The subjects analyzed in this study were part of a mother/infant cohort followed for HIV-1 infection. Children were designated as rapid or slow progressors according to clinical assessment of outcome and time of survival. Infants 1449, 2669, 2873, and 2617 were considered rapid progressors since they died within the first year of life, due to apparent HIV-related complications. Slow progressors (infants 1984, 1084 and 1690) were followed for more than four years, and remained clinically asymptomatic for the duration of the study (Table 1). All children were anti-retroviral naïve throughout the study.

HIV-1 isolation was unsuccessful from all baseline (birth) samples and all infants were HIV PCR negative at birth, suggesting that they were infected either intrapartum or postpartum. HIV-1 env sequences were amplified from infant PBMC at different postpartum timepoints, as indicated in Table 1. Because the amount of sample from these children was limited, priority was given to virus isolation in lieu of PCR when necessary (e.g., infant 1084, viral isolation was positive by 4 month and the first PCR was performed 6 month after birth). A portion of the env gene from V1–V5 was amplified by PCR, cloned, and sequenced in order to longitudinally characterize Env genetic diversification and evolution.

We sequenced a total of 711 infant clones (23 – 48 sequences per timepoint) derived from PBMC genomic DNA. When all sequences were aligned and included in a single phylogenetic analysis, sequences from each mother-infant pair formed a monophyletic group, indicating that maternal and infant sequences were epidemiologically linked (data not shown). Viral subtype determinations showed that all cases were subtype C in Env, except for mother-infant-pair 1449 which was a subtype A/C recombinant.

In all infants, the initial viral populations contained a reduced repertoire of env sequence variants when compared to the maternal population. These samples exhibited a large fraction of unique haplotypes, but with low nucleotide diversity, as would be expected in populations increasing in effective size from a limited set of founders (Table 1). Haplotype diversity (H/N in Table 1), an index of the number and relative frequency of unique sequences, ranged between 0.9 and 1.0, its maximum value, but average genetic distances within each sample remained low throughout the study (DNA % in Table 1). Mean genetic distance (DNA% in Table 1) were lower at the earliest time points, where they ranged from 0.3 to 1.2%, while for the latest, mean genetic distances ranged from 0.5 to 4.9 %. Representative phylogenetic analyses from a rapid progressor (1449) and a slow progressor (1984) are shown in Figure 1. Results from the different phylogenetic analyses for each mother-infant-pair were congruent among themselves, despite differences in the methods or weighting schemes used. In all cases, the results suggest that infections were established by highly homogeneous populations, with little phylogenetic structure among early sequences. In the case of fast-progressors, the diversity observed in different longitudinal samples taken from the infant was low relative to the mother, as indicated by the shorter branches leading to infant sequences when compared to the mother. A similar pattern can be observed for the earlier sequences of slow-progressors. Later time-point samples display longer branches in the phylogeny, as mutations accumulate and infant Env sequence diversity increases. It is important to note that trees from rapid and slow progressor were indistinguishable when analyses were restricted to sequences collected within 12 months after birth.

Similar patterns of variation were observed at the amino acid level, although levels of polymorphism were higher relative to variation at the nucleotide level. Mean amino acid differences (AA% in Table 1) within the initial populations ranged from 0.6 to 2.4 % for the earliest samples, and from 1.0 to 8.9% for the later time-point samples. Mean genetic distance (DNA % in Table 1) within contemporaneous sequences were lower in rapid progressors than in slow progressors (Table 1), but this difference was not statistically significant. There was a trend towards increased levels of genetic diversity as time progressed, with some refractory periods. Accordingly, we observed the highest levels of genetic diversity (DNA% in Table 1) in samples collected at the latest time points in slow progressors (Table 1, 48- month samples from infants 1084, 1690 and 1984). However, the rates of change in genetic diversity and genetic divergence were similar for all patients (data not shown), although sample sizes precluded statistical tests of this observation.

