recent cross-species transmission of a simian immunodeficiency virus (SIV) from one or more nonhuman primates, and represents the current conventional wisdom (Dietrich et al., 1989; Doolittle, 1989; Allan et al., 1991; Fox, 1992; Myers et al., 1992, 1993; Hirsch et al., 1993; Temin, 1993; Myers and Korber, 1994). However, some have suggested that certain rural African populations of humans may have been infected with an immunodeficiency virus for many decades, centuries, or even millennia (Montagnier, 1985; Hahn, 1990; McClure, 1990), and Ewald (1991, 1994) has described an evolutionary model in which virulent strains are placed at a selective advantage by higher rates of sexual partner change.

Understanding HIV origins is of general interest to systematists as well. Viruses evolve by descent with modification like any other group of organisms, and systematists will become increasingly involved in attempts to understand their complex histories as more of their DNA sequences become available. Systematists working on viruses need to consider distinctive features of viral evolution, including extremely high rates of molecular sequence evolution, subsequent high levels of within-population sequence variability (variously described as yielding species swarms or quasispecies), evolutionary rates that vary depending on the species of host and type of cell infected, potential for recombination when representatives of different viral lineages infect the same host cell, and potential biases in the sampling of host species.

Our objectives here are to (1) assess the evidence used in support of the "new virus" hypothesis, (2) present our own phylogenetic analyses of representative viral taxa, (3) estimate the most-parsimonious evolution of the character "virus host" in the study taxa, and (4) comment on methodological issues in the systematics of viruses.

Although it is not possible to reject either hypothesis, we conclude that any consensus favoring the "new virus" hypothesis is not justified on the basis of current evidence, and that the "old virus" hypothesis remains a viable alternative.

Phylogenetic Analyses of HIV Origins

Previous Studies

HIVs and SIVs are retroviruses, a group characterized by the ability to reverse-transcribe RNA into DNA. HIV and SIV genomes are about 10 kilobases in size and contain at least nine recognizable genes. Previous phylogenetic analyses of primary HIV and SIV lineages have used, variously, the pol and gag genes. The pol gene encodes reverse transcriptase and endonuclease enzymes, whereas gag encodes capsid protein (gag p24), which forms a shell around the viral RNA, and several internal proteins (gag p15, gag p18) functioning in viral reproduction (Stine, 1993; Hahn, 1994). The regions of pol encoding reverse transcriptase and endonuclease comprise the most slowly evolving regions of the genome (McClure et al., 1988). Previous studies have found HIVs and SIVs to form a monophyletic group, with immunodeficiency viruses from the domestic cat (FIV), sheep (VISNA), and horse (EIAV) being closely related to the primate immunodeficiency viruses (Doolittle et al., 1989; Yokoyama, 1991).

Because HIVs are parasites, the question of their origins includes two issues: specifically, (1) the phylogeny of the viruses, and (2) the history of virus transmission among host species. Parsimony may be used in addressing both of these questions. Parsimony is used in estimating the history of host shifts for viral taxa by minimizing the number of ad hoc assumptions of host shift on a phylogenetic tree. This minimization of ad hoc assumptions of host shift is justified on the same logical basis as the minimization of ad hoc assumptions of character convergence (homoplasy) in phylogenetic analyses using a parsimony criterion (see Farris, 1983). Further, most viruses are narrowly host-specific and unable to survive immune system surveillance in new host species. This is also consistent with a parsimony approach to assessing virus transmission history. It is important to note, however, that patterns of the phylogeny for extant viruses and their history of transmission need not be congruent. For example, the appearance of a virus from a particular host species as basal in a phylogeny could be the result of extinction of lineages from the true early host, which "gave" the virus to the current host recently, or a lack of sampling from the true early host species. Also, phylogenetic trees for viral taxa do not indicate whether a particular host species was a virus donor or a virus recipient. The trees simply denote hypothesized lineage-splitting events among viruses. For these reasons, phylogenetic trees for viruses must be interpreted with caution in assessing the history of virus host shifts.

Use of phylogenetic analyses to support the "new virus" hypothesis is common in the literature. For example, Myers et al. (1993:126) present a phylogenetic hypothesis for HIVs and SIVs based on a 648-nucleotide base section of the gag p24 gene (Fig. 1) and say that this "phylogenetic tree analysis... strongly supports the hypothesis of the simian origin of AIDS (the "new virus" hypothesis)". However, we see no such support in the tree itself. We note that there are both HIVs and SIVs on either side of the first bifurcation event within the tree, and that various HIVs could have either descended from or given rise to various SIVs. For example, sooty mangabey viruses (ancestral to SIVsmmh4 and SIVsmpbj in Fig. 1) might have given rise to HIV2s (HIV2rod, HIV2d205) or, conversely, an ancestral HIV2 lineage may have given rise to the sooty mangabey viruses. The tree itself is consistent with either scenario. Phylogenetic trees show purported sister relationships among extant lineages, but do not denote ancestor desendent relationships among those extant lineages.

Figure 1. Phylogenetic tree topology from Myers et al. (1993) based on immunodeficiency viral gag p24 sequences for 648 nucleotide positions. Tree is rooted at the midpoint of the greatest patristic distance. Host species and abbreviations for viruses are defined in the Appendix.

Rather, support for the "new virus" hypothesis is non-phylogenetic and circumstantial, being rooted in unsupported assumptions that (1) surveys of HIV presence in human blood samples collected before 1980 are based on reliable and sufficiently large samples, (2) high virulence denotes recency of the infection of the host species, (3) all HIVs are virulent, giving rise to AIDS, (4) SIVs are not virulent in their natural hosts. Regarding the first point, researchers point out that HIV seropositivity assays for human blood samples collected prior to the 1980s are largely negative, and that assays for blood samples collected during the 1980s in particular, show increasing levels of positivity (Grmek, 1990; see Myers et al., 1993 and references therein). This is consistent with the notion of the first human infection occurring during the middle portion of this century, but does not refute the alternative hypothesis that HIVs were present in one or more small and possibly isolated human populations not represented in pre-1980 or even pre-1959 blood samples (1959 is the date of collection for the earliest known seropositive sample). As dramatic as the seropositivity surveys are, their obvious geographic and quantitative sampling limitations compromise their ability to delineate the timing of infection. In a review of retrospective seropositivity surveys, Myers et al. (1993b) discuss the data available, which comes from just seven geographic locations across Africa involving hundreds or thousands of human blood samples. Not surprisingly, vast regions of Africa with millions of human inhabitants are not represented in retrospective seropositivity surveys. HIVs are lentiviruses which are characterized as a group by their potential for long periods of latency with no visible effects on hosts. HIVs present, though perhaps inconspicuous, in small isolated human populations or demes could readily have been missed by limited pre-1959 (or even pre-1980) blood samplings. Current high levels of virulence, particularly for HIV1s, may have been generated by recent changes in host behavior relating to virus transmission opportunities (Ewald, 1994).

