Introduction Ten years before the onset of the HIV pandemy, the ‘Red Queen’ metaphor was proposed as a framework for understanding the selective pressures that may drive both genetic and phenotypic diversification during the co-evolution of predators and preys [1]. As Lewis Caroll's Alice has to keep running with the Red Queen just to remain in the same place, so do viruses and their hosts each have to keep evolving new weapons, defenses, and counterattacks just to remain in the same place, in other words to persist. In such an evolutionary arms races context, virulence is a notion viewed from the perspective of the virus, and pathogenesis from the perspective of the host [2]. Persistent infection without disease is a feature of the simian lentiviruses in their old-world non-human primate hosts of origin [2,3]. In contrast, virulence and pathogenesis leading to AIDS is a consequence of recent cross-species transfers of these lentiviruses into humans. Hence, AIDS pathogenesis seems to result from maladapatative virus/hosts interactions, rather than from intrinsic requirements for virus persistence or transmission. However recent they may be, these pathogenic HIV/host interactions evolve very rapidly, due to high mutation and recombination [4] rates of HIV, and to the recent additional artificial selective pressure exerted in the form of HAART. On a global scale, with around 40 million people infected by HIV, most of whom live in poor countries with no access to any antiviral therapy [5], the pandemy emerges as a disastrous, ongoing confrontation between infected people of ever broadening genetic diversity, and viruses with intra-subtype diversity as large as 20% and inter-subtype diversity as high as 35% [6]. Significant progress has been made this year in the elucidation of the complexity of HIV/host interactions, with both depressing and uplifting implications. Briefly, a series of findings has revealed that host genetic polymorphism may not only influence the progression rate to disease, through complex epistatic effects between different genetic alleles, but also the very shaping of HIV diversity at a population level. Secondly, a previously unsuspected level of refinements has been uncovered in the viral subversion of host cell machinery, of molecular control of cell survival and cell death, and in viral escape from immune attacks. Last but not least, findings reported this summer and fall have suggested that ancestral, innate cellular anti-viral defenses, that we share with plants and insects, may be involved in our fight against HIV infection, and that pharmacological use of such ancient weapons may allow the development of radically novel anti-HIV therapeutic strategies. Host genetic polymorphism and disease progression The progression rate to AIDS is influenced by host single gene polymorphism affecting either HIV-1 co-receptor usage (the chemokine receptors and chemokines) or the presentation by infected cells of viral epitopes to cytotoxic T lymphocytes (CTLs) [the human leukocyte antigen class I (HLA-I)] [7]. CTLs seem crucial in the containment of HIV-1 replication in vivo [8] and HLA-1 polymorphism may affect both the effectiveness of the CTL response and the likelihood of selection of CTL-escape variants. For example, HLA B 57 [7] and HLA B27 [9] alleles are associated with slower progression to AIDS, due to the fact that two successive mutations in an immunodominant HIV-1-Gag epitope appear to be required for HIV-1 to both escape CTL response on these HLA backgrounds and conserve viral fitness [9]. Conversely, a single amino acid sequence difference in the HLA-B35 allele (HLA-B*35-Px) is associated with accelerated progression to AIDS [10]. Newly identified single gene polymorphisms involving chemokines and chemokine receptors also include single nucleotide polymorphisms (SNPs) in RANTES, associated with rapid progression to disease [11], and a CCR5’-cis regulatory region polymorphism associated with either accelerated or delayed progression to AIDS [12]. However, recent findings [13,14] indicate that polymorphism in one locus may have only very restricted — or even misleading — predictive value. Indeed, CCR5 polymorphism associated with different susceptibilites to HIV-1 infection or progression rates to disease in North American or European people do not show any predictive value in African people in Uganda [13]. This may result from (i) differing predominant HIV-1 subtypes, (ii) differing immune activation environments, such as parasite infections, or (iii) different genotype frequencies at loci that may act antagonistically — epistatically — with these variant CCR5 alleles. The importance of the latter possibility has been revealed by the first identification, this summer, of an epistatic inter-action between alleles present at two unlinked, highly polymorphic loci both involved