Positive Darwinian selection is indicated when the estimated ratio of non-synonymous changes to synonymous changes (dN/dS) >1. We observed high dN/dS values for the env gene (Table 1), suggesting that positive selection was occurring in the infant env genes. The values ranged from 0.41 to 1.37, with a mean of 0.78 (Table 1). The significance of this finding is that higher dN/dS values have been linked to longer survival, and presumably, a higher dN/dS value is a consequence of a stronger and/or broader immune response 2055. In the three slow progressor infants there was at least one time point where the dN/dS > 1; whereas a dN/dS > 1 was detected in only one of the rapid progressor infants (1449), but it is possible that this is a function of the duration of infection. Indeed, while we observed a higher level of non-synonymous substitutions in slow progressors (mean = 0.89) versus rapid progressors (mean = 0.78), this difference was not statistically significant.

To temporally and positionally visualize where non-synonymous changes occurred relative to 'constant' and 'variable' domains, as defined in subtype B, we compared the infant amino acid sequences to an alignment of HIV-1 HXB2 and the solved SIV glycoprotein structure by Chen et al. 56. One representative infant from each group is shown in Figure 2. For clarity and ease of comparison to the rapid progressor infant 1449, we have separated the early time points from the complete analysis of slow progressor infant 1984. Inspection of the variation from both rapid and slow progressors revealed several common regions of the env sequence with high levels of non-synonymous variation and indicated that the C2 domain was the least variable, whereas the most variable areas were the V1–V2 loop, the 3' end of the C2 region, the V3 loop, the 5' end of C3, and the variable loops V4 and V5 (Figure 2). In addition, the variable loops V1–V2, V4 and V5 concentrated most of the indels observed. Comparison of 1449 and 1984 at similar time points (Figure 2, top and middle panels), revealed changes located in corresponding regions (e.g. V4 and V5), but there were also changes unique to either 1449 or 1984 (e.g. the 5' end of V1–V2 in 1449). Unfortunately, this study is not able to establish whether the unique mutations observed in 1449 are associated with rapid disease progression. There is also an accumulation of non-synonymous changes with time, particularly evident in 1984 where changes at many positions are cumulative, implying continued selection operating on positions over an extended time period (Figure 2, middle and lower panels). Whether this indicates immunological pressure or functional constraints for fitness remains to be determined. In contrast, for 1449 (Figure 2, top panel), a number of changes appear at only one time-point with no previous evidence of selection at that position.

Taken together, the higher diversity associated with later time points, in combination with the observed accumulation of amino acid substitutions in putatively exposed regions of the glycoprotein indicate that selective pressures, including humoral immunity, may be playing a substantial role in driving Env evolution.

The number of putative N-linked glycosylation sites (PNGS) and Env domain length have been hypothesized to modulate HIV-1 sensitivity to neutralization and to impact likelihood of transmission 5758. According to this hypothesis, shorter variants with fewer PNGS are expected in the earlier time-points, they have higher transmission fitness as the immune response of the recipient is still not developed; longer V1–V5 forms with more PNGS are expected to evolve at later time-points in response to increased and prolonged immune pressure. Longitudinal data including range and median values for Env V1–V5 length and PNGS are presented in Table 1, and the trend in median values in Figure 3. Rapid progressors exhibit a large range in both PNGS and V1–V5 length, with minor longitudinal changes during a period of up to 8 months postpartum. The number of PNGS is positively correlated with sequence length in these cases. Slow progressors show a tendency to increase (1984 and 1084) or decrease (1690) the number of PNGS with time, but the range of variation falls within the range observed for fast progressors (Figure 3, top panel). The same pattern is observed for longitudinal variation in V1–V5 length (Figure 3, bottom panel). Overall, no clear trend was observed as would be suggested by the predictions 5758, and the values for these parameters did not differ between fast and slow progressors.

