PACT prevents aberrant activation of PKR by endogenous dsRNA without sequestration

Abstract

The innate immune sensor PKR for double-stranded RNA (dsRNA) is critical for antiviral defense, but its aberrant activation by cellular dsRNA is linked to various diseases. The dsRNA-binding protein PACT plays a critical yet controversial role in this pathway. We show that PACT directly suppresses PKR activation by endogenous dsRNA ligands, such as inverted-repeat Alu RNAs, which robustly activate PKR in the absence of PACT. Instead of competing for dsRNA binding, PACT prevents PKR from scanning along dsRNA-a necessary step for PKR molecules to encounter and phosphorylate each other for activation. While PKR favors longer dsRNA for increased co-occupancy and scanning-mediated activation, longer dsRNA is also more susceptible to PACT-mediated regulation due to increased PACT-PKR co-occupancy. Unlike viral inhibitors that constitutively suppress PKR, this RNA-dependent mechanism allows PACT to fine-tune PKR activation based on dsRNA length and quantity, ensuring self-tolerance without sequestering most cellular dsRNA.

Fig. 1. PACT restricts aberrant activation of PKR, thereby maintaining cellular homeostasis.

A ATP bioluminescence assay showing cell viability in PACT KO-sensitive (HCC1806, MDA-MB-157, NCI-H727, NCI-H2286, SW1271) and -insensitive (ZR-75-1, HCC1428, NCI-H1437, NCI-H2023, NCI-H1650) cell lines after control gene-KO (ctrl) or PACT-KO using 2 different guide RNAs. The control (ctrl) genes are AAVS1 (sg1) and Chr2.2 (sg2). Values represent means ( ± SD) of 3 biological repeats. P values were based on one-way ANOVA test. B Crystal violet staining assay showing cell viability in 3 PACT KO-sensitive cells (HCC1806, NCI-H727, and NCI-H2286) after control gene-KO vs. PACT-KO. C Crystal violet staining assay with HCC1806 cells after control gene-KO and PACT-KO complemented with increasing amounts of wild-type PACT. PACT used here was engineered to be resistant to PACT sgRNA 1 targeting. D Western blot analysis showing levels of PKR and eIF2α phosphorylation in PACT KO-sensitive (top panel) and -insensitive (bottom panel) cells after control gene-KO or PACT-KO. Vinculin was used as a loading control. The experiment was repeated 3 times independently with similar results. E ATP bioluminescence assay in HCC1806 cells in PKR-sufficient and PKR-KO cells in control gene-KO and PACT-KO backgrounds. Values represent means ( ± SD) of 3 biological repeats. P values were based on two-way ANOVA test. F A representative crystal violet staining for samples in E. G ATP bioluminescence assay in HCC1806 cells in control gene-KO and PACT-KO treated with indicated concentrations of ISRIB. Values represent means ( ± SD) from 4 technical replicates across two independent experiments (n = 2). P values were based on two-tailed student’s t-test. ****p = <0.0001, ***p = <0.001, **p = <0.01, *p = <0.05, not significant (ns) is for p > 0.05. Source data and exact P values are provided as a Source Data file.

Fig. 2. PACT and PKR share endogenous dsRNA ligands, but PACT inhibits PKR without blocking PKR’s dsRNA binding.