in immune response regulation: one HLA-I allele, and one natural killer (NK) cell killer immunoglobulin-like receptor (KIR) allele [14]. While CTLs are crucial components of the adaptive immune response, NK cells are effectors of the innate immune responses. HLA-I molecules control both adaptive and innate immunity, presenting viral peptides to CTLs through their T-cell receptors and stimulating or inhibiting NK cells through their KIR receptors. While the HLA-B BW4–80I allele by itself exerts no effect on AIDS progression, the KIR 3DS1 allele by itself is associated with a moderate increased rate of AIDS progression [14]. However, individuals with both KIR 3DS1 and HLA-B BW4–80I genotypes show a marked delay in disease progression [14]. This suggests the importance of both epistatic genetic interactions and crucial interactions between the adaptive and innate immune response in the control of HIV infection and pathogenesis. Host genetic polymorphism and the evolution of HIV-1 diversity Ongoing HIV1 diversification is not only due to high mutation rates, but also to recombinations [4,15] that may be common in infected individuals at the level of each single infected cell [4]. While it has been debated whether HIV variations should be considered as adaptive evolution or mere genetic noise [16,17], recent work suggests that (i) CTL-mediated selection of viral escape mutants might occur early after infection [18], (ii) CTL-escape HIV-1 mutants can be transmitted between individuals [19], and (iii) such a transmission may be favored between people sharing given HLA-I alleles involved in the presentation of immunodominant epitopes to CTLs. For example, vertical transmission from mothers to children sharing the HLA-B27 allele cause the development, in children, of HIV escape variants mutated in the CTL epitope that is immunodominant in more than 85% HLA-B27 positive adults [19]. Certain drug-resistant viruses seem to pay a price in terms of pathogenicity [20], replicating without preventing CD4 T-cell reconstitution [21]. Do CTL-escape variants also pay a price in terms of pathogenicity? Recent findings suggest that selection — and maybe transmission — of CTL escape mutants is an integral part of pathogenesis [22]. Indeed, HLA-I polymorphism was found to be associated with HIV-Pol protein sequence polymorphism, and at a population level, HLA-I alleles and viral Pol sequences were predictive of viral load [23]; strongly suggesting that HLA-I-dependent immune responses drive both HIV-1 evolution and pathogenesis. HLA-dependent viral selection during transmission may determine both viral diversity in individual hosts and viral consensus sequences in given populations sharing particular HLA-I alleles. Also, HLA-I polymorphism may influence the probability of selecting for HAART escape mutants, because drug resistant viruses may have different abilities to expand depending on the constraints exerted by the HLA-I-dependent immune response [23]. Finally, on a different evolutionary time scale, natural selection acting on immune response regulatory genes might result in the establishment of stable, non pathogenic equilibria, such as the interactions between simian lentiviruses and their natural old-world non-human primate hosts. Accordingly, recent findings suggest that chimpanzees have experienced an ancient, severe polymorphism reduction in some orthologues of their HLAI loci. It is thus possible that their present natural resistance to AIDS is related to the fact that they are the late offsprings of rare survivors of an HIV-like (SIVcpz) pandemy in the distant past [24]. HIV-1 transmission: the Trojan horse strategy Prior to any confrontation with immune-selective pressures, the primary problem faced by HIVs upon transmission is gaining entry into permissive cells in a new host. New findings concerning HIV-1 transmission through a mucosal barrier have been obtained using primary intestinal epithelial cells in vitro [25]. In contrast with previously used intestinal tumoral cell lines, these primary cells are not permissive for HIV infection and do neither express CD4 nor CXCR4. Rather they express glycosphingolipid galactosylceramide (GalCer) and CCR5, allowing the selective entry of R5 viruses through binding of their Env to GalCer and CCR5. Entry does not seem to involve viral fusion and is followed by transcytosis, allowing the effective transfer of infectious viral particles at the basal pole of the epithelial cells [25]. This suggests that some mucosal barriers might counter-select X4 viruses for transmission [25,26]. More generally, HIV-1 transmission may depend on the initial subversion of cell populations not permissive for infection, but used as ‘Trojan horses’ allowing the viral journey towards permissive cells. Once mucosal barriers have been circumvented, HIV1 will use another ‘Trojan horse’, subverting