Since co-receptor usage switches to an X4-utilization phenotype with disease progression in some adults and children, we evaluated co-receptor usage and phenotype of viral isolates from the two groups. We found that all viral isolates exclusively used CCR5 as a co-receptor (Table 1), exhibited macrophage-tropism, and did not infect T cell lines or form syncytia in vitro. The only exception was the 48-month isolate from infant 1690. For 1690, R5 co-receptor tropism was maintained until 42 months; after this time, the viral isolate exhibited dual X4/R5 co-receptor usage (Table 1), and infected both macrophages and MT-2 T lymphoblasts, where it formed syncytia (data not shown). To date, this is the only X4-utilizing virus isolated from our cohort, implying that while X4-utilizing subtype C HIV-1 can develop in patients, such development is uncommon and disease pathogenesis is not dependent on such phenotypic switches. To test whether the characterized subtype C Env sequences possessed co-receptor usage properties consistent with those defined for the virus isolated by co-culture, we generated Env chimeras by introducing the subtype C V1–V5 region into a subtype B NL4-3 Env expression vector. The chimeric Env constructs were then used to make pseudoviruses for evaluation of co-receptor usage in Ghost cell lines that express different co-receptors. All chimeras tested exhibited CCR5 tropism and lacked appreciable X4 tropism (data not shown). These findings are consistent with those obtained from experiments using primary isolates.

Since maternal anti-HIV antibodies are transmitted from mother to infant, it is possible that they play a role in the selection of transmitted viruses and affect the disease course in the child. Therefore, we evaluated maternal and infant neutralizing antibody (Nab) titer at birth against the first infant viral isolate. The level of Nab was determined from the rapid (infants 1449, 2669 and 2873) and slow progressors (infants 1084 and 1984), as well as one slow progressor (infant 1157) described previously 33. For rapid progressors, the first viral isolation was 2 months after birth, whereas the first viral isolates in the slow progressors are from 4 months (1084 infant) or 6 months (infant1157 and 1984). Our results (Table 2) indicate that the level of infant baseline Nab against infant first viral isolates was lower than the maternal baseline, implying that only a subset of the maternal neutralizing antibody was acquired by their infants. Comparison of the baseline Nab level between the corresponding mother and infant from each pair indicated that there is a direct correlation between the level of maternal Nab and the level of Nab passively transferred to their infants. Mothers with the low baseline Nab transferred the least Nab to their infants. But the level of Nab in either the maternal or the infant baseline plasma failed to differentiate rapid and slow progressors. For example, 88% neutralization by maternal baseline plasma was observed in one rapid (1449) and one slow (1157) progressor, respectively; whereas, in other cases, maternal baseline plasma from both rapid and slow progressors failed to effectively neutralize the earliest infant virus (infant 2873 vs.1984). Similarly, the neutralization capacity of the infants' plasma at birth against their first viral isolates does not differentiate the two groups. For example, both 2873 (rapid) and 1984 (slow) lack detectable Nab for their first viral isolates at birth.

To further characterize the infant antibody responses, we quantified neutralization by autologous sera from various timepoints for the first and last viral isolates from both groups. The neutralization profiles of two representatives from each group are shown in Figure 4. For the rapid progressors (1449 and 2669), we observed variability in the baseline neutralizing antibody activities acquired from the mother (Figure 4A). In 1449, the initial activity against the earliest virus (78% neutralization) declined, prior to the initiation of a de novo infant humoral immune response near the time of the first virus isolation, which rose thereafter. Whereas, in 2669, the maternal transfer was less effective (only 45 % neutralization), but the infant de novo neutralizing response was evident by two months since the neutralization was higher than baseline. The de novo development and maintenance of effective neutralization against the 2-month viral isolates appeared early in both cases and increased in activity until the end of follow-up (Figure 4A). In contrast, neutralizing antibodies against the late viral isolates (8 months for 1449; 6 months for 2669) were lower in magnitude and decreased throughout the disease course (Figure 4A).

