Our prior studies showed that DUX4 inhibited ISG induction in response to lentiviral infection and suppressed induction of MHC class I proteins in response to IFNγ (type II interferon) (Chew et al., 2019; Geng et al., 2012). To determine whether DUX4 broadly inhibited ISG induction by IFNγ, we used the MB135-iDUX4 cell line, a human skeletal muscle cell line with an integrated doxycycline-inducible DUX4 (iDUX4) transgene (Jagannathan et al., 2016; see Figure 1—figure supplement 1 for schematics and sequences of the transgenes used in this study). Doxycycline induction of DUX4 expression in the MB135-iDUX4 cell line has been validated as an accurate cell model of the transcriptional consequences of DUX4 expression in FSHD muscle cells (Jagannathan et al., 2016) and in the early embryo (Hendrickson et al., 2017; Whiddon et al., 2017). Using a stringent eightfold induction cut-off (log2 fold-change >3), RNA-seq showed that IFNγ treatment induced 113 genes, whereas the expression of DUX4 suppressed ISG induction by IFNγ more than fourfold for 76 (67%) of these genes and more than twofold for 102 genes (90%) (Supplementary file 1).

Informed by the RNA-seq results, we used RT-qPCR to measure the response of four ISGs that represent different components of the response to immune signaling: the RNA helicase IFIH1; the interferon-stimulated exonuclease ISG20; the chemoattractant CXCL9; and the MHC II chaperone CD74. IFNγ induction of all four genes was robustly blocked by DUX4 expression while a DUX4-target gene ZSCAN4 was strongly induced, indicating that the ISG suppression did not represent a universal block to gene induction (Figure 1A, MB135-iDUX4, and Figure 1—figure supplement 2A [for this and subsequent constructs, Figure 1—figure supplement 2 shows RT-qPCR data from additional independent cell lines together with protein expression and nuclear localization]), whereas doxycycline treatment in the absence of iDUX4 did not suppress induction of the ISG panel (Figure 1A, MB135 parental). In contrast to DUX4, a paralog in the DUX family, DUXB, did not suppress induction of the ISG panel by IFNγ (Figure 1A, MB135-iDUXB).

To determine whether DUX4 also inhibits ISG induction by other innate immune signaling pathways, we transfected the MB135-iDUX4 cells with three different innate immune stimuli: poly(I:C), a long dsRNA mimic to stimulate IFIH1 (MDA5); RIG-I ligand, a short 5’ppp-dsRNA to stimulate DDX58 (RIG-I); or cGAMP, a signaling component of the cGAS dsDNA sensing pathway. Additionally, we stimulated the cells with interferon-beta (IFNβ, type I interferon), which primarily signals through JAK-STAT pathways via a STAT1-STAT2-IRF9 complex, as opposed to the STAT1 homodimers induced by IFNγ. For all signaling pathways, DUX4 suppressed the induction of a subset of the panel of ISG genes induced by each ligand (Figure 1B). One exception, CXCL9 was induced by IFNβ, poly(I:C), and the RIG-I ligand but not suppressed by DUX4. cGAMP did not induce CXCL9 or CD74, precluding evaluation of the role of DUX4 in regulating these ISGs. These results indicate that DUX4 can modulate the activity of multiple signaling pathways. However, because these pathways converge on common nodes, such as the induction of interferon, additional studies are needed to determine whether DUX4 inhibits unique components in each pathway or a common component responsible for ISG upregulation across pathways. We decided to focus further efforts on identifying the mechanism behind the suppression of IFNγ-mediated transcription as this pathway was most broadly suppressed by DUX4.

There are two conserved regions of the DUX4 protein, the N-terminal homeodomains (aa19-78, aa94-153) and an ~50 amino acid region at the end of the C-terminal domain (CTD) that is required for transcriptional activation by DUX4 (aa371-424) (Choi et al., 2016; Geng et al., 2011; Leidenroth and Hewitt, 2010). A mutation in the first homeodomain, F67A, significantly diminishes DUX4 DNA binding and target gene activation (Wallace et al., 2011). When expressed in MB135 cells, iDUX4-F67A minimally activated the DUX4 target gene ZSCAN4, yet still suppressed ISG induction by IFNγ (Figure 2A and B and Figure 1—figure supplement 2B). A second construct, iDUX4aa154-424, has the N-terminal homeodomain region replaced by a cassette containing a 3x FLAG tag and two nuclear localization signals (3xFLAG-NLS). The iDUX4aa154-424, hereafter called iDUX4-CTD, was completely transcriptionally silent yet equally suppressed activation of ISGs (Figure 2A and C and Figure 1—figure supplement 2C). RNA sequencing analysis using the same criteria to characterize ISG suppression by the full-length DUX4 demonstrated that the F67A mutant suppressed 70% of induced genes by more than twofold, or 41% of induced genes by more than fourfold, whereas the iDUX4-CTD showed 90 or 52% suppression, respectively (Supplementary file 1). Together, these data indicate that DUX4 transcriptional activity is not necessary to suppress IFNγ-mediated gene induction.

