Transcriptional Mapping of T Cell Memory Reveals That Chronic Viral Infection Arrests Memory Differentiation.

Human T cells responding to cancer and chronic viral infection are dysfunctional and ineffective. In contrast, T cells responding to acute viral infection provide highly-effective long-term immunity. The basis for this difference is not known. However understanding the T cell defects in cancer and chronic viral diseases is critical to optimizing immunotherapy for these diseases. We hypothesized that it is impaired differentiation of naive T cells into memory T cells that causes the dysfunctional response in cancer and chronic viral infection, because of a high, persistent antigen load. To test this hypothesis, we studied CD8 T cell memory differentiation in 2 murine models of lymphocytic choriomeningitis virus infection: an acute infection in which virus-specific T cells are fully functional; and a chronic infection model in which T cells are dysfunctional. In order to define how the defective T cells in chronic infection differ from fully-functional T cells in acute infection, we generated gene expression profiles from naive (day 0), effector (day 7 [d7]) and memory (day 30 [d30]) T cells using oligonucleotide microarrays. In acute infection, these differentiation states were so distinct that 100% of naive, effector and memory T cell samples could be correctly classified, using a k-Nearest Neighbor prediction classifier (kNN) with leave-one-out cross validation. Next, we determined where d30 T cells from chronic infection lay relative to the “differentiation-space” of acute infection. None of the d30 T cell samples from the chronic model were classified as memory by kNN prediction; instead, all were misclassified as effectors, i.e. at an earlier stage of differentiation. However, d30 T cells from chronic infection were not simply persistent effectors. Direct comparison of chronic d30 T cells with d7 effectors revealed marked differences in gene expression pattern. This suggests that the dysfunctional chronic d30 cells had adopted a discrete differentiation state similar to, but distinct from, effectors. Lastly, we identified biologically-related sets of genes that were enriched in naive-to-memory differentiation in acute, but not chronic infection, using gene-set enrichment analysis. In the acute model, Stat-5 target genes were significantly enriched in memory cells relative to naive cells (p=0.004 by permutation testing). In contrast, Stat-5 target genes were far less enriched in d30 T cells from the chronically-infected model. Our results show that T cells exposed to high, persistent antigen-load fail to complete normal memory differentiation and are arrested in a distinct, effector-like phase, which is characterized by reduced Stat-5 signaling. We are currently mapping memory T cell differentiation in humans to determine if a similar molecular basis explains the T cell defects seen in cancer and chronic viral infection. Mapping the memory differentiation states of antigen-specific T cells will provide a novel approach to interrogating the efficacy of immunotherapies for cancer and chronic viral infections in humans.