Similarly, the autologous plasma neutralization of slow progressor infant early (4 or 6-month) and late (48-month) viral isolates was evaluated, and two representatives (1984 and 1084) are shown in Figure 4B. In the slow progressor 1084, a substantial amount of Nab was detected at birth, but decayed to zero by six months. Subsequently, the child developed an effective neutralizing response against both the earliest virus and the contemporaneous (12 and 48-month) viruses. In contrast, slow progressor 1984 received no detectable Nab from the mother, but mounted an effective neutralizing response by 6 months whose magnitude was directly correlated with the timepoint of virus isolation, with 6-month virus being more effectively neutralized than 12 and 48-month viruses. It is apparent that in slow progressors there are infants who passively acquired neutralizing activity (1084), while others (1984) did not. Therefore, it is unlikely that rapid progression is due to receipt of lower maternal Nab, or that slow progression is due to acquisition of high level of maternal Nab or the development of a higher or more durable de novo humoral response.

In order to determine whether there are differences in the rates of replication among the viral isolates from rapid and slow progressors, the replication of the first viral isolates (slow progressor only) and last viral isolates (all 7 infected children) in PBMC was determined (Figure 5). The titer (TCID₅₀/ml) of the last viral isolates from all rapid progressors (4, 6 or 8-month after birth) displayed steady increase after 5 or 9 days incubation and peaked by 9 (infant 2669 and 2873), 13 (infant 2617) or 17 (infant 1449) days (Figure 5A). For slow progressors, the first viral isolates (6-month for 1984, 4-month for 1084) displayed similar replication kinetics compared to the rapid progressors. However, when comparing the first and last viral isolates from the slow progressors, the late viruses (48-month for infants1984, 1084 and 1690) showed a slightly more rapid replication kinetics than the early viruses, with a peak value by 13 days, while the late viruses peaked by 9 days (Figure 5B).

Seven mother/infants pairs 1449, 2669, 2873, 2617, 1984, 1084 and 1690 were recruited into the study. Venous blood was obtained from the mother before delivery and from the infant within 24 hours of birth. Follow-up blood specimens were obtained at 2, 4, and 6 or 8 months (for rapid progressors) and at regular intervals through 48 months (for slow progressors). The baseline HIV-1 serological status of the mother was determined by two rapid assays, Capillus (Cambridge Biotech, Ireland) and Determine (Abbott laboratories, USA). Positive serological results were confirmed by immunofluorescence assay (IFA), as previously described 70. HIV-1 infection in the infants was detected by sequential viral isolation from the infants' peripheral blood mononuclear cells (PBMC), as previously described 33, and by PCR of the HIV-1 provirus env gene from genomic DNA. The first timepoint for positive PCR from each infant is indicated in Table 1.

The syncytium-inducing (SI) or non-syncytium-inducing (NSI) phenotype was determined by infecting MT-2 cells as described 33. Virus was scored as 'SI' if syncytia and increasing level of p24 antigen were observed within a 10-day period and as 'NSI' if syncytia failed to form within that time. To further define the viral tropism, viral replication was assessed in primary monocyte-derived macrophages (MDM) and in the MT-2 T-cell line, as described 33.

Co-receptor usage was defined using Ghost cell lines that express specific co-receptors (Ghost-CXCR4 cells [CXCR4], Ghost-CCR5 cells [CCR5] and Ghost-CCR3 cells [CCR3])(NIH AIDS Research and Reference Reagent Program, Division of AIDS, NIAID, NIH from Dr. Vineet KewalRamani and Dr. Dan Littman) with standard procedures as recommended by the supplier. Uninfected control produced only one to two GFP-expressing cells per well. HIV-1 strains NL4-3 and SF128A were used as positive controls for CXCR4 and CCR5 utilization, respectively.

Virus sensitivity to neutralization by maternal or infant plasma-derived antibody (1:20 dilution of plasma) was determined by infections of TZM-bl cells (NIH AIDS Research and Reference Reagent Program catalog no. 8129, TZM-bl) as described 33.

