HIV‐1 Genetic Diversity Surveillance in the United States

Few systematic efforts have been made to characterize the prevalence of HIV types, HIV-1 groups, and HIV-1 group M subtypes within a specified geographic area, and few US studies like the one by Sides et al. [1] in this issue of the Journal of Infectious Diseases have been published. Their findings are important: they demonstrate a 95% prevalence of non-B HIV-1 subtypes; a predominance of subtypes C, A, and CRF02_AG/A1; and a substantial diversity of subtypes among the 87 HIV-infected African-born individuals in the targeted surveillance portion of their Minnesota study. African-born individuals contribute 20% of new HIV reports in Minnesota, resulting in a genetically diverse HIV epidemic. These findings, from a state where African-born individuals make up <1% of the population, suggest that it may be time to consider HIV genetic diversity surveillance in the United States As Sides et al. note, several reports and abstracts have noted the appearance of HIV-1 group O, HIV-2, and non-B HIV-1 subtypes in the United States. These studies do not contradict the general belief that these strains contribute minimally to the US HIV epidemic. As reviewed by Sides et al., reports of HIV-2 and HIV-1 group O have been confined to small numbers of cases in a few articles. For non-B HIV-1 subtypes, studies that have included prevalence estimates have reported a non-B subtype prevalence of <5% among predominantly US-born populations with HIV infection. With the exception of scattered small analyses of travelers, military personnel, and partners of foreign-born individuals, US studies have reported HIV strains other than HIV-1 subtype B to be associated with origination in countries where other strains predominate, as Sides et al. now report in Minnesota The United States has no national estimates of HIV genetic diversity, and no prevalence studies have been performed in most areas of the country. Those studies that have been performed might not have been large or representative enough to detect a low prevalence of unusual strains. The fact that varied strains of HIV, including HIV-2, are found not only in Africa and Asia but also in the Philippines [2] and Latin American countries should alert us that the United States may see the increases in locally transmitted viral diversity presently being seen in many European countries. It can be assumed that strains previously associated with foreign birth, military service abroad, or travel will gradually become more prevalent among US-born nontravelers, but we do not know the extent to which this has occurred already. The study by Sides et al. shows that non-B subtypes may contribute substantially to an HIV epidemic in unexpected areas of the country, but their sample of US-born individuals newly identified as HIV infected was not large enough to detect local transmission if it exists at a low level Does the United States need a genetic-diversity surveillance system to alert us when such transmission begins to occur in particular areas? Would national and local HIV genetic-diversity estimates serve as a basis for public-health action, as surveillance data should? Surveillance of genetic diversity would inform our use of laboratory tests for HIV diagnosis, screening, and patient monitoring and would ensure that appropriate specimen panels would be used in the development of new tests for use in the United States. A surveillance system could support the selection of optimal treatment regimens; allow the epidemiologic investigation of transmission patterns and unusual strains; increase our understanding of local HIV epidemics; support vaccine development and other preventive measures; and contribute sequences to research databases Reports from HIV screening and diagnostic tests could underestimate HIV genetic diversity in the United States, because the technology in use may not detect or distinguish among variant strains. All 4 HIV screening EIAs approved for use in the United States [1] can diagnose HIV-1 regardless of subtype, but they vary in their ability to detect HIV-1 group O. Only 2 of 4 can detect HIV-2 infection [3–6 ], and a survey by the Association of Public Health Laboratories found that, in 2004, 63% of US laboratories used an HIV-1 viral lysate diagnostic test that does not detect HIV-2 [7]. If HIV-2 were being transmitted in areas where older technology is being used, we might not recognize the phenomenon. Blood donations in the United States are screened to detect antibodies to HIV-1 groups M and O and HIV-2, but the qualitative nucleic-acid test used to screen for very recent infection will detect HIV-1 groups M and O but not HIV-2 [8]. The yield for HIV-2 is so low—12 positive results among 37 million tests in the first years of screening—that the addition of HIV-2 is not considered to be cost-effective [9]. Most rapid HIV diagnostic tests [10, 11], which are beginning to be used more widely, will detect HIV-2 and all HIV-1 subtypes but do not distinguish among them. No fourth-generation antibody/antigen HIV diagnostic tests [12] are licensed