Portal de Periódicos da CAPES

The public health significance of HIV-1 subtypes

2001; Lippincott Williams & Wilkins; Volume: 15; Linguagem: Inglês

10.1097/00002030-200100005-00009

1473-5571

Iain Tatt, K. L. Barlow, Angus Nicoll, Jonathan P. Clewley,

HIV/AIDS drug development and treatment

Introduction Since the first descriptions of HIV/AIDS, HIV-1 has emerged as the most significant pathogen worldwide, with approaching 60 million infections to the end of 2000. Over 18 million deaths from AIDS have occurred, and HIV/AIDS is now classed among the top five causes of global mortality [1]. As with other infectious pathogens that exhibit a high degree of genetic variation, such as influenza, the extensive genetic variability of HIV-1 poses significant challenges for disease control and epidemiological surveillance [2]. The genetic heterogeneity of HIV-1 isolates is one of the major characteristics of the virus and the epidemic. This diversity, generated by the processes of mutation and recombination [3], has led to the development of a subtype nomenclature for the classification of isolates [4]. The major (M) group of HIV-1, responsible for the majority of infections in the epidemic, contains several recognized non-recombinant subtypes and sub-subtypes, and at least 10 circulating recombinant forms (CRFs) [4]. These subtypes show uneven regional distribution, and past attempts have been made to use them to describe the global molecular epidemiology of HIV-1 [5,6]. In this context, undertaking systematic sampling and monitoring of populations is important to determine the value of these studies. We review here the public health and clinical significance of HIV-1 genetic variability, and the methods and rationale for public health surveillance of subtypes. Diversity of subtypes Epidemiological and genetic evidence suggest that the two main types of HIV (HIV-1 and HIV-2) are zoonoses (human infections originating from animals), acquired via several independent transmission events from the chimpanzee (Pan troglodytes) and sooty mangabey (Cercocebus atys), respectively [7,8]. While HIV-1 appears to have been present in humans since around the 1930s [9], modern sociodemographic changes together with chance introductions into vulnerable populations have created optimal conditions for its rapid spread [10]. HIV-1 genetic variability is generated by the lack of proof-reading ability of the reverse transcriptase (RT) [11], the rapid turnover of HIV-1 in vivo[12], host selective immune pressures [13], and recombination events during replication [3]. This variability has allowed HIV-1 to be classified into the M group of sequences as well as outlier (O), and non-M, non-O (N) groups [14-16]. Recently, guidelines for the definition of subtypes and CRFs have been proposed that take into account the genetic diversity and evolution of the epidemic [17]. A new subtype or CRF may be defined by the existence of at least three fully characterized genomes, obtained from epidemiologically unlinked individuals. Phylogenetic analyses should group these sequences approximately equidistantly from all other subtypes. At the time of writing, group M consists of nine 'full-length' subtypes, designated A-D, F-H, J and K [4,14]. Further analysis of complete genome sequences from viruses initially described as belonging to subtype E (now classified CRF_01AE) and subtype I has revealed these viruses to be recombinant strains, and not distinct subtypes as initially proposed [18,19]. More detailed phylogenetic analysis has also led to the re-classification of some subtype groupings. For example, subtype F was originally classified into three sub-subtypes (F1, F2 and F3) based on gag and env phylogenetic comparisons [20]. However, subsequent full-length genome analysis has led to the re-classification of sub-subtype F3 as subtype K [21]. Groups O and N currently represent only a small number of characterized strains, although env and gag sequence analysis of group O isolates has demonstrated that they belong to four distinct phylogenetic clusters of similar diversity to the group M subtypes [22]. Inter-subtype and intra-subtype recombination may occur within dual-infected individuals to generate novel recombinant strains, even between highly diverse viruses such as those belonging to different groups [23,24]. There are currently 10 CRFs responsible for sub-epidemics worldwide (http://hiv-web.lanl.gov/CRFs/CRFs.html). In 1995, it was estimated that recombinant viruses accounted for up to 15% of global infections [25], and we have shown that up to 20% of heterosexual infections in the United Kingdom may be attributable to recombinant viral strains [26]. Additional CRFs and mosaics continue to be characterized, such as the recent discovery of a C/D CRF (CRF10_CD) found among perinatally infected infants in Tanzania [27], and an A/J recombinant infecting pregnant women in Cameroon [28]. Given the continued global spread of HIV-1, it is likely that recombinants will play an increasing role in the generation of HIV-1 diversity [6,23]. Genetic sequencing is the definitive method for subtyping [29]. Despite advances in long-polymerase chain reaction (PCR) technology [30], practical limitations have historically led to a subtype being assigned based on single genomic regions, most commonly env, gag, or pol[31]. There are several available techniques for subtype assignation, and deciding which is appropriate will often depend on the objectives of the surveillance strategy (Table 1).Table 1: HIV-1 subtype classification techniques.Serological assays have proved useful for screening populations of limited subtype diversity, especially where one subtype predominates [32,33]. However, such assays are not applicable where multiple subtypes co-circulate as they can lead to subtype misclassification [34-36]. In such circumstances, PCR-based subtyping approaches, most commonly the heteroduplex mobility assay, have been widely applied [37-41]. In addition, by obtaining partial sequences from different gene regions, most commonly gag and env, recombinant prevalences can be estimated and potentially interesting genomes identified [26,39-41]. Epidemiology and global distribution of subtypes There have been numerous descriptions of the HIV-1 genetic diversity within populations [5,6,42]. However, most studies have derived estimates using unsystematic, opportunistic sampling frames with relatively small numbers of specimens [43-46]. As such, current understanding of the distribution of HIV-1 subtypes may inadequately represent the worldwide diversity of the virus [6]. In particular, reporting bias can distort the global description through an over-representation of unusual subtypes and limited surveillance of populations with a more homogeneous subtype profile. Equally, the complex distribution of HIV across different transmission modes and sexual groups, coupled with the fact that transmission rates may differ within groups even within a single country or locality, complicates the surveillance of HIV infection. The greatest genetic diversity is observed among HIV-1 strains from Africa [47] where transmission occurs primarily via heterosexual, iatrogenic (blood transfusion) and mother-to-child routes [48]. Distinct geographic subtype patterns are seen within Africa, with subtypes A and D most prevalent in Eastern Africa [49,50], subtype A in Western Africa [51,52], and subtype C in South Africa [53], although the epidemic in the latter region appears to be increasing in diversity, with subtypes A, B, D and recombinant forms now prevalent [39,54](Fig. 1). Phylogenetic analysis of the gag gene from isolates obtained from Senegal, Cameroon, Gabon, and Djibouti has suggested that between 60 and 84% of subtype A viruses circulating in Africa may in fact be similar to CRF_02AG [55], and not 'pure' subtype A viruses as previously described.Fig. 1: Geographical distribution of HIV-1 major (M) group subtypes. For clarity, only the most prevalent subtypes are shown. CRF, circulating recombinant form.In Asia, an epidemic spread of HIV-1 is occurring with fewer co-circulating subtypes than in Africa. The epidemics in India and China are dominated by subtype C, the most prevalent subtype worldwide, which is thought to account for approximately 50% of infections [56,57]. Subtypes A, B, and recombinant strains have also been identified [56]. In contrast, two strains, subtypes B and CRF_01AE, co-circulate within the injecting drug user (IDU) population in Thailand [58]. These two subtypes have also been shown to co-circulate among Thai fishermen [59]. The original spread of HIV-1 in North America, Western Europe, and Australia was of subtype B strains among the MSM and IDU populations [14,60]. Heterosexual spread is increasing in these regions with non-B subtypes and recombinant strains observed, mostly among individuals with epidemiological links to areas where multiple subtypes co-circulate [26,43,44,61-65]. Most South American viruses appear to be subtype B, although there also the epidemic is increasing in diversity with subtypes A, C, and F, and recombinant viruses being reported [66,67]. The continued global expansion of subtypes has created the potential for recombinant virus generation and transmission, and the number of reports of recombinant infections is increasing. In part, this is due to improvements in surveillance methodology, and recombinant strains have been detected in many countries [26,31,55,67-71]. Intersubtype recombinant forms now appear to be circulating among IDU in China (B/C) [56], North Vietnam and Southeast Asia (CRF_01AE) [72], Argentina (B/F) [73], and Russia (A/B) [70]. Public health implications of genetic diversity The genetic variability of infectious pathogens, and any consequent phenotypic variation, poses a significant challenge to disease control and surveillance [74,75]. Differences between subtypes in their viral properties can be argued from several lines of evidence. HIV-1 and HIV-2 show approximately ∼50% similarity at the nucleotide level [14] and differ in their pathogenicity and transmissibility [76,77]. While both are zoonoses [7,8], neither virus appears harmful to their natural primate hosts [78], and yet they are pathogenic when transferred to foreign primate host species [79]. In addition, the extraordinarily rapid rate of HIV-1 evolution in