A Gel-shift assay with 112 bp dsRNA binding to increasing concentrations (0–4000 nM) of PACT wild-type or 4KE (K84E + K85E + K177E + K178E) mutant. Right: specific binding curve using Hill equation assuming 8 binding sites (see “Methods”). Wild-type plot represents mean ± SD from 3 experiments. B Crystal violet assay showing cell viability (HCC1806) after control/PACT-KO, complemented with increasing expression of PACT wild-type or 4KE. Bottom: WB analysis. The experiment was repeated 3 times. C Volcano plots showing RNA peaks enriched by PKR (left) and PACT (right) fCLIP in HCC1806 cells. Peak identification criteria: (Log₂FC (HA-fCLIP/input) ≥1 with HA-PKR & ≤0 without HA-PKR) and (Log₂FC (FLAG-fCLIP/input) ≥1 with FLAG-PACT & ≤0 without FLAG-PACT). See also Supplementary Fig. 3C, F. Mean from 3 (PKR) and 2 (PACT) replicates. P values were calculated in Deseq2 (two-sided) and adjusted by Benjamini-Hochberg method. D Snapshot of an IR-Alu sequence enriched in PKR (red) and PACT (blue) fCLIP-seq. E Venn diagrams showing overlap of total (left) and IR-Alu/near IR-Alu peaks (right) enriched in PKR and PACT fCLIP. Near IR-Alu denote peaks flanked by or adjacent to IR-Alus (<300 bp). F HA-PKR fCLIP intensity before (x-axis) and after PACT KO (y-axis). Plot shows the average fold change of area under the curve (AuC) for peaks from (C). RNA levels (AuC) were normalized to total non-rRNA reads, and fCLIP/input ratio calculated for normalized AuC. See Supplementary Fig. 4A and Methods for additional details. Mean from 2 biological repeats. Correlation coefficient (R) = 0.71. G fCLIP-RT-qPCR validation of 4 peaks from (F) using endogenous PKR. RNA levels were normalized to 18S rRNA before calculating fCLIP/input ratio. Mean ( ± SD) from 3 replicates. P values from a two-tailed t-test. (ns, not significant; P > 0.05). H A-to-I editing levels in PKR fCLIP and input samples (HCC1806) before and after PACT KO, measured as A-to-G mismatch counts per million reads (CPM). Mean ( ± SD) of 4 biological repeats. P values from one-way ANOVA followed by Tukey multiple comparisons test. ****p < 0.0001; (ns) p > 0.12. Source data and exact P values are provided as a Source Data file.

Fig. 3. PKR scans along dsRNA, resulting in dsRNA length-dependent activity.

A In vitro PKR kinase assay monitoring PKR autophosphorylation with increasing concentrations (0.125, 0.25, 2.5 ng/μl) of 112 bp dsRNA with 4 different sequences (see supplementary Table 3) and GC contents. [³⁵S]-IRF3 served as loading control (*). Wild-type full-length PKR was used unless stated otherwise. Results were reproduced in 2 replicates. B PKR kinase assay with increasing dsRNA concentrations (0.0625–15 ng/μl) for lengths 112, 62, 42, and 21 bp. Results reproduced in 3 replicates. C, D Quantification of PKR activity from A displayed as heatmap and x-y graph. Data represent mean ( ± SD) of 3 biological repeats. E PKR activity with 112 bp dsRNA, perfect duplex Alu dsRNA, and IR-Alus from NICN1 and PSMB2 3’-UTRs (see Supplementary Fig. 5D) at 0.0625, 0.25, and 1 ng/μl. Values are mean ( ± SD) of 3 biological replicates. F Schematic of single-molecule FRET experiments to monitor scanning motions of PKR on dsRNA. Cy5-labeled PKR (acceptor, 50 nM) was added to a chamber containing Cy3-labeled dsRNA (donor) immobilized on the surface. Cy3 and Cy5 fluorescence were monitored upon Cy3 excitation using sm-TIRF microscopy. Time course traces of Cy3/Cy5 intensities and FRET (right panel) and zoomed-in trace for “On” state (lower panel). PKR K296R was used to ensure analysis of PKR movement before its activation. *indicates transient or abortive binding––such events were excluded from the autocorrelation analysis in G. See Supplementary Fig. 6 for additional traces and description of identifying “On” events. Created in BioRender. torres, c. (2025) https://BioRender.com/g96x368. G Autocorrelation analysis of FRET fluctuations for PKR in the “On” state on 112, 59, 42 bp dsRNAs and “Off” state on 112 bp dsRNA. See Methods for the autocorrelation function. (H) Diffusion rates (mean ± SD) calculated from exponential fitting of the autocorrelation curves of PKR (see Methods) with different dsRNA lengths: 0.130 ± 0.00657 s (42 bp), 0.149 ± 0.00657 s (59 bp) and 0.230 ± 0.0101 s (112 bp); n = 143 (42 bp), 69 (59 bp) and 236 (112 bp) diffusion events. Source data is provided as a Source Data file.

