Portal de Periódicos da CAPES

Genetic recombination and its role in the development of the HIV-1 pandemic

2002; Lippincott Williams & Wilkins; Volume: 16; Linguagem: Inglês

10.1097/00002030-200216004-00002

1473-5571

Rafael Nájera, Elena Delgado, Lucía Pérez-Álvarez, Michael M. Thomson,

HIV/AIDS Research and Interventions

Introduction Genetic recombination is part of the normal mechanisms of retroviral replication and, as such, plays an important role in the generation of viral diversity. Generation of recombinant retroviruses requires that two viruses infect a single cell, either simultaneously, by a single transmission event, or sequentially, in multiple transmission event, or sequentially, in multiple transmission events. In HIV-1, recombination can occur between different strains of the same subtype (intrasubtype recombination), different or different groups (intergroup recombination). The importance of recombination in shaping HIV-1 global diversity has been increasingly recognized in recent years. Several factors have contributed to this trend, including new surveys carried out in previously unexplored or insufficiently sampled areas, more frequent phylogenetic analysis of multiple genome segments [including protease and reverse transcriptase (RT)] used to detect drug resistance mutations, and growing numbers of full-length HIV-1 genomes analyzed in multiple geographical areas. The cumulating picture emerging from these studies indicates that HIV-1 recombinant forms are much more prevalent, geographically spread, and diverse in the global pandemic than previously known. This applies not only to circulating recombinant forms (CRFs) (recombinant forms indentified in at least three epidemiologically-linked individuals), but also, and even more notoriously, to unique recombinant forms (URFs) (recombinant forms found in a single individual or in a single epidemiologically-linked cluster). The recurrent finding of URFs in high proportions in areas where multiple HIV-1 genetic forms co-circulate attests to the high frequency of dual infections with diverse variants, which can either occur simultaneously or, more probably, sequentially, as suggested by the report of the first well documented case of HIV-1 superinfection in humans [1]. In the present review, basic concepts on the mechanisms of HIV-1 recombination, recent developments in the molecular epidemiology of HIV-1 recombinant forms, and their implications for vaccine development and therapeutic strategies are discussed. Molecular mechanisms of HIV-1 recombination Retroviruses and other RNA viruses exhibit high mutation rates due to the lack of proof-reading activity of the viral polymerases, the short replication times, and the large population sizes. Consequently, they are present in the host organism as mixtures of genetically diverse but related populations termed quasispecies [2]. These represent all possible mutants, as illustrated by the presence of antiretroviral drug resistance-associated mutations in treatment-naïve patients, the so-called ‘natural resistance’ [3,4]. Retroviruses are also known for their high recombinogenic potential [5,6], deriving from the fact that recombination constitutes an intrinsic part of their normal replication cycle. Recombination can mediate the repair of defective retroviral genomes [6–8], can increase viral diversity, and can accelerate the spread of beneficial mutations among viral quasispecies [9]. The increased variation potential mediated by recombination confers on retroviruses the capability to respond rapidly to changing selective pressures, either immunological [10,11] or pharmacological [12,13], through the prompt generation of the fittest variants possessing the adequate set of mutations to elude those pressures [14]. Recombination in retroviruses requires the co-packaging in each virion of two RNA genomes [5,15]. Recombination between genomes packaged in two different virions is possible only by productive infection of a single cell by both virions, which allows for the production of heterozygous particles. In a subsequent cycle of infection, a recombinant genome can be generated through alternate jumps of RT between both co-packaged genomes (Fig. 1).Fig. 1.: Generation of recombinant retroviruses. Two virions with genetically dissimilar genomes infect a single cell, either sequentially or simultaneously. Both viral genomes undergo reverse transcription and the proviruses become integrated in the cell chromosome. Heterodimers of genomic transcripts derived from both integrated proviruses are packaged in viral particles. In a second replication cycle, heterozygous viruses produce recombinant proviral genomes by alternate template jumps during reverse transcription.Synthesis