Breastfeeding transmission remains a leading cause of pediatric HIV infections worldwide. Uncovering predictors of time to viral rebound (TTR) in a preclinical breastfeeding transmission model could identify clinical biomarkers to guide remission strategies and clinical trial design.
At 4 wks of age, rhesus macaques (RMs) were orally infected with SHIV.C.CH505.375H.dCT and placed on daily ART at 4-7 dpi (Early, n=10), 2 wpi (Intermediate, n=10), or 8 wpi (Late, n=10). Analytical treatment interruption (ATI) was performed after 1 yr. Peripheral blood (PB), lymph nodes (LN), and rectal biopsies (RB) were collected for viral and immune (cellular and humoral) measures. The Cox Proportional Hazards model with LASSO regularization was used to identify and rank parameters that predicted TTR. Variables were pruned for multi-linearity using a Pearson correlation cutoff of 0.9. Leave-one-out Cross Validation (LOOCV) deviance, the error of a fitted model in predicting out-of-sample data points, was used for selection of predictor variables.
Viral rebound occurred in 26/30 infant RMs within 7-98 d of ATI, with TTR significantly delayed in the Early ART group (p<0.001) compared to the Intermediate and Late ART groups (median TTR 84, 18, and 17 d post-ATI, respectively). Intact SHIV genome levels were undetectable in PB and LN CD4+ T cells prior to ATI in 10/10 Early ART animals. Reservoir burden pre-ATI was comparable between Intermediate and Late ART groups (median 6 and 17 intact genomes/million CD4+ T cells, respectively). Through multivariate regression modeling of 82 variables, we constructed a predictor inclusion rank table where peak plasma viral load best described TTR (LOOCV deviance 4.03), with the strength of predictions increasing with successive inclusion of pre-ART SHIV RNA in PB CD4+ T cells and pre-ATI levels of CD69+Perf+ NK cells, intact SHIV genomes in LN CD4+ T cells, SHIV DNA in RB CD4+ T cells, GzmB+ CD8+ T cells, and ADCP (LOOCV deviance 3.69). Individually, peak plasma viral load and PB CD4+ T cell SHIV RNA were associated with accelerated TTR (HR 1.63 and 1.23, respectively), whereas GzmB+ CD8+ T cells were associated with delayed TTR (HR 0.81).
This work provides novel insight into predictors of TTR in a preclinical NHP model of pediatric HIV-1 infection. Predictors identified in our model should be explored further to determine if they can serve as readily-measurable biomarkers to screen children with HIV-1 being considered for ATI.