Regarding the other three assumptions listed above, Doolittle (1989:339) notes the circumstantial view in pointing out that "primary hosts all seem to be healthy" whereas the likely secondary hosts are not healthy. Myers et al. (1992:373) stated,

"given the pervasiveness of SIVs in diverse African monkey populations, and the relative newness of HIV in human populations, the hypothesis of a recent simian origin of human AIDS through one or more events of cross-species transmission has gained widespread acceptance over the past few years."
However, if the "relative newness of HIV in human populations" is "given" how can one reach any other conclusion? Often, authors simply note the sister relationship between viruses isolated from different host species and presuppose the direction of infection to be from nonhumans to humans. The assumptions mentioned should not be accepted uncritically and are discussed below. Other researchers have been more cautious in inferring direction of infection from phylogenetic analyses. For example, Hirsch et al. (1989:391) noted that their
"data cannot exclude the possibility that HIV2 from a human was passed to a sooty mangabey and subsequently evolved as SIVsm... [and that] sequences of older HIV2 and SIV isolates (from mangabeys or other species) are required to resolve these issues."

Although direction of infection cannot be inferred from a sister relationship between two viruses from different host species, accurate tree topology is crucial in estimating the number of host shifts that have occurred, and the most-parsimonious sequence of host shift events. Many aspects of HIV and SIV relationships are poorly resolved, particularly the earliest divergences involving five lineages: (1) HIV1s/SIVcpz, (2) HIV2s/SIVsm/SIVmac, (3) SIVmnd, (4) SIVagms, and (5) SIVsyk. HIVs, as currently named, are clearly not monophyletic. The two primary HIV types, HIV1 and HIV2, each include representative strains (or taxa) that are more closely related to one or more SIVs than they are to other HIVs. As a corollary, SIVs are also not monophyletic. Whether HIV1s are monophyletic and HIV2s are monophyletic has been less clear. SIVcpz, previously placed as sister to all HIV1s (Huet et al., 1990), may belong inside an HIV1 clade when divergent HIV1s are included in analyses. Similarly, SIVs from sooty mangabeys and macaques are often, though not always, placed within the HIV2 clade.

Clearly, there have been multiple host species shifts by the viruses, and, given a nonsister relationship between the HIV1 and HIV2 types, proponents of the "new virus" hypothesis must invoke at least two recent and independent human infections with quite different viruses, each capable of causing AIDS. HIV1s and HIV2s are about 40% different in nucleotide sequence over the entire genome and differ in presence of some accessory genes. HIV1 and SIVcpz contain two open reading frames, termed vpr and vpu, whereas HIV2/SIVmac/SIVsm lack the vpu gene but contain another gene termed vpx. Rather than requiring at least two recent human infections with quite different viruses, the "old virus" hypothesis requires only one cross-species infection of humans.

Current Study

Methods.---We based our phylogenetic analysis of HIVs and SIVs on the relatively slowly evolving pol gene and gag p24 region of 28 viral isolates obtained from the Los Alamos HIV sequence data base (Myers et al., 1993) and GenBank (NCBI Entrez release 7.0). These include isolates from both primary HIV types (HIV1 and HIV2), wild-caught African primates (SIVcpz from chimpanzee, Pan troglodytes; SIVagms from African green monkeys, Cercopithecus aethiops and C. pygerythrus; SIVsyk from Sykes' monkey, Cercopithecus mitis albogularis; SIVsms from sooty mangabeys, Cercocebus atys; and SIVmnd from mandrill, Mandrillus sphinx), captive primates (SIVmacs from Macaca mulatta, M. nemistrina, M. arctoides), and a domestic cat (FIV from Felis cattus We combine regions from both the pol and gag genes in emphasizing congruence among characters as support for sister-group relationships (Kluge, 1989). Inferred amino acid sequences were aligned using Clustal V (Higgins et al., 1992) with fixed and floating gap penalties set to 10. Those sequence regions that did have gaps were excluded from phylogenetic analyses, leaving unambiguously aligned regions spanning 2,763 nucleotide base positions in pol and 698 base positions in gag p24. The amino acid alignment served as a template for aligning the corresponding nucleic acids.

Phylogenetic analysis was conducted on the aligned, edited sequences using PAUP, 3.1.1 (Swofford, 1993). Given the large number of terminal taxa and large size of the data matrices, we used a heuristic search algorithm with the tree bisection-reconnection branch swapping procedure. Because the heuristic search does not explore all possible topologies to find the shortest tree, we repeated the search 100 times for each analysis. Each search was initiated using a different randomly constructed starting topology, reducing the possibility that the algorithm will find a local parsimony optimum rather than the universal optimum for a particular data set.

Inaccuracies in phylogenetic analyses stem from an inability to discriminate homologous similarity (due to descent) from homoplasious (convergent or parallel) similarity. Two steps for molecular systematists in making this discrimination are choosing genes that are not saturated with change (having multiple substitutions at individual base positions), and using a data-set-dependent, a priori weighting scheme to place greater weight on those characters whose rates of change are relatively slow, as similarities among such characters will tend to include less homoplasious similarity (Mindell and Honeycutt, 1990; Hillis et al., 1993). Toward this end, we have calculated the number of third codon position transition and transversion changes for pol and gag p24 DNAs between representative HIV and SIV lineage pairs (Table 1). We expect any homoplasious similarity to be found particularly in third codon position transitions. Third codon positions tend to have faster rates of change due to the greater number of synonymous substitutions that are possible there, relative to first and second codon positions, and a tendency for transition substitutions to accumulate more rapidly than transversions has long been known (Brown et al., 1982; Graur, 1985).

Table 1.  Pairwise differences between viral taxa based on
1,153 third codon positions for DNA sequences from the
pol and gag p24 regions combined (see Appendix
for sequence sources).  Numbers of inferred transitions 
are below the diagonal, and numbers of inferred transversions
are above the diagonal.

               1    2    3    4    5    6    7    8    9

1. HIV1eli    ---    9  305  310  307  302  336  320  411
2. HIV1ndk     51  ---  308  313  310  303  341  321  412
3. HIV2ben    301  309  ---   15  312  297  328  327  414
4. HIV2d194   305  309  146  ---  319  300  333  338  417
5. SIVagmtyo  299  312  278  293  ---  137  337  299  431
6. SIVagm3    294  295  293  303  328  ---  323  282  426
7. SIVsyk     287  284  299  307  287  270  ---  338  403
8. SIVmndgb   280  276  290  295  291  288  285  ---  404
9. FIV14      251  244  282  282  266  248  290  243  ---

If DNA sequence characters are saturated with change, the number of inferred changes will not increase as divergence time increases between taxon pairs. That is, the correspondence between time and increasing amounts of sequence divergence will break down. We can make such comparisons among our study taxa by noting that divergences within subsets of HIV1s (excluding HIV1ant70 and HIV1mvp) and within HIV2s (excluding HIV2d205 and HIV2uc1) are more recent than divergences among the primary HIV/SIV lineages (HIV1, HIV2, SIVagm, SIVmnd, SIVsyk), which in turn are more recent than the divergence of their common ancestor from FIV (Doolittle et al., 1989; Yokoyama, 1991). Table 1 indicates that pol and gag p24 third codon position transitions are relatively saturated with change, as pairwise comparisons with FIV show no more changes than do comparison among the primary HIV/SIV lineages. Conversely, third position transversions are relatively unsaturated with change, as comparisons with FIV consistently show more changes than do other comparisons. Interestingly, third position transition comparisons with FIV are actually smaller than more recent divergences among HIV/SIV lineages, as would be expected when more slowly accumulating transversions begin to overwrite transitions. Similar comparisons to those in Table 1 for codon positions one and two (data not shown) indicate nonsaturation for both transitions and transversions at those positions. Thus, we give third position transitions a weight of zero, a priori, in our phylogenetic analyses to reduce the confounding effects of nonhomologous similarity.