of the double-stranded proviral DNA by RT involves two obligatory template jumps, necessary for long terminal repeat duplication at both genome ends (reviewed in [16]) (Fig. 2). In the first jump, the minustrand, strong-stop DNA (ssDNA) intermediate, initiated at the tRNA annealed to the primer binding site, is translocated from the 5’ end to the 3’ end of the genome. This jump is possible by the presence of repeated sequences at both extremities of the genome, and can be intermolecular or intramolecular. The second jump involves the transfer of the plus-strand ssDNA, initiated at the 3’ polypurine tract, from the 3’ end to the 5’ end of the genome. This jump is mediated by base pairing of the primer binding site of plus-strand ssDNA (generated by copying the 3’ end of the tRNA attached covalently to the minus-strand DNA) to its complementary sequence at the minus-strand DNA, and is almost always intramolecular.Fig. 2.: Generation of a recombinant double-stranded provirus during reverse transcription. (a) Proviral DNA synthesis starts at the primer binding site (PBS), near the 5’ end of the genome, of one of the co-packaged strands, primed by the 3’ segment of a cellular tRNA. Polymerization proceeds until reaching the 5’ end of the genome, generating the minus-strand ssDNA. (b) This DNA intermediate anneals to the 3’ end of the same or the other co-packaged strand thanks to the presence of repeated (R) sequences at both ends of the genome. (c) RT resumes minus-strand DNA polymerization, while the RNAse H carried by RT degrades the RNA hybridized to the growing DNA chain. Alternate template jumps along polymerization of minus-strand DNA generates a recombinant genome. (d) An RNA oligonucleotide located 5’ of the U3 segment, resistant to RNAse H cleavage, the polypurine tract (PPT), primes the synthesis of the plus-strand DNA, which continues through the end of the minus-strand DNA copying 18 nucleotides of the covalently attached tRNA, thus regenerating the PBS. (e) This DNA intermediate, termed the plus-strand ssDNA, undergoes a second template jump, which is almost always intramolecular, by annealing of the PBS to its complementary sequence in the minus-strand DNA. (f) The synthesis of both strands is completed using each other as template. In HIV-1 and other lentiviruses, a second PPT located at the center of the genome, the central PPT, is also used as primer for plus-strand DNA synthesis (omitted from the figure). LTR, Long Terminal Repeat.Besides the obligatory strong stop DNA jumps, strand switches may occur along internal genome segments [5,17,18], which in heterozygous virions would generate a recombinant genome. In HIV-1, it has been reported that an average of two to three jumps occurs during each replication cycle [19,20]. Experimental evidence supports internal interstrand jumps occuring predominantly, if not exclusively, during minus-strand DNA synthesis [15,17,19–21]. In the forced copy-choice model, breaks in the RNA would force RT to switch templates, thus restoring the continuity in the genome to generate a viable progeny [22]. A related model proposes that the low processivity of retroviral RT would result in pauses along RNA-dependent DNA polymerization, which would promote template switching without the need of breaks [23]. In agreement with this model, in vitro studies indicate that pauses in polymerization enhance template switching [24,25]. During pauses, which may be favored by secondary structures [26], nucleotide misincorporations [27], or low dNTP concentrations [24,28,29], the extent of RNAse H cleavage of the original (donor) template beneath the stalled RT is increased [27]. This process generates a longer single-stranded DNA segment, which would become free to anneal to the opposite (acceptor) strand, positioning it near the polymerase site, thus facilitating template switching [28,29]. In contraposition to this model, other authors postulate that pause-independent strand transfers might be predominant [30], which might be enhanced by interstrand proximity mediated by secondary structure interactions [32]. The frequency of recombination may also be highly dependent on sequence homology [33]. In a recent study, for a sequence difference of 25% or more, HIV-1 recombination directed by sequence homology was not more frequent than that which was homology independent [33]. Multiple methods are available for the analysis of recombinant sequences. Those that have been used more frequently in HIV-1 are similarity [34,35] or diversity [36,37] plots and bootscanning [38]. More precise mapping of breakpoints can be achieved by informative site analysis [39]. An HIV-1 intersubtype recombinant form is identified when its phylogentic relationships