The DUX4-CTD contains a pair of (L)LxxL(L) motifs, LLDELL and LLEEL, that are conserved across the DUXC/DUX4 family (Leidenroth and Hewitt, 2010). DUX4 transgenes with mutations in the first motif, deletion of the second motif, or both (iDUX4mL1, iDUX4dL2, iDUX4mL1dL2) (see Figure 1—figure supplement 1B for sequences of these mutants) failed to activate the DUX4 target ZSCAN4 (Figure 2A). iDUX4ml1dl2 and iDUX4dl2 both lost the ability to suppress the panel of ISGs, whereas iDUX4mL1 showed partial activity, suppressing three of the four ISGs (Figure 2D and Figure 1—figure supplement 2D), indicating that these (L)LxxL(L) motifs are necessary for both ISG suppression and for transcriptional activation by DUX4.

To test sufficiency, we generated two additional C-terminal fragments of DUX4 (Figure 2C). The first, iDUX4-CTDmL1dL2, contains the CTD of iDUX4mL1dL2 with its N-terminal HDs replaced with the 3xFLAG-NLS cassette. Similar to iDUX4mL1dL2, iDUX4-CTDmL1dL2 did not block the panel of ISGs (Figure 2C and Figure 1—figure supplement 2E). The second construct, iDUX4aa339-424, contains only the C-terminal 85 aa residues including both (L)LxxL(L) motifs, and maintained ISG suppression, though not as strongly on the IFIH1 and ISG20 genes (Figure 2C and Figure 1—figure supplement 1F). In summary, these data support a model in which the DUX4-CTD is both necessary and sufficient to suppress a major portion of the ISG response to IFNγ.

As an unbiased method to identify proteins that interact with the C-terminal region of DUX4, we conducted two experiments using liquid chromatography mass spectroscopy (LC-MS) to identify proteins that co-immunoprecipitated with DUX4-CTD constructs expressed in MB135 myoblasts. We used the DUX4-CTD because the prior experiments showed that it contained the regions necessary and sufficient to suppress ISG induction. In the first experiment, we used MB135iDUX4-CTD cells either untreated, treated with doxycycline alone, or with both doxycycline and IFNγ. In the second experiment, we used MB135iDUX4-CTD and MB135iDUX4mL1dL2 cells both treated with doxycycline and IFNγ compared to these two cell lines untreated and combined as a control. Proteins with a minimum of two peptide spectrum matches (PSMs) in at least one sample that were identified in both experiments were assigned to 1 of 10 categories (see ‘Materials and methods’) to separate candidate interactors from other categories that might be co-purified because of obligate interactions (e.g., proteasome or ribosome) or might be less likely to be relevant to immune responses (e.g., cytoskeletal proteins). Candidate interactors were then ranked based on the total PSMs for that protein across all samples. (It is important to note that the ‘bait’ constructs were expressed at low levels in the samples not treated with doxycycline and that the immunoprecipitation concentrated this background, which might account for some of the candidate proteins appearing in the untreated samples.) STAT1 and DDX3X, two key regulators of innate immune signaling, ranked at the top of the list of candidate DUX4-CTD interactors, together with several other proteins implicated in modulating innate immune signaling (Figure 3, left panel, and Supplementary file 2). Western blot analysis using independent biological samples from a co-IP experiment with MB135-iDUX4-CTD and MB135-iDUXB (as a control) validated the DUX4-CTD interactions with DDX3X, STAT1, PRKDC, YBX1, HNRNPM, PABPC1, NCL, CDK4, and HNRPU (Figure 3, right panel).

Because of its central role in IFNγ signaling, we elected to focus on the interaction of STAT1 with DUX4. To map the region(s) of the DUX4-CTD necessary to interact with STAT1, we expressed a truncation series in MB135 cells (all with an N-terminal 3xFLAG tag and NLS and all treated with IFNγ): iDUX4-CTD (aa154-424), iDUX4aa154-372, iDUX4aa154-308, and iDUX4aa154-271. The region of DUX4 between amino acids 271 and 372 was necessary for co-IP of STAT1, whereas the region between 372 and 424 containing the (L)LxxL(L) motifs might enhance DUX4-CTD binding to the phosphorylated forms of STAT1 (Figure 4A).

