Widespread Shortening of 3' Untranslated Regions and Increased Exon Inclusion Are Evolutionarily Conserved Features of Innate Immune Responses to Infection

Abstract

The contribution of pre-mRNA processing mechanisms to the regulation of immune responses remains poorly studied despite emerging examples of their role as regulators of immune defenses. We sought to investigate the role of mRNA processing in the cellular responses of human macrophages to live bacterial infections. Here, we used mRNA sequencing to quantify gene expression and isoform abundances in primary macrophages from 60 individuals, before and after infection with Listeria monocytogenes and Salmonella typhimurium. In response to both bacteria we identified thousands of genes that significantly change isoform usage in response to infection, characterized by an overall increase in isoform diversity after infection. In response to both bacteria, we found global shifts towards (i) the inclusion of cassette exons and (ii) shorter 3' UTRs, with near-universal shifts towards usage of more upstream polyadenylation sites. Using complementary data collected in non-human primates, we show that these features are evolutionarily conserved among primates. Following infection, we identify candidate RNA processing factors whose expression is associated with individual-specific variation in isoform abundance. Finally, by profiling microRNA levels, we show that 3' UTRs with reduced abundance after infection are significantly enriched for target sites for particular miRNAs. These results suggest that the pervasive usage of shorter 3' UTRs is a mechanism for particular genes to evade repression by immune-activated miRNAs. Collectively, our results suggest that dynamic changes in RNA processing may play key roles in the regulation of innate immune responses.

Fig 1. Gene expression and isoform proportion differences in response to bacterial infection.

(A) Principal component analysis of gene expression data from all samples (PC1 and PC2 on the x- and y-axis, respectively). (B) IL24, a gene with significant changes in isoform usage upon infection with Listeria and Salmonella. For each IL24 isoform in response to infection, plotted are the average relative isoform usages across samples (left panel, top) with their isoform structures (left panel, bottom) and corresponding fold changes (right panel; log₂ scale; with standard error bars). Isoforms are ordered by relative abundance in non-infected samples, and colored circles (right panel) correspond to colors in barplot (left panel). (C) Number of genes showing only DIU, only DGE, and both DIU and DGE upon infection with Listeria and Salmonella, (11,353 genes tested).

Fig 2. RNA processing changes in response to bacterial infection.

(A) Proportion of events for RNA processing category that are significantly changing after infection with Listeria (left), Salmonella (middle), or variation between non-infected samples as a control (right). Numbers indicate the number of significant changes per category. (B) Significantly Gene Ontology categories for genes with any significant RNA processing change (FDR ≤ 10%). (C) Distribution of ΔΨ values for each RNA processing category. Negative values represent less inclusion, while positive values represent more inclusion, as defined by the schematic exon representations.

Fig 3. 3’RNA sequencing shows increased usage of upstream polyadenylation sites upon infection.

(A) Meta-gene distributions of 3’RNA-seq read densities at the upstream polyA sites (core regions, left) and downstream polyA sites (extended regions, right) of Tandem 3’ UTRs after infection with Listeria or Salmonella (top and bottom, respectively). Shown are the read distributions for non-infected samples across all Tandem 3’ UTRs (black) and infected samples at Tandem 3’ UTRs that significantly change after infection (yellow) or show no change after infection (blue), as called by the RNA-seq data. (B) Distribution of ΔΨ values calculated from 3’RNA-seq data for Tandem 3’ UTRs. We observe significant shifts (P < 2.2 × 10⁻¹⁶ for both Listeria and Salmonella) towards negative ΔΨ values in Tandem UTRs that are identified as significantly changing in RNA-seq data (yellow) relative to Tandem UTRs without any change after infection (blue).

Fig 4. Directed shifts in RNA processing persist across time, stimulus conditions, and closely related species.

(A) Correlations between ΔΨ values after 2 hours of infection (x-axis) and ΔΨ values after 24 hours of infection (y-axis), presented as density plots per RNA processing category in contrasting colors. Plotted are events that are significant after 2 hours of either Listeria or Salmonella infection. (B) Distributions of ΔΨ values for skipped exons (purple) and TandemUTRs (blue) following infection of whole blood cells with lipopolysaccharide, assessed in human (top), chimpanzee (middle) and rhesus macaque (bottom) individuals (N = 6 per species). We observe prominent global shifts in isoform distributions in human and macaque (SE_human P = 0.003, TandemUTR_human P = 2.4×10⁻¹³; SE_macaque P = 0.05, TandemUTR_macaque P = 8.8×10⁻⁶) with a more modest trend observed in chimpanzee, potentially due to poor transcript annotations in the chimpanzee genome (SE_chimpanzee P = 0.26, TandemUTR_chimpanzee P = 0.002). All p-values were calculated using a Student’s t-test testing deviation from a mean of zero.

Fig 5. Relationship between RNA processing changes and gene expression changes.

(A) Distribution of fold changes in gene expression (y-axis, log₂ scale) for genes with significant skipped exon changes after infection (purple) and genes with no change after infection (gray). (B) Distribution of Spearman correlations between ΔΨ and fold change in gene expression across 60 individuals per gene (solid line), for genes with only one annotated alternative event and significantly changed SE usage (purple; n_L = 46 genes and n_S = 97 genes) or tandem UTR usage (blue; n_L = 36 genes and n_S = 86 genes). Dotted lines show distribution of the correlation coefficients after permuting ΔΨ values.

Fig 6. Tandem 3’ UTR shortening allows evasion of regulation by miRNAs.

(A) Distribution of frequency of miRNA target sites per nucleotide in the extended regions of Tandem UTRs that either show no change after infection (grey) or significantly change after infection (blue). (B) Significantly enriched miRNA target sites (FDR ≤ 10%, |FC| > 1.5) in the extended regions of significantly changing Tandem UTRs after infection with Listeria-only (top), with Salmonella-only (middle), or both bacteria (bottom). For each bacteria, the barplots in the left panels show the fold enrichment (x-axis, log₂ scale) of target sites in the extended regions. White bars represent non-significant enrichments. Panels on the right show the fold change in miRNA expression (x-axis, log₂ scale with standard error bars) after either 2 hours of infection (light colors) or 24 hours of infection (dark colors).