with different subtypes switches along the genome. Detection of intrasubtype recombination is difficult unless both parental viruses are identified or they group in distinct recognized phylogenetic clusters. For a complete characterization of the mosaic structure, full genome analysis is required. A more practical alternative for a large number of samples is to analyze discrete segments in separate regions of the genome. Commonly, segments of gag and env (most frequently, the V3 loop region) or of pol and env have been used for phylogenetic analysis. HIV-1 recombinant forms in the global pandemic Although recombination in an HIV-1 virus isolated from an infected individual was suggested as early as 1988 [40], corresponding to the African isolate MAL, the recognition of the importance of recombination in shaping HIV-1 diversity in the global pandemic is a relatively recent development. The classification of HIV-1 in distinct genetic subtypes was originally based on clustering in phylogenetic trees of env and sag sequences [41–43], with five of the nine presently recognized subtypes having been indentified in these early studies The first case of intersubtype recombination in infected individuals was reported in 1994 in Brazil in two sexual partners harboring a BF recombinant virus [44]. One year later, a high frequency (10%) of HIV-1 intersubtype recombinant virses was reported to be found in sequence databases [45]. In 1996, the full genome analysis of isolates from a genetic form circulating in Thailand and Central Africa indicated that it was an intersubtype recombinant form, later designated CRF01_AE [36,46]. Its recombinant nature had been suggested previously by analysis of partial sequences [46]. In 1998, the second CRF (an AG recombinant) was identified by analysis of full genome sequences from Nigeria and Djibouti [47]. This CRF (currently designated CRF01_AG) has recently been shown to be the most common HIV-1 variant circulating in West Africa and some West-Central African Countries [48,49]. The current nomenclature of CRF and the criteria for their definition were adopted in 1999 [50]. According to these criteria, to define a CRF, three epidemiologically unlinked viruses with coincident mosaic structures and consistent phylogenetic clustering must be characterized, at least two of them in near full-length genomes (>=8 kb). CRFs are designated with consecutive numbers according to the order of discovery, followed by the parental subtypes, or cpx (from complex) if there are more than two parental subtypes. Presently, 13 CRF have been identified [51,52] (excluding CRF09, whose characterization still remains unreported); eight of them originated in Africa, two in Europe, two in Asia, and one in America. Those with the greatest epidemic importance, representing the most prevalent HIV-1 genetic forms in some areas, are CRF01_AE in Southeast Asia, CRF02_AG in West and West-Central Africa, CRF03_BC in China, and CRF12_BF and related recombinants in Argentina and Uruguay. In a recent survey on the global distribution of HIV-1 genetic forms, 18% sampled infections corresponded to recombinant viruses. The globally most prevalent recombinant forms were CRF02_AG (representing 31% infections in West Africa) and CRF01_AE (representing 63% infections in Southeast Asia and 24% in East Asia and the Pacific) [53]. Recent developments in the molecular epidemiology of HIV-1 recombinant forms Recent developments related to HIV-1 recombinant forms in the past 2 years include the description of four new CRF, the recognition of the geographical spread in new areas of previously identified CRF, the finding of high frequency of URF in areas where multiple genetic forms co-circulate, and the first documentation of HIV-1 superinfection in humans. Recently identified CRF CRF12_BF has been identified in Argentina and Uruguay [54,55]. Its existence was proposed in a previous study, in which BF recombinant viruses were found to be widely circulating in Argentina, with coincident breakpoints in pol suggesting a common ancestry [56]. One of the most notable findings in Argentina and Uruguay, revealed by full-length genome analysis, is that URF related to CRF12_BF, most of them apparently derived from secondary recombination with subtype B viruses, appear to be more prevalent than genuine CRF12_BF viruses [54,55] (Fig. 3a). It has been proposed that CRF12_BF might have originated in Brazil [54]. This hypothesis is deduced from the facts that nonrecombinant F subtype viruses have not been detected in Argentina or Uruguay [54,56] and that F subtype segments of CRF12_BF are phylogenetically related to F subtype viruses from Brazil [54]. However, no CRF12_BF viruses have been found in Brazil [54,57], which