Prevention of SIVmac251 reservoir seeding in rhesus monkeys by early antiretroviral therapy

The precise time when the viral reservoir is seeded during acute HIV-1 infection remains unclear. We previously demonstrated that the viral reservoir was seeded by day 3 following SIVmac251 infection in rhesus monkeys. Here we report the impact of initiating ART on day 0 (6 h), 1, 2, or 3 following intrarectal SIVmac251 infection in 20 rhesus monkeys (N = 5/group). After 6 months of daily suppressive ART, antiretroviral drugs were discontinued, and viral rebound was monitored. 0% (0 of 5), 20% (1 of 5), 60% (3 of 5), and 100% (5 of 5) of animals that initiated ART on days 0 (6 h), 1, 2, or 3, respectively, showed viral rebound following ART discontinuation and correlated with integrated viral DNA in lymph node CD4+ T cells. These data demonstrate that the viral reservoir is seeded within the first few days of infection and that early ART initiation limits the viral reservoir.


Viral highlighter analysis.
The env codon sequence alignments were manually reviewed after automatic alignment with Gene Cutter (hiv.lanl.gov/content/sequence/GENE_CUTTER/cutter.html) and iterated using highlighter tools.

Viral dynamics modeling.
We examined the dynamics of reservoir seeding and rebound using relationships predicted from viral dynamics models (2). If each newly infected cell has some fixed probability of entering into latency, then the size of the latent reservoir should be proportional to the cumulative number of infections occurring before ART initiation [3]. Area-under-the-curve (AUC) viral load was used as a measure of the total amount of infections occurring, which is a reasonable approximation for acute SIV infection since CD4 + T cell levels do not change consistently or significantly. 2 To calculate AUC viral load before ART in animals with detectable values, we fit the kinetics in each animal to a simple model of early exponential viral growth: Here v0 is the effective infectious viral load from which the infection starts (v(0)), and a is the growth rate of viral load during acute infection. Values of viral load below the detection limit (50 copies/mL) were treated as censored data, and fitting was done using a maximum likelihood approach that assumes that the observed viral load is log-normally distributed around the true viral load, and that measurements include an error with variance + , . AUC was then as was calculated as: To estimate AUC for animals treated after 3 or fewer days for which no detectable viral load was observed, we used the average values of log10 (v0) and a from the remaining animals and the relevant # :;< value (6 hours, or 1, 2, or 3 days).
We then examined the relationship between pre-ART AUC viral load and measures of the long-lived reservoir. Because HIV/SIV DNA levels decay rapidly upon initial ART initiation and are assumed to be dominated at that time by labile forms, we used SIV DNA levels in the PBMC, LNMC, and GMMC after 6 months of ART (immediately prior to ART cessation). We implemented a regression method that accounted both for errors in both variables (AUC and SIV DNA) and censored data (SIV DNA values below the detection limit of 3 copies/10 6 cells). Log values of both variables were fit to a linear relationship to estimate the scaling coefficient between the variables. Spearman rank-order correlations were also performed to test the strength of the relationship. The fitted relationship was used to predict the exact values of undetectable SIV DNA values.
The probability of rebound ever occurring upon ART cessation, as opposed to "cure", is predicted by viral dynamics models (3) to follow the simple relationship: Where I J; is the total body number of latently infected cells, 3 is the rate (per day) at which any individual cell reactivates from latency (1/ 3 is the average time until reactivation), K is the probability that any individual reactivated cell will successfully establish infection, and depends on both the mean and variance of secondary infections and potentially on the probability of immune escape, and L is the death rate (per day) of latently infected cells ( ln 2 /L is the half life of cells in the reservoir).
We examined the ability of different measures of the reservoir to predict the probability of rebound using a relationship inspired by this equation: Where we assume that the parameters 3, L, and K are the same for all animals and can be captured, along with the scaling between total reservoir size I J; and reservoir metric W J; , with the constant X. Fitting was done using a maximum likelihood approach and the following reservoir metrics: pre-ART AUC VL (observed or inferred), and preinterruption SIV DNA in PBMC, LNMC, and GMMC (actual value if above the detection limit and inferred value based on previously fitted relationship to pre-ART AUC VL).  Supplementary figure 3b  Figure 3. Highlighter alignment of env sequences from macaques after ART stop and viral rebound. Highlighter analyses of (a) nucleotide and (b) amino acid sequences are shown as compared to the inoculum sequences. A total of 112 sequences (day 1: 11, day 2: 28 and day 3: 73) were generated by SGA and compared with 37 sequences from the inoculum and to the consensus sequence of the SIVmac251 challenge stock. Nucleotide polymorphisms are indicated by a colored tic mark (adenine in green, cytosine in blue, guanine in orange and thymine in red). APOBEC-mediated G-to-A mutations are indicated by purple diamonds, and gaps (deletions) are indicated by gray tics (compressed). Amino acids that differ from the consensus sequence of the SIVmac251 challenge stock are indicated in color. Amino acid substitutions found in more than one sequences and only in plasma SIV isolated from macaques between 28 and 42 days post-ART stop are shown.