To determine whether phosphorylation of STAT1 enhanced interaction with DUX4, we co-expressed the FLAG-tagged iDUX4-CTD with an MYC-tagged iSTAT1 or STAT1 mutants Y701A or S727A, wherein doxycycline would induce expression of both the DUX4 and STAT1 transgenes, and performed an αFLAG co-IP to look for STAT1 interaction. The αMYC signal of the IFNγ-treated samples suggests that our iSTAT1, iSTAT1-Y701A, and iSTAT1-S727A transgenes are expressed at similar levels, and yet the wild-type STAT1 and STAT1-S727A showed enhanced binding to the CTD with IFNγ treatment while the STAT1-Y701A did not (Figure 4B). Furthermore, immunofluorescence showed that DUX4-CTD expression, which was highly restricted to the nucleus, did not alter the localization of STAT1 in either untreated cells (low levels throughout the cell) or IFNγ-treated cells (high levels in the nucleus) (Figure 4—figure supplement 1A and B). Similarly, there was no detectable pSTAT1-Y701 present in the nuclei of untreated cells, but there was a strong pSTAT1-Y701 signal in IFNγ-treated nuclei (Figure 4—figure supplement 1C). Furthermore, the distribution of total STAT1 in the immortalized MB135iDUX4-CTD cells was similar to that in primary human fibroblasts and both immortalized and primary MB135 myoblasts, indicating that the immortalization did not alter the distribution of total STAT1 (Figure 4—figure supplement 1D). Proximity ligation assay (PLA) indicated close interaction between the iDUX4-CTD and endogenous pSTAT1-Y701 in the nuclei of MB135 cells treated with doxycycline and IFNγ (Figure 4C), and PLA similarly showed an interaction of total STAT1 and DUX4-CTD in primary human fibroblasts (Figure 4—figure supplement 2). Therefore, the interaction between DUX4-CTD and STAT1 is enhanced by phosphorylation of STAT1-Y701, and we can observe this interaction within the nuclei of DUX4-CTD-expressing cells.

Chromatin immunoprecipitation (ChIP) was performed on MB135-iDUX4-CTD cells to assess STAT1 binding to ISG promoters. Compared to a gene-desert region where there should not be STAT1 binding (h16q21), there was a robust induction of STAT1 binding following IFNγ treatment at the promoters of several ISGs (GBP1, IDO1, CXCL10) with previously characterized STAT1 binding sites (Rosowski et al., 2014; Figure 5A, left four panels). Treatment with IFNγ following induction of DUX4-CTD diminished STAT1 occupancy at all three ISGs, and paired RT-qPCR confirmed that the DUX4-CTD robustly suppressed the RNA induction by IFNγ (Figure 5A, right panel). We used CUT&Tag (Cleavage Under Target & Tagmentation) (Kaya-Okur et al., 2019) to assess Pol-II occupancy genome wide and found that DUX4-CTD blocked Pol-II recruitment to ISGs without affecting occupancy at other genes (Figure 5B).

The majority of EWSR1 fusion-negative small blue round cell sarcomas have a genetic re-arrangement between CIC and DUX4 that creates a fusion protein containing the carboxyterminal (L)LxxL(L) motif region of DUX4 (Graham et al., 2012; Kawamura-Saito et al., 2006). We confirmed that the Kitra-SRS sarcoma cell line expresses a CIC-DUX4 fusion mRNA containing the terminal 98 amino acids of DUX4 as previously described (Nakai et al., 2019). Compared to MB135 myoblasts, Kitra-SRS cells showed absent-to-low induction of ISGs when treated with IFNγ and control siRNAs. In contrast, siRNA knockdown of the CIC-DUX4 fusion in the KitraSRS cells resulted in a substantially increased IFNγ induction of ISGs, whereas knockdown of CIC in the MB135 cells did not alter ISG induction (Figure 6A). To confirm that the CIC-DUX4 fusion was suppressing ISG induction, we expressed a doxycycline-inducible CIC or the Kitra-SRS CIC-DUX4 fusion protein in MB135 cells and showed that the CIC-DUX4 fusion, but not CIC, suppressed IFNγ induction of ISGs IFIH1, CXCL9, and CD74, although not ISG20 (Figure 6B). Furthermore, PLAs were consistent with an interaction of both total STAT1 and phosphorylated STAT1-Y701 with the CIC-DUX4 fusion in the nuclei of Kitra-SRS cells (Figure 6C).

As DUX4 expression has been shown to dampen MHC I activation in multiple cancer lines (Chew et al., 2019), we decided to test the effect of the CIC-DUX4 fusion on MHC I expression in Kitra-SRS cells via flow cytometry. We again treated Kitra-SRS cells and MB135 parental myoblasts with either control siRNAs (siCTRL) or siRNAs targeting CIC and DUX4 (siCIC-DUX4), with or without IFNγ stimulation. We found that knockdown of the endogenous CIC in MB135 myoblasts had no effect on MHC I response to IFNγ (54.9% in siCTRL+ IFNγ compared to 60.7% in siCIC-DUX4+ IFNγ, Figure 6—figure supplement 1A). In contrast, knockdown of the CIC-DUX4 fusion protein in Kitra-SRS cells almost doubled the number of MHC I-positive cells with IFNγ treatment (48.1%) compared to cells treated with siCTRL and IFNγ (27.9%, Figure 6—figure supplement 1B).

