Examination of a Second Region of the HIV Type 1 Genome Reveals Additional Cases of Superinfection

H uman immunodeficiency virus type 1 (HIV-1) superinfection occurs when an individual infected with one strain of HIV acquires a second strain, indicating a failure of the initial immune response to protect against subsequent infection. Previous studies from our laboratory and others have demonstrated that HIV-1 superinfection is occurring at a high rate, perhaps approaching the rate of initial infection. ^{1 –3} It is important to determine whether the rate of superinfection is significantly different from the rate of initial infection in order to assess whether the immune response induced in natural HIV-1 infection is ever protective.

There is evidence that examining multiple regions of the HIV-1 genome allows the detection of more cases of HIV-1 superinfection than examining just one region, and this would be predicted given the high rate of HIV-1 recombination. ⁴ For example, in a previous study conducted by our group in a well-characterized cohort of HIV-infected high-risk women in Mombasa, Kenya, ⁵ we identified seven cases of HIV-1 superinfection among 36 women by screening two well-separated regions of the genome (in gag and env). ³ If we had studied only gag or only env, we would have found only four or only five cases, respectively. Another study of a different group of 21 women from the same Mombasa cohort identified three cases of HIV-1 superinfection by examining just one gene (env). ¹ We hypothesized that examining an additional region of the HIV-1 genome would reveal additional cases of HIV-1 superinfection in this group and thus provide a better measure of the overall frequency of superinfection in this cohort.

We examined gag sequences from 14 of the 21 women in the Mombasa cohort in whom superinfection was not detected in env in a previous study, ¹ and for whom peripheral blood mononuclear cell (PBMC) DNA was available. The methods for DNA extraction, quantification, and amplification have been described previously. ³ Briefly, we quantified the HIV proviral copy number by real-time polymerase chain reaction (PCR) and used single-copy nested PCR to amplify an ∼660-bp region of gag p17 and p24 (HXB2 positions 796–1500) from proviral DNA. PCR products were treated with Exo-Sap (Amersham Biosciences) and directly sequenced. Sequence alignments were manually edited using MacClade, ⁶ and phylogenetic trees were constructed by maximum likelihood using a general time reversible model in GARLI. ⁷

For each individual, we examined a median of six gag sequences (range 1–13), each from an independent PCR, from an early time point and a late time point in infection. The early time point was a median of 117 days postinfection (DPI) (range 21–692) and the late time point was a median of 1250 DPI (range 430–2238). Gag sequences from 12 of the 14 individuals formed monophyletic clusters on a phylogenetic tree, indicating single infection (Fig. 1). For each of the other two individuals, sequences from the early time point clustered separately from the sequences from the late time point, suggesting superinfection (Fig. 1).

OPEN IN VIEWER

FIG. 1.

Maximum-likelihood tree of gag sequences from 14 individuals. All sequences from each of the two superinfection cases (QA252 and QC858) are included, with branch labels indicating the individual and the days postinfection from which the sequence was sampled. Two randomly selected gag sequences from each of the other 12 individuals in this study are also included: one sequence from the early time point and one sequence from the late time point (unlabeled branches). Reference sequences from the LANL database are used to identify subtypes (labeled branches). Bootstrap values greater than 70% are shown.

In the first superinfection case (QA252), gag sequences from the early time point belonged to subtype D while those from the late time point belonged to subtype A (Fig. 1). We obtained samples from intervening time points and sequenced the gag region to verify the presence of the second viral strain and to determine the timing of superinfection. We found the subtype D strain present at 864, 976, 1046, and 1487 DPI and the subtype A strain present at 1487 DPI (Fig. 2a). Thus, this case of intersubtype superinfection most likely occurred between 1046 and 1487 DPI. As shown in Fig. 2a, this individual's viral load between the time points immediately before (1046 DPI) and immediately after (1487 DPI) superinfection increased from 3.7 to 4.6 log₁₀ copies/ml and subsequently showed an overall increasing trend.