Two thousand TCID₅₀ (Tissue Culture Infectious Dose) of the viral isolates from early or late time points were added to a T-25 flask containing 2 × 10⁷ PHA-stimulated PBMC from a pool of two HIV-1 seronegative blood donors. The subtype B laboratory isolate SF 128A 71 was used as a control. After incubation at 37°C for 6 hours, cells were washed 3 times with PBS and replenished with fresh medium. All infected cultures were sampled and supplemented with fresh culture medium at days 1, 5, 9, 13, 17 and 21-postinfection. Virus titer (TCID₅₀/ml) was measured by infection of TZM-bl cells (NIH AIDS Research and Reference Reagent Program catalog no. 8129, TZM-bl).

Genomic DNA was purified from uncultured patient PBMC using the Puregene kit (Gentra Systems). Primers used for amplification of the subtype C env gene were designed based on a reference alignment of all HIV-1 subtypes obtained from the Los Alamos Sequence Database. Nested PCR was used to amplify a 1100 bp fragment spanning the V1–V5 region of the env gene from samples collected at birth through 48 months. In order to minimize the PCR bias, Env V1–V5 region were amplified in duplicates in the first-round PCR of each maternal and infant sample. First-round PCR products were combined and used as template in a second-round PCR. Primers used in the first round PCR were: forward primer EnvF1: 5'-GATGCATGAGGATATAATCAGTTTATGGGA (corresponding to position 6533 to 6562 of HIV-1 HXB2 strain), reverse primer EnvR1: 5'-ATTGATGCT GCGCCCATAGTGCT (position 7828 – 7806). Internal primers used in the second round PCR were EnvF2: 5'-AGTTTATGGGACCAAAGCCTAAAGCCATGT (position 6552 to 6581) and EnvR2: 5'-ACTGCTCTTTTTTCTCTCTCCACCACTCT (position 7762 to 7734). First and second round PCR were carried out at 94°C for 2 min; 94°C for 30 sec, 55°C for 1 min, 68°C for 1 min for 35 cycles; 68°C for 7 min using a Perkin-Elmer thermocycler. Amplified fragments were cloned into the pGEM-T Easy vector (Promega) and sequenced in both directions with dideoxy terminators (ABI BigDye Kit). The V1–V5 region of the env gene was analyzed.

Subtype affiliation was determined by comparing patient env sequences to a reference panel from the HIV database. Maximum likelihood (ML), Minimum-Evolution (ME) and neighbor-joining (NJ) phylogenetic analyses were done to visualize how viral populations within patients change with time. Maximum likelihood searches were conducted in Treefinder 72 using an independent model of nucleotide substitution for each codon position (GTR+Γ). Minimum-Evolution and Neighbor-joining analyses were performed in MEGA 73 using the Kimura 2 parameter nucleotide substitution model to estimate genetic distances. Support for the nodes in ME and NJ was evaluated by running 1000 bootstrap pseudoreplicates.

For each infant, samples were grouped according to the time of collection, and analyzed independently. To keep a consistent set of sites in the analyses, positions where putative insertions or deletions were inserted for alignment purposes in any timepoint were excluded from population level comparisons. Variation in the pattern of genetic diversity for each time-point was explored using MEGA3.1 73, DNAsp ver 4.10 74 and Datamonkey 75. The number and location of putative N-linked glycosylation sites (PNGS) was estimated using N-GlycoSite from the Los Alamos National Laboratory database. Within each infant, temporal changes in viral diversity and viral divergence from the earliest population were calculated. Viral diversity estimates correlate with the effective population size of viral populations and were estimated based on the average number nucleotide differences between sequences, whereas viral divergence was calculated as the average genetic distance relative to the earliest population collected.

The overall dN/dS (non-synonymous changes relative to synonymous changes) values for each contemporaneous data set were calculated using Datamonkey 75, and used to estimate the relative strength of natural selection in each set of sequences. In addition, the number of synonymous and non-synonymous substitutions observed per site in infant Env (V1–V5) domains was calculated by Datamonkey, and the distribution of non-synonymous substitutions along the sequence was plotted to identify regions that accumulate most variations ('hot spots'). These regions were mapped on the structural predictions from Chen et al. 56 using their alignment of the SIV Env structure with HIV.