DUX4 expression in cultured FSHD muscle cells is often described as low; however, this is due to the high heterogeneity caused by strong expression in a small population of cells (Rickard et al., 2015; Snider et al., 2010). In cultured FSHD myotubes, approximately 5% of the myotubes might express DUX4 in their nuclei. To determine whether endogenous DUX4 suppresses IFNγ signaling, we assessed induction of IDO1 by IFNγ in FSHD myotubes. Differentiation of FSHD myoblasts into multinucleated myotubes results in distinct populations of DUX4-expressing and DUX4-negative myotubes in the same culture, allowing for side-by-side evaluation of DUX4-positive and DUX4-negative muscle cells in the same culture. We determined the IFNγ induction of IDO1 as a representative ISG based on its low basal expression in skeletal muscle and our prior demonstration that it is suppressed in the MB135-iDUX4-CTD cells (see Figure 5A, right panel). Treatment with IFNγ produced a reliable IDO1 signal within the nucleus and cytoplasm of individual myotubes that did not express DUX4, whereas DUX4-positive myotubes did not show IDO1 expression in response to IFNγ; we quantified these differences by measuring the mean fluorescent intensity (MFI) of αIDO1 in DUX4+ versus DUX4- nuclei and the MFI of αDUX4 in IDO+ versus IDO1- nuclei and found significant differences by unpaired t-test (Figure 7A). Therefore, similar to our MB135-iDUX4 studies, endogenous DUX4 expressed at a physiological level is sufficient to prevent ISG induction by IFNγ.

Dux, the mouse ortholog of human DUX4, is expressed at the equivalent developmental stage to human DUX4 (Hendrickson et al., 2017), activates a parallel transcriptional program (Whiddon et al., 2017), and contains the (L)LxxL(L) motifs that we have shown to be necessary for ISG repression by human DUX4. In fact, the mouse Dux sequence contains a 60 amino acid triplication of the (L)LxxL(L)-containing region (Figure 7—figure supplement 1). Accordingly, we introduced a doxycycline-inducible mouse Dux transgene into human MB135 cells (MB135-iDux) and found that the full-length Dux protein repressed the panel of ISGs even more robustly than the full-length or CTD portion of human DUX4 (Figure 7B, left). Similar to human DUX4, Western analysis confirmed the co-immunoprecipitation of STAT1 and both phosphorylated pSTAT1-Y701 and pSTAT1-S727 with mouse Dux (Figure 7B, right). These data demonstrate that the suppression of ISG induction and interaction with phosphorylated STAT1 is conserved in the DUXC family.

Cell types used: primary human fibroblasts, primary human myoblasts, and immortalized human myoblasts. Myoblasts were obtained from the Fields Center for FSHD Research at the University of Rochester Medical Center. Fibroblasts were obtained from the Fred Hutchinson Cancer Center lab of Dr. D. Miller. Cells were used directly from source. Myoblast identity confirmed by muscle gene expression. Periodic mycoplasma testing did not identify mycoplasma contamination.

All myoblast experiments were conducted in immortalized MB135 (Homo sapiens, female, control, Fields Center for FSHD and Neuromuscular Research at the University of Rochester Medical Center), primary MB135 myoblasts (‘MB135 1°,’ H. sapiens, female, control, Fields Center for FSHD and Neuromuscular Research at the University of Rochester Medical Center), or MB200 (H. sapiens, male, FSHD2 subject, Fields Center for FSHD and Neuromuscular Research at the University of Rochester Medical Center) cell lines, respectively, cultured in Ham’s F-10 Nutrient Mix (Gibco) supplemented with 15% fetal bovine serum (Hyclone Cat# SH3007103), 100 U/100 µg/ml penicillin/streptomycin (Gibco Cat# 15-140-122), 1 µM dexamethasone (Sigma Cat# D4902), and 10 ng/ml recombinant human basic fibroblast growth factor (PeproTech Cat# G5071). To differentiate the myoblasts to myotubes, media were changed to DMEM supplemented with 10 ug/ml insulin (Sigma Cat# I1882) and 10 ug/ml transferrin (Sigma Cat# T-0665). Cell lines containing doxycycline-inducible transgenes were additionally cultured with 2 µg/ml puromycin (Sigma Cat# P833). Transgenes were induced with 1 µg/ml of doxycycline (Sigma Cat# D9891) for 4 hr prior to other treatments for a total of 20 hr. The Kitra-SRS cells (RRID:CVCL_YI69) were provided by Dr. H. Otani and Osaka University (Nakai et al., 2019) and were cultured in DMEM supplemented with 10% fetal bovine serum (Hyclone Cat# SH3007103) and 100 U/100 µg/ml penicillin/streptomycin (Gibco Cat# 15-140-122). Biological replicates consisted of independent but parallel experiments, such as simultaneously stimulating three cell culture plates with IFNγ. Technical replicates consisted of repeat measurements of the same biological sample, such as loading the same biological sample in triplicate for analysis by RT-qPCR.