the cellular machinery and physiological function of dendritic cells, the immune system sentinels that reside in all tissues, take up antigens, and present them to lymphocytes following migration into lymphoid organs. HIV-1-Env binding to the C-type lectin DC-SIGN [27], expressed by most dendritic cells, does not result in viral fusion, but in rapid internalization of intact viral particles in a low pH nonlysosomal compartment, where they retain infectivity for several days [28]. Once in the lymphoid organs, the uninfected, virus harboring dendritic cells will interact with CD4 T lymphocytes and transmit infection in a very efficient way. Also, a subset of DC-SIGN expressing dendritic cells circulating in the peripheral blood efficiently infects CD4 T lymphocytes [29], providing a potential mechanism for rapid viral dissemination throughout the body. Because DC-SIGN allows transmission of both R5 and X4 viruses [28,30], dendritic cells might not be involved in the R5 transmission bottle-neck. However, a subset of dendritic cells are permissive for R5 — but not X4 — HIV1 infection [30], leading to Nef-mediated upregulation of DC-SIGN surface levels and to enhanced R5 virus transmission to CD4 T lymphocytes [30]. DC-SIGN is also expressed by brain endothelial cells, where it may favor viral passage through the blood/brain barrier [31], and a related receptor, DC-SIGNR [32], expressed on placental capillaries, may be involved in vertical transmission. Together, these findings suggest the importance of the DC-SIGN and DC-SIGN R family in the delivery of infectious viral particles to permissive cells, and provide new targets for prophylactic and therapeutic intervention. Overcoming epigenetic restrictions to infection Recent findings indicate that the mere surface expression of the HIV-1 receptor and coreceptors may be insufficient to allow effective infection of permissive cells. For example, the inclusion of CD4 molecules in ‘rafts’ — cholesterol-enriched cell membrane micro-domains that are moving platforms with embedded proteins involved in cell signaling and intercellular contacts [33] — while not necessary for HIV-1 Env binding to CD4, seems required for HIV-1 infection [34]. Also, oxidation of the disulfide bond of the D2 extracellular domain of CD4, while not required for HIV-1-Env binding to the D1 domain of CD4, is required for CD4-dependent HIV-1 infection [35]. Both findings [34,35] may explain previously described anti-viral effects of anti-cholesterol and anti-oxidant drugs, and suggest potential new targets for therapy. Another epigenetic restriction to infection occurs at a post-entry level: infection is abortive in quiescent CD4 T lymphocytes in vitro [36,37]. However, this may not be true in vivo, as suggested by recent studies using either tonsil organ culture [38] or the addition of the IL-7 cytokine [39]. Moreover, HIV-1 may by itself modify the microenvironment in a way that favors permissivity. Indeed, Nef induces the release by infected macrophages of soluble factors attracting and activating CD4 T lymphocytes [40], and the Env of R5 HIV-1 viruses induce a cascade of cell signaling in quiescent, uninfected CD4 T lymphocytes that might allow their infection [41]. However, the main strategy used by HIV is probably the above mentioned hijacking of dendritic cells as ‘Trojan horses’ for viral delivery, which exploits the physiological pathway of quiescent CD4 T lymphocyte activation, allowing their infection in the absence of additional stimuli [28,29,30]. Because dendritic cells are the sole antigen-presenting cells able to activate antigen-specific, naive, quiescent CD4 T lymphocytes, this may also explain why most infected CD4 T lymphocytes detected in vivo belong to a subset of HIV-1-specific cells [42,43]. Viral subversion of host cell machinery and defenses Additional, post-entry strategies may facilitate the insertion of HIV-1 proviral DNA in the chromosomes of activated, but non-cycling cells [44]. While the nuclear envelope disassembles during mitosis, the nuclear pores of non-cycling cells seem to small to allow entry of the pre-integration complex (PIC) containing the proviral DNA [45]. The HIV-1-Vpr protein, contained in incoming virions, induces transient nuclear envelope herniations and breaks that may allow PIC entry [46]. Once proviral integration has been achieved, other interactions with the host cell machinery are required for the production of infection-competent viral progeny [44]. For example, HIV-1-Gag protein assembly into immature capsids requires its association with the cellular protein HP68, a RNAse L inhibitor, that might be subverted as a chaperone [47]. The selective localization of Gag and subsequently of viral budding within ‘rafts’ microdomains at the plasma membrane [48], may favor both viral assembly and transmission, since rafts seem