Unbiased reconstruction of a mammalian transcriptional network mediating pathogen responses

Abstract

Models of mammalian regulatory networks controlling gene expression have been inferred from genomic data but have largely not been validated. We present an unbiased strategy to systematically perturb candidate regulators and monitor cellular transcriptional responses. We applied this approach to derive regulatory networks that control the transcriptional response of mouse primary dendritic cells to pathogens. Our approach revealed the regulatory functions of 125 transcription factors, chromatin modifiers, and RNA binding proteins, which enabled the construction of a network model consisting of 24 core regulators and 76 fine-tuners that help to explain how pathogen-sensing pathways achieve specificity. This study establishes a broadly applicable, comprehensive, and unbiased approach to reveal the wiring and functions of a regulatory network controlling a major transcriptional response in primary mammalian cells.

Fig. 1. A systematic strategy for network reconstruction

The strategy consists of four steps, from left, (1) state measurement using arrays, (2) selection of regulators and response signatures, (3) network perturbation with shRNAs against each regulator followed by measurement of signature genes, and (4) network reconstruction from the perturbational data.

Fig. 2. Gene expression response to pathogen components

A, mRNA profiles of the 1800 genes whose expression was induced to at least 1.7-fold from baseline level in both duplicates of at least one time point in CD11c+ dendritic cells stimulated with the indicated pathogen component across a time course of 0, 0.5, 1, 2, 4, 6, 8, 12, 16, 24 hours (tick marks; GRD – gardiquimod). Replicates were collapsed and genes hierarchically clustered. Rows – genes, columns – experiments, red – induced from baseline, blue – repressed from baseline, white – unchanged from baseline. B, Model illustrating the differential networks controlling the anti-viral (‘polyI:C-like’) and inflammatory (‘PAM-like’) programs.

Fig. 3. Gene regulatory programs controlling the response to pathogen components

A,B, A strategy to minimize false discovery (FDR) calls of significant changes in an output target gene resulting from knockdown of a regulator gene. A, The first FDR procedure (top) compares the expression of the gene following a perturbation with a regulator shRNA (right) to its expression upon perturbation with 32 non-targeting shRNAs (left). The dashed lines identify the gene-specific FDR-based thresholds for induction (blue line) and repression (red line). A discrete vector of significant calls (bottom) is derived from the raw data (blue – regulator represses the target gene; red – regulator induces the target gene). B, A second FDR procedure (top) compares the expression of the target gene to that of eight control (target) genes upon perturbation with the same shRNA. In the example shown, the gene's induction (left) was significant relative to the variation in expression among the control target genes resulting in a high score (bottom, dark blue), but its repression did not significantly differ from the controls, resulting in a lower score (bottom, weaker red). C. All significant (95% confidence) relations between regulators (columns) and targets (rows), colored as in B. The gray bar (right) represents the NMF-based calls for each target gene; black – anti-viral program; dark gray – inflammatory program; light gray – control genes. The bottom bar shows the degree of effect by the regulator on each program as determined by the NMF projection of the regulator's perturbation profile. (Yellow – positive effect, Green – negative effect; NMF scores are mean-normalized).

Fig. 4. The core regulatory circuits controlling the inflammatory and anti-viral responses

A. The antiviral sub-network shows regulatory relations between the core anti-viral regulators (blue nodes, top), their targets (boxes, bottom), each other, and inflammatory regulators (green node, top right). The two top regulators, Stat1 and 2, activate all anti-viral targets (dashed blue arrows). The second-tier regulators activate sub-sets of targets (dashed purple arrows). B, Examples of feed forward loop classes identified in the network, with fraction of each class. C. A core regulatory network of the inflammatory and anti-viral programs, consisting of the most distinct regulators, and their relation to ligands and receptors (top). Pointed arrows – induction; blunt arrows – repression; green ovals – inflammatory regulators; blue ovals – antiviral regulators. Example target genes are noted. D. nCounter expression profiles for the target genes (rows) upon perturbation with shRNAs against a subset of viral regulators (columns) and followed by stimulation with LPS (left) or polyI:C (right). All values are normalized by expression in cells infected with a control shRNA and under the same stimulus (shCtl).

Fig 5. The polycomb component Cbx4 selectively restricts IFNB1 production under bacterial perturbations

A,C LPS (red), polyI:C (blue) and PAM (green)-induced expression of ifnb1 (A) and cbx4 (C) derived from data in Fig. 2A. B. by nCounter) in response to LPS in DCs perturbed by control shRNAs or shRNAs targeting each of 125 regulators (format as in Fig. 3B). D. Ifnb1 mRNA levels (by qRT-PCT) following LPS treatment in unsorted mouse bone marrow DCs perturbed with an shRNA against Cbx4 (black) or a control shRNA (grey); signals are relative to t=0. E. ifnb1 mRNA levels (by nCounter) at 6h post-LPS or PAM in bone marrow DCs perturbed with an shRNA against Cbx4 or one of three control shRNAs. F. Model for bacterial-specific repression of ifnb1 by Cbx4: both polyI:C and LPS induce ifnb1 expression early (left), but only LPS induces Cbx4, which then represses the ifnb1 locus at a later time (right, top).