Antagonism among DUX family members evolved from an ancestral toxic single homeodomain protein

Abstract

Double homeobox (DUX) genes are unique to eutherian mammals and normally expressed transiently during zygotic genome activation. The canonical member, DUX4, is involved in facioscapulohumeral muscular dystrophy (FSHD) and cancer, when misexpressed in other contexts. We evaluate the 3 human DUX genes and the ancestral single homeobox gene sDUX from the non-eutherian mammal, platypus, and find that DUX4 activities are not shared with DUXA or DUXB, which lack transcriptional activation potential, but surprisingly are shared with platypus sDUX. In human myoblasts, platypus sDUX drives cytotoxicity, inhibits myogenesis, and induces DUX4 target genes, particularly those associated with zygotic genome activation (ZGA), by binding DNA as a homodimer in a way that overlaps the DUX4 homeodomain crystal structure. DUXA lacks transcriptional activity but has DNA-binding and chromatin accessibility overlap with DUX4 and sDUX, including on ZGA genes and LTR elements, and can actually be converted into a DUX4-like cytotoxic factor by fusion to a synthetic transactivation domain. DUXA competition antagonizes the activity of DUX4 on its target genes, including in FSHD patient cells. Since DUXA is an early DUX4 target gene, this activity potentiates feedback inhibition, constraining the window of DUX4 activity. The DUX gene family therefore comprises cross-regulating members of opposing function, with implications for their roles in ZGA, FSHD, and cancer.

Highlights: Platypus sDUX is toxic and inhibits myogenic differentiation.DUXA targets overlap substantially with those of DUX4.DUXA fused to a synthetic transactivation domain acquires DUX4-like toxicity.DUXA behaves as a competitive inhibitor of DUX4.

Figure 1.. Platypus sDUX displays functional overlap with DUX4

a. Evolution of the DUX gene family from an ancestral single homeobox gene in the common ancestor to eutherian mammals. b. Sequence dendrogram of DUX homeodomains. DUX family members are shaded yellow (DUXA), green (DUXB), and blue (DUXC); and sDUX is shaded red. c. Morphology of LHCN-M2isDUX human myoblasts in the absence (left) or presence of 100 ng/mL dox for 4 days to induce sDUX (right). Scale bar, 100 um. d. EdU incorporation in LHCN-M2isDUX human myoblasts after 6 hrs in the absence (blue) or presence (red) of 500 ng/mL dox to induce sDUX. e. Annexin V incorporation in LHCN-M2isDUX human myoblasts after 24 hrs in the absence (blue) or presence (red) of 500 ng/mL dox to induce sDUX. f. Sequence comparison of the sDUX homeodomain with HD1 and HD2 of the 3 human DUX family members, DUXC from elephant and mouse, and mouse DUXBL. Known or predicted DNA binding regions are highlighted in blue, and residues known or predicted to have specific nucleotide base contacts are underlined. Yellow shading indicates the E and R residues responsible for TAAT or TGAT specificities of DUX4 HD1 and HD2, resp. Asterisks indicate positions with a conserved residue, colons indicate conservation of strongly similar biochemical properties, periods indicate conservation between amino acids of weakly similar properties. g. RTqPCR for ZSCAN4, LEUTX and PRAMEF1 in LHCN-M2iDUX4 parent and LHCN-M2-isDUX myoblasts induced with 200 ng/mL doxycycline for 14 hours. Data are presented as mean ± SEM; *p<0.05, ***p<0.01**, ****p<0.0001 by one-way ANOVA, n=3. h. Luciferase assays in comparing activity of DUX4 and sDUX on different versions of the double homeodomain motif in 293T cells exposed to 500 ng/mL dox for 24 hrs. The core HD binding sites are underlined. DUX4 and DUX cores are separated by 3 nt while the PAX7 recognition motif cores are separated by 2 nt. Copy number of motifs indicated by 1x or 2x. i. Viability assay for 293T cells after 48 hrs of expression (100 ng/mL dox) of full length sDUX vs. sDUXΔC, lacking the C-terminal 74 amino acids. j. RTqPCR for DUX4 target genes LEUTX and TRIM43 after expression of sDUX or sDUXΔC in the same cells in panel I (500 ng/mL, 24 hrs). k. Representative western blots with acetylation specific-antibodies in human myoblasts expressing sDUX and treated with the p300/CBP-specific HAT inhibitor, iP300w. LHCN-isDUX cells were induced with 200 ng/ml doxycycline and concurrently treated with 0.3 μM iP300w for 12 hours. l. RTqPCR for sDUX target genes RFPL4B and ZSCAN4 in the presence or absence of iP300w on the cell presented in panel k.