concentrated at points of intercellular contacts. Prior to viral budding, HIV-1-Env is also expressed at the cell surface, through the hijacking of physiological secretory granules containing the immuno-inhibitory CTLA-4 cellular protein [49]. Whether such cell surface co-expression of Env and CTLA-4 may favor immune evasion of virus producing cell remains to be investigated. Another important level of HIV-1-mediated host cell subversion has been recently identified [50], suggesting that present-day HIV-1/host combats occur on a back-ground of ancestral evolutionary arms races [51]. The incorporation in the viral particles of HIV-1-Vif accessory gene product, that binds genomic HIV-1-RNA, is crucial for allowing infected cells to produce infection-competent viral particles. Vif acts by inhibiting the anti-HIV-1 effects of a previously unsuspected cellular anti-viral weapon, the CEM-15 protein. CEM-15 acts in cells already infected by restricting production of infectious particles, perhaps by altering genomic HIV-1-RNA through an RNA editing activity [50]. Finally, such virus/host interactions may occur in more cell populations than previously suspected, resulting in the establishment of additional, long-lived reservoirs, persisting after effective, prolonged HAART. Such reservoirs might include a subset of circulating NK cells expressing CD4, CCR5 and CXCR4 [52], and more surprisingly, kidney epithelial cells harboring tissue-specific HIV-1 quasi-species [53,54]. Thus, effective HIV-1 control may require a focus on HIV-1 subversion of cells other than CD4 T cells and macrophages in tissues other than lymphoid organs. Viral subversion of cell survival and cell death One ancestral target of evolutionary arms races between viruses and their hosts is the control of cell survival and death [55]. Briefly, rapid cell death induction upon viral entry is an ancient effective first line host-defense strategy. Conversely, many viruses repress premature death in cells they require for replication or persistence. Subsequent host immune attack involves CTL- and NK-mediated killing of infected cells, and several viruses have evolved counterattacks that may favor either evasion or immune effector cell killing [56,57]. HIV-1-specific CD8 T cells were recently found to lack terminal effector differentiation markers [58,59] and to be more sensitive to death signaling induced by CD95 ligand (CD95L), a TNF family member, than the CMV- and EBV-specific CD8 T cells for the same HIV-1-infected persons [59]. Because CD95L expression is upregulated in HIV-1-infected cells [56,60], HIV-1-specific effector CD8 T cells might die upon encounter with their target cells. This may explain both why CD8 T-cell turnover is so high in HIV-1-infected persons, and why infusion of autologous HIV-1-specific CD8 T cells leads to their rapid disappearance in vivo despite initial homing to lymphoid organs [61,62]. HIV-1 viral particles, as well as several HIV proteins [56,60,63,64,65,66,67], including Tat, Env, Vpr, Vpu and Nef, can induce death in either infected or uninfected T cells. Nef seems to play a particularly complex role in HIV-mediated subversion of cell survival and death [56,63,68]. Nef favors immune evasion, and hence survival, of infected cells through selective downregulation of the HLA-I molecules that allow CTLs to detect viral peptides, while not interfering with the HLA-I molecules that inhibit NK cell killing. Nef also causes the expression of CD95L on the surface of infected cells, favoring death of HIV-1-specific CTLs, while protecting the infected cell against death ‘from without’ induced by the TNF family ligands expressed on its surface, or on the surface of neighboring cells [69]. Finally, Nef also represses death signaling ‘from within’ by pro-apoptotic Bcl2 family protein [70] that may be activated in the infected cell by HIV-1 proteins such as Vpr and Vpu [64,71]. The induction of apoptosis depends on caspase protease activation in the dying cell [55]. While it seems that most uninfected cells in HIV-infected persons die through an apoptotic process [71,73], how HIV-infected cells die once viral progeny has been produced remains unclear. Indeed, recent discordant findings have identified either apoptosis and caspase activation [64,74] or necrosis and lack of caspase activation [75,76] as the phenotype of dying, productively infected CD4 T cells. The latter finding, if confirmed, has interesting implications. First, as other viruses [55,56], HIV-1 may encode yet unidentified caspase inhibitors. Second, a previous, puzzling finding in vivo, in HIV-1-infected persons indicated that all dying cells detected were uninfected, while no infected cell detected was dying [72]. Because the assay used was specific for apoptosis (and not for cell death [72]), a possible interpretation is that infected cells indeed die in