ASPSCR1-TFE3 reprograms transcription by organizing enhancer loops around hexameric VCP/p97

Abstract

The t(X,17) chromosomal translocation, generating the ASPSCR1::TFE3 fusion oncoprotein, is the singular genetic driver of alveolar soft part sarcoma (ASPS) and some Xp11-rearranged renal cell carcinomas (RCCs), frustrating efforts to identify therapeutic targets for these rare cancers. Here, proteomic analysis identifies VCP/p97, an AAA+ ATPase with known segregase function, as strongly enriched in co-immunoprecipitated nuclear complexes with ASPSCR1::TFE3. We demonstrate that VCP is a likely obligate co-factor of ASPSCR1::TFE3, one of the only such fusion oncoprotein co-factors identified in cancer biology. Specifically, VCP co-distributes with ASPSCR1::TFE3 across chromatin in association with enhancers genome-wide. VCP presence, its hexameric assembly, and its enzymatic function orchestrate the oncogenic transcriptional signature of ASPSCR1::TFE3, by facilitating assembly of higher-order chromatin conformation structures demonstrated by HiChIP. Finally, ASPSCR1::TFE3 and VCP demonstrate co-dependence for cancer cell proliferation and tumorigenesis in vitro and in ASPS and RCC mouse models, underscoring VCP's potential as a novel therapeutic target.

Fig. 1. AT3 interacts with VCP/p97 in the nucleus of ASPS and RCC tumor cells through its ASPSCR1 portion.

a Plot of enrichment scores (defined in Methods) for homologous proteins identified on mass spectroscopy proteomics following ASPSCR1 immunoprecipitation (IP) of nuclear lysates from human FU-UR-1 cells (n = 8 + 2 controls IgG) and mouse ASPS tumors (n = 7 + 3 controls IgG) b IPs followed by western blots (WB) showing reciprocal AT3 (by ASPSCR1 or TFE3 antibodies) and VCP interactions in multiple tumor cell circumstances, but not control HCT116 cells (input = 10% of sample; FT = flow-through; IP-IgG and FT-IgG lanes are mock IPs with non-specific IgG; each IP-WB was repeated on a biologically independent sample; all uncropped gel images provided in Source Data file in Supplemental Information). c IP-WBs demonstrating reciprocal interactions in multiple tumor contexts between AT3 and MATR3 or MTA2 (n = 2 biological repeat not shown). d ASPSCR1, TFE3, and AT3 constructs oriented left-to-right:amino-to-carboxy. Gray background coded by ASPSCR1; slashes indicate amino acids 1-311. White background coded by TFE3; slashes indicate exon 4 through 3’-terminus; black coded by exon 3, the putative activation domain. e FLAG-IP-WB for MATR3, MTA2, and VCP in HEK cells transfected with FLAG-tagged constructs. Expression varied on n = 3 biological repeats: more AT3.2 was expressed in the iteration of the experiment depicted. f Fluorescence photomicrographs of a human ASPS tumor and control clear cell sarcoma (CCS) and Ewing sarcoma (ES) tumors stained with DAPI, TFE3 (detecting AT3), and VCP antibodies. (panel = 100 µm square; n > 4 fields per each of n = 4 tumors/controls). g Fluorescence images of HEK293T co-transfected with VCP-GFP and either ASPSCR1-TFE3-mRFP or mRFP alone. (panel = 30 µm square). Nuclear (nuc.) and cytoplasmic (cyt.) fraction western blots (WBs) for VCP (n > 4 fields per n = 2 biological repeated experiment). h Fluorescence photomicrographs of proximity ligation assay (PLA) between TFE3 and VCP antibodies in the ASPS tumor and controls, with an additional IgG-PLA control. (panel=100 µm square; n > 4 fields per each of n = 4 tumors/controls). i Quantitative fluorescent PLA signal per nucleus was measured in (79, 57, 143, 29, 105, 81, 83, 57, 216, 271, 215, 244, 84, 59, 195, 65 nuclei for the samples listed in order.).

