Hammering out HIV-1 incidence with Hamming distance

In epidemiology, the rate at which new infections occur with HIV-1 (or any other pathogen) in a population over a certain period of time is called incidence. Determining incidence is a very important measure for monitoring epidemics, future projections of the infected individuals, and evaluating success of vaccinations and preventive measures. It is also important to calculate appropriate sample size in any prospective vaccination trial. For measuring incidence, it is important to identify incident (recent) infections and distinguish them from pre-existing (chronic) ones. This distinction is of particular importance in the case of HIV-1, as the existing infections with the virus cannot and do not prevent subsequent infections with the same or different viral strains [1]. In line with the importance of detecting incident HIV-1 infections, several clinical tests have been developed [2]. The tests are based upon sequential expression of several HIV markers including viral RNA, viral antigens and various characteristics (e.g. antigen specificity, isotype, avidity, etc.) of virus-specific antibodies at precise predetermined times after the infection. The tests, however, are very cumbersome, time-consuming, intrusive, resource-intensive and difficult to perform on a large number of samples. Furthermore, they lack standardization, reproducibility and broader applicability to all infecting viral clades. Therefore, developing novel HIV-1 incidence assays remains a public health priority. In the original research paper presented in this issue, Park et al.[3] have devised a novel sequence-based assay to estimate HIV-1 incidence with high sensitivity and specificity. In order to develop this test, the authors performed a meta-analysis of about 6000 published full-length env gene sequences amplified from 182 HIV-infected individuals with incident/recent infections (being infected for less than 1 year) and 43 individuals with chronic infections (being infected for more than 1 year). All the infected individuals were from different geographic locations in North America, Caribbean and Africa. The authors noted that neither env gene diversity measured by average Hamming distance nor its variance could differentiate between the two types of the infections. The main difficulty with these parameters was in differentiating incident infections with multiple founder strains (see below) from chronic ones. The authors found and capitalized on a specific signature of incident infections in the distribution of Hamming distance frequencies between all possible pairs of env gene sequences: the 10% quantile (Q10), the Hamming distance that divides lower 10% of the Hamming distance frequency distribution from upper 90%, can distinguish between incident and chronic infections. The statistic is much lower in incident infections compared with that of chronic ones. The authors have shown that an appropriately selected Q10 from the isocost line of a receiver operating curve can measure HIV-1 incidence with high sensitivity and specificity [3]. It is gratifying to note that the test can distinguish incident infections from chronic ones whether the formers originate from a single or multiple founder strains. It is noteworthy that due to the presence of a genetic bottle neck, a majority (80%) of the heterosexual HIV-1 transmissions occurs with single viral strains. Genital tract inflammation and the use of contraceptives, however, favor HIV-1 transmission with multiple viral strains. They also occur more frequently in men having sex with men and intravenous drug users [4]. They evolve more rapidly due to inter-strain recombination events and progress to AIDS faster [5]. The success of the Q10 statistic in identifying incident infections is due to its focus on the presence of similar (identical or closely related) rather than on dissimilar HIV gene sequences. The viral gene sequences diversify in an infected individual giving rise to a swarm of viral ‘quasispecies’. In the beginning, these sequences are more or less similar to one another and can be detected in incident infections due to their relatively small Hamming distances. They occupy lower tail end of the Hamming distance frequency distribution. As the infection progresses with time, the similar sequences disappear exponentially, are replaced with dissimilar ones, which have relatively large Hamming distance and occupy upper parts of the distribution. The parameter Hamming distance was named after Richard Hamming, who originally described it in 1950 to correct errors in telecommunication codes [6]. Hamming distance measures points of variation between two strings. For example, the two six-base sequences ‘AGTTCA’ and ‘AGCTCA’ have a Hamming distance of 1, as they differ in base composition at one (third) position when compared with each other. Since its original description, Hamming distance has been used widely in mathematical modeling in evolutionary biology. Early in the course of an acute infection, there is little selective pressure from the host immune responses. Herein, the virus expands exponentially and accumulates random mutations. The Hamming distance frequencies of the viral env gene sequences follow a Poisson distribution if the infection had occurred with a single viral strain. The departure from the distribution has been used to determine multiplicity of the infection in the infected individuals [7,8]. Such tests, however, cannot be used in late stages of the acute infection as well as in chronic ones due to selective pressure of the immune responses on viral evolution [9]. Of the potential limitations of the Hamming distance Q10 statistic, the authors list late stages of the HIV infection, when a decline in the viral sequence diversity occurs. To this we add that the test performance (sensitivity and specificity) also needs to be verified separately in the so-called long-term nonprogressors and rapid progressors as well as of various opportunistic infections. An interesting question would be to know the potential effects of antiretroviral therapy on the test performance. It is noteworthy that the authors used the sequence data that was obtained before the advent of next-generation sequencing (NGS) technologies. The earlier sequencing methods were not able to detect less frequent viral sequences. The NGS assays can detect viral gene sequences as rare as 1% [10]. It can only be hoped that they would be of help in refining the test. To the best of our knowledge, this is the first time that viral sequence variation has been used for determining incidence of any pathogen. In the context of HIV, the novel test is clade-independent, is not affected by the viral load and requires only a drop of blood. With recent advances in sequencing technologies, it can be easily adapted to high-throughput clinical settings. The study by Park et al.[3] has indeed provided a much needed new tool for estimating HIV-1 incidence in human populations. Acknowledgements Conflicts of interest There are no conflicts of interest.