implies that these recombinants either are not currently circulating or are minoritary in this country. According to infection dates and genetic distances, CRF12_BF is the oldest reported CRF of nonAfrican origin. CRF12_BF and related recombinants represented 65% infections in Argentina in a recent survey [54], and have also been identified in Bolivia [55], Chile, and Spain (unpublished data Delgado E, et al..Fig. 3.: Mosaic structures of recombinants analyzed in full-length or partial [protease (PR)-RT] sequences. (a) CRF12_BF and related unique BF recombinants from Argentina, and (b) CRF14_BG and unique BG recombinants from Galicia (Spain). LTR, Long Terminal Repeat.CRF14_BG has been identified in Northwestern Spain (region of Galicia) circulating in a minority of injecting drug users (IDU), representing the first reported CRF originated in Western Europe [58,59]. This CRF is probably also circulating in Portugal [59,60]. A subtype G strain circulating in Galicia and Portugal has been identified as parental of CRF14_BG [59,60]. Similar to findings in Argentina, although on a lesser scale, we have found a variety of URF derived from secondary recombination of CRF14_BG with subtype B viruses [59] (unpublished data Delgado E, et al.) (Fig. 3b). CRF06_cpx, previously characterized in only two isolates, has recently fulfilled the criteria for recognition as a CRF by the full-length genome characterization of two additional viruses [61]. Re-examination of its mosaic structure indicates that it is an A/G/J/K recombinant. Analysis of viruses characterized in partial sequences points to the wide georgaphical distribution of CRF06_cpx in West Africa [61]. In Niger, CRF06_cpx viruses represented 18% of HIV-1 infections in a recent survey [62]. A recent outbreak in Estonia among IDU also appears to be caused by CRF06_cpx viruses [63]. CRF11_cpx, an A/G/J/E recombinant, has been identified in Central Africa by analysis of full-length sequences of four viruses from Cameroon and the Central African Republic [64]. Their mosaic structure coincides with that of a previously characterized virus from the Democratic Republic of Congo [65]. Analysis of partial sequences suggests the presence of CRF11_cpx in Senegal and Gabon [65]. The existence of another complex CRF (CRF13_cpx) in Cameroon, derived from the same parental clades but differing in structure from CRF11_cpx, has been proposed on the basis of full-length sequences of two epidemiologically unlinked isolates [66]. Two new CRF have been proposed to be circulating in Cuba by analysis of partial pol and env sequences; Dpol/Aenv and Upol/Henv, detected in 21% and 7% of samples, respectively [67]. Their origin is probably African, since the parental viruses were not found in Cuba. In fact, the Upol/Henv viruses cluster both in pol and env, with a complex recombinant isolate from Cameroon (CM53379) [49] characterized in the full-length genome. Recombinant viruses phylogenetically related to the Central African isolate MAL, the first reported HIV-1 recombinant virus, have been identified in Norway [68], Spain [58], the United Kindgdom [69], and Cuba [67], suggesting the existence of a MAL-like CRF. High frequencies of URF Recent studies have revealed high prevalences of URFs in West Africa (Nigeria [70] and Niger [61], West-Central Africa (Democratic Republic of Congo [71] and Cameroon [49,72]), East Africa (Tanzania [73,74], Kenya [75,76], and Uganda [77–79]), South America (Argentina [55,56], Uruguay [56], and Brazil [57]), the Caribbean (Cuba [67]), Asia (Myanmar [80] and China [81]), and, among nonsubytpe B viruses, in Western Europe [Northwestern Spain (unpublished data Delgado E, et al.) and the United Kingdom [69]]. In rural Cameroon, six (27%) of 22 viruses characterized in full-length sequences were URF [72]. In East Africa, full-length genome analysis revealed that URF constituted more than one-half (five of nine) isolates in Tanzania [74], 39% in Kenya [75], and 31% in Uganda [77], involving subtypes circulating in each of these countries. These results are concordant with those of surveys in which partial segments were analyzed [73,76,78,79]. In Argentina, 14 (78%) of 18 BF recombinant viruses analyzed in full-length sequences were URF, related to but different from CRF12_BF [54,55] (Fig. 3a). Similar findings were obtained in Uruguay [56]. In Rio de Janeiro (Brazil), analysis of gag-pol and env sequences showed that seven of 78 viruses were BF recombinants, all with unique structures, apparently unrelated to CRF12_BF, while only two were of nonrecombinant F subtype [57]. In Cuba, analysis of partial pol and env sequences revealed that 14 (13%) of 105 samples represented URF, formed by recombination between four genetic forms