OPEN IN VIEWER

FIG. 2.

Follow-up for superinfection cases QA252 (a) and QC858 (b). For each superinfection case, a plot of the individual's plasma viral load (log₁₀ copies/ml) through time is shown. Sequences were sampled at time points marked by large black circles. Small black circles indicate time points at which viral load was measured but sequences were not examined. Arrows indicate the early and late time points used in the initial screen, and the box indicates the first time point at which the superinfecting strain was detected. The number of sequences of each strain detected at each time point is indicated in a table below the graph.

In the second superinfection case (QC858), gag sequences from both the early and late time point belonged to subtype A but formed separate clusters on a phylogenetic tree (Fig. 1). In this case, the maximum genetic divergence of a late sequence compared to the most recent common ancestor of the early sequences was 11.9%, which is substantially higher than the divergence seen in singly infected individuals and is comparable to the divergence seen in other cases of superinfection. ³ Upon examination of samples from three intervening time points, we detected only the initial strain (termed A) at 341 DPI and only the superinfecting strain (termed A*) at 440 and 914 DPI (Fig. 2b). Thus, this case of intrasubtype superinfection occurred between 341 and 440 DPI. As in the previous case, the viral load increased in this individual, from 4.3 to 4.9 log₁₀ copies/ml.

Overall, we detected two cases of HIV-1 superinfection among 14 women in whom superinfection was not detected by a previous examination of env genes. ¹ Combining the results of this study with the three cases found in the previous env study, we found a total of five cases of HIV-1 superinfection among 17 women with 64.6 years of follow-up. Therefore, the incidence of HIV-1 superinfection among this group of 17 was 7.7% per year. This is somewhat higher than the rate of HIV-1 superinfection that we detected in a previous study of a different group of 36 women from the same cohort, 3.7% per year. ³ However, individuals were selected for these two studies based on different criteria, the most important of which was the subtype of the initial HIV strain, as defined by env sequences. The women in the group of 36 were initially infected with subtype A viruses, the most prevalent subtype in this region and therefore the subtype to which these women would continue to be most frequently exposed. ⁸ By contrast, women in the group of 17 were initially infected with less prevalent subtype C or D viruses, based on env sequences. As expected, the majority of super-infections in both groups (10 out of 12) were with subtype A viruses. Therefore, it is possible that the difference in superinfection rates between the two groups reflects a higher rate of intersubtype superinfection compared to intrasubtype superinfection. We detected a total of eight intersubtype superinfections and four intrasubtype superinfections; however, our study size was too small to statistically compare their rates.

Combining both studies, we have examined both gag and env sequences for a total of 53 women from this cohort of high-risk women in Mombasa, Kenya. We detected a total of 12 cases of HIV-1 superinfection over 253.2 total person-years of follow-up, corresponding to an incidence of 4.7% per year. We did not have the statistical power to determine whether the rate of superinfection was different from the rate of initial infection in this cohort, which was ∼8% per year. ⁹

The results of this study reinforce our previous findings that HIV-1 superinfection occurs frequently, throughout the course of infection, and by closely related viruses. Moreover, our findings demonstrate the importance of examining multiple regions of the HIV-1 genome to detect cases of HIV-1 superinfection. Although this and other studies in the Mombasa cohort represent one of the most intensive HIV-1 superinfection screens to date, we did not have the statistical power to compare the rate of HIV-1 superinfection to that of initial infection or determine significant risk factors for HIV-1 superinfection. These results highlight the need for larger studies that examine multiple genome regions in order to define the relative frequency of superinfection compared to first infection and to characterize immune deficits and other risk factors that may be correlated with superinfection. Sequence analysis of multiple genome regions at multiple time points is very labor intensive; however, full-genome screens may be possible in the near future using rapid sequencing technologies, allowing a much better estimate of the frequency of HIV-1 superinfection.