Human DUX4 and mouse Dux truncation constructs were created by cloning synthesized, codon-optimized gBlock fragments into the pCW57.1 vector (Addgene plasmid #41393) downstream of the doxycycline-inducible promoter (replacing the GFP expression gene), or the pRRLSIN vector (Addgene plasmid #12252) downstream of the constitutive hPGK promoter. Lentiviral particles were created by transfecting 293T cells with a subcloned expression vector, the psPAX2 packaging vector (Addgene plasmid #12260), and the pMD2.G envelope vector (Addgene plasmid #12259) using Lipofectamine 2000 according to the manufacturer’s instructions (Invitrogen Cat# 11668019). Experimental cell lines were transduced and, when appropriate, selected using 2 µg/ml puromycin at low-enough confluence to allow for isolation of clonal lines using cloning cylinders. Transgenic clonal lines were validated for protein size, expression level, and localization by western blot and immunofluorescence.

Myoblasts were transfected with either (final concentrations) 10 µM 2’,3’-cGAMP (Invivogen, Cat# tlrl-nacga23), 2 µg/ml poly(I:C) (Sigma, Cat# P1530), or 1 µg/ml 3’ppp-dsRNA RIG-I ligand (Dan Stetson Lab, UW) using Lipofectamine 2000 (Thermo Fisher, Cat# 11668019) according to the manufacturer’s protocol or were stimulated with 1000U IFNβ (R&D Systems, Cat# 8499-IF-010-CF) or 200 ng/ml IFNγ (R&D Systems, Cat# 285-IF-100-CF) by addition directly to cell culture medium. After 16 hr of immune stimulation, RNA was collected from cells using the NucleoSpin RNA Kit (Macherey-Nagel, Cat# 740955) according to the manufacturer’s instructions. RNA samples were quantified by NanoDrop and 1 µg of RNA per sample was treated with DNase I Amplification Grade (Thermo Fisher, Cat# 18068015), and then synthesized into cDNA using the Superscript IV First-Strand Synthesis System (Thermo Fisher 18091050), including oligo dT primers (Invitrogen, Cat# 18418012). qPCR was run in 384-well plates on an Applied Biosystems QuantStudio 6 Flex Real-Time PCR System (ABI) and analyzed in Microsoft Excel.

RNA was extracted as described above from untreated, doxycycline-treated, IFNγ-treated, or doxycycline- and IFNγ-treated samples. RNA was submitted to the Fred Hutchinson Cancer Research Center Genomics Core for library preparation using the TruSeq3 Stranded mRNA kit (Illumina, Cat# RS-122-2001) followed by size and quality analysis by Tapestation (Agilent). Libraries were sequenced on a NextSeq P2-100 (Illumina).

Sequencing analysis was performed using R version 4.0.3 (R Development Core Team, 2020). Sequencing reads were trimmed using Trimmomatic (version 0.39) (Bolger et al., 2014), and aligned to the H. sapiens GRCh38 reference genome with the Rsubread aligner (Liao et al., 2019). Gene counts were analyzed using featureCounts (v2.0.1) (Liao et al., 2019) and the Gencode v35 annotation file. Normalization and differential expression analysis were done with DESeq2 (v1.26.0) (Love et al., 2014).

Cells were fixed for 10 min with 2% paraformaldehyde (Thermo Scientific) for DUX4/STAT1 and 4% paraformaldehyde for DUX4/IDO1, then permeabilized for 10 min with 0.5% Triton X-100 (Sigma), both at room temperature with gentle shaking. Cells were then blocked for 2 hr with PBS/0.3 M glycine/3% BSA at room temperature with gentle shaking. Primary antibodies were incubated at 4°C overnight at the following concentrations: rabbit anti-IDO1 [D5J4E] 1:100 (Cell Signaling Technology, 86630S, RRID:AB2636818), mouse anti-DUX4 [P2G4] 1:250 (Geng et al., 2011), rabbit anti-DUX4 [E5-5] 1:1000 (Geng et al., 2011), rabbit anti-DUX4 [E14-3] 1:1000 (Geng et al., 2011), mouse anti-FLAG [M2] 1:500 (Sigma #F1804, RRID:AB_262044), rabbit anti-STAT1 [EPR4407] 1:750 (Abcam #ab109320, RRID:AB_10863383), and rabbit anti-pSTAT1 Y701 [58D6] 1:400 (Cell Signaling Technology #9167). Cells were washed three times with 1× PBS containing 3% BSA, then secondary antibodies were incubated for 1 hr at room temperature: FITC-conjugated donkey anti-rabbit (Jackson ImmunoResearch #711-095-152, RRID:AB_2315776) or TRITC-conjugated donkey anti-mouse (Jackson ImmunoResearch #715-025-020, RRID:AB_2340764). Cells were washed once with 1× PBS containing 3% BSA then stained with DAPI (Sigma) 1:5000 for 10 min at room temperature and visualized.

