Human DUX4 and mouse Dux interact with STAT1 and broadly inhibit interferon-stimulated gene induction

Abstract

DUX4 activates the first wave of zygotic gene expression in the early embryo. Mis-expression of DUX4 in skeletal muscle causes facioscapulohumeral dystrophy (FSHD), whereas expression in cancers suppresses IFNγ induction of major histocompatibility complex class I (MHC class I) and contributes to immune evasion. We show that the DUX4 protein interacts with STAT1 and broadly suppresses expression of IFNγ-stimulated genes by decreasing bound STAT1 and Pol-II recruitment. Transcriptional suppression of interferon-stimulated genes (ISGs) requires conserved (L)LxxL(L) motifs in the carboxyterminal region of DUX4 and phosphorylation of STAT1 Y701 enhances interaction with DUX4. Consistent with these findings, expression of endogenous DUX4 in FSHD muscle cells and the CIC-DUX4 fusion containing the DUX4 CTD in a sarcoma cell line inhibit IFNγ induction of ISGs. Mouse Dux similarly interacted with STAT1 and suppressed IFNγ induction of ISGs. These findings identify an evolved role of the DUXC family in modulating immune signaling pathways with implications for development, cancers, and FSHD.

Keywords: DUX; DUX4; DUXC; STAT1; developmental biology; human; immunology; inflammation; interferon stimulated genes; interferon-gamma; interferons.

Figure 1.. DUX4 suppresses interferon-stimulated gene (ISG) induction.

(A) MB135 cells expressing doxycycline-inducible DUX4 (MB135-iDUX4), parental MB135 cells, or MB135 cells expressing doxycycline-inducible DUXB (MB135-iDUXB) were untreated, treated with IFNγ, or treated with doxycycline and IFNγ. RT-qPCR was used to evaluate expression of a DUX4 target gene, ZSCAN4, and ISGs IFIH1, ISG20, CXCL9, and CD74. Ct values were normalized to the housekeeping gene RPL27. Data represent the mean ± SD of three biological replicates with three technical replicates each. See Figure 1—figure supplement 2 for biological replicates in independent cell lines. (B) MB135-iDUX4 cells were untreated, treated with either IFNβ (type 1 IFN pathway), poly(I:C) (IFIH1/MDA5 pathway), 5’ppp-dsRNA (DDX58/RIG-I pathway), or cGAMP (cGAS/STING pathway), or treated with doxycycline and the same immune reagent. RT-qPCR was used to evaluate expression of IFIH1, ISG20, CXCL9, and CD74. Ct values were normalized to the housekeeping gene RPL27. Data represent the mean ± SD of three biological replicates with three technical replicates each (unpaired t-test; ****p<0.0001, ***p<0.001, **p<0.01, *p<0.05, ns p>0.05).

Figure 1—figure supplement 1.. Schematics of constructs cloned for use in this study.

(A) Schematic depiction of each transgene used in this study highlighting the N-terminal homeodomains (light gray in DUX4, no fill in DUXB, light green in mDux), DNA-binding HMG box (dark blue in CIC and CIC-DUX4), conserved C-terminal domain (medium gray in DUX4 and CIC-DUX4, medium green in mouse Dux), (L)LxxL(L) (black in DUX4 and CIC-DUX4, dark green in mouse Dux), mutations (* and black bar F67A, * replacement of LDELL with AAEAA), and 3xFLAG-NLS cassette regions (no fill). ‘3XFLAG’ refers to a triplicated FLAG tag sequence. ‘NLS’ refers to two independent NLS sequences, derived from SV40 and the eight amino acid NLS from SMCHD1 (PPKRMRRE), which have been mapped functionally (Hiramuki and Tapscott, 2018). (B) MUSCLE alignment of the terminal ~50 aa of the human DUX4, mutated human DUX4 (mL1, dL2, mL1dL2), and mouse Dux constructs used in this study.

Figure 1—figure supplement 2.. Biological replicates in independent cell lines for each DUX4 construct.

Additional subcloned MB135 cell lines of the iDUX4 (A), iDUX4-F67A (B), iDUX4-CTD (C), iDUX4mL1dL2 (D), iDUX4-CTDmL1dL2 (E), iDUX4aa339-324 (F), and iDux (G) treated with IFNγ ± doxycycline. RT-qPCR shows interferon-stimulated gene (ISG) expression graphed as a % of IFNγ-only. Data represent the mean ± SD of two or three biological replicates (see individual construct data points) with three technical replicates each (unpaired t-test, ****p<0.0001, ***p<0.005, **p<0.01, *p<0.05, ns = nonsignificant). Immunofluorescence panels show protein expression and nuclear localization using an antibody against the N-terminal (anti-DUX4 [E14-3]) or C-terminal (anti-DUX4 [E5-5]) residues of DUX4 as appropriate for the construct.