The relative support for each node within the minimal-length topology was evaluated using the support index (Bremer, 1988; Källersjö et al., 1992), which denotes the difference in length between the most-parsimonious tree and the shortest tree in which the particular node (clade) is not present. To estimate the support index for a particular clade, we constructed a constraint tree in which the clade is the only resolved relationship among the study taxa and then used 10 replicate heuristic searches, with random stepwise addition of taxa, to find the shortest fully-resolved topology in which that relationship was not present.

Figure 2. Most-parsimonious phylogenetic tree for 27 primate and 1 feline immunodeficiency viruses based on the combined DNA sequences from the pol (2,763 nucleotide positions) and gag p24 (698 nucleotide positions) genes. Third codon position transitions were given a weight of zero, a priori, to reduce homoplasious similarity within the data set (see Table 1). Numbers along nodes are support indices denoting the number of additional steps needed to break the node (Bremer, 1988; Källersjö et al., 1992). The tree length is 6,226 steps, excluding uninformative characters. Nucleotide sequences were obtained from the Los Alamos HIV database and GenBank (see Appendix for accession numbers, abbreviation definitions and virus host species). Isolates from wild-caught primates include those from chimpanzee (SIVcpz), sooty mangabeys (SIVsmm9, SIVsmmh4), African green monkeys (SIVagm3, SIVagm155, SIVagmtyo, SIVagm677), Sykes' monkey (SIVsyk) and mandrill (SIVmndgb). Isolates from captive primates include those from rhesus macaque (SIVmm239), pig-tailed macaque (SIVmne) and stump-tailed macaque (SIVstm). The feline immunodeficiency virus isolate FIV14 was treated as an outgroup to the primate immunodeficiency viruses. Nodes A and B denote taxa named primate immunodeficiency virus 1 (PIV1) and PIV2, respectively.

Results.---We found a single most-parsimonious tree in analysis of the combined pol and gag p24 DNAs, giving third codon position transitions zero weight, and we consider this our current best estimate of phylogenetic relationships for the viruses (Fig. 2). Nodes within the tree vary in their degree of support, based on the indices reported. SIVcpz is sister to six HIV1s, and two other HIV1s (HIV1ant70, HIV1mvp) are basal to this group. This placement of SIVcpz inside a larger HIV1 clade suggests that there was either one viral host shift from humans to chimp or two host shifts from chimp to humans. Myers et al. (1992:373) have suggested that HIV1ant70 "may ultimately be interpreted as yet another SIV form." This interpretation might seem to reduce the likelihood that a human to chimp transfer has occurred, in that no HIV1 would diverge "prior to" the divergence of SIVcpz. However, interpretation of HIV1ant70 as an SIV (and an aberrant colonist in humans) connotes that HIV1ant70 represents a rare divergent viral lineage in humans, and recent evidence shows this not to be the case. Nkengasong et al. (1993) found that blood samples from 16 humans from Cameroon and Gabon Africa reacted positively with HIV1ant70 peptides in enzyme-linked immunosorbent assays (ELISA), indicating HIV1ant70 to be endemic among HIV1 seropositive individuals in these two countries and, thus, more common than previously thought. Further indicating endemism, HIV1mvp (also from a Cameroonian) is placed as sister to HIV1ant70 at a strongly supported node in our parsimony analysis. On the basis of current evidence, we are left with the indication of human to chimp viral transfer as more parsimonious than the reverse (one host species shift versus two; Fig. 3). Publication of sequences from additional chimp SIVs will help resolve this issue.

Figure 3. Most parsimonious evolution of the character "viral host" based on our most parsimonious tree topology (presented in Fig. 2). Changes in viral host are shown (patterns and shadings of branches) invoking the fewest possible number of shifts. Branches shown as equivocal denote that two or more character states (viral hosts) are possible without altering the minimum number of changes invoked. Because species distinctions among African green monkeys are unclear, we have conservatively listed the four SIVagms shown (here and in the Appendix) as representing one host species, although we note that some might recognize three (SIVagm155, SIVagmtyo, SIVagm3) as being from Cercopithecus pygerythrus and one (SIVagm677) as being from C. aethiops. Including two distinct African green monkey species here, however, does not result in any additional changes in the character "viral host" elsewhere in the tree.

Our analysis shows HIV2s to be polyphyletic, by placement of two HIV2s (HIV2d205, HIV2uc1) as sister to a clade including HIV2s and SIVs from sooty mangabeys and macaques (Fig. 2), with the latter clade (excluding HIV2d205 and HIV2uc1) being moderately well supported. Like the situation with HIV1s and SIVcpz, we are left with the more parsimonious scenario indicating virus transmission from humans (HIV2s) to sooty mangabeys and macaques (Fig. 3). Our tree places SIVmnd and SIVsyk as basal to the other HIVs and SIVs, although those nodes have relatively moderate levels of support. SIVagms are sister to HIV2s/SIVsms/SIVmacs in agreement with most previous analyses, but differing from Doolittle (1989). Based on current evidence our topology is preferable due to its incorporation of more character evidence and a data-set-dependent, a priori weighting scheme which reduces effects of homoplasious similarity.

Our tree (Fig. 2) differs from that of Myers et al. (1993; Fig. 1) in not indicating HIV2 monophyly and in the relative placement of SIVmnd and SIVsyk. Myers et al. used unweighted gag p24 region sequences alone, which can be seen to include homoplasious similarity based on our pairwise comparisons (Table 1). Their exclusion of pol sequences weakens their analyses, as pol includes the most conserved, and hence most phylogenetically reliable, sequences in the genome. They also used midpoint rooting which should only be used as a last resort, when no suitable outgroup is available. The midpoint method places the basal node for any tree arbitrarily along the longest path connecting any pair of taxa. This assumes constant rates of character change across taxa without justification, and differences in rate sufficient to affect placement of the basal node will change sister relationships shown in the tree. Although they acknowledge the basal position of HIV1ant70 relative to other HIV1s and SIVcpz, they do not include HIV1ant70 (or its sister taxon HIV1mvp) in their analysis. This exclusion and their diagnosis of HIV2s as monophyletic, despite numerous analyses contradicting HIV2 monophyly (e.g., Dietrich et al., 1989; Gao et al., 1992; Myers et al., 1992; Barnett et al, 1993) allow Myers et al. (1993) to consistently favor the "new virus" hypothesis and ignore the alternative "old virus" hypothesis.