vivo, but through a non-apoptotic pathway. If this is the case, therapeutic strategies aimed at preventing caspase activation and apoptosis induction might allow to selectively prevent uninfected-bystander cell death without providing the infected cells with any survival benefit. While it is possible that bystander CD4 T-cell apoptosis might reduce to some extent HIV-1 ability to invade new cell targets, recent findings suggest that excessive uninfected-cell apoptosis may be detrimental not only through the direct cell loss caused, but also through the induction of additional pathogenic mechanisms. For example, ingestion of uninfected apoptotic cells by dendritic cells might down-regulate the immune response [77,78]. Also, ingestion of uninfected apoptotic cells by HIV-1-infected macrophages enhances their viral production [79]. And apoptotic cells may cooperate with HIV-1 particles in inducing death of uninfected neighboring CD4 T cells [80]. Thus, the very extent of apoptosis induced during HIV-1 infection may be itself provide an amplification loop in Aids pathogenesis. Riches of ancestral anti-viral weapons? Our defenses against infections involve the sequential mobilization of various layers of recognition and effector mechanisms of distinct evolutionary origin. Our adaptive immune system, that allows refined ‘self/nonself’ discrimination and memory, and depends on several millions of different antigen receptors expressed by our T and B lymphocytes, emerged in the common ancestors of present-day jawed vertebrates [81]. Our immediate and innate immune response to conserved microbial molecular patterns, however, involves Toll-like receptor family homologues that we share with insects and plants, and ancestral anti-microbial effector peptide family homologues that we share with insects, such as defensins [82]. Human defensins homologues, including α- and β-defensins and circular minidefensins can be secreted by several immune cell types and/or by epithelial cells [83]. Human defensins were previously reported to inactivate several human viruses [84], and this year, to suppress HIV-1 replication in vitro [85,86]. The anti-HIV-1 α-defensins -1, -2 and -3 are secreted by activated CD8 T cells from HIV-1-infected long term non – progressors [86], and might represent the elusive, soluble CD8 antiviral factors (CAF) [87] that inhibits replication of both R5 and X4 viruses [88]. These findings imply that HIV-1 most often achieves suppressing the expression of these ancestral defenses, but that pharmacological use of defensins may allow the development of new effective anti-HIV-1 therapy. There are still other ancestral anti-viral defenses, that do not depend on the activation of a specialized immune system, and that operate at the level of each single cell. In plants and invertebrates, RNA interference (RNAi) is involved in protection against transposons and viruses [89]. RNAi induces post-transcriptional gene silencing by forming silencing complexes containing enzymes and double stranded small interfering (si) RNAs of 21–25 base pairs, that cause degradation of target RNAs containing the same sequence and lead to the formation of new identical siRNAs, allowing an autocatalytic amplification process [89,90,91]. This year, siRNAs were shown to also induce gene silencing when introduced in mammalian cells [92,93], indicating that critical components of the silencing complex (and perhaps siRNA) are present and operational in mammals. Also, the use of soluble siRNA molecules or of siRNAs produced from transfection, targeting either host (CD4] genes or various HIV-1 genes was reported to allow inhibition of HIV-1 infection and/or replication in human cells in vitro [94,95,96]. Will similar approaches also work in vivo? Might RNAi easily select for HIV-escape variants? And, knowing that at least one invertebrate virus achieves RNAi-suppression [97], might HIV already possess — or rapidly evolve — similar counterattacks? While these questions remain to be addressed, these important findings reveal the previously unsuspected riches of anti-viral weapons that have emerged in our ancestors during their long evolutionary arms races against viruses. The further deciphering of these complex virus/host interactions may provide us with the promising opportunity to transform such ancestral weapons into novel therapeutic agents. Together with the ongoing drug and immunotherapy developments [98,99], they might allow us to gain an edge in our global fight against HIV. Acknowledgements The research programs developed in EMI-U 9922 are supported by institutional grants from INSERM and Paris 7 University, and by grants from Agence Nationale de Recherche sur le Sida (ANRS), Ensemble contre le Side (ECS/Sidaction), Paris 7 University — Valorisation, Fondation pour la Recherche Médicale (FRM) and Etablissement Françis des Greffes.