Figure 2.. DUX4 transcriptional activity is not necessary for interferon-stimulated gene (ISG) suppression, whereas the C-terminal domain (CTD) is both necessary and sufficient.

(A) MB135 cell lines with the indicated doxycycline-inducible transgene ± doxycycline were evaluated for ZSCAN4 expression by RT-qPCR as a measure of the ability of the construct to activate a DUX4-target gene. Ct values were normalized to the housekeeping gene RPL27. Data represent the mean ± SD of three biological replicates with three technical replicates each. (B–D) MB135 cell lines with the indicated doxycycline-inducible transgene were treated with IFNγ ± doxycycline. Light gray, N-terminal boxes, homeodomains; medium gray, C-terminal box, conserved region of CTD; black, C-terminal boxes, (L)LxxL(L) motifs; * indicates sites of mutation for F67A in HD1 and mutation of first LLDELL to AADEAA. See Figure 1—figure supplement 1 for additional description of 3XFLAG and NLS cassette. RT-qPCR was used to evaluate expression of IFIH1, ISG20, CXCL9, and CD74 and Ct values were normalized to the housekeeping gene RPL27, then normalized to the IFNγ-only treatment to set the induced level to 100%. Data represent the mean ± SD of three biological replicates with three technical replicates each (unpaired t-test; ****p<0.0001, ***p<0.001, **p<0.01, *p<0.05, ns p>0.05). See Figure 1—figure supplement 2 for additional cell lines.

Figure 3.. The DUX4 protein interacts with STAT1 and additional immune response regulators.

Left panel, representative candidate interactors identified by mass spectrometry of proteins that co-immunoprecipitated with the DUX4-CTD and their relative ranking in the candidate list (see Supplementary file 2 for full list). Right panel, validation western blot of proteins that co-immunoprecipitate with the DUX4-CTD in cell lysates from MB135 cells expressing doxycycline-inducible 3xFLAG-DUXB or 3xFLAG-DUX4-CTD, ± IFNγ treatment. Data represent biological duplicates. See Figure 3—source data 1 for uncropped/raw images.

Figure 4.. The DUX4-CTD preferentially interacts with pSTAT1-Y701.

(A) Western blot showing input and immunoprecipitated proteins from either 3xFLAG-iDUXB (DUXB) or a truncation series of the 3x-FLAG-iDUX4-CTD cells (iDUX4) precipitated with anti-FLAG and probed with the indicated antibodies. Serial deletions of the iDUX4-CTD were assayed as indicated. All samples were treated with IFNγ.An asterisk indicates the correct band for each FLAG-tagged construct. See Figure 4—source data 1 for uncropped/raw Western blots. (B) Input and anti-FLAG immunoprecipitation from 3xFLAG-iDUXB or 3x-FLAG-iDUX4-CTD cells co-expressing doxycycline-inducible 3xMYC-iSTAT1, -iSTAT1-Y701A, or -iSTAT1-S727A with or without IFNγ treatment and probed with the indicated antibodies. An ‘x’ indicates the endogenous (non-MYC tagged) STAT1 band. See Figure 4—source data 1 for uncropped/raw Western blots. (C) Proximity ligation assay (PLA) showing co-localization of endogenous STAT1 and pSTAT1 701 with the iDUX4-CTD compared to the interaction with the DUXB-CTD, in the nuclear compartment of IFNγ- and doxycycline-treated MB135 cells. Mean fluorescent intensity (MFI) of the nuclei in the PLA channel was measured for 10 images per cell line and treatment and plotted (unpaired t-test; ****p<0.0001).

Figure 4—figure supplement 1.. Expression of the DUX4-CTD does not prevent translocation of STAT1 to the nucleus.