Fig. 2. AT3 interacts with hexameric assemblies of VCP.

a WB for VCP after BN-PAGE of whole cell lysates from HEK293T cells after transfection with control GFP, ASPSCR1, AT3, truncated ASPSCR1 (trASPSCR1, the portion included in AT3) or truncated TFE3 (trTFE3, the portion included in AT3). VCP hexamer size is ~582 kD (n = 3 biological repeats performed with similar result). b Mass spectrometry results after FLAG-IP from HEK293T cells transfected with tagged ASPSCR1 or AT3, showing VCP predominant in both, other than the peptides aligned with the FLAG-tagged bait itself, noted in red as ASPSCR1 or both ASPSCR1 and TFE3. c Negative stain TEM micrographs of FLAG-IP eluates after ASPSCR1 or AT3 overexpression in HEK293T cells (bars = 20 nm; biological repeats n = 3). d Reference-free 2D class averages of negatively stained particles recovered from FLAG-AT3 co-IP reveal intact VCP hexameric assemblies. Left and middle, VCP top or pore views (3620 and 2774 particles in each class, respectively); right, VCP side view (182 particles; box length and width = 25 nm). e Schematic of constructs of each type of AT3 as well as CΔ, a truncated portion of the carboxy half of ASPSCR1, not included in AT3 (HA human influenza hemagglutinin tag; NLS nuclear localization signal). Orientation from top-to-bottom is amino-to-carboxy termini. f VCP WB after BN-PAGE following transfection of HEK293T cells with ASPSCR1, CΔ, or an empty vector control as well as the last of these transfections treated subsequently with VCP inhibitor CB-5083 (biological repeats n = 3). g VCP WB after BN-PAGE following IP for TFE3-FLAG or AT3-FLAG with or without co-transfected CΔ, along with denaturing gel blots of input proteins below (biological repeats n = 3). h Negative stain TEM micrographs of in vitro mixing of VCP with control green fluorescent protein (GFP) or CΔ, with AT3, added in the order of the molar ratios listed (bars = 20 nm; biological repeats n = 2).

Fig. 3. VCP co-localizes on chromatin to target loci of AT3.

a Heatmaps of overlapping enrichment in ChIP-seq with antibodies against AT3 and VCP in human FU-UR-1 cells, ASPS-1 cells, two human ASPS tumors, and 3 mouse ASPS tumors. Each heatmap is centered on peaks across the genome, sorted by AT3 enrichment, representing at least 2 replicates. b Example tracks with overlaid ChIP-seq enrichments of AT3 and VCP at selected highly enriched target genes (See also Supplementary Fig. 4a). Tracks scaled together 0 to 15 reads per million (RPM). c Graph depicting mean ± standard deviation as well as raw value points of fold-enrichment of double cross-linked VCP ChIP over input chromatin by quantitative polymerase chain reaction (qPCR) for two AT3-expressing cell lines (ASPS-1 and FU-UR-1) and two control cell lines (HCT116, colon cancer; ASKA, synovial sarcoma) at a panel of selected target genes and one negative control locus (CCND2). Two-tailed Student’s t-test generated p-values as indicated. d Similar graph of VCP ChIP enrichment over input for FU-UR-1 cells exposed for 48 h to control siSCR or one of two siTFE3s directed against AT3. e Similar graph of VCP ChIP enrichment over input for HEK293T cells transfected with GFP control, TFE3 control, ASPSCR1 control, AT3.1 or AT3.2.

