Fig. 3: Plasma proteomic predictors of virological dynamics and loss of HIV control.

A Dot plot of all available historical viral load measurements between transient and persistent HIV controllers. Y axis depicts log10 viral load and X-axis depicts years since first HIV RNA measurement. This analysis included 181 ART-naive HIV controllers (145 VC and 36 EC). The lines indicate the local regression (LOESS) fit; each dot represents an individual viral load measurement from one participant. B Kaplan-Meier loss-free follow-up. Vertical tick marks represent censored data points, indicating either end of follow-up or censoring by ART initiation in the absence of loss of virological control. The number at risk and cumulative number of events and censoring are reported. Loss of virological control is defined as reaching >10,000 copies/mL. This analysis included 181 ART-naive HIV controllers (145 VC and 36 EC). C, D Volcano plot of the differentially expressed proteins associated with (C) virological loss, defined as reaching >10,000 copies/mL or (D) substantial virus increase, defined as a > 1000 copies/mL increase in moving median of viral load or reaching >10,000 copies/mL. X-axis depicts hazard ratio of differentially expressed proteins and Y-axis depicts -log10(FDR P value), where a hazard ratio >1 means increased risk of (C) virological loss or (D) substantial virus increase. Differential expression analysis was performed using cox regression with age, sex, center and storage time as covariates. This analysis included 181 ART-naive HIV controllers (145 VC and 36 EC). E Forest plot of the hazard ratios or virological loss, substantial virus increase and immunological loss, with error bars representing 95% confidence intervals (CI) of proteins that were FDR significant in either C. and/or D., as determined with cox regression with age, sex, center and storage time as covariates. Virological loss was defined as reaching >10,000 copies/mL, substantial virus increase as a > 1000 copies/mL increase in moving median of viral load or reaching >10,000 copies/mL, and immunological loss as reaching CD4 < 500. Immunological loss was only assessed for individuals with available longitudinal CD4 data and CD4 counts >500 at the time of proteomics sample (n = 138). F Forest plot of the hazard ratios of experiencing a detectable viral load ( > 50 copies/mL) during follow-up for EC only (n = 36). Dots represent hazard ratios and error bars the 95% CI intervals. Black dots and lines represent all EC, whereas grey dots and lines are a selection of EC who were not experiencing a blip (plasma viral load 50–1000 copies/mL) at the time of proteomics measurement.