Cells were lysed on the plate with digitonin lysis buffer pH 7.4 (37.5 µg/ml digitonin, 25 mM Tris-HCl pH 7.5, 125 mM NaCl, 1 mM EDTA, 5% glycerol) supplemented with Pierce Protease Inhibitors EDTA-free (Pierce, Cat# PIA32955) and Pierce Phosphatase Inhibitors (Pierce, Cat# PIA32957), transferred to a centrifuge tube, and incubated for 10 min at 4°C with rotation. Centrifugation at 2500 rcf at 4°C for 5 min pelleted the nuclei, supernatant was discarded, and nuclei resuspended in 1 ml IP buffer pH 7.4 (25 mM Tris-HCl pH 7.5, 175 mM NaCl, 1 mM EDTA, 0.2% NP-40, 5% glycerol) and incubated for 1 hr at 4°C with rotation then spun at 21,000 rcf for 10 min at 4°C to pellet insoluble debris. Protein concentration was determined using the Pierce BCA Protein Assay Kit (Thermo Fisher, Cat# 23225). An equivalent amount of protein per sample was pre-cleared with Dynabeads Protein G beads (Invitrogen, Cat# 10003D) bound to rat anti-mouse IgG for IP (HRP) (Abcam #131368, RRID:AB_2895114) for 1 hr at 4°C with rotation. FLAG-tagged constructs were then immunoprecipitated with Dynabeads Protein G beads coupled to mouse anti-FLAG [M2] (Sigma #F3165, RRID:AB_259529) for 3 hr at 4°C with rotation. Beads were washed 3× with 1 ml IP buffer and eluted by adding 2× NuPage LDS Sample Buffer (Thermo Fisher, diluted from 4× with PBS) to the beads and heating for 10 min at 70°C.

For LC-MS, anti-FLAG immunoprecipitation was performed with beads cross-linked to mouse anti-FLAG [M2] (Sigma #F3165, RRID:AB_259529) and the proteins competitively eluted with FLAG peptide. Eluted protein samples were electrophoresed into a NuPage 4–12% Bis-Tris gel, excised, and processed by the Fred Hutchinson Cancer Research Center Proteomics Core. Samples were reduced, alkylated, digested with trypsin, desalted, and run on the Orbitrap Eclipse Tribid Mass Spectrometer (Thermo Fisher). Proteomics data were analyzed using Proteome Discoverer 2.4 against a UniProt human database that included common contaminants using Sequest HT and Percolator for scoring. Results were filtered to only include protein identifications from high-confidence peptides with a 1% false discovery rate. Proteins that were identified in at least one sample from both independent experiments with at least two PSMs in one sample were assigned to 1 of 10 categories: 1, candidates; 2, cytoskeletal associated; 3, cytoskeletal; 4, ribosome/translation associated; 5, proteasome associated; 6, membrane or extracellular; ER, golgi, or vesicle associated; 8, lipid metabolism; 9, chaperones; and 10, nuclear import or nuclear membrane associated. The proteins in category 1 were further investigated for interactions with DUX4. It should be noted that this category assignment process de-prioritized groups of proteins based on assignment to a cellular compartment or function (e.g., ribosome/translation proteins might associate with DUX4 as part of a translation complex rather than having a role in immune signaling) and it is possible that some of the proteins assigned to the non-candidate categories might be functional interactors with DUX4 and have an important biological role.