circulating in the island (B and G subtypes, and Dpol/Aevn and Upol/Henv recombinants) [67] (Fig. 4).Fig. 4.: Diverse mosaic patterns of HIV-1 recombinants from Cuba analysed in partial pol and env segments. Recombination patterns of CU68 (Upol/Henv) and CU40 (Dpol/Aenv) have been found in seven and 21 of 105 individuals, respectively, representing probable CRF. The rest represent unique recombinants.In Asia, high frequencies of URF have been reported in central Myanmar [80] and in the Yunnan province of China (71% viruses in the Dehong prefecture were unique BC recombinant forms) [81]. In Northwestern Spain (Galicia), among nonsubtype B viruses, analysis of partial pol and env sequences indicate that 72% (49 of 68) viruses are recombinant, 20 (41%) of which are URF. Among these, BG recombinants, originated by recombination of G subtype or CRF14_BG with B subtype viruses, are the most common (unpublished data Delgado E, et al.) (Fig. 3b). In the United Kingdom, 20% nonsubtype B infections have been reported to be recombinant, of which 72% did not belong to recognized CRF [69]. The geographical distribution of CRFs and VRFs is summarized in Table 1.Table 1: Geographical distribution of HIV-1 recombinant forms.Recent findings therefore reinforce the importance of recombination as a major mechanism for the generation of viral diversity in the global HIV-1 pandemic. The real prevalence of URF is probably underestimated by analysis of partial segments. A relatively large number of full-length genome sequences have been analyzed only in a few countries and, even there, the number is far from representative. In any case, the high prevalence of URF in many areas attests to the common occurence of dual infections with different genetic forms. In what proportions these derive from simultaneous or closely spaced infections, or from superinfections in established infections with mature immune responses is not known. The prevalence of recombination between HIV-1 viruses of the same genetic form (intraclade recombinants) is unknown due to the difficulty of their identification. Their detection is possible only if the parental viruses are characterized or if recombination occurs between viruses grouping in different phylogenetic clusters circulating in the same geogrphical area. Co-circulation of phylogenetically distinct viruses of the same subtype has been reported in Ethiopia (subtype C) [82], Brazil (subtype B) [83], Cuba (subtype B) [67], and Portugal (subtype G) [60], where the prevalence of intrasubtype recombination could be studied. Biological characteristics The V3 loop region of HIV-1 is a principal determinant of viral tropism, phenotype, and co-receptor use (Fig. 5). The association between the syncytium-inducing phenotype and the presence of positively charged amino acids in V3 loop positions 11 or 25 has been shown in HIV-1 strains corresponding to the A, B, C, D, E, (CRF-01 AE) and G subtypes [84], although several differences in some subtype have been defined [85–88].Fig. 5.: Biological characteristics of HIV-1 in subtypes and recombinants. Correlation between V3 loop net charge, syncytiuminducing (SI) phenotype, chemokine receptor usage, and tropism for different cell types in culture. (a) SI variants display high V3 loop net charge (≥ +5) and primarily use the chemokine receptor CXCR4 (T lymphotropic), exclusively or together with CCR5 (dual tropic). (b) Non-SI variants display low net charge (≤ +4) and primarily use the chemokine receptor CCR5. (c) Recombination in V3 between SI and non-SI variants could result in change of viral phenotype.The biological characteristics of recombinant HIV-1 strains isolated from peripheral blood mononuclear cells (PBMC) or organic fluids are not well known. In a recent study, co-receptor usage and syncytium phenotype of CRF02_AG-like viruses from Western Camernoon were related to the characteristics described in B subtype viruses, and correlated with the clinical status of the patients [89]. The phenotype of the newly defined CRF14_BG [59] has been described as syncytium inducing, with usage of CXCR4, CCR5, CCR3, and CCR2b co-receptors, showing a correlation with phenotype predicted from the V3 loop sequence and having a characteristic pattern of mutations in the V3 loop [90]. The nonsyncytium-inducing phenotype and CCR5 usage have been predicted based on V3 sequences of diverse unique recombinant strains that were detected in Northwestern Spain [91]. The preferential usage of CCR5 may favor the spread of these recombinants by heterosexual activity. The change of HIV-1 co-receptor use in 143 HIV-1 isolates of subtypes A, B, C, and D, and one recombinant form (CRF01_AE), from 24 vertically infected children, has been recently reported [92], showing a significant association between decreased CD4-cell counts and severity of disease and the emergence of CXCR 4-using viruses. In spite of the hypervariability of the V3 loop, the three amino acids Gly317-Pro318-Gly319 at the crown of the loop are highly conserved. The GPG motif might be conserved because it helps to create the conformation of the V3 loop required to induce syncytia and apoptosis. However, several changes at the GPG motif have been described in nonB subtypes. In one Romanian subtype F isolate [93], two identical sequences derived from PBMC and MT2 cultures had a GVGR tetrapeptide apex sequence. An unusual GIGK crown has been also described in a syncytium-inducing isolate of subtype C [88]. In another study [94], almost 40% of the Brazilian subtype B isolates had a GWGR motif at the tip of the V3 loop, showing biological properties similar to those observed in other subtype B strains. An unusual tetrapeptide V3 loop apex sequence (GRGR) was detected in proviral DNA clones from PBMC from a patient with a G + B dual infection, and also from biological clones of subtype B primary isolates [91]. The effect of single point mutations at the GPGR tip sequence on cellular tropism and co-receptor usage has been studied [95], showing that the cell tropism of HIV-1 to macrophages and brain-derived cells, and their use of the co-receptors were markedly affected by the tip sequence of the V3 domain. Little is known regarding the consecutive changes in biological properties of HIV-1 strains following reinfection and subsequent recombination, changes that could be different depending on the genetic and biological characterization of both viruses. Direct evidence in favor of molecular and biological interaction and synergism between two infecting HIV-1 subtype B strains from a co-infected patient has been reported, showing the predominance, both in vitro and in vivo, of one group of viruses, and suggesting the active role of viral synergism or a helper virus effect in modulating the pathogenesis of viral strains [96]. Dual infections and superinfection Two or more viruses can infect an individual, a phenomenon referred to as dual infection (or multiple infection if more than two viruses are involved). If these viruses infect the same cell, a recombinant virus may be generated. Dual infections may be simultaneous or sequential. In general, a dual infection occurring in a single transmission event is designated co-infection. On the other hand, superinfection has been used generally for sequential infections, involving at least two transmission events. However, some confusion has originated from the use by some authors of superinfection as a generic name for both simultaneous and sequential infections [97]. In an attempt to clarify there concepts, we propose re-infection as an alternative term to indicate a new infection occurring in a previously infected seropositive individual. As discussed earlier, the peculiar replication mechanism of retroviruses allows for genetic recombination to occur when a dual infection has taken place, which in HIV-1 can be intrasubtype, intersubtype, or intergroup. As recently reviewed [97], superinfection was first documented in vitro with multiple strains of HIV-1 [98,99], with the extent of superinfection increasing over time. It has been reported that the replication of superinfecting viruses in PBMCs from HIV-1-infected asymptomatic individuals with high CD4 cell counts, but not from patients with low CD4 cell counts, was suppressed by autologous CD8 cells [100]. This finding would support the notion that during the early asymptomatic stages of established HIV-1 infections the emergence of superinfecting viruses might be inhibited by CD8 cells. In a recent study, HIV-infected cells from lymphoid tissue of seropositive individuals were often multiply infected, and 6–18% of the viruses within multiply infected cells were recombinant strains [101]. Several studies of superinfection in animals [102,103] suggest that superinfection in with different HIV-1 strains may affect the course of the infection. In humans, several studies have documented dual infections with HIV-1 and HIV-2, with viruses of groups M and O, as well as with viruses of different subtypes (reviewed in [97]). Very recently [104], it has been found that splenocytes from two HIV-1-infected patients had a mean of three to four genetically different proviruses per cell and had evidence for huge numbers of recombinants and extensive genetic variation within a single cell (up to 34% amino acid difference), indicating that recombination is very frequent and important in shaping the evolution of HIV in individual patients. In humans, potential superinfection with strains of