Minimum Evolution of the Character "Virus Host"

We used our tree topology (Fig. 2) to infer the most-parsimonious pathway of cross-species infection within the primate immunodeficiency viruses. Host species for each of the 28 viral isolates was coded as a character state, and the minimum number of changes among alternative states were then distributed on the tree (Fig. 3; using MacClade; Maddison and Maddison, 1992). As mentioned above, within the HIV1/SIVcpz clade, human is shown as the ancestral host species. Similarly, within the HIV2/SIVsm/SIVmac/SIVagm clade, human is also shown as the ancestral host species. Basal character state for the entire HIV/SIV clade is "equivocal." That is, two or more different states (virus hosts) could be invoked without altering the number of changes on the overall tree. Recently, Myers and Korbin (1994) have been able to include a second SIV from a chimpanzee (cpzant) in their phylogenetic analyses (although this sequence is currently unpublished), and their analysis places cpzant as sister to the clade including the HIV1s and SIVcpz. Even with inclusion of cpzant in this position on our tree diagnosing change in the character "virus host" (Fig. 3) human remains the most-parsimonious ancestral host for the HIV1/SIVcpz/cpzant clade as well as for the HIV2/SIVsm/SIVmac/SIVagm clade.

Obviously, this analysis does not resolve the sequence of host species shifts, and we make no such claim. Inference from this analysis is confounded by sampling bias, as many more viruses have been sequenced from humans than from any other primate species. Human is shown as the ancestral host for the two clades mentioned above because two of the five most divergent primate immunodeficiency viruses are isolated from humans, whereas each of the other three divergent viral lineages is unique to a different host species. If, for example, further sampling of SIVs from African green monkeys or mandrills were to uncover taxa within each of those lineages as divergent as HIV1 and HIV2 types are from each other, that would alter the character-state changes as inferred in Figure 3. We point out, however, that some such sampling has been done for African green monkeys from disparate locales in Africa, and divergences as great as those seen between HIV lineages are not observed. SIVagms from western Africa (e.g., Senegal) and from eastern Africa (Kenya, Ethiopia; as included in our study), form a monophyletic group (Allan et al., 1991), in contrast to HIV1s from eastern and central Africa and HIV2s from western Africa which do not form a monophyletic group (Fig. 2). We also note that changing the tree topology in Figure 3, such that SIVagms are basal to the entire HIV1/HIV2 clade, does not change the diagnosis of ancestral host species within either the HIV1 or the HIV2 clade. Our point in presenting this analysis is simply to show that the current evidence does not support the "new virus" (new in humans) hypothesis.

Virulent Viruses Are New and Nonvirulent Viruses Are Old: An Oversimplification

It has long been thought that mutualistic associations between parasites and hosts are more stable evolutionarily than are parasitic or destructive ones (Smith, 1939; Burnet and White, 1972). Parasites that quickly kill their hosts will provide little opportunity for their progeny to successfully colonize new host individuals, and, hence, may go extinct. This observation is reflected in a widely claimed tendency for viruses to evolve toward avirulence (particularly found in medical texts), and the following quote from Dubos (1965), "Given enough time a state of peaceful coexistence eventually becomes established between any host and parasite."

This has led to an oversimplified prescription that virulent viruses are new and that nonvirulent viruses are old. It is becoming increasingly evident, however, that there can be great variation in the timing and direction of virulence change. Just as a virus can change in its effects from pathogenic to benign, it can also change from benign to pathogenic, depending on natural selection and the effects of changing replication rates on the fitness of the virus (Ewald, 1994). As described by May (1993:66),

"There is no generalization [regarding change in virulence for many or most viruses]. The virus may become less virulent, more virulent, or exhibit unchanging virulence; the virus may become less transmissible, more transmissible, or show unchanging transmissibility. All of this depends on the tradeoffs among virulence, transmissibility, and the cost of resistance, which are also constrained by the nature of the host-pathogen association."
Examples of viruses that have shown an increase in virulence at one time over another include influenza A (see Langmuir and Schoenbaum, 1976; Webster, 1993) and myxoma virus (Dwyer et al., 1990; Fenner and Kerr, 1994). Levin and Pimentel (1981) simulated the evolution of a simple system with one host species susceptible to two viral lineages, one of which is more virulent than the other. They found no general trend toward avirulence, and that increased virulence may be favored when it increases transmission rate.

In keeping with the older view of virulence, apparent mildness of SIV in sooty mangabeys and African green monkeys has been attributed to an old virus/host association. However, one need not invoke an old association to explain avirulence. The mildness could be attributed to relatively low rates of sexual partner change. Sooty mangabey females have apparent low rates of sexual partner change, restricting copulation to a few males during their estrus period, and not copulating during a prolonged period of maternal care (T. Butynski, pers. comm.). African green monkey females are sexually receptive only seasonally and in groups controlled by a single male (Fedigan and Fedigan, 1988). Thus, potential for rapid spread of SIV through these species appears limited, and viral strains with a rapid replication rate (compromising their host's immune system and health) will have little selective advantage.

Results of laboratory infections of chimps with HIV1 are also inconsistent with the supposition that low virulence denotes a long virus/host association. No AIDS-like disease has been observed among over 100 chimps that have been experimentally infected with HIV1, nor among the minority that have remained infected for 5 to 10 years (Fultz, 1993; Johnson et al., 1993). Further, in chimps in which HIV1s have become established and have increased in numbers, the capability for successful infection of chimp blood cells has increased (Gendelman et al., 1991; Watanabe et al., 1991), indicating a potential for virulence to increase over time.

The avirulence of SIVs in sooty mangabeys, chimps, mandrills, and other species also remains open to question. A severely ill individual would not last long in nature, compared to infected but asymptomatic or recovered individuals that could complete normal life spans. For this reason, snapshot seropositivity surveys of existing populations may underestimate the frequency of infections associated with severe illnesses. A highly virulent, molecularly cloned, SIV strain originally from a sooty mangabey (SIVsmpbj; Dewhurst et al., 1990) causes death in experimentally infected sooty mangabeys and macaques (Fultz, 1993), whereas the original parental virus caused a chronic AIDS-like syndrome in macaques and only asymptomatic infection in sooty mangabeys. This belies the notions that SIVs in their "natural" host species are exclusively avirulent and that they cannot become more virulent over time. The laboratory transmission that has favored increased virulence in this SIV variant is similar to that proposed for HIV. Rapidly reproducing and severe variants can be maintained if the rapid reproduction provides them with a fitness advantage over more slowly reproducing strains.

The "old virus" hypothesis holds that primitive HIVs may have had low virulence and were maintained in a population that displayed low levels of sexual partner change, perhaps in a rural area. This leads to a prediction that some early divergent, low virulence viral strains could still be extant in such populations, and viral isolates have been discovered. HIV2d205 and HIV2uc1 represent an early divergent lineage within HIV2/SIVsm clade and were obtained from asymptomatic individuals from rural Ghana and Ivory Coast, respectively. HIV2uc1 is entirely noncytopathic and readily neutralized by sera from HIV2 infected individuals (Barnett et al., 1993).