Immortalized MB135 myoblasts with doxycycline-inducible iDUX4-CTD (A) and iDUXB (B) transgenes were treated ± doxycycline and ± IFNγ, then fixed and stained for total STAT1 and either transgene. Both cell lines show good induction and nuclear translocation of STAT1 restricted to the +IFNγ condition, and good doxycycline-dependent induction of the transgene. (C) Phosphorylation of STAT1 and nuclear translocation of pSTAT1-Y701 is equally restricted to the +IFNγ condition in MB135iDUX4CTD and MB135iDUXBCTD cell lines. Immunofluorescence staining for pSTAT1-Y701 shows no phosphorylation or nuclear translocation of STAT1 in untreated cells, and strong localization of pSTAT1-Y701 to the nucleus only with IFNγ treatment. Mean fluorescent intensity (MFI) of nuclear pSTAT1-Y701 signal was measured from two fields per treatment per cell line and plotted (right). There was significant nuclear enrichment of pSTAT1-Y701 with IFNγ treatment compared to untreated cells in both immortalized transgenic cells lines (unpaired t-test, **** p<00001, ns = nonsignificant; n = 26 DUX4CTD untreated, n = 36 DUX4CTD + IFNγ, n = 43 DUXBCTD untreated, and n = 53 DUXBCTD + IFNγ). (D) Primary human foreskin fibroblasts (HFF 1°), primary MB135 myoblasts (MB135 1°), and immortalized MB135 myoblasts (MB135 immortalized) were treated ± IFNγ, then fixed and stained for total STAT1. Background staining noise was similar across cell lines in the untreated condition, and all cell lines showed good induction and nuclear localization of STAT1 in the + IFNγ condition.

Figure 4—figure supplement 2.. Primary human foreskin fibroblasts (HFFs) expressing transgenic DUX4CTD show increased interaction with STAT1 and reduced MHC I activation with IFNγ treatment.

Primary HFFs (HFF 1°) expressing no transgene (noTG), constitutive 3XFLAG-DUXBCTD, or constitutive 3XFLAG-DUX4CTD were treated with IFNγ, and then fixed and used in a proximity ligation assay (PLA) to determine interaction between the FLAG-tagged transgenes and total STAT1. Post-PLA the cells were re-stained with an aFLAG antibody to determine transgene-positive (TG+) nuclei based on fluorescence. Mean fluorescent intensity (MFI) in the PLA channel of TG+ nuclei from five images per cell line was calculated and plotted (right; ‘n’ = total number of TG+ nuclei analyzed). TG+ DUX4 CTD nuclei had significantly more interaction with total STAT1 compared to TG+ nuclei of the DUXBCTD cell line (one-way ANOVA with multiple comparisons, ****p<0.0001, **p<0.01, ns = nonsignificant). The PLA signal, which relies on presence and proximity of both the aFLAG and aSTAT1 antibodies, was higher in TG+ nuclei of both cell lines as expected.

Figure 5.. The DUX4-CTD decreases STAT1 occupancy at interferon-stimulated gene (ISG) promoters and blocks Pol-II recruitment.

(A, left four panels) Chromatin immunoprecipitation using anti-STAT1 or IgG from MB135-iDUX4-CTD cells untreated, IFNγ-treated, or IFNγ and doxycycline treated. Ab1: 50:50 mix of STAT1 antibodies Abcam ab239360 and ab234400; Ab2: Abcam ab109320. ChIP-qPCR analysis relative to a standard curve constructed from purified input DNA was used to determine the quantity of DNA per IP sample, which was then graphed as a % of input. Data represent the mean ± SD of two biological replicates with three technical replicates each (unpaired t-test; ****p<0.0001, ***p<0.001, **p<0.01, *p<0.05, ns p>0.05). (A, right panel) RT-qPCR of RNA from cells used for STAT1 ChIP showing induction of interferon-stimulated genes (ISGs) by IFNγ and suppression by DUX4-CTD. (B) CUT&Tag data showing the intensity of Pol-II signal across a 2000 bp window centered on the TSS of ISGs (left) or IFNγ-unchanged genes (right) in untreated, IFNγ-treated, or IFNγ and doxycycline-treated MB135-iDUX4-CTD cells.

Figure 6.. Endogenous DUX4 suppresses interferon-stimulated gene (ISG) induction in a sarcoma cell line expressing a CIC-DUX4 fusion gene.