Figure 2.. Structural basis of sDUX interaction with DNA

a. Binding of DUX4 HD1-HD2 and sDUX HD to the non-palindromic (TAAT---ATCA) and palindromic (TGAT---ATCA or TAAT---ATTA) motifs tested in EMSA. All DNA substrates were double-stranded DNA. Sequence of the fluorescently labeled strand is shown above each gel. b. Binding of DUX4 HD1-HD2 and sDUX HD to the TAAT and TGAT half-motifs or a non-palindromic motif with 2 nt spacing (N2) tested in EMSA. c. Binding of sDUX HD to the three different types of DNA substrates, highlighting distinct mobilities of dimeric vs. monomeric complex in EMSA. d. Crystal structure of sDUX homodimer bound to the non-palindromic target motif with 3 nt spacing, the structure of DUX4 double homeodomain with the same DNA substrate as reported previously , and their superposition. e. Differential interaction of sDUX Arg75 with the TAAT and TGAT motifs in the crystal structure. 2Fo-Fc electron density for the structure with Br-dU-containing DNA is shown as blue mesh, contoured at 0.9 σ.

Figure 3.. Toxicity and myogenic differentiation effects of DUX family members and VP64-fusions

a. Viability assay in LHCN-M2 human myoblasts after 96 hrs of expression of different DUX family proteins induced with 200 ng/mL doxycycline. b. Luciferase assays in 293T cells induced with 1 μg/mL dox for 24 hrs to express various DUX family proteins and VP64 derivatives on the 3 versions of DUX motif (N3 and P3) as well as a motif with spacing of 2 nt (P2-type). c. Heat map of the top DUX4 target genes showing expression in cells expressing various DUX family members and VP64 derivatives after 6 hours of induction (200 ng/mL dox). d. Heat map of the top myogenic genes affected by DUX4, and the effect of other DUX family members, and DUXA- and DUXB-VP64 activation domain fusions on this set. Genes downregulated by DUX4 shown above; genes upregulated by DUX4 shown below. Cells were induced by 200 ng/mL dox for 12 hours. e. RTqPCR for MYH7, a marker of differentiation, after differentiation with 72 hrs of expression of different DUX family proteins at 20 or 125 ng/mL dox. f. Representative immunostaining for sarcomeric myosin heavy chain on the same cells shown in panel e.

Figure 4.. Transcriptional profiles of DUX family members

a. Principal component analysis of gene expression profiles in the absence (open circles) and presence (closed circles) of 250 uM dox for 12 hrs. DUX4 expressing cells are strongly shifted along PC1 which accounts for 62% of the variance while DUXA-VP64 and sDUX expressing cells are more prominently shifted along PC2 which contains 8% of the variance. b. Heatmap representation of log2 transformed FPKM values for the top 50 differentially expressed genes among all factors. Expression of DUX4, DUXA, DUXB and sDUX is shown in the upper panel. c. Venn diagrams showing overlap in upregulated gene sets among total genes (left) and ZGA genes (right). Upregulated genes for each set were defined as having a greater than 2-fold change, an Benjamini-Hochberg adjusted p-value less than 0.05 and a mean FPKM value > 2.5 in the dox induced sample. d. Distribution of repeat classification among the differentially expressed retroviral elements for DUX4, DUXA-VP64 and sDUX. Differentially expressed elements were defined as being upregulated or downregulated greater than 2-fold and having a Benjamini-Hochberg adjusted p-value less the 0.05. e. Pearson correlations of log2 fold changes in expression upon induction with dox between each of the DUX factor. The upper left half of the matrix corresponds to conventional genes whereas the lower right half of the matrix corresponds to log2 fold changes for repetitive elements. f. Scatter plots of the log2 fold changes in conventional expression between DUX factors. Points representing individual genes are colored according to density to show the higher density of points near the center. A linear fit of each correlation is shown as a black line. g. Scatter plots similar to panel f for repetitive elements.