Issues in Systematics of Viruses

Conflicting Topologies for Viral Gene-trees May or May Not Indicate Recombination Among Viral Lineages

When two or more individual viruses penetrate a particular host cell and begin nucleic acid replication, the potential exists for recombination among the viral genomes due to a replicase enzyme slipping from one viral genome template to another (Coffin, 1979; Hu and Temin, 1990). Recombination among HIV variants has been found to occur in vitro (Clavel et al., 1989) and has been inferred or suggested to occur in vivo based on (1) observed viral sequences having a mixture of components from formerly distinct lineages (Howell et al., 1991) and (2) conflicting tree topologies based on phylogenetic analyses of different genes (Li et al., 1988; McClure et al., 1988; Gao et al., 1992; Myers et al., 1993). Conflicting phylogenies based on different genes, however, may also stem from differential success in phylogenetic analyses. That is, one tree might be accurate whereas the other is not, despite absence of any recombination. In analyzing different genes and different types of substitutions changing at different rates, systematists often find different data sets supporting different trees. This may stem from differential success in distinguishing homologous from homoplasious similarity (distinguishing signal from noise) in the different data sets (Farris, 1983; Swofford and Olsen, 1990; Hillis, 1991; Mindell, 1991). In considering recently diverged taxa, this might also stem from the confounding effects of within-population variation on analyses among higher level taxa (Neigel and Avise, 1986; Avise, 1989). Within-population variation for retroviruses can be extreme, depending on which gene regions are considered (Zarling and Temin, 1976; Holmes et al., 1992). Systematists working on viruses will need to consider these possibilities prior to invoking recombination to explain such conflicts in gene tree topologies.

High Extinction Rates and Sampling Problems

Because of their short generation times, large numbers of progeny, and high mutation rates, viruses have a great capacity for rapid diversification. Consequently, there is also a great capacity for lineage extinction events, and this has implications for studies of phylogeny and the history of host shifts. Inclusion of fossil taxa has been seen to alter inferred phylogenetic relationships in studies of plants and animals (Doyle and Donoghue, 1987; Gauthier et al., 1988) and the same can be expected to occur in analyses of viruses. Obtaining more samples of extant taxa and, where possible, extinct taxa on the basis of viral sequences from preserved tissues will help in understanding the effects of this taxon inclusion/exclusion problem. Peter Houde and colleagues (ms in review) at New Mexico State University are working on amplifying and sequencing SIVs from primate museum study skins and, in the process, of identifying new host species and minimum dates for host species infection.

Clear distinction must be made between phylogeny of the viruses and the history of their distribution, as the two need not be congruent. Divergent lineages such as SIVsyk and SIVmnd may appear basally in a phylogenetic analysis (as in Fig. 2), without Sykes' monkey or mandrill being old (early) host species. That basal appearance could be the result of the extinction of lineages from the true early host, which "gave" the virus to current hosts relatively recently, or of a lack of sampling from the true early host species. Just as the true phylogeny for any set of taxa is unknowable (unless directly observed) and can only be inferred, the true history of viral host-shifts can also only be inferred. For this reason, attempts to determine the natural or ancestral host of a virus will always be susceptible to biases from "unobserved" host shifts, related to high extinction rates for viral lineages and the inevitably small samples available for analysis.

Rate Heterogeneity

RNA viruses like the primate immunodeficiency viruses, with base substitution rates averaging 10-3 per site per year, often have rates of evolution exceeding that of their eukaryotic host species by a million fold or more (Holland, 1992). This is a result of the high error rate of the viral-encoded reverse transcriptase and the lack of misincorporation repair mechanisms. Although this rapid rate of viral sequence change is not qualitatively different from that encountered by systematists working on other taxa, there are several sources of rate variability among viruses that are not currently recognized in other taxa. Retroviruses undergo replication involving three different enzymes with variable error rates. In the viral stage (in the host cell cytoplasm), retroviral RNA is transcribed into retroviral DNA by reverse transcriptase having a high error rate, as mentioned above. In the proviral stage (in the host cell nucleus), retroviral DNA is replicated by the host cell's DNA polymerase which is less error prone and entails efficient mutation repair mechanisms. Subsequently, the proviral DNA is transcribed back into RNA by the host cell's RNA polymerase. The error rate for cellular RNA polymerase is not well known, though it may be similar to that of reverse transcriptase (Coffin, 1991). Thus, there is the potential for closely related viral lineages to differ in their rates of change, due to experiencing different amounts of high error (reverse transcriptase and cellular RNA polymerase) and relatively low error (cellular DNA polymerase) replication. These differences will tend to vary with changing virulence, as low virulence entails longer proviral times and fewer replication cycles, and high virulence entails greater amounts of low fidelity reverse-transcription. A further consequence of the proviral stage is the opportunity for recombination with cellular genes and the possible addition of new sequences into the retrovirus genome (Bishop and Varmus, 1985).

We expect that rates of retroviral change may vary depending upon the particular host species infected, given that different animal species may show different rates of molecular sequence evolution (e.g., Britten, 1986; Li and Tanimura, 1987; Avise et al., 1992; Martin and Palumbi, 1993) and that retroviruses use the host's replication machinery. Rates might also vary depending on the particular cell type infected. This follows from observations of correlation between rates of sequence change and metabolic rate (rate of oxygen metabolism) and of differences in metabolic rate for different cell and tissue types. Underlying the correlation with metabolic rate is apparent DNA damage due to oxygen-derived free radicals (Joenje, 1989; Shigenaga et al., 1989). Oxidative damage potentially influences rates of sequence evolution across all taxa; however, the generally fast rate of retroviral evolution accentuates these and other effects to a greater degree than is seen in other organisms.

Viral sequences also show patterns of rate heterogeneity correlated with codon position and transition/transversion differences as seen in other organisms (Graur, 1985; Table 1). We have sought to account for the effects of some of these in our current analyses with an a priori weighting scheme. The effect on phylogeny of other rate heterogeneity sources mentioned (three different replication enzymes, host and cell specific effects) are poorly known at present, although potentially significant. In light of the fast pace of primary sequence evolution and subsequent low levels of sequence similarity among many viral taxa, the more slowly evolving features of secondary and tertiary structures for encoded proteins may prove useful for alignment and phylogenetic analyses in the future (see Johnson et al., 1990; Eickbush, 1994).

Estimates of lineage divergence times assume rate constancy over time and will be distorted to the extent that rate heterogeneity exists for the characters analyzed. Not surprisingly, this has given rise to incongruent estimates by different researchers. Estimates for divergence time between HIV1s and HIV2s range from 40 (Smith et al., 1988) to 600-1200 years ago (Eigen and Nieselt-Struwe, 1990).

Naming Virus Clades Rather Than Grades

Up to this point the names used for primate immunodeficiency virus taxa have been based on the host species in which the viruses have been found. What these represent, then, are viral grades based on their distribution. Named grades are less desirable than named clades given the primary purpose of taxonomy to communicate results of evolutionary history (phylogenetic analysis) using a system of names. As more viral taxa become known and are added to phylogenetic analyses, viral taxonomy can be revised to provide a more accurate history of their evolution. Such a revision can discourage misconceptions or premature conclusions regarding lineage origins. For example, the association by name of HIV1s and HIV2s suggests (to systematists) a common origin for them to the exclusion of other immunodeficiency viruses, and as discussed above, this appears not to be the case. Similarly, the taxon "SIVs" gives the unsupported impression that all SIVs are more closely related to each other than they are to various HIVs. de Queiroz and Gauthier (1992) have described useful conventions for naming taxa, the most basic of which is that all names refer to clades.