Fig. 4. AT3:VCP interaction localizes to the promoters and enhancers of target genes.

a Graph and annotation histogram for AT3 ChIP-seq peaks called in 2 human ASPS tumors within highly enriched regions (definition Supplementary Fig. 4a). RPM heatmaps for distal (putative enhancer; 1404 peaks) and proximal promoter peaks (1079 peaks; H3K4me1 window ±10 kb, others ±3 kb), and enrichment plots (H3K27me3 non-significant), then mean differential expression of nearest genes in human tumors compared to muscle samples by RNAseq (dataset GSE54729, FDR = false discovery rate q-value), noting genes also annotated in mouse highly enriched AT3 regions and correlation human to mouse differential expression tumors over muscle. b Spearman rank correlation coefficients between AT3 ChIP-seq enrichment genome-wide compared to VCP, H3K27ac and H3K27me3, (For human tumors, S = 2.81 × 10¹³, 3.24 × 10¹³, 8.45 × 10¹³, respectively; all p-values < 2.2 × 10⁻¹⁶). c Dominant motifs in peaks co-enriched for AT3, H3K27ac, and RNAPOL2. d Example ChIP-seq tracks from human tumors (n = 2, pooled) showing the region and 10-fold-magnified view (genomic distance scale noted for each). Vertical track scales are AT3: 0–4, VCP: 0–2, H3K27ac: 0–5, H3K4me1: 0–1, H3K36me3: 0–1, RNAPOL2: 0–6, H3K27me3: 0–0.2. e Example ChIP-seq tracks for Ctsd and Hif1a in mouse AT3-induced tumors. Vertical track scales AT3: 0–10, VCP: 0–4, H3K27ac: 0–9, H3K4me1: 0–2, H3K36me3: 0–2, RNAPOL2: 0–10, H3K27me3: 0–2. f Plots of mean fold-change expression in human ASPS tumors over muscle (25th to 75th percentile boxes and error bars of standard deviation, outliers depicted individually; n = 5, 3, respectively) for each number of peaks associated by nearest-gene annotations. (two-tailed heteroscedastic t-test p-value comparing 1–2 peaks to 3 or more). g Differential gene expression following 48-h siRNA depletion of AT3 in ASPS-1 cells. Blue dots represent genes annotated by highly enriched regions (from Supplementary Fig. 4a). Genes noted in black are those annotated by AT3 peaks generally. Gray dots have no associated AT3 ChIP-seq peaks. (Wald test p-values, Benjamini-Hochberg correction.) h Enrichr plot of Reactome gene sets enriched in both FU-UR-1 and ASPS-1 direct target genes of highly enriched areas and downregulated on AT3 depletion in each cell line. Specific Reactomes were assigned to general categories.

Fig. 5. VCP presence, hexamerization, and ATPase activity impact AT3-related transcription.

a Correlation of fold-change (FC; to control) FU-UR-1 RNA-seq transcription comparing siRNAs that deplete VCP or AT3 among target genes (highly enriched AT3 region genes with >2FC by AT3 depletion. (n = 3 × 2 siRNAs each for VCP and AT3, pooled.) b Similar targets correlation for ASPS-1 (also n = 3 × 2 pooling). c RT-qPCR for target genes (controls HPRT, VCP) for HEK293T expressing TFE3, type 1 or 2 ASPSCR1::TFE3 (AT3.1, AT3.2), each with or without exogenous VCP (VCP^ex) overexpression. (n = 3, pooled; mean ± standard deviation; homoscedastic two-tailed t-tests comparing added VCP^ex to each alone.) d Genes plotted had at least 8-fold upregulated expression by RNA-seq for TFE3, AT3.1, and AT3.2 relative to control GFP in the absence of CΔ in HEK293T cells. The plot indicates log transformed p-values and FC comparing each transcription factor with co-transfected CΔ to that without CΔ. Co-transfected CΔ to  HEK293T cells changes TFE3 targets in both directions, mostly insignificantly. CΔ significantly downregulates most AT3.1/AT3.2 targets (n = 3 each group, homoscedastic two-tailed t-tests). e Less pronounced correlation between the differential expression by RNA-seq in ASPS-1 from CB-5083 relative to vehicle at 48 h (n = 3 for each) to siRNAs depleting VCP relative to control at 48 h (n = 3 × 2 pooled for each). f The same for FU-UR-1 cells (for CB-5083 experiment, n = 2 for each). g Plot of log-transformed p-values and FCs of CB-5083 or DMSO treated HEK293T cells transfected with TFE3. The genes included were 8-fold upregulated by TFE3, AT3.1 and AT3.2 over GFP control transfection. Most significant changes showed further upregulation by CB-5083. (n = 3, pooled, homoscedastic two-tailed t-tests). h Genes upregulated by CB-5083 added to TFE3 are mostly downregulated by CB-5083 added to AT3.1 or AT3.2. (n = 3, pooled, homoscedastic two-tailed t-tests). i Schematic working model where VCP hexamers interact with chromatin via AT3 to activate transcription, abrogated by VCP depletion, hexamer disassembly, or enzymatic inhibition, the last with mixed impacts. j RT-qPCR for target genes after TFE3, AT3.1, or AT3.2 expression in HEK293T with co-transfection of control, VCP, or VCP^E305Q (hexamer-assembling, enzymatically inactive mutant), which increases some targets’ expression (n = 3, pooled; mean ± st.dev.; homoscedastic two-tailed t-tests).