humans, estimated at around 0.0024 substitutions per base pair per year [9], coupled with the recombination events that may introduce large-scale genetic changes, creates the conditions for increased viral evolution. Interventions such as antiviral drug therapies may further accelerate genetic change through the preferential selection of resistant strains [80]. As we will describe, such variability may impact on vaccine modelling and diagnostic testing, the generation of antiviral resistance, and aid the surveillance of transmission patterns within the epidemic. Selection of candidate vaccines The development of a protective vaccine remains the greatest challenge in the global battle against HIV-1. It has been hindered by the lack of a suitable animal model, the limited understanding of the markers of protective immunity, and the intracellular mode of transmission and persistent nature of HIV infection in the immune system [81]. Several approaches have been applied to the development of vaccines with varying degrees of success, including epitope vaccines [82,83], live vector constructs containing various combinations of env, gag, pol, and nef[84,85], DNA vaccine constructs [86,87], live attenuated nef-deletion vaccines [88], and whole-killed viral constructs [89]. The genetic variability of HIV-1 complicates vaccine development [90]. The ability of HIV-1 to mutate rapidly during the course of infection can result in the generation of escape mutants that evade host immune recognition [91]. The generation of a quasi-species of genome sequences over the duration of infection [92], coupled with the extent of diversity of the genome, raises the probability that an immunogen specific to a single viral subtype may not elicit an immune response that targets all viral strains. Despite this, approaches that utilize well-conserved viral proteins such as the regulatory proteins Tat and Rev, or accessory proteins such as Nef, Vif, Vpr and Vpu, may prove less susceptible to mutations than approaches that utilize structural proteins such as Env. For example, the HIV-1 regulatory protein Tat has been shown to possess limited antigenic polymorphism across geographically diverse strains representing subtypes A, B, C, D, F, and G [93], and to elicit both a humoral and cellular specific immune response in animal models [86,94]. The majority of candidate vaccines currently undergoing trials have been developed using subtype B viral strains (most commonly HXB2 or MN), representing the predominant type in North America and Europe. Their ability to elicit a cross-subtype response is not yet fully understood. Some studies have suggested neutralizing antibodies to be subtype specific [95,96], while others have indicated that cross-clade cytotoxic T-lymphocyte responses may occur [97,98]. Trials of a subunit vaccine containing gp120 env subunits from subtypes B (MN) and CRF_01AE (CM244) among IDU in Thailand have suggested dual-antibody responses may be generated [99]. However, the extent to which epitopes are conserved across subtypes remains unclear. Whatever vaccine strategy is adopted, information on local subtype prevalence, incidence, and genetic diversity within the population will prove crucial, as will close monitoring for vaccine failures. Impact on HIV testing It is essential that commercially available assays are capable of diagnosing infection in all HIV-1-positive individuals, including those infected with less prevalent, more diverse subtypes. If this is not the case, accurate patient diagnosis, serosurveillance, and clinical assessment are all threatened. Equally, reports that some viral subtypes can 'escape' HIV testing may damage public confidence in health services or even the belief that HIV is pathogenic. The genetic variability of HIV-1 has been shown to affect the accuracy and sensitivity of diagnostic assays [100-113]. The principal method of HIV diagnosis is the detection of HIV antibodies, but the antigenic epitopes used in such tests may not detect more diverse viral strains. Early diagnostic assays using synthetic peptides or recombinant antigens failed to identify group O and HIV-2 infections, a feature overcome by the incorporation of antigens appropriate for these variants into later generations of tests [114]. A recent study of six United States Food and Drug Administration licensed HIV-1 immunoassays suggested that all were capable of detecting infections with subtypes A-G, J and group O [102]. Cross-reactivity has also been observed between highly divergent group N sera and group M test antigens [15,115]. The detection of newly acquired HIV-1 infections is of paramount importance in the control and management of the epidemic. Such estimates allow primary prevention strategies, and aid in the treatment of infected individuals. Recent HIV-1 seroconversions have been found with the use of serological tests based on sensitive versus less-sensitive (detuned) antibody reactivities: early antibodies are present at lower levels and are not detected in the less-sensitive detuned assays [112,113,116]. However, subtype-specific differences in this type of assay have been described, with subtypes B and E giving different