RBM39 shapes innate immunity by controlling the expression of key factors of the interferon response

Abstract

Background and aims: The contribution of innate immunity to clearance of viral infections of the liver, in particular sensing via Toll-like receptor 3 (TLR3), is incompletely understood. We aimed to identify the factors contributing to the TLR3 response in hepatocytes via CRISPR/Cas9 screening.

Methods: A genome-wide CRISPR/Cas9 screen on the TLR3 pathway was performed in two liver-derived cell lines, followed by siRNA knockdown validation. SiRNA knockdown and indisulam treatment were used to study the role of RNA-binding motif protein 39 (RBM39) in innate immunity upon poly(I:C) or cytokine treatment and viral infections. Transcriptome, proteome, and alternative splicing were studied via RNA sequencing and mass spectrometry upon depletion of RBM39.

Results: Our CRISPR/Cas9 screen identified RBM39, which is highly expressed in hepatocytes, as an important regulator of the TLR3 pathway. Knockdown of RBM39 or treatment with indisulam, an aryl sulfonamide drug targeting RBM39 for proteasomal degradation, strongly reduced the induction of interferon-stimulated genes (ISGs) in response to double-stranded RNA (dsRNA) or viral infections. RNA sequencing (seq) and mass spectrometry identified that transcription and/or splicing of the key pathway components IRF3, RIG-I, and MDA5 were affected by RBM39 depletion, along with multiple other cellular processes identified previously. RBM39 knockdown further restrained type I and type III IFN pathways by reducing the expression of individual receptor subunits and STAT1/2. The function of RBM39 was furthermore not restricted to hepatocytes.

Conclusion: We identified RBM39 as a regulatory factor of cell intrinsic innate immune signaling. Depletion of RBM39 impaired TLR3, RIG-I/MDA5, and IFN responses by affecting the basal expression of key pathway components.

Keywords: IFNs; IRF3; RBM39; STAT1; STAT2; splicing.

Figure 1

A genome-wide CRISPR/Cas9 screen identifies potential TLR3-related factors. (A) Schematic of the lentiviral truncated BID (tBID) death reporter system. The tBID expression is driven by the IFIT1 promoter, while the expression of the neomycin selection marker (NeoR) is under the control of the constitutive PGK promoter. Working principle: (1) dsRNA is taken up via endocytosis and transported to the endosome. (2) The endosomal TLR3 recognizes dsRNA and triggers signaling via TRIF, resulting in the activation of transcription factors IRF3 and NF-κB. (3) Activated transcription factors induce the expression of tBID by binding the IFIT1 promoter. (4) tBID associates with Bcl-2 proteins BAX and BAK to form a complex that permeabilizes the outer mitochondrial membrane and mediates the release of cytochrome c (Cyt c) into the cytoplasm. (5) Cyt c binds APAF1, triggering its oligomerization and binding to Pro-Caspase-9. Activated Caspase-9, in turn, activates Caspase-3 and -7, inducing apoptosis. (B) Workflow of the CRISPR/Cas9 screen. PH5CH and Huh7-Lunet-TLR3 stably expressing the death reporter and Cas9 were transduced with the genome-wide lentiviral CRISPR sgRNA library and selected with puromycin for stable expression. TLR3 stimulation with poly(I:C) induced the expression of tBID and thus apoptosis in case of intact TLR3 signaling. Then, surviving TLR3-deficient cells were enriched and collected for next-generation sequencing (NGS). (C) Illustration of the filtering steps to generate the final hit list. For each experiment independently, the ~2,000 most enriched gRNAs were determined in the NGS dataset and matched to their respective target genes, creating a list of 2,154 initial genes of interest. Subsequently, it was only focused on protein-coding genes. To ensure reproducibility, only genes with enriched gRNAs in three out of four repetitions were considered; this also eliminated genes showing enrichment in only one of the cell lines used. Lastly, genes were filtered to have at least five out of six available gRNAs enriched in at least one repetition to remove hits driven by potential off-target effects of just a subset of gRNAs. (D–F) A total of 50 candidate genes were selected and subjected to siRNA silencing for 48 h; then, the cells were stimulated with 50 μg/mL poly(I:C) supernatant feeding for 8 h. IFIT1 mRNA was measured by RT-qPCR normalized to GAPDH and expressed as relative expression to siRNA non-targeting (siNT)-treated samples. Knockdown of the candidates upregulated (D) or downregulated (E) IFIT1 mRNA expression in Huh7-Lunet-TLR3 cells. Knockdown of RBM39, PCF11, PTPRT, and KDM2A in PH5CH cells reduced IFIT1 mRNA expression in PH5CH cells (F). siUNC93B1, siTLR3, and siTRIF were used as positive controls. The data are from three biological replicates (n = 3); error bars indicate standard deviation (SD).

