Direct phenotypic analysis of the in vivo latent HIV reservoir is complicated by the need to reactivate these cells ex vivo to identify them, which changes the phenotypes of the latent cells. We used CyTOF to quantitate the levels of 43 different proteins on reactivated cells from ART-suppressed, HIV-infected individuals, and implemented a bioinformatics approach to trace each reactivated cell to its original latent state.
PBMCs (n=7), rectosigmoid biopsies (n=7), and lymph node aspirates (n=2) from treated individuals were phenotyped by CyTOF immediately after cell isolation, or stimulated with PMA/ionomycin or LRAs and then phenotyped. Reactivated cells were traced back to their original pre-stimulation state using the bioinformatics approach PP-SLIDE (Cavrois et al, Cell Reports 2017). Markers identified as preferentially expressed on latent cells were validated by sorting the cells and then conducting viral outgrowth assays and proviral sequencing.
Latent cells were non-randomly distributed amongst memory CD4+ T cells. Markers preferentially expressed on latent cells included those that were shared between donors (PD1, CCR5, CD2, CD49d, Ox40) and donor-specific ones (CXCR5, TIGIT, CCR6, CD28, CD7). Markers differentially expressed between latent cells in blood vs. tissues, and between latent cells reactivatable by different stimulation methods, were identified. Analysis of longitudinal samples suggested the phenotype of latent cells is stable over time. Multiparameter sorting revealed that donor-shared surface markers identified by CyTOF markedly enriched for latent cells with replication-competent HIV: Tfh, already highly enriched for replication-competent HIV, was further enriched by 3 orders of magnitude using such markers. Viral sequencing analysis revealed the enriched cells to be largely clonally expanded.
We have validated CyTOF phenotyping of reactivated latent cells paired with bioinformatics analysis by PP-SLIDE as an effective way to chart the in vivo blood and tissue HIV latent reservoir. Our results demonstrate that 1) latent cells are not randomly distributed amongst memory CD4+ T cells, 2) the phenotypes of latent cells are stable over time, 3) LRAs can target different latent cells than PMA/ionomycin, 4) there are shared as well as donor-specific surface markers of latent cells, and 5) sorting of cells based on surface markers identified by CyTOF markedly enriches for clonally-expanded latent cells with replication-competent HIV.