for use in the United States; if surveillance were to show that variants were becoming more common where older diagnostic tests are in use, the capability to distinguish among variants might be demanded Anomalies between clinical symptoms and HIV serological results—or, in diagnosed HIV infections, between apparent clinical stage and viral load—may be reported to the US Centers for Disease Control and Prevention (CDC) as potential “cases of public health importance.” These anomalies have occasionally been found to result from infection with HIV-2, HIV-1 group O or N, or non-B HIV-1 group M subtypes [13] (R. Phelps, HIV Incidence and Case Surveillance Branch, Division of HIV/AIDS Prevention–Surveillance and Epidemiology, National Center for HIV, STD, and TB Prevention, CDC, personal communication), but only a handful of such cases are investigated each year. Investigations occur only when health departments request the CDC’s help; whether this system would detect an increasing number of such cases is uncertain. Older tests of viral load do not accurately characterize numbers of copies of non-B HIV-1 subtypes [14], but newer tests cope well [15]. The proportion of patients for whom viral-load monitoring is performed by use of the older generation of tests is unknown. No viral load test available in the United States will adequately quantify the load of HIV-2 or HIV-1 group O [16–18 ]. Surveillance information that shows an increase in the prevalence of non-B HIV-1 subtypes might prompt clinicians to use new technology; the presence of HIV-2 or HIV-1 group O would require alternative means to monitor disease progression and treatment outcomes for some patients The US Food and Drug Administration has approved HIV drug-resistance genotyping kits only for HIV-1 subtype B. Studies have indicated that the Celera Diagnostics ViroSeq HIV-1 diagnostic system will adequately sequence diverse non-B subtypes [19–24 ], by use of specimens from patients with high viral loads, although some primers fail or produce ambiguous results. Some problems have been reported with the Bayer Diagnostics TRUGENE HIV-1 kit [19, 21–23 ]. Specialist laboratories with their own home-grown methods and the large commercial laboratories generally have alternate sets of primers that successfully deal with non-B subtypes [21, 25]. Large increases in international funding for HIV diagnosis and treatment in countries where HIV-1 subtype B is relatively rare may facilitate the rapid development of new laboratory tests that perform well with diverse strains, but manufacturers may not think it worthwhile to apply for licensing in the United States unless surveillance statistics indicate a need Surveillance information on the HIV strains circulating in a particular area could influence decision-making in the selection of antiretroviral drug regimens. Clinicians should be alerted to the presence of HIV-2 and HIV-1 group O, because naturally occurring major mutations in reverse transcriptase render these strains resistant to the nonnucleoside reverse-transcriptase inhibitors [26, 27]. No alternative regimens are recommended at present for the treatment of infection with non-B HIV-1 subtypes, although antiretroviral drugs have been developed with HIV-1 subtype B used as the genetic backbone. A higher prevalence of protease polymorphism and minor drug-resistant mutations are found among HIV-1 non-B subtypes [28], but there is no evidence that these influence the initial treatment response. Large, well-designed studies have so far shown an equally good response when highly active antiretroviral therapy (HAART) regimens are used, regardless of HIV-1 subtype [29–33 ]. HIV drug resistance does not seem to emerge at higher rates, and the same major resistance mutations tend to develop [34–38 ], although it appears that different pathways to resistance are taken in different proportions according to subtype [39]. A reasonable working assumption is that all HIV-1 subtypes will respond equally well to first-line HAART, but in-depth studies have not been performed for all subtypes, and more information is needed on whether the genetic backbones of different subtypes could affect outcomes with second-line or salvage therapy once resistance to 1 or 2 drug classes has evolved. If more detailed studies on specific non-B subtypes lead to specific recommendations for differential treatment, genetic-diversity surveillance information could support guidelines for pretreatment testing for subtype in certain geographic areas Molecular epidemiological studies have been important in tracing the introduction of new HIV types, groups, and subtypes into various countries and regions; genetic diversity surveillance would provide a more systematic method of evaluating the movement of various strains across geographic regions. Such studies have also been used to evaluate clusters where potential links exist without obvious modes of HIV transmission [40–42 ]. To date, no HIV strain has been identified that is analogous to tuberculosis strain W, a highly virulent and multidrug-resistant strain of Mycobacterium tuberculosis