different subtypes may emerge in IDU sharing injection equipment with direct intravenous inoculation of high viral doses or in persons practicing unprotected sex. Dual infection is a necessary first step for viral recombination and would explain the high rates of unique recombinant HIV-1 genomes in areas where multiple subtypes are circulating. Although several reports of possible HIV-1 superinfection or re-infection have been presented [105,106], evidence based on molecular virology and epidemiological data has only been documented very recently [1,107]. A recently published paper [107] reports two superinfections in IDU in Bangkok. In both cases, the superinfecting strains, which were subtype B and CRF01_AE variants, respectively, were detected by molecular and serologic analyses 3 and 21 weeks after the complete seroconversion, which occurred with a CRF01_AE and with a subtype B strain, respectively. However, the methods used cannot completely exclude the possibility that the transmission of both subtypes occurred simultaneously or very close in time to each other. A very recent abstract [108] describes the case of a patient who, after 1000 days of good response to treatment, had a rebound of viremia from less than 5000 copies/ml to more than 50 000 copies/ml. Sequencing of the original and the new viruses, both of subtype B, showed a genetic difference of more than 12%, suggesting that the patient had been sequentially superinfected or re-infected. A history of sexual contacts consistent with re-infection was also documented. The conclusive evidence of HIV-1 superinfection in humans has been just published [1]. The reported case is that of a man included in the QUEST trial, initially infected with a CRF01_AE strain following multiple unprotected sexual contacts with male partners. At the time of the second viremia rebound, from less than 200 copies/ml to more than 200 000 copies/ml, following discontinuation of treatment and coinciding with new unprotected sexual contacts, a subtype B isolate was found both in PBMC DNA and plasma RNA by analysis of protease, RT, gag and the C2V3 portion of the envelope genes. The possibility of acquisition of both clades from the beginning was ruled out by a polymerase chain reaction assay using subtype-specific primers. The subtype B superinfection led to a rapid progression of the disease, in contrast to the observation in experimental superinfection in monkeys. We have also studied a case of probable superinfection (unpublished data Pérez-Alvarez L. et al.); a female IDU, who shared injection equipment with her sexual partner, another HIV-1-infected patient from the cohort. The analysis of near full-length HIV-1 genome sequences from plasma-derived RNA samples obtained in August 1999 from both subjects showed that both were closely related nonrecombinant subtype G viruses. In August 2001, the V3 sequences from PBMC had switched to subtype B, and a segment of pol (protease-RT) was BG recombinant, with the G subtype segment of pol clustering very closely in phylogenetic trees with the G subtype virus characterized in the previous sample. Similar results were obtained in sequences from cultured primary isolates and from a third sample collected 1 year later. These results are strongly suggestive of superinfection with a subtype B virus, with generation of a BG mosaic form through recombination with the subtype G virus present initially. At the time of the detection of the recombinant virus, the patient was treatment naïve, and had a CD4 cell count of 490/μl and a plasma viral load of 45 700 copies/ml. The recombinant virus was detected after the patient had changed sexual partners, although phylogenetic analyses indicated that the new sexual partner was not the source of the superinfecting virus, but we need to consider that she was sharing, during this time, injection materials with some intravenous drug users. As a consequence of superinfection and recombination, more virulent viruses, single-drug-resistant or multidrug-resistant variants, or viruses with altered tropism may emerge in the infected patient. To know the impact of superinfection on disease progression, it will be necessary to perform prospective studies in superinfected patients. Implications of HIV-1 recombination for vaccine and therapeutic strategies Recombination and resistance to antiretroviral drugs In vitro, recombination may link drug resistance mutations in HIV-1, leading to increased resistance to one antiretroviral [12] or to the generation of multidrug-resistant viruses [13]. It is also conceivable that recombination might facilitate the acquisition of mutations compensating for fitness loss caused by drug resistance mutations. The high frequency of recombination