(A, left panel) RT-qPCR of the indicated genes in MB135 parental or Kitra-SRS that express a CIC DUX4-fusion gene containing the DUX4 CTD. Cells were transfected with control or CIC- and DUX4-targeting siRNAs. Ct values were normalized to the housekeeping gene RPL27. Data represent the mean ± SD of three biological replicates with three technical replicates each (unpaired t-test; ****p<0.0001, ***p<0.001, ** p<0.01,*p<0.05, ns p>0.05). (A, right panel) Western blot showing lysates from MB135 or Kitra-SRS cells treated with control or CIC- and DUX4-targeting siRNAs ± IFNγ and probed with the indicated antibodies. See Figure 6—source data 1 for uncropped/raw western blots. (B) RT-qPCR of the indicated genes in MB135 with an inducible CIC (MB135-iCIC) or an inducible CIC-DUX4 fusion gene (MB135-iCIC-DUX4). Cells were untreated, IFNγ-treated, or IFNγ and doxycycline-treated. Ct values were normalized to the housekeeping gene RPL27, then normalized to the IFNγ-only treatment to set the induced level to 100%. Data represent the mean ± SD of three biological replicates with three technical replicates each (unpaired t-test; **p<0.01, ns p>0.05). (C) Proximity ligation assay (PLA) of KitraSRS cells showing association of the endogenous CIC-DUX4 fusion protein with either total STAT1 or phosphorylated STAT1-Y701 exclusively when cells were treated +IFNγ. Mean fluorescent intensity (MFI) was quantified from 200 nuclei per condition and plotted for both pairs of antibodies (unpaired t-test; ****p<0.0001).

Figure 6—figure supplement 1.. Knockdown of the CIC-DUX4 fusion protein in Kitra-SRS cells rescues upregulation of MHC I in response to IFNγ.

MB135 parental myoblasts (A) and Kitra-SRS sarcoma cells (B) were treated with control siRNAs or a combination of siRNAs targeting CIC and DUX4, then left untreated or treated with IFNγ. While knockdown of the endogenous CIC in MB135 cells had no effect on MHC I upregulation in IFNγ-treated cells, knockdown of the CIC-DUX4 fusion protein in Kitra-SRS cells increased MHC I-positive cells from 27.9% to 48.1%.

Figure 7.. Conservation of interferon-stimulated gene (ISG) repression in facioscapulohumeral dystrophy (FSHD) myoblasts and ISG repression and STAT1 interaction by mouse Dux.

(A) FSHD MB200 myoblasts were differentiated into myotubes, which results in the expression of endogenous DUX4 in a subset of myotubes. Cultures were treated ± IFNγ, and DUX4 and IDO1 were visualized by immunofluorescence. Representative images of untreated and IFNγ-treated (two fields, F1 and F2) cells are shown, with white arrows highlighting DUX4+ myotubes that lack IDO1 signal. Mean fluorescent intensity (MFI) of the αDUX4 and αIDO1 nuclear signal was measured in the IFNγ-treated cells only. Data represent the mean ± SD of nuclear MFI from three images, total nuclei per condition listed as ‘n’ (unpaired t-test; **p<0.01). (B, left panel) RT-PCR of the indicated genes in MB135-iDux cells untreated or treated with IFNγ ± doxycycline. Ct values were normalized to the housekeeping gene RPL27, then normalized to the IFNγ-only treatment to set the induced level to 100%. Data represent the mean ± SD of three biological replicates with three technical replicates each (unpaired t-test; ****p<0.0001, **p<0.01). (B, right panel) Western blot showing input and immunoprecipitated proteins from either 3xFLAG-iDux or 3x-FLAG-iDUXB cells ± IFNγ precipitated with anti-FLAG and probed with the indicated antibodies. See Figure 7—source data 1 for uncropped/raw Western blots. (C) A model supported by the data showing how the DUX4-CTD might prevent STAT1 ISG induction. (Top) In the absence of the DUX4-CTD, pSTAT1 Y701 (black ‘P’) dimerizes, translocates to the nucleus, binds its GAS motif in the ISG promoter, acquires secondary phosphorylation at S727 (gray ‘P’), and recruits a stable transcription complex that includes Pol-II to drive transcription of ISGs. (Bottom) In the presence of the DUX4-CTD, STAT1 is phosphorylated, translocates to the nucleus, and binds its GAS motif as evidenced by the pSTAT1 S727 in complex with the CTD. However, diminished steady-state occupancy of STAT1 at the ISG promoters and absence of Pol-II recruitment indicate that the STAT1-DUX4-CTD complex does not stably bind DNA and fails to recruit Pol-II and the pre-initiation complex. The (L)LXXL(L) motifs (black bars in DUX4-CTD) are necessary to interfere with transcription suppression and likely prevent STAT1 from interacting with a factor in the pre-initiation complex or recruit a co-repressor.

Figure 7—figure supplement 1.. Mouse Dux contains a triplication of the (L)LxxL(L)-containing region.

(A) Mouse Dux protein sequence with homeodomains in bold and (L)LxxL(L) motifs underlined. (B) Alignment of a partial triplication of the mouse Dux protein with aa258-440 aligning with aa441-623 and aa624-650 aligning with aa258-284.