We can begin by recognizing clades in Figure 2 as taxa. The clade that is descendent from the hypothetical common ancestor at node A in Figure 2 includes all the known HIV1s and SIVcpz, and can be called primate immunodeficiency virus 1 (PIV1). The clade that is descendent from the hypothetical common ancestor at node B in Figure 2 includes all the known HIV2s and SIVs from sooty mangabeys and macaques, and can be called PIV2. Other taxa may be recognized in a similar fashion as the need arises. Members of the taxon PIV1 have an apparent synapomorphy (shared derived character) in the presence of the vpu accessory gene, whereas members of PIV2 uniquely possess the accessory genes vpr and vpx in combination (Gibbs and Desrosiers, 1994). In a nonphylogenetic taxonomy such characters might have been used to define taxa. However, in our proposed phylogenetic taxonomy, such characters are used in diagnosing clades, but not in defining them (determining inclusion or exclusion of species or taxa). Rather, taxon names are defined in terms of common ancestry and relationship.


Evidence currently available does not support the popular view (the "new virus" hypothesis) that HIVs (or PIVs to use our term introduced above, Figure 2) have recently colonized humans and that PIVs in humans are recent descendants from one or another of the PIV lineages known from nonhumans. Phylogenetic trees show only sister relationships for extant taxa, not ancestor-descendant relationships for extant taxa. Use of our phylogenetic hypothesis and a parsimony criterion to estimate the fewest number of host species shifts (that is, to diagnose changes in the character "viral host") indicates humans to be the ancestral host species for a clade including SIVcpz from chimpanzee and for a clade including SIVsms from sooty mangabeys. We specifically do not claim that the latter analysis resolves the issue of ancestral host, however, in light of potential sampling biases. Our point is to show that current evidence does not support the "new virus" hypothesis. Support for the "new virus" hypothesis then devolves to unjustified assumptions that pre-1959 human blood samples testing negative for PIV presence successfully represent all human populations and demes potentially harboring PIVs, and that new viruses are virulent and old viruses are mild. Small human populations with dormant PIVs may readily have been missed by limited sampling, and the assumption that new viruses are virulent and old viruses are mild ignores the ability of natural selection to affect an increase, a decrease or stasis in virulence over time. Even if the latter assumption were valid, inferred newness of PIV infection of humans is contradicted by discovery of noncytopathic HIV2uc1 and relatively low virulence (longer latency and asymptomatic periods) of PIV2s in rural human populations having relatively low rates of sexual contact among individuals.

Retroviral evolution challenges systematists with a variety of distinctive and potentially confounding features, including (1) extremely fast rates of molecular sequence evolution (due to short generation times, large numbers of progeny, and low fidelity replication), (2) evolutionary rate heterogeneity within and among virus sequences (due to potential host specific and cell-type specific rate differences, and variable use of three different replication enzymes having variable error rates, and (3) potential for genetic recombination among different lineages infecting the same cell, complicating character homology determinations. Improved understanding of these features and greater sampling of primate host species will enhance future studies of immunodeficiency virus phylogeny, and may entail revision of current hypotheses of relationship.


We would like to thank J. J. Bull, M. J. Donoghue, D. M. Hillis, T. D. Kocher, M. M. Miyamoto, G. Myers, and T. W. Scott for valuable discussion and comments on various drafts of this manuscript. DPM's work was supported by the National Science Foundation (BSR-9019669), and JWS's work was supported by an Alfred P. Sloan Post-doctoral Fellowship.


Allan, J. S., M. Short, M. E. Taylor, S. Su, V. M. Hirsch, P. R. Johnson, G. M. Shaw, and B. H. Hahn. 1991. Species-specific diversity among simian immunodeficiency viruses from African green monkeys. J. Virol. 65:2816-2828.

Avise, J. C. 1989. Gene trees and organismal histories: A phylogenetic approach to population biology. Evolution 43:1192-1208.

Avise, J. C., B. W. Bowen, T. A. Lamb, B. Meylan, and E. Bermingham. 1992. Mitochondrial DNA evolution at a turtle's pace: Evidence for low genetic variability and reduced microevolutionary rate in the testudines. Mol. Biol. Evol. 9:457-473.

Barnett, S. W., M. Quiroga, A. Werner, D. Dina and J. A. Levy. 1993. Distinguishing features of an infectious molecular clone of the highly divergent and noncytopathic human immunodeficiency virus type 2 UC1 strain. J. Virol. 67:1006-1014.

Bishop, J. M., and H. Varmus. 1985. Function and origins of retroviral transforming genes. Pages 249-356 in RNA tumor viruses: Molecular biology of tumor viruses, (R. Weiss, N. Teich, H. Varmus, and J. Coffin, eds.). 2nd edition, Volume 2 [supplements and appendixes] Cold Spring Harbor Laboratory, Cold Spring Harbor, New York.

Bremer, K. 1988. The limits of amino-acid sequence data in angiosperm phylogenetic reconstruction. Evolution 42:795-803.

Britten, R. J. 1986. Rates of DNA sequence evolution differ between taxonomic groups. Science 231:1393-1398.

Brown, W. M., E. M. Prager, A. Wang, and A. C. Wilson. 1982. Mitochondrial DNA sequences of primates: Tempo and mode of evolution. J. Mol. Evol. 18:225-239.

Burnet, F. M., and D. O. White. 1972. Natural history of infectious disease, 4th edition. Cambridge Univ. Press, Cambridge, England.

Clavel, F., M. D. Hoggan, J. R. Willey, K. Strebel, M. A. Martin, and R. Repaske. 1989. Genetic recombination of human immunodeficiency virus. J. Virol. 63:1455-1459.

Coffin, J. M. 1979. Structure, replication, and recombination of retrovirus genomes; some unifying hypotheses. J. Gen. Virol. 42:1-26.

Coffin, J. M. 1991 Retroviridae and their replication. Pages 645-708 in Fundamental virology (B. N. Fields and D. M. Knipe, eds.). Raven Press, New York.

De Queiroz, K., and J. Gauthier. 1992. Phylogenetic taxonomy. Annu. Rev. Ecol. Syst. 23:449-480.

Dewhurst, S., J. E. Embretson, D. C. Anderson, J. I. Mullins, and P. N. Fultz. 1990. Sequence analysis and acute pathogenicity of molecularly cloned SIVsmm-PBj14. Nature 345:636-640.

Dietrich, U., M. Adamski, R. Kreutz, A. Seipp, H. Kuhnel, and H. Rubsamen-Waigmann. 1989. A highly divergent HIV-2 related isolate. Nature 342:948-950.

Doolittle, R. F. 1989. Immunodeficiency viruses: The simian-human connection. Nature 339:338-339.

Doolittle, R. F., D. -F. Feng. M. S. Johnson, and M. A. Mcclure. 1989. Origins and evolutionary relationships of retroviruses. Q. Rev. Biol. 64:1-30.

Doyle, J. A., and M. J. Donoghue. 1987. The importance of fossils in elucidating seed plant phylogeny and macroevolution. Rev. Palaeobot. Palynol. 50:63-95.

Dubos, R. 1965. Man adapting. Yale Univ. Press, New Haven, Connecticut.

Dwyer, G., S. A. Levin, and L. Buttel. 1990. A simulation model of the population dynamics and evolution of myxomatosis. Ecol. Monogr. 60:423-447.

Eickbush, T. H. 1994. Origin and evolutionary relationships of retroelements. Pages 121-157 in The evolutionary biology of viruses (S. S. Morse ed). Raven Press, Ltd., New York.

