Gene expression variability across cells and species shapes innate immunity

Abstract

SummaryAs the first line of defence against pathogens, cells mount an innate immune response, which is highly variable from cell to cell. The response must be potent yet carefully controlled to avoid self-damage. How these constraints have shaped the evolution of innate immunity remains poorly understood. Here, we characterise this programme’s transcriptional divergence between species and expression variability across cells. Using bulk and single-cell transcriptomics in primate and rodent fibroblasts challenged with an immune stimulus, we reveal a striking architecture of the innate immune response. Rapidly diverging genes, including cytokines and chemokines, also vary across cells and have distinct promoter structures. Conversely, genes involved in response regulation, such as transcription factors and kinases, are conserved between species and display low cell-to-cell variability. We suggest that this unique expression pattern, observed across species and conditions, has evolved as a mechanism for fine-tuned regulation, to achieve an effective but balanced response.

Extreme heterogeneity of influenza virus infection in single cells

“…Influenza virus infection can trigger innate-immune sensors that lead to the transcriptional induction of type I and III interferons, and subsequently of anti-viral interferon-stimulated genes ( Killip et al, 2015; Iwasaki and Pillai, 2014 ; Crotta et al, 2013 ). However, activation of the interferon response is stochastic and bi-modal at the level of single cells ( Chen et al, 2010 ; Shalek et al, 2013 , 2014 ; Perez-Cidoncha et al, 2014 ; Bhushal et al, 2017 ; Hagai et al, 2017 ). We therefore hypothesized that we might see two sub-populations of infected cells: one in which the interferon response inhibited viral transcription, and another in which the virus was able to express high levels of its mRNA by evading or blocking this response.…”

“…Only one of the hundreds of virus-infected cells expresses any detectable interferon, despite the fact that a number of cells fail to express the Influenza-virus interferon antagonist NS1. It is known that interferon activation is stochastic at the level of single cells in response to both synthetic ligands ( Shalek et al, 2013 , 2014 ; Bhushal et al, 2017 ; Hagai et al, 2017 ) and actual infection ( Rand et al, 2012 ; Perez-Cidoncha et al, 2006 ; Avraham et al, 2015 ; Killip et al, 2017 ). But interferon expression is a prominent transcriptional signature of high-MOI Influenza virus infection of bulk cells, including in the epithelial cell line and at the time-points used in our experiments ( Geiss et al, 2002 ; Sutejo et al, 2012 ).…”

“…The Influenza virus genome consists of eight negative-sense RNA segments, and many infected cells fail to express one more of these RNAs ( Heldt et al, 2015 ; Dou et al, 2017 ) or their encoded proteins ( Brooke et al, 2013). In addition, activation of innate-immune responses is inherently stochastic ( Shalek et al, 2013 , 2014 ; Bhushal et al, 2017 ; Hagai et al, 2017 ), and only some Influenza-infected cells express anti-viral interferon genes ( Perez-Cidoncha et al, 2014 ; Killip et al, 2017 ). However, the extent of cell-to-cell variation in these and other host and viral factors remains unclear, as does the association among them in individual infected cells.…”

scTenifoldNet: a machine learning workflow for constructing and comparing transcriptome-wide gene regulatory networks from single-cell data

“…Next, we show the use of scTenifoldNet to a scRNAseq data set from human dermal fibroblasts [52]. In the original paper, Hagai et.…”

“…al. [52] focused on single-cell transcriptional responses induced by the stimulus of polyinosinic-polycytidylic acid (polyI:C), a synthetic double-stranded RNA (dsRNA) (Fig. 6A).…”