window periods of infection (155 and 270 days, respectively) [113]. Also, the 3A11-LS (Abbott, Abbott Park, Illinois, USA) assay has been shown to misdiagnose late-stage infections in a small proportion of subtype E infections [113], and to misclassify infections from some patients with late-stage AIDS [112]. At present, such detuned assays are based on modified commercial assays that utilize a subtype B derived antigen. Therefore, for them to be confidently applied in settings where non-B subtypes are common, appropriate cutoffs and window periods for all HIV-1 subtypes must be determined. In some circumstances, such as for the diagnosis of infection among infants born to HIV-positive mothers or for early HIV diagnosis, it is more appropriate to detect viral antigens directly rather than anti-HIV antibodies [117]. Recently, a comparison of four commercially available p24 antigen assays has shown the sensitivity for detection of diverse samples to vary between assays [118]. The most recent diagnostic assays (fourth generation) combine the detection of anti-HIV antibodies and p24 antigen within a single test. These have been shown to narrow the window period of infection (the time between infection and reactivity in a screening test) by up to 9 days [119,120]. Such assays have therefore been recommended as initial screening tests for early infections. The use of PCR for proviral DNA detection appears more sensitive than p24 antigen assays, but here too diverse subtypes have been shown to give false-negative results [106,107,121]. In summary, therefore, commercial diagnostic tests for both HIV antibodies, HIV-1 proteins such as p24, and for HIV-1 DNA and RNA by PCR may all have reduced sensitivity for certain variants of HIV-1. Assays that quantify the viral load within an individual play a key role in monitoring disease progression and aiding prognosis. High viral load during the acute phase and before seroconversion has been linked with more rapid disease progression [122], decreased numbers of peripheral CD4 T cells [123], increased sexual and vertical transmission [124,125], and failure of antiviral therapy [126]. Currently, four types of assay are available for viral load quantification: Amplicor HIV-1 Monitor v.1.5 (Monitor; Roche Diagnostics, Branchburg, New Jersey, USA) [127], nucleic acid sequence-based amplification (NASBA; Organon Teknika, Boxtel, The Netherlands) [128,129], HIV-1 Quantiplex branched DNA (bDNA; Chiron Diagnostics, Emeryville, California, USA) [130,131], and the LCx HIV RNA Quantitative assay (LCx; Abbott) [132]. The methodology employed for viral load quantification differs between these assays. Both the Monitor and NASBA assays use amplification of a region of the gag gene from viral RNA, with the Monitor assay generating and then amplifying DNA whereas NASBA amplifies multiple copies of viral RNA. Both the bDNA and LCx assays target the pol gene, with the bDNA assay utilizing sequential hybridization of multiple probes to target sequences and the LCx assay employing a competitive RT-PCR approach. The ability of these assays to reliably quantify viral load from non-B subtype infections has been examined in several studies [100,104,105,109,132-135]. Subtype-associated differences in sensitivity, particularly for subtypes A and G, have been described [100,104,133]. For example, a study comparing the quantification of RNA from samples of subtypes A-H showed a failure of detection of RNA in some samples of subtype A and G with early versions of the Monitor and NASBA assays [133]. Current versions of the LCx, bDNA, and Monitor assays have been shown to reliably quantify HIV-1 RNA from subtype A-G infections [109]. The use of PCR may lead to errors in viral load quantification due to mismatches between the primer and the template [109,110]. Primer mismatches have been linked to the failure to quantify viral load from group O infections by the Monitor assay, infections that were detected by the LCx assay [109]. Such problems may be reduced, however, by the continual updating of primers/probes and assay conditions that take into account the genetic diversity of HIV-1. For example, the addition of new primer sets to the original Monitor assay was shown to increase its ability to reliably quantify subtypes A and E [105]. Diagnostic assays that utilize primers located in the conserved pol gene, such as the LCx assay, have also been shown to amplify examples of the highly divergent group N viruses [136]. In areas where subtype variability is greatest, therefore, an assay that targets a highly conserved target region, such as pol, may provide the most reliable estimates of viral load. Alternatively, a non-PCR approach such as the bDNA assay may also provide accurate results [105]. There is a need for well-characterized viral genome sequences of different subtypes to accurately compare the performance of various tests. A quality control panel containing HIV-1 isolates initially classified as subtypes A-G on the basis of phylogenetic analysis of their gag and/or env genes is available for this purpose [110]. However, further analysis of the gag, pol, and env regions from this panel has shown it to contain