found in areas where multiple genetic forms co-circulate warns against the uncontrolled use of antiretroviral drugs, which could foster the emergence and widespread circulation of multidrug-resistant strains generated through recombination. Implications of recombination for vaccine development The frequent finding of URF reflects the common occurrence of dual infections, and recent reports show that superinfection in established infection is possible. Documentation of superinfection by different genetic forms implies that the immune responses elicited by one HIV-1 genetic form may not protect against infection with another genetic form. This, together with reported preferential intraclade over cross-clade cytotoxic T lymphocyte (CTL) and antibody immune reactivities (reviewed in [109]), these findings would support the use of vaccines with immunogens genetically matching HIV-1 variants circulating in each area, or, preferably, multivalent vaccines incorporating the local variants among its components [109–111]. Another implication concerns the type of immune responses elicited by a vaccine. CTL-inducing vaccines may control viral replication, at least temporally, but do not generally induce sterilizing immunity. It has recently been reported that initial control of viral replication by CTL-inducing vaccines may be lost over time as CTL escape mutations develop [112]. CTL escape mutants can be transmitted and remain stable over many years [113], and may have contributed to the generation of the global HIV-1 diversity [114]. The circulation of variants carrying stable escape mutations, which can accumulate and become linked through recombination, may diminish the efficacy of vaccines. A related point is the possible relaxation in risk behavior avoidance derived from the misperception of being protected from infection by a vaccine that, in fact, is nonsterilizing. This would favor further superinfections, with possible generation of recombinants carrying multiple escape mutations. The implication is that the search for a vaccine with sterilizing activity must remain among the highest priorities, which presently only is conceivable through the induction of protective antibody responses. Conclusions The importance of HIV-1 viral recombination in the HIV-1 epidemic is increasingly being recognized as a result of recent studies. The lack or low numbers of full-length genome sequences from many areas and the difficulty of detecting intrasubtype recombinants may imply that the prevalence of recombinant forms is probably underestimated. The recent report of superinfection in established infection in humans is consistent with the notion that superinfection may contribute significantly to the high frequency of unique recombinants found in many areas, with important public health implications [115]. Follow-up of patients experimenting treatment failure (as judged by viral load response), without a logical explanation, such as selection of resistance mutants or nonadherence to treatment, should provide the basis for investigating the possibility of superinfection. Treatment interruptions may constitute a risk period in which superinfection could occur with higher probability, especially during chronic infection, because of the low levels of CD8 responses and the low plasma drug concentrations. Further studies on full-length genome characterization will provide a better estimation of the prevalence and diversity of both intersubtype and intrasubtype recombinant forms, and continued surveillance is needed to assess the temporal trends in their incidence. These studies are particularly important for the planning and interpretation of the results of vaccine trials. As discussed, the information obtained from these studies might have important implications for vaccine and therapeutic strategies [111]. Acknowledgments The authors thank the Health and Social Services of the Government of Galicia (Dr José Marí Hernández Cochón), the Spanish National AIDS Plan (Dr Francisco Parras), and the WHO-UNAIDS Vaccine Initiative (Dr José Esparza and Dr Saladin Osmanov) for continuing encouragement and financial support for the studies described in the present paper. Addendum Since the preparation of this manuscript the report mentioned in reference 108, has been published [116]. This paper represents the second fully documented case of superinfection and the first to show a very important point, that ‘virus-specific CD8+ T-cell responses that control replication of one strain of HIV-1 may not be sufficiently potent to prevent superinfection with a second virus of the same clade’, emphasizing the importance of reinfection or superinfection for the pathogenesis of the disease, therapy and vaccine development.