Figure 5.. Chromatin changes induced by DUX family members

a. Principal component analysis of ATAC-seq data in the absence (open circles) and presence (closed circles) of 250 uM dox for 12 hrs. Called peaks from each factor were combined into a composite set of genomic locations that had changes in DNA accessibility in one or more samples. A matrix of the number of reads within each region across all the samples was used to calculate the principal components. b. ATAC-seq coverage tracks for each DUX factor at the ZSCAN4 locus in the absence (gray) or presence (colored according to panel a) of dox. Each track shown is a combination of two biological replicates normalized to the total number of reads per million. The scale for each pair of tracks is shown at the right and the gene structures is shown below. The red box corresponds to peaks that appear for some DUX factors upon the addition of dox. c. ATAC-seq coverage similar to panel B for the PRAMEF12 locus. d. The DNA sequences from the top 500 peaks were used to define de novo motifs for DUX4, DUXA-VP64 and sDUX. The top motif and its p-value are shown. The DUX4 and DUXA-VP64 motifs are similar to the previously defined motifs for DUX4 and DUXA repectively (https://jaspar.genereg.net/). e. The de novo motifs identified from the top 500 DUX4, DUXA-VP64 and sDUX peaks were identified across each dataset using a 95% scoring threshold. Classification of each sequence is based on the nucleotide at the 2nd and 10th position. f. The number of sequence motifs identified in panel e per ATAC-seq peak were counted to show that peaks frequently contain multiple copies of the DUX motif in all three datasets. g. Heatmaps of ATAC-seq coverage for peaks within 10 kb of an upregulated gene. Peaks were classified into one of seven groups depending on whether they were unique to a particular dataset or shared between two or more datasets. Groups that contained more than 100 peaks were randomly subsampled to 100 peaks. Coverage was normalized to total counts per million reads of each dataset and then plotted on the scale defined by the dox induced dataset shown below the heatmaps.

Figure 6.. DUXA binds many DUX4 targets and inhibits DUX4 transcriptional activity

a. Tornado plots aligning on peak centers for called DUX4 peaks (upper panel) and DUXA peaks (lower panel) for both input and IP samples for both experiments. Overlap of DUX4 and DUXA peaks (peak calling: FDR <= 0.1, minimum treatment reads >= 1, enrichment >= 1, log10(qvalue) <= −0.5; not overlapping blacklisted regions like satellite repeats or misassemblies (ENCODE). b. ChIP-seq coverage tracks for DUXA and DUX4 surrounding the ZSCAN4 (upper) and PRAMEF12 (lower) genes. Input tracks are shown in gray and IP tracks are colored uniquely. Each track shown is a combination of two normalized biological replicates. The scale for each track is shown at the right and the gene structure is shown below. Arrows indicate shared peaks. c. Venn diagrams showing overlap in peaks (left) and in bound genes (right). Promoter = gene TSS + upstream 5000 nt + downstream 5000 nt; overlap of ≥ 1 nt. d. RTqPCR for DUX4 target genes in LHCN-M2iDUX4 immortalized human WT myoblasts co-expressing either DUXAiresEGFP or EGFP control, and cells in which overexpressed DUXA was knocked down by siRNA or control non targeted siRNA. Cells were induced with 200 ng/mL doxycycline for 8 hours. Data are presented as mean ± SEM; ***p<0.001, ****p<0.0001 by two-way ANOVA, n=3. e. RTqPCR for DUX4 target genes in immortalized human FSHD myoblasts engineered for dox-inducible for DUXA (M008-iDUXA). Cells were induced with 200 ng/mL doxycycline for 48 hours in proliferation conditions (Prolif.). For differentiation (Diff.), cells were cultured in differentiation medium for 3 days. DUXA was induced (200 ng/mL dox) over the last 24 hours of differentiation. Data are presented as mean ± SEM; ****p<0.0001 by two-way ANOVA, n=3.