four intersubtype recombinants and two group O viruses, highlighting the need for complete genome characterization of putative reference strains [109]. The increasing genetic diversity of HIV-1 means that the evaluation and modification of nucleic-acid based diagnostic viral load and detection assays needs to be an ongoing process to ensure reproducible quantification for all HIV-1 strains. Genetic markers of antiviral resistance The emergence and transmission of HIV-1 resistance to antivirals in areas where those therapies are commonly used poses a serious challenge to individual patient management [137]. Detecting the phenotypic expression of resistance by growth of virus is difficult, and therefore genetic markers associated with resistance are monitored by sequencing. A review of the genetic basis of antiviral resistance is beyond the scope of this article and is available elsewhere [80]. It is reasonable to expect subtype-associated differences in antiviral resistance given that diverse viral forms such as HIV-1 group O have been shown to possess inherent resistance to non-nucleoside reverse transcriptase inhibitors [138]. The high level of genetic diversity of HIV-1, coupled with the fast turnover of virions, has been shown to lead to the rapid generation of drug-resistant strains [139]. The introduction of sequence-based assays as an aid to the clinical management of infection, as well as studies of antiviral resistance in treatment-naïve populations of heterogeneous subtype, has allowed a more complete picture of subtype-associated mutation patterns to emerge. Naturally occurring drug resistance mutations have been demonstrated among drug-naïve individuals infected with subtypes F and G [140,141], and mutations associated with resistance to protease inhibitors (PRI) have been shown in PRI-naïve individuals to be more prevalent among non-B subtypes than among subtype B strains [142]. Furthermore, individuals infected with subtypes A-D and CRF_01AE may possess naturally occurring mutations at sites that contribute to PRI resistance in subtype B viruses [143]. However, contradictory evidence is provided by several studies that have demonstrated no difference in the key resistance mutation profiles among individuals infected with a range of subtypes and recombinant virus [26,144]. Therefore, further research is required to determine what, if any, role the subtype may play in the development of primary antiviral resistance. Subtype-associated differences may exist among the secondary (minor) drug resistance mutations. The M36I protease resistance mutation is present in subtype C viruses from untreated individuals, both in the United Kingdom and elsewhere [26,145,146]. Secondary mutations contributing to both protease (M36I and M46L) and RT (A98S and T200A) inhibitor resistance have also been found to be more common among non-B subtypes compared with subtype B viruses [144]. A 6-year study of drug resistance in primary HIV-1 infections in the United Kingdom has shown that, among persons infected with a subtype B virus, the risk of being infected with a drug-resistant virus has increased with time [137]. Monitoring the transmission of drug-resistant variants may therefore become necessary if the treatment of newly infected individuals is not to be compromised. Molecular markers of transmission patterns The characterization of circulating viral strains can perform an important role in the tracking of the epidemic. For example, the genetic variability of HIV-1 can be used to study transmission patterns, both to indicate individual transmission events [147-149] and to monitor the more general spread of the epidemic. Recent individual transmission events can be detected with great precision, although it is often difficult to determine the direction of infection. In one recent Scottish case, research data from genetic typing led to the criminal prosecution of a person who infected his partner when seemingly aware of his HIV-positive status. Epidemiological studies incorporating subtype characterization has allowed detection of the introduction of distinct HIV-1 subtypes into China, Thailand, the Philippines and several African countries [56,150-152]. In some instances, subtype may be associated with a particular route of transmission, such as in Thailand where subtypes B (in MSM and IDU) and CRF_01AE (in heterosexuals) were independently introduced over a similar time period, although these subtype patterns are now less distinct [58,150,153]. Phylogenetic evidence has also dated the introduction of the founder subtype CRF_01AE virus into the heterosexual Thai population to around 1986-1987 [9]. Biological variation: transmissibility The possibility that HIV subtype may influence viral transmissibility and pathogenicity has been suggested [42,154,155]. Many other factors influence HIV-1 transmission including co-infection with other sexually transmitted diseases, host-immune and protective factors (as diverse as HIV co-receptor mutations and male circumcision) and the risk behaviours of the individual [155]. These variables confound attempts to determine the effect of subtype on viral transmissibility. For example, the