Panel A: Pol sequence alignment for participant 13 (non-perinatally infected). The reference sequence (top) was arbitrarily chosen from among those recovered from CD4+T-cells sampled at the earliest timepoint. Sites of HLA-driven adaptation in Pol (defined in [42]) are highlighted, with red, orange and blue denoting adapted (inferred escaped), possibly adapted and susceptible forms, respectively. Optimally-described CTL epitopes restricted by host HLA alleles are shaded in grey. The proportion of HLA-associated sites exhibiting adapted or possibly adapted forms is reported after each sequence. Note the three codons (257, 264 and 277 in this alignment, denoting RT codons 158, 165 and 178), all within HLA-restricted CTL epitopes, where adapted and susceptible forms co-exist within the reservoir. Panel B: Average inferred immune escape burden in Pol (calculated as the median “% adapted” value of all sequences for each participant), stratified by group. P-value calculated using Student’s T-test. Panel C: Percent of HLA-matched optimal CTL epitopes in Pol exhibiting within-host sequence variation, stratified by group. As the data for non-perinatally infected participants are non-normally distributed, the p-value is calculated using the Mann-Whitney U-Test. Panels D, E: Same as panels B and C, but for Nef. Panel F: The B*07-restricted RM9 epitope (Nef codons 71–79) as an example of reservoir immune escape complexity within and between hosts. Letter size is proportional to within-host amino acid prevalence, with red and blue denoting adapted and susceptible forms [42], respectively (all other residues are grey). Note that participant 11 was perinatally-infected, illustrating that HLA-adapted and susceptible forms of a given CTL epitope can co-exist in the reservoir, even in persons who initiated cART two decades afer infection. Panel G: Additional examples of HLA-restricted optimal epitopes in Pol and Nef where adapted and susceptible forms co-exist within an individual’s reservoir.