Figure 2

RBM39 is crucial for the TLR3 pathway. (A) Simplified schematic of the TLR3-IRF3 pathway. (B–E) RBM39 rescue experiment. PH5CH cells expressing RBM39.Esc or empty vector were transfected with siRBM39 or siNT/siTRIF as controls for 48 h and then supernatant-fed with 50 µg/mL poly(I:C) for 6 h. IFIT1 (B), ISG15 (C), MxA (D), and CXCL10 (E) mRNA was measured by RT-qPCR; RBM39 protein expression was measured by immunoblotting (B). (F–H) PH5CH cells were transfected with siRNA for 48 h and then fed with 50 μg/mL poly(I:C) in the supernatant for 6 h. Secreted IFN-β protein (F) was measured by ELISA. Simplified schematic of the TLR3-NF-κB pathway (G). TNFAIP3 (H) and IL6 (H) mRNA was quantified by RT-qPCR. (I) Chemical structure of indisulam. (J) PH5CH cells were treated with 1 µM indisulam or DMSO as control for 48 h and further stimulated with 50 µg/mL poly(I:C) in the supernatant for 6 h. RBM39 degradation was measured by western blot (left). IFIT1 mRNA expression was measured by RT-qPCR (right). Relative expression to siNT control was shown; mRNA fold change was normalized to GAPDH. Data are obtained from three biological replicates (n = 3); error bars refer to SD (all panels). Statistical significance was assessed through Welch’s unpaired t-test. ns = not significant, * = p < 0.05, ** = p < 0.01, *** = p < 0.001.

Figure 3

RBM39 participates in the RIG-I/MDA5 response. (A) Simplified schematic of the RIG-I/MDA5-IRF3 pathway. (B–D) Huh7.5 cells with ectopic RIG-I (B) or MDA5 (C) expression and A549 cells (D) were transfected with siRBM39 or siNT/siMAVS as control for 48 h and then transfected with 0.5 µg/mL poly(I:C) for 6 h. IFIT1 mRNA was measured by RT-qPCR and is shown relative to siNT. (E) A549 cells were treated with indisulam or DMSO as control for 48 h and then transfected with 0.5 µg/mL poly(I:C) for 6 h. RBM39 degradation was determined by immunoblotting (left). IFIT1 mRNA was quantified by RT-qPCR (right). (F, G) A549 cells were transfected with siRBM39 or siNT/siMAVS as control (F) or treated with 1 µM indisulam or DMSO as control (G). At 48 h after treatment, the cells were infected with Sendai virus (MOI = 1) for 24 h. IFIT1 mRNA (left) and Sendai virus RNA copies (right) were measured through RT-qPCR. mRNA fold change was normalized to GAPDH. The data were obtained from three biological replicates (n = 3); error bars refer to SD. Statistical significance was assessed through Welch’s unpaired t-test. * = p < 0.05, ** = p < 0.01, *** = p < 0.001, **** = p < 0.0001.

Figure 4

RBM39 controls the basal expression level of IRF3. (A) PH5CH cells or PH5CH RBM39.Esc cells were transfected with siRBM39 or siNT/siTRIF as controls for 48 h and then fed with 50 µg/mL poly(I:C) for 6 h. Phosphorylated IRF3 (p-IRF3), IRF3, RBM39, and β-actin protein expression levels were assessed by western blot (left). Quantification of three biological replicates is shown (right). (B, C) PH5CH cells (B) and primary human hepatocytes (PHH) (C) were treated with increasing concentrations of indisulam as indicated. IRF3, RBM39, and β-actin protein expression levels were detected through immunoblotting (left). Cell viability was measured via CellTiter-Glo luminescent cell viability assay (right) and normalized to DMSO control. Quantification of three independent experiments was expressed as average and is shown under each band. Protein expression was normalized to β-actin; relative protein levels to siNT or DMSO control are shown. The data shown are from three biological replicates (n = 3); error bars indicate SD. Statistical significance was assessed through Welch’s unpaired t-test. ns = not significant, **** = p < 0.0001.