co-existence of the two epidemics in Thailand and some growth properties of subtype CRF_01AE in Langerhans cells [156] led to the erroneous hypothesis of a greater transmissibility of subtype CRF_01AE viruses. However, the laboratory findings could not be substantiated [157,158] and the epidemiological evidence supports founder effects for the historical presence of CRF_01AE in different populations [159,160]. There is now crossover between these two subtypes and populations. Studies of per-sex-act transmission probabilities in populations where HIV-1 subtype B predominates have suggested lower transmission rates than those where non-B subtypes are most prevalent [161,162]. Recently, a study in Tanzania suggested maternal subtype may play a role in vertical transmission, with subtypes A, C, and recombinant viruses being more likely to be perinatally transmitted than subtype D, possibly due to the use of the co-receptor CCR5 [163]. A consistent subtype-associated difference in transmissibility has yet to be proven, however, with similar rates of both sexual and vertical transmission, irrespective of subtype, having been observed [164,165]. Indeed, the co-existence of many HIV-1 subtypes rather than just one or a few would suggest that subtype alone is unlikely to be the main factor in viral transmissibility. In conclusion, this suggests that any differences are more likely to be due to factors other than viral subtype, such as higher prevalences of other sexually transmitted infections where non-B subtypes predominate [166]. The lack of conclusive evidence for an association between subtype and transmission does not preclude the existence of such an effect, but implies that recognizing it will require studies utilizing epidemiological, molecular, immunological and statistical methods in those areas where multiple subtypes co-circulate. Biological variation: pathogenicity The rate of disease progression among HIV-1-infected individuals varies widely [155,167,168]. As well as antiretroviral treatment, progression is influenced by host factors including chemokine co-receptor and human leukocyte antigen genotype [13,169], the route of transmission and age at time of infection [170], the presence of another sexually transmitted disease [171], ethnicity and individual socio-economic conditions [172], and viral characteristics [173]. The influence of subtype on viral pathogenicity has been investigated in several studies. It has been suggested that the chemokine co-receptor may differ between subtypes with, for example, the X4 (CXCR4) phenotype (previously called rapid/high or syncytium inducing) being under-represented among subtype C isolates [174-176]. While the X4 phenotype has been associated with disease progression among individuals infected with subtype B, there is no evidence that subtype C viruses are less pathogenic than other subtypes. A link has been suggested between the presence of an additional NF-κB site among subtype C viruses and enhanced viral transcription and replication [175,176]. However, sequence analysis of subtype C isolates from Zambia, India, Tanzania, South Africa, Brazil, and China has revealed that this extra site is present in only a small number of subtype C viruses [177]. This finding emphasizes the need to analyse large numbers of sequences before drawing conclusions about differences in subtype-specific biological properties. Studies of the relationship between HIV-1 subtype and disease progression have produced contradictory results. One study, comparing Swedish patients infected with subtype B strains and Africans resident in Sweden infected predominantly with non-B subtype virus, found no difference in the rate of CD4 cell loss or disease progression over a 2-year period [178]. Similarly, two further studies of disease progression, one comparing subtype B-infected native Israelis and subtype C-infected Ethiopian immigrants in Israel [179] and another comparing people infected with either subtype B or CRF_01AE in Thailand [180], have found no differences in rates of progression. However, a cohort study of female sex workers in Senegal indicated that individuals infected with non-A subtypes progressed faster to disease than those infected with subtype A [181]. Several studies have indicated a more rapid rate of disease progression among people infected in Africa compared with those infected in Western Europe and the United States [167,181,182]. However, a lack of association between host ethnicity (taken as a marker for viral subtype) and disease progression has been shown in several studies [171,172,178]. It is therefore unclear to what extent such differences may be explained by environmental factors such as the quality of medical services, access to antiretroviral regimens, and the prevalence of concomitant infections, as opposed to viral and/or host factors [155]. Epidemiological surveillance strategies The impact of the genetic diversity of HIV-1 on public health measures (summarized in Table 2) necessitates surveillance strategies that systematically sample and characterize representative strains from populations at varying risks of infection. However, to date, most available