Activation of Clustered IFNγ Target Genes Drives Cohesin-Controlled Transcriptional Memory

Abstract

Cytokine activation of cells induces gene networks involved in inflammation and immunity. Transient gene activation can have a lasting effect even in the absence of ongoing transcription, known as long-term transcriptional memory. Here we explore the nature of the establishment and maintenance of interferon γ (IFNγ)-induced priming of human cells. We find that, although ongoing transcription and local chromatin signatures are short-lived, the IFNγ-primed state stably propagates through at least 14 cell division cycles. Single-cell analysis reveals that memory is manifested by an increased probability of primed cells to engage in target gene expression, correlating with the strength of initial gene activation. Further, we find that strongly memorized genes tend to reside in genomic clusters and that long-term memory of these genes is locally restricted by cohesin. We define the duration, stochastic nature, and molecular mechanisms of IFNγ-induced transcriptional memory, relevant to understanding enhanced innate immune signaling.

Keywords: GBP5; cohesin; epigenetics; gene regulation; immunological priming; interferon γ; signaling; stochastic gene expression; transcription; transcriptional memory.

Graphical abstract

Figure 1

Long-Term Transcriptional Memory of Genes Following Priming with IFNγ (A) Principle of cytokine-induced transcriptional memory. (B) Experimental outline of the transcriptional memory experiment. (C) Plot showing differential rates of reinduction relative to priming as measured by RNA-seq. Average read counts for three replicate experiments, as outlined in (B), were assembled for each gene. The log2 fold change of the gene read count following reinduction over those following priming is plotted. “0” indicates no change, whereas positive values indicate increased expression upon reinduction. Data were ranked according to the mean expression level for both of the conditions and all replicates. Genes that show strong transcriptional memory or tolerance (reduced expression upon reinduction) are labeled. Red dots represent genes with a p-adj value below 0.1. (D) HeLa cells were primed and reinduced according to the regimen outlined in (B). GBP5 mRNA levels were quantified by RT-qPCR and normalized to ACTB expression. Error bars, SD; n = 3 biological replicates. (E) HeLa cells and human male primary fibroblasts were subjected to the IFNγ treatment regimen outlined in (B), processed for western blotting, and probed for GBP5 protein expression. α-Tubulin (α-TUB), loading control; ^∗, cross-reacting band. See also Figure S1.

Figure 2

Memory of Prior IFNγ Induction Is Reversible but Persists for up to 14 Days in Continuously Cycling Cells (A) Scheme outlining the experiment to determine the duration of IFNγ-mediated transcriptional memory. (B) Cells subjected to the IFNγ treatment regimen outlined in (A) were processed for western blotting and probed for GBP5 protein expression. α-TUB, loading control. (C) Cells as in (B) but processed for RT-qPCR of GBP5 and HLA-DRA mRNA. Expression levels were normalized to ACTB and internally to cells after priming. Error bars, SD; n = 3 biological replicates; numbers represent the p values.

Figure 3

Priming Results in Increased Frequency of Activation and Enhanced GBP5 Expression upon Reinduction (A) Alternative solutions for achieving IFNγ transcriptional memory. (B) Scheme describing a single-cell RNA-seq transcriptional memory experiment to distinguish between the alternatives from (A). (C) Representation of the single-cell RNA-seq data from HeLa cells for the GBP5 gene from the experiment shown in (B). Each dot represents the expression level of the GBP5 gene for one cell in the naive (N = 90), priming (N = 89), and reinduction state (N = 91). (D) Binned representation of the single-cell RNA-seq data for the GBP5 gene from the experiment shown in (C). (E) Scheme outlining the GBP5 promoter trap cell line, in which EGFP was inserted into exon 2 downstream of the GBP5 translation start site. One allele was targeted, and the remaining allele remained unperturbed. Purple, 5′ UTR; blue, GBP5 coding sequence. (F) EGFP::GBP5 cells were subjected to the IFNγ treatment regimen outlined in Figure 1B, processed for fluorescence western blotting, and probed for GBP5 and EGFP expression. α-TUB, loading control. Tubulin-normalized fluorescence intensities are plotted. (G) EGFP::GBP5 cells were subjected to the IFNγ treatment regimen as outlined in (B) and processed for cytometry. Cell frequencies as a function of EGFP fluorescence intensity are plotted. Red dotted lines are fiducial marks based on untagged cells and are used to define the cutoff for cell percentages expressing or not expressing. See also Figure S2.

