Fig 2: DHX9 is overexpressed in breast cancer and suppresses PKR activationa Schematic showing the domain structure of PKR, ADAR1, DHX9 and other helicases identified by proximity labeling in Fig. 1, dsRBD refers to the dsRNA Binding Domain. b Pearson and Spearman correlation coefficients for the correlation between ADAR1 expression at the RNA level and the expression of each indicated helicase at the RNA level, data from breast tumors within TCGA. Scatterplots showing the correlation between ADAR1-p110 c, or ADAR1-p150 d, and DHX9 expression in normal breast or breast tumors. e Expression of DHX9 at the RNA level in normal breast, non-TNBC or TNBC tumors. f Waterfall plot showing DHX9 dependency of breast cancer cell lines using data from DepMap, ER = estrogen receptor positive cell lines, ERRB2 = HER2 positive cell lines. g Representative immunoblot following knockdown of DHX9 with two different shRNAs in four TNBC cell lines. Immunoblots for other replicates and uncropped blots can be found in Source Data Figures. h Foci formation assay for the same cells used in g for immunoblot. i Quantification of PKR phosphorylation as determined by the immunoblot in g. Quantification of protein expression for other proteins of interest can be found in Extended Data Fig. 3a–e. Bars represent the average of at least three replicates, error bars are +/− SD. * p <0.05, ** p <0.01, *** p < 0.001. P-values determined by Dunnett’s test.

Fig 3: DHX9 and ADAR1 redundantly suppress dsRNA sensing in ADAR1-independentcell lines Representative immunoblot showing the phenotype of ADAR1 and/or DHX9 knockdown in MCF-7 a, or SK-BR-3 f. Immunoblots for other replicates and uncropped blots can be found in Source Data Figures. Protein abundance from the immunoblot in a and f was normalized by total protein abundance by quantification of the Stain Free gel in b and g respectively. Fold change of PKR phosphorylation at Thr-446 in MCF-7 d or SK-BR-3 i as determined by the immunoblots in a or f respectively. Quantification of protein expression for other proteins of interest can be found in Extended Data Fig. 4a–l. Representative foci formation phenotype of ADAR1 and/or DHX9 knockdown in MCF-7 c or SK-BR-3 h, quantification of relative foci area is shown in e or j respectively. k Analysis of rRNA integrity upon knockdown of ADAR1 and/or DHX9 in MCF-7 or SK-BR-3. Additionally, panel k shows the effect of poly(I:C) (p(I:C)) transfection on rRNA integrity in MCF-7. Bars represent the average of at least three replicates, error bars are +/− SD. * p <0.05, ** p <0.01, *** p < 0.001. P-values determined by one-way ANOVA with post-hoc Tukey. Comparisons between the two different shRNAs targeting DHX9 (shDHX9–3 and shDHX9–5) were not included for clarity.

Fig 4: Rescue of PKR activation by ADAR1-p110, ADAR-p150, DHX9 and DHX9 mutantsa Schematic showing the domain structure of ADAR1 isoforms, DHX9 and mutants of DHX9, dsRBD refers to the dsRNA Binding Domain. Representative immunoblot showing the phenotype of ADAR1 and DHX9 knockdown with DHX9, DHX9K417R or dsRBD-EGFP overexpression b, or ADAR1 isoform overexpression f. Immunoblots for other replicates and uncropped blots can be found in Source Data Figures. Fold change of PKR phosphorylation at Thr-446 upon ADAR1 and DHX9 knockdown with DHX9, DHX9K417R or dsRBD-EGFP overexpression d, or ADAR1 isoform overexpression h, quantified from immunoblots in b and f respectively. Quantification of protein expression for other proteins of interest can be found in Extended Data Fig. 7b–f and 8a–f. Representative foci formation phenotype of ADAR1 and DHX9 knockdown with DHX9, DHX9K417R or dsRBD-EGFP overexpression c, or ADAR1 isoform overexpression g. Quantification of relative foci area is shown in e or i, respectively. Bars represent the average of at least three replicates, error bars are +/− SD. * p <0.05, ** p <0.01, *** p < 0.001. P-values determined by Dunnett’s test.

Fig 5: Validation of putative protein-protein interactions identified by proximity labelingImmunoprecipitation of DHX9 a-c, DDX54 d-e or DDX17 f-g followed by immunoblot for ADAR in breast cancer cell lines. Immunoblot of immunoprecipitation eluates and inputs from SK-BR-3 a, d and f, MCF-7 b, e and g, and MDA-MB-231 c. Input represents 5% of the lysate used for immunoprecipitation. The IgG lanes represent immunoprecipitation eluates from pulldown with anti-rabbit IgG antibody. The lanes labeled DHX9, DDX54 and DDX17 indicate the eluates from immunoprecipitation with antibodies against those proteins respectively. The IgGHC label indicates the band corresponding to the IgG heavy chain from the antibody used for immunoprecipitation. Uncropped immunoblots for panels a-g can be found in Source Data Figures. Immunofluorescence for ADAR1 and DHX9 h, DDX54 i, or DDX17 j in SKBR3 or MCF-7.