Eigen, M., and K. Nieselt-Struwe. 1990. How old is the immunodeficiency virus? AIDS 4 (suppl. 1):s85-s93.

Ewald, P. W. 1991. Transmission modes and the evolution of virulence, with special reference to cholera, influenza and AIDS. Human Nature 2:1-30.

Ewald, P.W. 1994. The evolution of infectious disease. Oxford Univ. Press, New York.

Farris, J. S. 1983. The logical basis of phylogenetic analysis. Pages 7-36 in Advances in cladistics (N. W. Platnick and V. A. Funk, eds.). Columbia Univ. Press, New York.

Fedigan L., and L. M. Fedigan. 1988. Cercopithecus aethiops: a review of field studies. Pages 389-411 in A primate radiation: Evolutionary biology of the African guenons (A. Gautier-Hion, F. Bourliere, J. -P. Gautier, and J. Kingdon, eds.). Cambridge Univ. Press, Cambridge, England.

Fenner, F., and P. J. Kerr. 1994. Evolution of the poxviruses, including the coevolution of virus and host in myxomatosis. Pages 273-292 in The evolutionary biology of viruses (S. S. Morse ed). Raven Press, Ltd., New York.

Fox, C. 1992. Possible origins of AIDS. Science 256:1259-1260.

Fultz, P. N. 1993. Nonhuman primate models for AIDS. Clin. Infect. Dis. 17(Suppl. 1):S230-S235.

Gao, F., L. Yue, A. T. White, P. G. Pappas, J. Barchue, A. P. Hanson, B. M. Greene, P. M. Sharp, G. M. Shaw, and B. H. Hahn. 1992. Human infection by genetically diverse SIVsm-related HIV-2 in west Africa. Nature 358:495-499.

Gauthier, J., A. G. Kluge, and T. Rowe. 1988. Amniote phylogeny and the importance of fossils. Cladistics 4:105-209.

Gendelman H. E., G. D. Ehrlich, L. M. Baca, S. Conley, J. Ribas, D. C. Kalter, M. S. Meltzer, B. J. Poiesz, and P. Nara. 1991. The inability of human immunodeficiency virus to infect chimpanzee monocytes can be overcome by serial viral passage in vivo. J. Virol. 65:3853-3863.

Gibbs, J. S. and R. C. Desrosiers. 1994. Auxiliary proteins of the primate immunodeficiency viruses. Pages 137-158 in Human retroviruses (B. R. Cullen, ed.). Oxford Univ. Press, New York.

Graur, D. 1985. Pattern of nucleotide substitution and the extent of purifying selection in retroviruses. J. Mol. Evol. 21:221-231.

Grmek, M. D. 1990. History of AIDS: emergence and origin of a modern pandemic. Princeton Univ. Press, Princeton.

Hahn, B. H. 1990. Biologically unique SIV-like HIV-2 variants in healthy West African individuals. Pages 31-38 in Retroviruses of human A.I.D.S. and related animal diseases (M. Girard and L. Valette, eds.). Fondation Marcel Mérieux, Lyon.

Hahn, B. H. 1994. Viral genes and their products. Pages 21-43 in Textbook of AIDS medicine (S. Broder, T. Merigan, and D. Bolognesi, eds.). Williams & Wilkins, Baltimore, Maryland.

Higgins, D., A. J. Bleasby, and R. Fuchs. 1992. Clustal V: Improved software for multiple sequence alignment. CABIOS 8:189-191.

Hillis, D. M. 1991. Discriminating between phylogenetic signal and random noise in DNA sequences. Pages 278-294 in Phylogenetic analysis of DNA sequences (M. M. Miyamoto and J. Cracraft, eds.). Oxford Univ. Press, New York.

Hillis, D. M., M. W. Allard, and M. M. Miyamoto. 1993. Analysis of DNA sequence data: Phylogenetic inference. Methods Enzymol. 224:456-487.

Hirsch, V. M., G. A. Dapolito, S. Goldstein, H. Mcclure, P. Emau, P. N. Fultz, M. Isahakia, R. Lenroot, G. Myers, and P. R. Johnson. 1993. A distinct African lentivirus from Sykes' monkeys. J. Virol. 67:1517-1528.

Hirsch, V. M., R. A. Olmsted, M. Murphey-Corb, R. H. Purcell, and P. R. Johnson. 1989. An African primate lentivirus (SIVsm) closely related to HIV-2. Nature 339:389-392.

Holland, J. J. (ed.). 1992. Genetic diversity of RNA viruses [Current topics in microbiology and immunology, 176]. Springer-Verlag, New York.

Holmes, E. C., L. Q. Zhang, P. Simmonds, C. A. Ludlam, and A. J. Leigh Brown. 1992. Convergent and divergent sequence evolution in the surface envelope glycoprotein of human immunodeficiency virus type 1 within a single infected patient. Proc. Natl. Acad. Sci. USA 89:4835-4839.

Howell, R. M., J. E. Fitzgibbon, N. Noe, Z. Ren, D. J. Gocke, T. A. Schwartzer, and D. T. Dubin. 1991. In vivo sequence variation of the human immunodeficiency virus type 1 env gene: Evidence for recombination among variants found in a single individual. AIDS Res. Hum. Retroviruses 7:869-876.

Hu, W. S., and Temin H. M. 1990. Genetic consequences of packaging two RNA genomes in one retroviral particle: Pseudoploidy and high rate of genetic recombination. Proc. Natl. Acad. Sci. USA 86:9253-9257.

Huet, T., R. Cheynier, A. Meyerhans, G. Roelants, and S. Wain-Hobson. 1990. Genetic organization of a chimpanzee lentivirus related to HIV-1. Nature 345:356-358.

Joenje, H. 1989. Genetic toxicology of oxygen. Mutat. Res. 219:193-208.

Johnson, B. K., G. A. Stone, M. S. Godec, D. M. Asher, D. C. Gajdusek, and C. J. Gibbs, Jr. 1993. Long-term observations of human immunodeficiency virus-infected chimpanzees. AIDS Res. Hum. Genetics 9:375-378.

Johnson, M. S., A. Sali, and T. L. Blundell. 1990. Phylogenetic relationships from three-dimensional protein structures. Methods Enzymol. 183:670-690.

Källersjö, M., J. S. Farris, A. G. Kluge, and C. Bult. 1992. Skewness and permutation. Cladistics 8:275-287.

Kluge, A. G. 1989. A concern for evidence and a phylogenetic hypothesis of relationships among Epicrates (Boidae, Serpentes). Syst. Zool. 38:7-25.

Langmuir, A. D., and S. C. Schoenbaum. 1976. The epidemiology of influenza. Hosp. Pract. 11:49-56.

Levin, S. A., and D. Pimentel. 1981. Selection for intermediate rates of increase in parasite-host systems. Am. Nat. 117:308-315.

Li, W. -H., T. Gojobori, and P. M. Sharp. 1988. Rates and dates of divergence between AIDS virus nucleotide sequences. Mol. Biol. Evol. 5:313-330.

Li, W. -H., and M. Tanimura. 1987. The molecular clock runs more slowly in man than in apes. Nature 326:93-96.

Maddison, W. P., and D. R. Maddison. 1992. MacClade: Analysis of phylogeny and character evolution. Sinauer, Sunderland, Massachusetts.