ChIP was performed as previously described (Nelson et al., 2006) with the following modifications: cells were plated and allowed to grow to 70–80% confluence, then treated with doxycycline and/or IFNγ in combination as labeled in Figure 5. Cells were fixed with 1.42% formaldehyde for 15 min at room temperature with shaking. Fixation was quenched with 125 mM glycine, and cells were scraped into Falcon tubes and collected by centrifugation. Cells were lysed to isolate nuclei for 10 min on ice using IP buffer (150 mM NaCl, 50 mM Tris-HCl pH 7.4, 5 mM EDTA, 1% Triton X-100, 0.5% NP-40) containing Pierce Protease Inhibitors EDTA-free (Pierce, Cat# PIA32955) and Pierce Phosphatase Inhibitors (Pierce, Cat# PIA32957) added fresh. Pelleted nuclei were sonicated on a Diagenode Bioruptor on ‘Low’ for 10 min as 30 s on/30 s off, followed by four rounds of sonication on ‘High’ for 10 min each as 30 s on/30 s off (50 min total sonication) in IP Buffer + 0.5% SDS. For immunoprecipitation, 500 ng of chromatin was set aside per condition as an ‘Input’ and 4 µg of antibody was added to 10 µg of chromatin in an equal volume of IP Buffer + 0.5% SDS across samples. ‘STAT1 Ab1’ consisted of a 50:50 mix of rabbit anti-STAT1 [EPR21057-141] (Abcam #ab234400) and rabbit anti-STAT1 [EPR23049-111] (Abcam ab#239360). ‘STAT1 Ab2’ was rabbit anti-STAT1 [EPR4407] (Abcam #ab109320, RRID:AB_10863383). For an IgG control, we used purified Rabbit Polyclonal Isotype Control Antibody (BioLegend #CTL-4112). IP Buffer was added to lower the percentage of SDS <0.1%, and tubes were incubated with rotation overnight at 4°C. During this time, Protein-A Agarose Fastflow beads (Millipore, Cat# 16-156) were washed twice with IP Buffer and then blocked in IP Buffer containing 2% BSA by rotating overnight at 4°C. After clearing the chromatin as described, beads were aliquoted to fresh tubes and the top 90% of chromatin was transferred to the tubes containing the blocked bead slurry. Tubes were rotated for 1 hr at 4°C. Beads were washed five times with cold IP Buffer containing 0.1% SDS, two times with cold IP Buffer containing 500 mM NaCl, and two times with cold PBS. DNA was isolated as described in the original protocol and used as a template in qPCR. Input DNA was used to create a standard curve. qPCR primers for the h16q21 gene desert region and the ISGs were previously published (Maston et al., 2012; Rosowski et al., 2014).

MB135iDUXBCTD, MB135iDUX4CTD, HFF1°, HFF1°-DUXBCTD, and HFF1°-DUX4CTD cells were plated onto Millicell EZ Slide 8-well glass slides (Millipore PEZGS0816) and treated with IFNγ/dox or IFNγ-alone as described in figures. KitraSRS cells were plated onto standard TC dishes and treated ±IFNγ, then trypsinized and scraped from dishes into DMEM to quench the trypsin and pelleted by centrifugation. Pelleted KitraSRS cells were resuspended in 1× PBS and immediately spun onto slides at 1900 rcf for 1 min, after which they were treated identically to the slide-plated cells. All cells were fixed for 10 min with 4% paraformaldehyde (Thermo Scientific), permeabilized for 10 min with 0.5% Triton X-100 (Sigma), and then blocked for 2 hr at room temperature with PBS/0.3 M glycine/3% BSA. Primary antibodies were diluted in PBS/3% BSA and incubated with samples overnight at 4°C. For PLA of cell lines expressing FLAG-tagged transgenes, mouse anti-FLAG [M2] (F1804) (1:4000) was used in combination with either rabbit anti-STAT1 [EPR4407] 1:1000 (Abcam ab109320) or rabbit anti-pSTAT1 Y701 [58D6] 1:1000 (Cell Signaling Technology #9167). For the Kitra-SRS cells expressing endogenous CIC-DUX4, rabbit anti-CIC 1:500 (Invitrogen, PA5-83721) was used in combination with either mouse anti-STAT1 [1/Stat1] 1:1000 (Abcam ab281999) or mouse anti-pSTAT1 Y701 [M135] 1:1000 (Abcam ab29045). Samples were washed three times for 10 min with 1× Wash Buffer A (10 mM Tris, 150 mM NaCl, 0.05% Tween, adjusted pH to 7.4), and then incubated with Duolink In Situ PLA Probe Anti-Rabbit PLUS (Sigma, Cat# DUO92002) and Duolink In Situ PLA Probe Anti-Mouse MINUS (Sigma, Cat# DUO92004) diluted 1:5 in PBS/3% BSA for 1 hr in a humidity chamber at 37°C. Samples were washed three times for 10 min with 1× Wash Buffer A, and then treated with ligase from the Duolink In Situ Detection Reagents Green kit (Sigma, Cat# DUO92014) for 30 min in a humidity chamber at 37°C. Samples were washed three times for 10 min with 1× Wash Buffer A, and then treated with polymerase from the Duolink In Situ Detection Reagents Green kit for 1 hr and 40 min in a humidity chamber at 37°C. Samples were washed two times for 10 min with 1× Wash Buffer B (200 mM Tris, 100 mM NaCl, adjusted pH to 7.5) and then once for 1 min with 0.01× Wash Buffer B. Samples were mounted with Prolong Glass Antifade Mountant with NucBlue (Thermo Fisher, Cat# P36983), and then visualized with a fluorescent microscope using FITC and DAPI filters.