Fig. 6. Chromatin conformation enhancer loops depend on AT3:VCP.

a Correlation of AT3 and VCP enrichments aggregated for each target gene in ASPS-1 cells from all narrowly defined peaks localized to the promoter or a looped enhancer (normalized by H3K27ac-HiChIP loop score). b Similar FU-UR-1 plot. c Correlation between log₂fold-change expression from siRNA VCP depletion over control in FU-UR-1 cells, with AT3 depletion or CB-5083 VCPi for the 50 strongest promoter-targeted genes and 50 strongest enhancer-targeted genes. d Principal component (PC) analysis according to the top one percent (n = 99, by AT3 aggregate enrichment) of genes by RNA-seq after siSCR, siTFE3, siVCP, or combinations in FU-UR-1 cells. e K-means cluster heatmaps of PC1-contributing genes (coefficients greater than 0.05, n = 54) in FU-UR-1 cells, showing expression after application of siRNAs or CB-5083 in FU-UR-1 cells as well as expression in human and mouse ASPS tumors compared to skeletal muscle. f Example tracks of H3K27ac-HiChIP loops around a highly targeted gene in baseline FU-UR-1 cells (above, black) and FU-UR-1 cells exposed to siVCP for 6 days (below, cyan; n = 2 siRNAs against VCP). g Negative control example tracks show retained/strengthened H3K27ac-HiChIP loops after filtering out VCP-ChIP-seq-peak-associated loops. h Comparison dot diagrams of the change in VCP-peak-associated or not-associated H3K27ac-HiChIP loops at baseline compared to 6 days of VCP depletion with siVCP (n = 2 for each). i Normalized loop score (NLS) aggregated per promoter for VCP-associated genes in FU-UR-1 cells at baseline (from 6b) or 6 days VCP depletion. Genes with no change or increased NLS per promoter are designated as VCP-associated or not. j Plot of differential H3K27ac ChIP-seq enrichment in FU-UR-1 cells with siRNA-depleted AT3 (left) or VCP (right) at loci defined as VCP-peaks and HiChIP loop ends that are top-ranked for AT3 enrichment as well (above) versus the lowest AT3 enriched HiChIP loop H3K27ac ChIP peaks that are not coincident with VCP peaks (below). (n = 3, biologically independent samples, Wald test p-values, Benjamini-Hochberg correction.).