subtype data has been derived from studies that employ unsystematic, opportunistic sampling frames (Table 3).Table 2: Role of systematic surveillance of HIV-1 subtypes.Table 3: Sources of specimens for HIV-1 subtype surveillance.The surveillance strategy chosen for monitoring HIV-1 diversity will depend on the particular objectives of the programme (Table 3). Prospective cohort studies have been applied in several countries; however, these are costly and often not entirely representative of persons at highest risk of infection [59,151]. Therefore, a number of countries have begun to sample blood collections for subtyping of all, or a sample of, newly diagnosed individuals (C. Archibald, R. Janssen, personal communication, 2000). This strategy may not be applicable in resource-poor settings where the majority of new infections are occurring, however, as it may miss trends in undiagnosed infections. The implementation of unlinked, anonymous surveillance of population groups at varying risks of infection has been shown to produce minimally biased estimates of subtype prevalence (Table 3)[183-185]. By these means, for example, it is known in the United Kingdom that approximately 25% of HIV-1-seropositive individuals are infected with a non-B subtype [185]. In addition, up to 20% of the infections in heterosexuals in the United Kingdom are with a recombinant non-B subtype virus [26]. While such studies have proved useful in estimating incidence and in detecting the presence of new subtypes and recombinants, they can be costly and logistically difficult to undertake. Obtaining samples that are representative of populations is desirable but often difficult to achieve, and a pragmatic approach has been to utilize consistent sampling frames so as to detect important changes over time [46,59,186]. Conclusion The genetic variability of HIV-1 determines the way in which the epidemic is monitored, and is also a significant factor in the management of individual infections. HIV-1 diversity is likely to play a role in the generation of secondary anti-viral resistance, although to date there is less convincing evidence of significant differences in transmissibility or pathogenicity of the subtypes. Numerous epidemiological studies have provided a picture of the global subtype distribution and these show increasing evidence of the importance of HIV-1 recombinants. Knowledge of the subtype distribution is of importance in the development of vaccines and is essential to ensure that tests used to diagnose and guide clinical and therapeutic monitoring of HIV infections remain fully sensitive and specific. Viral sequencing has proved useful in determining epidemiological linkage between infected individuals, and subtype surveillance serves as a guide to transmission patterns at broader levels. From this, a comprehensive picture of the worldwide pandemic can be obtained by combining epidemiological, molecular, and demographic data. Epidemiological surveillance protocols therefore need to incorporate molecular techniques that are appropriate to the subtype prevalence in the population of interest, and which also retain the ability to detect the introduction of novel subtypes. The reliance of the majority of current algorithms on subtyping a single genomic region diminishes their ability to detect unusual and new recombinant strains that may be biologically important. Sequencing multiple loci may also be epidemiologically informative for detecting transmission patterns. The development of algorithms that utilize several regions of the genome, and where necessary incorporate sequencing, will play an important role in future surveillance of the epidemic. To obtain a comprehensive picture of the diversity of HIV within a population, attempts should be made to prospectively sample a representative cross-section of individuals, incorporating groups at varying risk of infection, as well as cases with diverse modes of transmission. It is important that sampling strategies include and identify recent infections so as to provide a measure of current transmission patterns. However, recent seroconvertors are not always easy to detect, particularly in those groups not recognized to be at high risk of HIV infection, where diagnosis may not occur until late in the course of infection. The use of serological tests for HIV incidence, the so-called detuned antibody assays, may make the detection of recent infections easier. The collection of epidemiological and demographic data at source for each specimen is also necessary to enhance the future understanding of transmission patterns based on the subtypes of HIV-1. Acknowledgements The authors thank Philip Mortimer, John Parry, and Gary Murphy for their critical appraisals of this review. K.L.B. is supported by the Department of Health, and I.D.T. is supported by a University of Cambridge, Department of Clinical Medicine and Clinical Veterinary Medicine Simms Fund Scholarship. They also thank the many contributors to the Unlinked Anonymous HIV Prevalence Monitoring Programme throughout England and Wales, and especially the HIV Surveillance Group, Sexually Transmitted and Blood Borne Virus Laboratory, CPHL.