Cells were transfected with 50 pmol total siRNAs using the Lipofectamine RNAiMax Transfection Reagent (Thermo Fisher, Cat# 13778150) according to the manufacturer’s protocol and allowed to sit overnight (16 hr). Cells were changed to fresh growth medium the next morning and allowed to recover during the day, then transfected again with 50 pmol total siRNAs in Lipofectamine RNAiMax Transfection Reagent at the end of the day and left overnight (16 hr). Cells were changed to fresh growth medium the next morning and then used for downstream experiments. The control siRNA (siCTRL) was siOn-Target (Dharmacon, Cat# D-001810-01). The siRNAs targeting CIC-DUX4 were FlexiTube siRNAs Hs_DUX4_11 (QIAGEN, Cat# SI04268453), Hs_CIC_6 (QIAGEN, Cat# SI04275656), and HS_CIC_8 (QIAGEN, Cat# SI04368469).

CUT&Tag was performed as previously described (Kaya-Okur et al., 2019) with the following modifications: MB135-iDUX4-CTD myoblasts were plated and allowed to grow to 70–80% confluence. Cells were left untreated, treated with 200 ng/ml IFNγ for 16 hr, or pre-treated with 1 µg/ml doxycycline for 4 hr then had IFNγ added directly to cell media for an additional 16 hr. Fresh cells were harvested and washed in PBS, crosslinked with 0.1% formaldehyde for 90 s, then counted and 1.25e6 cells were aliquoted per reaction tube. Drosophila S2 cells were spiked-in at a genomic ratio of 1:10. Nuclei were prepared from cells in Buffer NE1 (20 mM HEPES-KOH pH 7.9, 10 mM KCl, 0.1% Triton X-100, 20% glycerol, 0.5 mM spermidine, Pierce Protease Inhibitors EDTA-free [PIA32955]) on ice for 10 min and then bound to concanavalin A-coated beads for 10 min. Rabbit anti-phospho Rbp1 CTD (Ser5) [D9N5I] (Cell Signaling Technology, Cat #13523) diluted 1:50 was bound overnight at 4°C in 25 µl per sample of Antibody Buffer (20 mM HEPES-KOH pH 7.5, 150 mM NaCl, 0.5 mM spermidine, 0.01% digitonin, 2 mM EDTA, 1× Roche cOmplete mini EDTA-free protease inhibitor). Anti-rabbit secondary antibody (EpiCypher, Cat#13-0047) diluted 1:100 was bound in 25 µl per sample of Wash150 Buffer (20 mM HEPES-KOH pH 7.5, 150 mM NaCl, 0.5 mM spermidine, 1× Roche cOmplete mini EDTA-free protease inhibitor) for 30 min at room temperature. pAG-Tn5 pre-loaded adapter complexes (EpiCypher, Cat# 15-1017) were added to the nuclei-bound beads for 1 hr at room temperature in 25 µl of Wash300 Buffer (20 mM HEPES-KOH pH 7.5, 300 mM NaCl, 0.5 mM spermidine, 1× Roche cOmplete mini EDTA-free protease inhibitor), then beads were washed and resuspended in Tagmentation Buffer (Wash300 Buffer + 10 mM MgCl₂) and incubated at 37°C for 1 hr in a thermocycler with heated lid. Tagmentation was stopped by addition of EDTA, SDS, and proteinase K. DNA was extracted by Phenol-Chloroform and amplified by PCR using CUTANA High Fidelity 2× PCR Master Mix (EpiCypher, Cat#15-1018) and cycling conditions: 5 min at 58°C; 5 min at 72°C; 45 s at 98°C; 14 cycles of 15 s at 98°C, 10 s at 60°C; 1 min at 72°C. PCR products were cleaned up using AMPure XP beads (Beckman Coulter, Cat# A63880) at a ratio of 1.3:1 according to the manufacturer’s instructions.

CUT&Tag data were aligned to the GRCh38 patch 13 human genome e following the Benchtop CUT&Tag v3 protocol (Kaya-Okur et al., 2019). Subsequent to alignment we calculated 1× genome coverage normalization with read centering and read extension using deepTools’ bamCoverage (Ramírez et al., 2016) then mapped the resulting coverage tracks to regions of interest using bedtools’ map function (Quinlan and Hall, 2010). Coverage graphs were plotted using ggplot2 from the tidyverse package in R (Wickham et al., 2019).

Plasmids used in this study will be deposited with Addgene or are available through request to the corresponding author.