that spread across the United States and was tracked by molecular epidemiological methods [43]. Surveillance databases could be useful in ascertaining the extent to which HIV strains of interest are spreading geographically. Although very large databases from commercial genotyping laboratories exist in the United States and have been queried for the presence of particular strains, they include only specimens sent by clinicians for clinical testing. The lack of a representative surveillance system means that existing cases could be missed, particularly among groups with limited access to care and treatment Support for research is another potential use for genetic-diversity surveillance. As Sides et al. state, the relationships between viral diversity and transmissibility, infectivity, and pathogenicity remain uncertain; better genetic-diversity surveillance information would allow the evaluation of some hypotheses and would generate others. Vaccine development is frequently focused on HIV-1 subtypes known to be common in a particular geographic area. Subtype surveillance in all parts of the world, including the United States, could be important in supporting optimal vaccine strategies Calls for HIV genetic diversity surveillance are not new. In 1996, Hu et al. [44] recommended the creation of worldwide HIV genetic-diversity surveillance networks to monitor the molecular epidemiological characteristics of HIV “by systematically collecting isolates of representative HIV strains from different populations with various transmission risk factors.” They emphasized that such efforts should include continual representative sampling on a long-term basis and noted the inherent biases in the use of cross-sectional convenience samples for prevalence estimates In 2002, Osmanov et al. [45] reported global and regional estimates of HIV genetic diversity based on mathematical modeling, using the large number of studies based on convenience sampling in many countries’ results and on “expert opinion.” The experts’ opinions coincided well with the modeling-based estimates; because many of the experts were involved in the studies from which the estimates derived, this is not surprising. Continued use of the estimates in this article is also not surprising; it is still the best available synthesis of most HIV genetic-diversity studies in the literature up to the year 2000. The authors’ discussion of the limitations of the data on which their estimates were based is seldom quoted, however, and their recommendations for representative surveillance were not put into effect Why have genetic-diversity surveillance networks not been put into place? HIV-2 surveillance could be easily performed, given that approved, sensitive EIAs can distinguish between the 2 types. The New York State Health Department Laboratory has tested routinely for HIV-1 and HIV-2 since 1998 and has detected an average of 38 cases of HIV-2 infection annually among ∼100,000 specimens screened [46]. Other laboratories should follow suit. Sides et al.’s study, although it does not deal with HIV-2 infection, shows that, even in a state with few foreign-born residents, HIV strains associated with other countries may still contribute substantially to the HIV epidemic. HIV-2 now makes up nearly 5% of HIV infections in Portugal [47]; >60% of those cases are among persons not from West Africa. Although there is little evidence for a similar phenomenon in the United States, we should be prepared to detect an increase if it occurs A reason for the lack of HIV-1 subtype surveillance is the expense and relative complexity of the laboratory testing involved and the lack of an inexpensive and easy screening test. Successful surveillance systems often, although not always, simply capture laboratory results that have been produced routinely for diagnostic or clinical purposes; however, HIV genetic sequencing is not required for a diagnosis of HIV infection, nor is there a clinical indication for HIV sequencing when all HIV-infected individuals enter care (although this may change if baseline HIV drug-resistance genotyping becomes more widespread) What we know about HIV genetic diversity in the United States continues to be based largely on groups of isolates that have accumulated in laboratories or clinical sites or from cross-sectional or prospective studies in clinical centers. Selection bias is inevitable in all convenience sampling. The many reported studies have been invaluable in detecting the importation of new strains and in showing that certain strains predominate among patients in certain clinics and have been helpful in exploring research questions. But the limitations of current prevalence estimates are clear: studies without a sampling strategy designed to represent the HIV-infected population of a geographic area are not likely to be representative In a geographic area where nearly all diagnoses of HIV infection and care take place in 1 center, such as in British Columbia [48], the sentinel center approach can achieve a representative sample. A network of sentinel centers in which services are