Fig. 7. AT3:VCP is a targetable functional dependency in cancer cells in vitro.

a Flow-cytometry for CFSE dye depletion by proliferation after siSCR, siTFE3, siVCP and combinations in FU-UR-1 cells. (siRNA#1 for each is represented by a solid line; siRNA#2 for each is represented by a dotted line.) b Clonogenic assays after siRNAs applied to both the FU-UR-1 cell line and the ASPS-1 cell line (black diamonds represent an siTFE3 that targets a portion of TFE3 not included in AT3). c CFSE dye depletion assay combining siRNA depletion of AT3 by siTFE3#1 (solid line) and siTFE3#2 (dotted line) or control siSCR with CB-5083 or vehicle, showing that AT3 depletion or CB-5083 diminishes proliferation (AT3 depletion slightly more than CB-5083), but that their combination further diminishes proliferation (pushes distribution further to the right). (siRNA#1 for each is represented by a solid line; siRNA#2 for each is represented by a dotted line.) d The same for siVCP#1 (solid line) and siVCP#2 (dotted line) or control siSCR, with CB-5083 or vehicle, showing the same phenomenon. e Schematic of the Rosa26-LSL-AT3-IRES-eGFP allele that is activated by Cre-recombinase-mediated excision of a stop sequence between the AT3 coding sequence and the native Rosa26 promoter. This allele, when activated in living mice led to the fully penetrant development of ASPS tumors. f Flow-cytometry for CFSE dye depletion by proliferation in GFP+ sorted mouse embryonic fibroblasts (MEFs) that bear an ASPSCR1-TFE3 allele that is induced by Cre-mediated recombination to remove a stop sequence between the promoter and the cDNA coding sequence. (Cells counted at day 0 were 9,073; ells counted at day 4 were 20,258 for TATCre added alone, 20,241 for TATCre and VCP-GFP added, 19,982 for VCP-GFP added alone, and 14,346 for GFP control added alone).

Fig. 8. AT3:VCP is a targetable functional dependency in cancer cells in vivo.

a Time to tumorigenesis (100 mm³ calculated size) of FUUR-1 cells, transfected and selected for IPTG-inducible shRNAs, xenografted into NRG mice (1 million cells per flank in Matrigel). Drinking water IPTG began on day 0, three days after xenograft injection. (n = 4 for each transfected population of cells, in which each shRNA has n = 3. Mean ± standard deviation, two-tailed homoscedastic t-test p-values. Open diamonds represent tumors <100 mm³ at 21 days, counted that days for statistics). b Growth curves for the same transfected xenografts (n = 12 per shRNA, mean ± standard deviation, two-tailed homoscedastic t-test p-value). c Final mass for each transfected xenograft tumor harvested (day 21, mean ± standard deviation, two-tailed homoscedastic t-test p-value). d Growth of FU-UR-1 xenografts (NRG mice) treated with CB-5083 at 100 mg/kg/day for 4 days per week versus vehicle control (harvested day 8, due to gavage toxicity) or at 50 mg/kg/day for 4 days per week versus vehicle control (n = 10 tumors per group; homoscedastic t-test p-values). e Immunofluorescence Ki-67 photomicrographs of FU-UR-1 xenografts harvested day 8 after treatment with 100 mg/kg CB-5083 or vehicle control, administered on days 1–4 and 8 (magnification bars, 50 µm). f Patient-derived xenograft growth data by luciferase activity on weekly measurements (n = 10, Tukey’s test p-values) g Fractional growth curves for individual tumors in genetically induced mouse ASPSs, randomized to treatments at volumes >10 mm³ by serial MRI (n = 13, CB-5083; n = 15, vehicle). h RT-qPCR for target genes (defined as mouse AT3-ChIP-seq targeted and having reduced expression of homologues in both FU-UR-1 and ASPS-1 cells with AT3 depletion) in control tumors (OS osteosarcoma, SS synovial sarcoma) or genetically induced mouse ASPS tumors treated with vehicle or CB-5083 for 50 mg/kg for 4 days prior to harvest. (n = 4 tumors per group per gene; homoscedastic two-tailed t-test p-values).