Figure 4

GBP5 and HLA-DRA mRNA Levels in Primed Cells Return to the Pre-induced Levels of Naive Cells, and Ongoing Transcription Is Not Required for Maintenance of Transcriptional Memory (A) Scheme to measure long-term transcriptional output of memory genes following priming. (B) HeLa cells were subjected to the IFNγ treatment regimen outlined in (A) and processed for RT-qPCR of GBP5 and HLA-DRA mRNA. Signals were normalized to ACTB expression and internally to naive cells. Error bars, SD; n = 3 biological replicates; numbers represent the p values. (C) Representation of processed RNA-seq data in HeLa cells analogous to data in Figure 1C but for primed over naive cells. (D) Outline of the triptolide-based RNA Pol II inhibition experiment. (E) HeLa cells were subjected to the IFNγ and triptolide treatment regimen outlined in (D) and processed for RT-qPCR of GBP5 mRNA. Signals were normalized to ACTB expression. Error bars, SD; n = 3 biological replicates. See also Figure S3.

Figure 5

Short-Term Maintenance of H3.3, H3K4me2, and H3K79me2 on the GBP5 and GBP4 Genes following IFNγ Priming (A) Outline of the experiment to measure chromatin status following IFNγ-mediated priming. (B) HeLa cells ectopically expressing H3.3-SNAP-HA were subjected to the IFNγ treatment regimen outlined in Figure 1B with a 2- and 7-day window of memory, processed for western blotting, and probed for GBP5 protein expression. α-TUB, loading control. (C) Representation of data for processed chromatin accessibility (ATAC-seq) and occupancy of the indicated chromatin-associated factors (ChIPmentation) for the time points indicated in (A). Sequenced reads were mapped to the human genome (hg38), and coverage data are displayed with internal scaling between naive, primed, 2- and 7-day IFNγ washout samples. Two proximally positioned memory genes, GBP4 and GBP5, are shown. (D) Enlarged presentation of the GBP4 and GBP5 genes.

Figure 6

Cohesin Negatively Regulates Memory for Most Genes in the GBP and HLA Clusters but Not for Genes Outside of Those Clusters (A) Outline of a transcriptional memory experiment (analogous to Figure 1B) combined with auxin-mediated transient depletion of SCC1 during priming, analyzed by RNA-seq. (B) HeLa Kyoto SCC1-EGFP-AID osTIR1-positive cells were subjected to the IFNγ and auxin treatment regimen outlined in (A) and processed for RNA-seq. The Log2 fold change for memory genes between reinduction (with or without auxin) and priming (without auxin) was plotted. “0” indicates no memory. The genes shown were selected based on a p-adj value below 0.001 as determined by the DESeq2 software (Love et al., 2014). (C) Top: representation of the genomic structure of the GBP locus. Bottom: individual gene plots from the data described in (B). Error bars, SD; n = 3 biological replicates. (D) Top: representation of the genomic structure of the HLA locus. Bottom: data presented as in (C) but for the HLA cluster genes. (E) Data presented as in (C) but for the CD74 and AKNA genes. See also Figures S4–S7.

Figure 7

Cohesin Bound at the GBP Cluster Inhibits Establishment of Transcriptional Memory within the Cluster but Not for Distal Genes (A) Representation of processed data for occupancy of SCC1 (ChIPmentation) and chromatin accessibility (ATAC-seq; see also Figure 5C) during the IFNγ long-term memory assay described in Figure 5A. Data are plotted as in Figure 5C. The results are shown for the GBP cluster; TAD boundaries (Figure S5A) and prominent cohesin sites are indicated. (B) Enlarged presentation of three selected cohesin sites. (C) Approximately 1-kb regions encompassing cohesin-bound sites boxed as A, B, and C in (A) were deleted by CRISPR-Cas9. Two independent polyclonal populations were generated for site A, single polyclonal populations for sites B and C, and a monoclonal clone for site C. Mutant cells were subjected to the transcriptional memory experiment outlined in Figure 1B. mRNA levels were quantified by RT-qPCR for the indicated genes and normalized to ACTB expression. Error bars, SD; n = 3 (n = 6 for site A) biological replicates. n.s., non-significant (p > 0.05). See also Figure S7.