Martin, A. P., and S. R. Palumbi. 1993. Body size, metabolic rate, generation time, and molecular clock. Proc. Natl. Acad. Sci. USA 90:4087-4091.

May, R. M. 1993. Ecology and evolution of host-virus associations. Pages 58-68 in Emerging viruses (S. S. Morse, ed.). Oxford Univ. Press, New York.

Mcclure, M. 1990. Where did the AIDS virus come from? New Sci. 126(1723):54-57.

Mcclure, M. A., M. S. Johnson, D. -F. Feng, and R. F. Doolittle. 1988. Sequence comparisons of retroviral proteins: Relative rates of change and general phylogeny. Proc. Natl. Acad. Sci. USA 85:2469-2473.

Mindell, D. P. 1991. Aligning DNA sequences: Homology and phylogenetic weighting. Pages 73-89 in Phylogenetic analysis of DNA sequences (M. M. Miyamoto and J. Cracraft, eds.). Oxford Univ. Press, New York.

Mindell, D. P., and R. L. Honeycutt. 1990. Ribosomal RNA in vertebrates: Evolution and phylogenetic applications. Ann. Rev. Ecol. Syst. 21:541-566.

Montagnier, L. 1985. Lymphadenopathy -associated virus: From molecular biology to pathogenicity. Ann. Intern. Med. 103:689-693.

Myers, G., D. Macinnes, and B. Korber. 1992. The emergence of simian/human immunodeficiency viruses. AIDS Res. Hum. Retroviruses 8:373-386.

Myers, G., B. Korber, S. Wain-Hobson, R. F. Smith, and G. N. Pavlakis (eds.). 1993. Human retroviruses and AIDS 1993. Theoretical Biology and Biophysics, Los Alamos National Laboratory, Los Alamos, New Mexico.

Myers, G., K. Macinnes, and L. Myers. 1993. Phylogenetic moments in the AIDS epidemic. Pages 120-137 in Emerging viruses (S. S. Morse, ed.). Oxford Univ. Press, New York.

Myers, G., and B. Korber. 1994. The future of human immunodeficiency virus. Pages 211-232 in The evolutionary biology of viruses (S. S. Morse ed). Raven Press, Ltd., New York.

Neigel, J. E., and J. C. Avise. 1986. Phylogenetic relationships of mitochondrial DNA under various demographic models of speciation. Pages 515-534 in Evolutionary processes and theory (E. Nevo and S. Karlin, eds.). Academic Press, New York.

Nkengasong, J. N., M. Peeters, M. Van Den Haesevelde, S. S. Musi, B. Willems, P. M. Ndumbe, E. Delaporte, J. L. Perret, P. Piot, and G. Van Den Groen. 1993. Antigenic evidence of the presence of the aberrant HIV1ant70 virus in Cameroon and Gabon. AIDS 7:1536-1537.

Shigenaga, M. K., C. J. Gimenco, and B. N. Ames. 1989. Urinary 8-hydroxy-2'-deoxyguanosine as a biological marker in in vivo oxidative DNA damage. Proc. Natl. Acad. Sci. USA 86:9697-9701.

Shope, R. E., and A. S. Evans. 1993. Assessing geographic and transport factors, and recognition of new viruses. Pages 109-119 in Emerging viruses (S. S. Morse, ed.). Oxford Univ. Press, New York.

Smith, T. 1939. Parasitism and disease. Princeton Univ. Press, Princeton, New Jersey.

Smith, T. F., A. Srinivasan, G. Schochetman, M. Marcus, and G. Myers. 1988. The phylogenetic history of immunodeficiency viruses. Nature 333:573-575.

Stine, G. J. 1993. AIDS update 1993. Prentice Hall, Englewood Cliffs, New Jersey.

Swofford, D. L. 1993. Phylogenetic analysis using parsimony, Version 3.1.1. Computer program distributed by the Illinois Natural History Survey, Champaign.

Swofford, D. L., and G. J. Olsen. 1990. Phylogeny reconstruction. Pages 411-501 in Molecular systematics (D. M. Hillis and C. Moritz, eds.). Sinauer, Sunderland, Massachusetts.

Temin, H. M. 1993. The high rate of retrovirus variation results in rapid evolution. Pages 219-225 in Emerging viruses (S. S. Morse, ed.). Oxford Univ. Press, New York.

Yokoyama, S. 1991. Molecular evolution of human immunodeficiency viruses and related retroviruses. Pages 96-111 in Evolution at the molecular level (S. K. Selander, A. G. Clark, and T. S. Whittam, eds.). Sinauer, Sunderland, Massachusetts.

Watanabe, M., D. J. Ringler, P. N. Fultz, J. J. Mackey, J. E. Boyson, C. G. Levine, and N. L. Letvin. 1991. A chimpanzee-passaged human immunodeficiency virus isolate is cytopathic for chimpanzee cells but does not induce disease. J. Virol. 65:3344-3348.

Webster, R. G. 1993. Influenza. Pages 37-45 in Emerging viruses (S. S. Morse, ed.). Oxford Univ. Press, New York.

Zarling, D. A., and H. M. Temin. 1976. High spontaneous mutation rate of an avian sarcoma virus. J. Virol. 17:74-84.

Received 4 February 1994; accepted 17 October 1994

APPENDIX. Immunodeficiency virus abbreviations, host species, and database sequence accession numbers.
Virus Abbreviation Host Species Accession Number
FIV14 Felis catus M25381, M25729 HIV1ant70 Homo sapiens L20587, M31171 HIV1eli Homo sapiens K03454 HIV1ndk Homo sapiens M27323 HIV1jrcsf Homo sapiens M38429 HIV1lai Homo sapiens K02013 HIV1mal Homo sapiens K03456 HIV1mn Homo sapiens M17449 HIV1mvp5180 Homo sapiens L20571 HIV1rf Homo sapiens M17451, M12508 HIV2ben Homo sapiens M30502 HIV2d194 Homo sapiens J04542, X52223 HIV2d205 Homo sapiens X16109, X61240 HIV2nihz Homo sapiens J03654 HIV2rod Homo sapiens M15390 HIV2st Homo sapiens M31113 HIV2uc1 Homo sapiens L07625 SIVagm3 Cercopithecus aethiops M30931 SIVagm9 Cercopithecus tantalus L19254 SIVagm40 Cercopithecus tantalus L19252 SIVagm49 Cercopithecus tantalus L19253 SIVagm155 Cercopithecus aethiops M29975 SIVagm677 Cercopithecus aethiops M66437 SIVagm692 Cercopithecus pygerythrus M29974 SIVagmtyo Cercopithecus aethiops X07805 SIVcpz Pan troglodytes X52154 SIVmm142 Macaca mulatta M16403, YOO277 SIVmm239 Macaca mulatta M33262 SIVmm251 Macaca mulatta M19499, M15897 SIVmndgb Mandrillus sphinx M27470, X15781 SIVmne Macaca nemistrina M32741 SIVsmmh4 Cercocebus atys X14307 SIVsmm9 Cercocebus atys M80194 SIVsmpbj Cercocebus atys M31325 SIVstm Macaca arctoides M83293 SIVsyk Cercocebus mitis albogularis L06042