Single-cell transcriptomics reveals gene expression heterogeneity but suffers from stochastic dropout and characteristic bimodal expression distributions in which expression is either strongly non-zero or non-detectable. We propose a two-part, generalized linear model for such bimodal data that parameterizes both of these features. We argue that the cellular detection rate, the fraction of genes expressed in a cell, should be adjusted for as a source of nuisance variation. Our model provides gene set enrichment analysis tailored to single-cell data. It provides insights into how networks of co-expressed genes evolve across an experimental treatment. MAST is available at https://github.com/RGLab/MAST.Electronic supplementary materialThe online version of this article (doi:10.1186/s13059-015-0844-5) contains supplementary material, which is available to authorized users.

CD8+ T cells only require a brief stimulation with antigen in vitro to divide and differentiate into effector and memory cells upon transfer in vivo. The efficiency of clonal expansion and the functional characteristics of memory cells derived from briefly stimulated cells are poorly defined. We developed a system that allowed us to examine programming entirely in vivo. This was achieved by rapidly killing peptide-pulsed DCs carrying a diphtheria toxin receptor transgene with timed injections of diphtheria toxin without altering the course of an accompanying infection. The magnitude of clonal expansion, but not the functionality of the effector cells, correlated directly with the duration of antigen exposure. Furthermore, memory T cells were capable of mounting a secondary response, regardless of the length of antigen encounter during the primary response. These results indicate that the duration of initial antigen encounter influences the magnitude of the primary response, but does not program responsiveness during the secondary challenge.

Using transgenic mice that express a constitutively active version of STAT5b, we demonstrate that STAT5 plays a key role in governing B cell development and T cell homeostasis. STAT5 activation leads to a 10-fold increase in pro-B, but not pro-T, cells. Conversely, STAT5 signaling promotes the expansion of mature αβ T cells (6-fold increase) and γδ and NK T cells (3- to 4-fold increase), but not of mature B cells. In addition, STAT5 activation has dramatically divergent effects on CD8+ vs CD4+ T cells, leading to the selective expansion of CD8+ memory-like T cells and CD4+CD25+ regulatory T cells. These results establish that activation of STAT5 is the primary mechanism underlying both IL-7/IL-15-dependent homeostatic proliferation of naive and memory CD8+ T cells and IL-2-dependent development of CD4+CD25+ regulatory T cells.

While the specificity and development of natural killer (NK) cells have been intensely studied, little is known about homeostasis of the mature NK population. Here we show that mouse NK cells undergo homeostatic proliferation when transferred into NK-deficient Rag−/− γC−/− hosts. Normal NK functional activity is maintained during this process, although there are some changes in NK phenotype. Using cell sorting, we demonstrate that mature (Mac-1hi) NK cells undergo homeostatic proliferation in an NK-deficient environment, yet immature (Mac-1lo) NK cells also proliferate in such hosts. We find that mature NK cells survive but do not proliferate in hosts which possess an endogenous NK pool. However, we go on to show that mature NK survival is critically dependent on interleukin (IL)-15. Surprisingly, NK survival is also compromised after transfer of cells into IL-15Rα−/− mice, implying that IL-15 responsiveness by bystander cells is critical for NK maintenance. These data imply that, similar to T cells, homeostasis of the NK pool is much more dynamic than previously appreciated and this may be relevant to manipulation of NK cells for therapeutic purposes.