Unsupervised learning techniques reveal heterogeneity in memory CD8+ T cell differentiation following acute, chronic and latent viral infections

CD8+ T lymphocytes are critical for the control of gammaherpesvirus latency. To determine how memory CD8+ T cells generated during latency differ from those primed during acute or chronic viral infection, we adoptively transferred naive P14 CD8+ T cells into uninfected recipients, and examined surface proteins, cytokines and transcription factors following infection with the Armstrong (acute) or Clone 13 (chronic) strains of lymphocytic choriomeningitis virus (LCMV), or murine gammaherpesvirus 68 (MHV68) expressing the LCMV epitope DbGP33-41. By performing k-means clustering and generating self organizing maps (SOM), we observed increased short-lived effector-like, CD27lo CD62Llo and Bcl-6lo CD8+ T cells following latent infection. In addition, we found that memory CD8+ T cells from latent primed mice underwent less expansion following adoptive transfer and antigen rechallenge. Data from cluster models were combined and visualized by principal component analysis (PCA) demonstrating memory CD8+ T cells from latent infection occupy an intermediate differentiation space. •We developed a novel MHV68 strain expressing a LCMV epitope.•We compared memory CD8+ T cell differentiation during different viral infections.•Flow cytometry data were analyzed using unsupervised machine learning techniques.•Latent primed memory CD8+ T cells occupy an intermediate differentiation space.