Fig. 4. PACT inhibits PKR by restricting its motion on dsRNA, leading to length-dependent inhibition.

A In vitro PKR kinase assay as in Fig. 3B, with increasing PACT concentrations (0–250 nM). [³⁵S]-IRF3 served as a loading control (*). Results were reproduced in 3 replicates. B Heatmap of relative PKR activity from gels in A, representing means of 3 biological repeats. C Relative PKR activity with 0.25 ng/μl (left) and 5 ng/μl (right) of 112, 62, and 42 bp dsRNA, with increasing PACT concentrations (0–250 nM). Mean ( ± SD) of 3 biological repeats. D Schematic of single-molecule FRET experiments monitoring PKR diffusion on 112 bp dsRNA in the presence of PACT. Cy5-labeled PKR (acceptor, 50 nM) was mixed with unlabeled PACT (50 nM) and added to a chamber with immobilized Cy3-labeled dsRNA (donor). Cy3 and Cy5 fluorescence and Cy3-Cy5 FRET were monitored upon Cy3 excitation using Single-molecule TIRF (sm-TIRF) microscopy. PKR K296R was used to ensure analysis of PKR movement before its activation. Created in BioRender. torres, c. (2025) https://BioRender.com/g96x368. E FRET histograms for PKR without PACT (blue) and with PACT (purple). To compare the range of PKR motions with and without PACT, we focused our analysis on traces showing a high FRET state ( > 0.8). Without PACT, PKR exhibits diffusion across a broad FRET range (0.2–0.8), while its movement is more restricted with PACT, showing FRET values between 0.6 and 0.8. See “Methods” for details on histogram quantification. F Autocorrelation analysis of FRET signal for PKR without PACT (blue) and with PACT (purple). Scanning times were calculated by single exponential curve fitting (see “Methods”). 236 and 93 diffusion events were analyzed for PKR without and with PACT, respectively. G Residence time of PKR on dsRNA (“on” time) in the presence and absence of PACT (50 nM). Average values (in seconds) are shown on the right. 236 and 117 diffusion events were analyzed for PKR without and with PACT, respectively. Data represent mean values ± SD. P values were calculated by a paired two-tailed t-test. (*P < 0.05). Source data and exact P values are provided as a Source Data file.

Fig. 5. PACT’s inhibition of PKR involves weak but direct and specific protein-protein interactions.

A In vitro PKR kinase assay using 112 bp dsRNA (0.25 ng/μl) in the presence of 0, 20, 50, 100, 250 nM PACT, TRBP, RIG-I ∆CARDs, ADAR1 DRBDs. Right: SDS-PAGE showing purity of the proteins. The results were reproduced in 3 independent replicates. B dsRNA binding curve of PACT, TRBP, RIG-I ∆CARDs, ADAR1 DRBDs, derived from the native gel-shift assay in Supplementary Fig. 10A monitoring 112 bp dsRNA binding. For curve fitting, Hill equation was used with the assumption of 8 protein binding sites per 112 bp dsRNA (see Methods). C Pulldown of PACT using PKR-GST or GST. All proteins were purified and treated with benzonase to remove potential nucleic acid contaminants. Catalytically dead K296R PKR was used. The results were reproduced in 3 independent replicates. D In vitro PKR kinase assay using an increasing concentration of PKR in the absence of dsRNA. Benzonase was added during the reaction (in addition to during purification) to ensure that the observed PKR activity is RNA-independent. The results were reproduced in 2 independent replicates. E RNA-independent PKR kinase assay with 1 μM PKR in the presence of 0, 1.25, 2.5, 5 μM PACT, GST, MBP. The results were reproduced in 3 independent replicates. F RNA-independent PKR kinase assay with 1 μM PKR in the presence of 0, 1.25, 2.5, 5 μM PACT wild type, ∆D1, ∆D2; and 0, 2.5, 5 μM ∆D3. G Correlation between the levels of PKR inhibition and PACT phosphorylation from F. Values are mean ( ± SD) of 2 biological replicates. H Pulldown of PKR or PKR∆DRBDs using PACT D1D2-GST or GST. Catalytically dead K296R mutant was used for both PKR and PKR∆DRBDs. The results were reproduced in 3 independent replicates. I RNA-independent PKR kinase assay with 1 μM PKR or 2 μM PKR∆DRBDs in the presence of 0, 1.25, 2.5, 5 μM PACT. Higher concentration of PKR∆DRBDs was due to its lower activity. The results were reproduced in 3 independent replicates. Source data are provided as a Source Data file.