provided for nearly all HIV-infected persons in the country, such as the Swiss HIV cohort network [31], also provides meaningful estimates. In the United States, where a large variety of sites provide HIV services in most areas, where diagnosis and care may not take place in the same site, and where many individuals who have been diagnosed with HIV infection are not receiving care, genetic-diversity prevalence estimates from clinical centers are problematic. Studies based in individual clinics may not represent the HIV-infected population in a geographic area, and sample sizes in clinic-based studies may also be insufficient. For a clinical trial, sample-size calculations are based on the ability to detect an effect; in a prevalence study, the sampling strategy and sample size must be designed to allow a sufficiently precise estimate of the factors of interest in the reference population. Sides et al. show an understanding of the need for a sufficiently large and representative sample in their discussion of their study of US-born persons with a new diagnosis of HIV infection in Minnesota. Because the 1-site sample represented only 13% of persons with a new diagnosis of HIV infection and overrepresented men who have sex with men, compared with the total population of those with new diagnoses, the authors could not conclude that the absence of non-B subtypes in their sample was representative. Suggested by their discussion, although not made explicit, is the possibility that, in the future, a larger and more representative multisite strategy could be developed on the basis of lessons learned in the pilot study Individuals newly identified as HIV infected are a potentially useful reference population, because they are well characterized by HIV reporting in most US states. Age, sex, race/ethnicity, and risk are reported, and checking takes place within and between states to ensure that a case is reported only once. A system that uses specimens from persons newly identified as HIV infected within a calendar year has the advantage of a well-characterized denominator. Disadvantages include the fact that persons receive diagnoses at different points in disease progression, so that estimates do not reflect only the HIV type currently being transmitted. Another disadvantage is the difficulty of acquiring suitable specimens at the time of diagnosis. In Canada, 1 central laboratory amplifies and sequences HIV diagnostic serum specimens from most provinces in the country: of 1634 specimens collected between 1998 and June 2002, only 1312 (80%) could be sequenced [49]. Seven percent were HIV-1 non-B subtypes, including A, C, D, and several recombinants Until 2004, national US HIV surveillance did not include HIV genetic-diversity surveillance. This changed after a 4-state feasibility pilot project demonstrated that relevant regions of the HIV pol gene routinely could be amplified and sequenced from routinely collected HIV diagnostic serum samples. HIV drug-resistance surveillance remnant diagnostic serum samples was formally incorporated into the HIV surveillance system in July 2004. Sequencing of the pol gene can also be used to screen for HIV-1 subtype (with some limitations on the evaluation of recombinant strains) [50]. HIV drug-resistance and HIV-1 subtype surveillance by use of HIV diagnostic serum samples is currently being supported by CDC cooperative agreements in 23 states and large cities. HIV drug resistance and subtype evaluation will be considered to be a routine part of HIV testing in participating centers, with results returned to providers designated by participants. Whether it will be possible to develop a nationally representative surveillance system is uncertain, given the resources required and the potential limitations of diagnostic sera. Sides et al.’s example should generate more interest in local surveillance efforts, and high priority should be given to areas where foreign-born persons make up ⩾10% of new HIV cases. Given the variation in refugee settlement and travel patterns of residents in the United States, we may see very different local epidemics in contiguous geographic areas Our understanding of HIV genetic diversity and the relationship of diversity to drug resistance owes much to analyses of large databases of sequences by such investigators as Shafer and other researchers at Stanford University [51,–,54]. Such work has been facilitated by public availability of HIV sequences from published and unpublished studies downloaded to GenBank [55] and from there to public HIV sequence databases [56], but this is still limited by some authors’ unwillingness to share the sequences on which their publications are based. Sequences from HIV genetic-diversity surveillance could add substantially to these databases, and, in turn, new findings from analyses of publicly available sequences will suggest new uses for genetic-diversity surveillance information. Experts who understand the importance of representative HIV genetic-diversity data should join in the discussion of how surveillance can best be performed and why such data are important