Figure 5

Global proteomic, transcriptomic, and splicing analysis upon modulation of RBM39 abundance. PH5CH cells were treated with siRBM39 vs. siNT or 1 µM indisulam vs. DMSO after 48 h. The proteome was measured by mass spectrometry. Transcriptome and splicing were measured through RNA-sequencing. (A, B) Volcano plot of the proteomic data for RBM39 knockdown (A) or indisulam-treated (B) samples; the numbers of down- and upregulated factors are shown under the plot. Factors with a fold change >2 or <0.5 and a p-value <0.05 were highlighted with orange (A) or green (B). All factors with a p-value <0.05 but lower than twofold change are shown in black. IRF3, RBM39, RIG-I, MDA5, STAT1, and STAT2 were highlighted. (C, D) Volcano plot of DEGs in siRBM39- (C) and indisulam-treated (D) samples; the numbers of down- and upregulated factors are shown under the plot. DEGs with a fold change >2 or <0.5 and a p-value <0.05 were highlighted with red (C) and blue (D). All factors with a p-value <0.05 but lower than twofold change are shown in black. IRF3, RBM39, RIG-I, MDA5, STAT1, STAT2, IFNAR1, IFNAR2, IFNLR1, and IL10RB were highlighted. (E) RNA-seq data is presented as differential expressed gene (DEG) and different transcript usage (DTU). The overview of DEG and DTU numbers in each condition and overlaps are indicated. The RNA-seq data are from three biological replicates (n = 3), and the proteomic data are from four replicates (n = 4). The proteomic data was analyzed via Perseus1.6.15.0. DEG and DTU analyses of individual genes were performed using DESeq2 and DRIMseq, respectively. Statistical significance was evaluated by “two-sample tests” (Student’s t-test with permutation-based FDR 0.05 and 250 number of randomizations) for proteomics and with the Wald test and corrected for multiple testing according to Benjamini–Hochberg for RNAseq analysis.

Figure 6

RBM39 controls the basal expression of IRF3 and RIG-I and MDA5. (A) DEG and DTU analysis of IRF3 mRNA. (B) Simplified schematic of the TLR3 pathway. (C) IRF3 was ectopically expressed in PH5CH IRF3 KO cells, and cells were transfected with siRBM39 or siNT/siTRIF. At 48 h after knockdown, the cells were fed with 50 µg/mL poly(I:C) in the supernatant for 6 h. IFIT1 mRNA expression levels were measured by RT-qPCR. (D, E) DEG and DTU analysis of RIG-I (D) and MDA5 (E) mRNA. (F) Simplified schematic of RIG-I/MDA5-IRF3 signaling. (G) A549 cells overexpressing IRF3 were transfected with siRBM39 or siNT/siMAVS as controls for 48 h. After the knockdown, the cells were transfected with 0.5 µg/mL poly(I:C) in the supernatant for 6 h. IFIT1 mRNA expression levels were measured by RT-qPCR. (H) Simplified schematic of the RIG-I/MDA5-NF-κB pathway. (I) A549 cells were silenced with siRBM39 or siNT/siTRIF for 48 h and then transfected with 0.5 µg/mL poly(I:C) for 6 h. TNFAIP3 and IL6 mRNA expression levels were measured by RT-qPCR. The data shown are from three biological replicates (n = 3); error bars indicate SD. The statistical significance of RT-qPCR data was assessed through Welch’s unpaired t-test, and for the transcriptomics data it was calculated with the Wald test and corrected for multiple testing according to Benjamini–Hochberg. ns = not significant, * = p < 0.05, ** = p < 0.01, *** = p < 0.001, **** = p < 0.0001.

Figure 7

RBM39 impacts on IFN-JAK-STAT signaling. (A–D) DEG and DTU analysis of IFNAR2 (A), IL10RB (B), STAT1 (C), and STAT2 (D) mRNA. (E) Simplified schematic of the type I and type III IFN pathways. (F, G) PH5CH cells were transfected with siRBM39 or siNT as control for 48 h and then stimulated with IFNα2 (F) or IFNλ1 (G) for 24 h. IFIT1 mRNA was measured by qPCR. The relative expression to siNT control is shown. mRNA fold change was normalized on GAPDH. The data were obtained from three biological replicates (n = 3); error bars refer to SD. The statistical significance of RT-qPCR data was assessed through Welch’s unpaired t-test. Transcriptomics data was calculated with the Wald test and corrected for multiple testing according to Benjamini–Hochberg. ** = p < 0.01, **** = p < 0.0001.