Fig. 6. PACT DRBD2 utilizes the surface distinct from the dsRNA interface to inhibit PKR.

A Structures of DRBDs in complex with dsRNA or protein partners, viewed from the same perspective relative to the DRBDs. Structure of PACT DRBD3 dimer is from this study. All other structures were previously published. The comparison shows the conserved protein:protein interface involving β3 away from the dsRNA binding surface. Inset: NMR structure of PACT DRBD3 showing its homo-dimerization by joining their β-sheets (β1- β2- β3) through a parallel β3: β3 interaction. B Sites of mutations on the putative protein:protein interface of PACT DRBD1 and DRBD2. Bottom: dsRNA binding curves for PACT wild type, DRBD1 mutants T76R, C77R, and DRBD2 mutants K173E, K187E, K191E, derived from EMSA in Supplementary Fig. 10E. C RNA-independent PKR kinase assay with 1 μM PKR in the presence of 5 μM PACT wild type, T76R, C77R, K173E, K187E, K191E. Right: quantification from 5 independent experiments. Data are presented as mean values ± SD. P-values were calculated by one-way ANOVA followed by Tukey multiple comparisons test. **p ≤ 0.0021, *p ≤ 0.0332; (ns) p > 0.12. P values: (buffer vs. wild type = 0.0019, wild type vs. K173E = 0.0034, wild type vs. K187E = 0.0093, wild type vs. K191E = 0.0231). D Correlation between the levels of PKR inhibition and PACT phosphorylation from C. Values are mean ( ± SD) of 5 biological repeats. E Crystal violet staining assay showing cell viability in NCI-H727 cells after control gene-KO (ctrl) and PACT-KO complemented with expression of PACT wild type or increasing amounts of C77R, K173E, K187E, and K191E. Source data are provided as a Source Data file.

Fig. 7. Model for PKR activation and PACT-mediated inhibition.

A Heatmaps showing dsRNA length and concentration dependance for PKR activity in the absence and presence of PACT. PACT restricts PKR selectively within a specific range of the dsRNA length-concentration space, resulting in the selective elevation of PKR’s activity threshold for longer dsRNAs (such as IR-Alus). Notably, PACT has minimal impact on PKR when stimulated with high concentrations of long dsRNA, suggesting that PACT acts as a homeostatic regulator without compromising PKR’s inherent activity. B Model of PKR activation by dsRNA and its inhibition by PACT. PKR activation involves dsRNA scanning, molecular collisions, and subsequent trans-autophosphorylation on long dsRNA. PACT also scans along dsRNA, restricting PKR’s movement, preventing molecular collision between PKR molecules. This inhibition requires more than PACT’s dsRNA-binding activity; it also involves a direct, albeit weak, interaction with PKR’s kinase domain, facilitated when both PKR and PACT are on the same dsRNA. As a result, PACT’s inhibition of PKR is more robust when dsRNA is longer and less abundant––conditions that promote co-occupancy––explaining how PACT modulates its inhibitory function based on dsRNA length and concentration. Created in BioRender. torres, c. (2025) https://BioRender.com/g96x368.