Introduction

Solid malignancies associated with structural variants involving nuclear protein in testis member 1 (NUTM1 or NUT) are a rare and under-recognized group of tumors. Originally they were identified as an extremely aggressive subset of poorly differentiated squamous cell carcinomas, termed NUT carcinomas (NC), predominantly located in the midline of the head, neck, and mediastinum, and refractory to chemotherapy [1]. In NCs NUTM1 is fused to the genes for bromodomain (BRD)-containing chromatin-interacting proteins BRD4 (70–80% of cases) or BRD3 (10%), or genes for BRD-associated proteins including NSD3, ZNF532, and ZNF59 in another 10% of cases [1]. However, the identification of tumors with novel NUTM1 fusion partners has expanded the morphologic and genetic spectrum of NUTM1-rearranged neoplasms (NRNs) [reviewed in [2]]. Some NRNs have small round blue, spindle, rhabdoid and/or plasmacytoid cell morphology, lack evidence of squamous or epithelial differentiation, and are probably best classified as sarcomas [3]. One class of these sarcoma or sarcoma-like NRNs harbors NUTM1 fusions to members of the MAX Dimerization (MAD) gene family [3,4,5,6,7,8].

Here we describe a small round cell gastroesophageal junction malignancy harboring a novel fusion between NUTM1 and MXI1 (also known as MXD2), a member of the MAD family of MYC antagonists. We present in vivo and in vitro evidence to support a fundamentally different oncogenic mechanism for MXI1-NUTM1 when compared to a typical NC-associated fusion, BRD3-NUTM1. We discuss the implications of our findings to the diagnosis, clinical course and treatment of NRNs.

Materials and methods

Targeted RNA sequencing

Tumor RNA was extracted from FFPE slides using the AllPrep DNA/RNA FFPE kit (Qiagen, Hilden, Germany). Targeted fusion analysis was performed using Anchored Multiplex PCR (AMPTM) [9] with the Archer FusionPlexTM Oncology Research Panel (ArcherDx, Boulder, CO). cDNA libraries were prepared and sequenced on a MiSeq sequencer (Illumina, San Diego, CA) as per the manufacturer’s recommendations. FASTQ files were processed using ArcherDx analysis software to annotate gene fusions and variants (http://analysis.archerdx.com/). For RT-PCR and Sanger sequencing, cDNA was synthesized from RNA using the SuperScript™ VILO™ cDNA Synthesis Kit (Invitrogen, Carlsbad, CA). PCR was performed using AmpliTaq Gold (ThermoFisher Scientific) with priming sites located in MXI1 exon 5 (5′-GCTCGAGAATTTGGAACGAG-3′) and NUTM1 exon 3 (5′-CTGGTGGGTCAGAAGTTGGT-3′). PCR product was analyzed on a 2% agarose gel and purified with ExoSAP-IT (ThermoFisher Scientific). Sequencing reactions were performed using the BigDye Terminator v3.1 Cycle Sequencing Kit (ThermoFisher Scientific) and sequencing was conducted on a 3730xl DNA Analyzer (Applied Biosystems, Foster City, CA). Sequencing results were analyzed using Mutation Surveyor software (SoftGenetics, State College, PA). TruSight RNA Fusion and TruSight RNA Pan-Cancer Panel (Illumina, San Diego, CA) cDNA libraries were both prepared and sequenced with a MiSeq sequencer (Illumina, San Diego, CA) as per the manufacturer’s recommendations. FASTQ files were analyzed with the RNA-Seq Alignment application on the BaseSpace Sequence Hub (Illumina, San Diego, CA).

Targeted DNA sequencing

Targeted sequencing comprised the analysis of 2.34 Mb genomic regions implicated in cancer, including sequencing of the entire coding regions of 386 genes as well as selected intronic regions (including NUTM1) of genes known to be involved in oncogenic fusions. KAPA Hyper (Roche) libraries were prepared and target enriched using SureSelectXT hybridization (Agilent) [10]. Pooled libraries were sequenced at 500× mean coverage on an Illumina NextSeq sequencer (Illumina, San Diego, CA, USA) using paired 75 bp reads. Variants were detected using combined VarScan [11], Mutect [12], and ugSomatic [13] variant callers. Copy number variations were detected using CNspector [14].

Immunohistochemistry

Diagnostic immunohistochemistry (IHC) was performed on sections from FFPE primary tumor tissues as well as on cell blocks of in vitro cultured cells (see below). See Supplementary material for full methods.

Transcriptome analysis by RNA sequencing (RNAseq)

RNA was extracted from FFPE tissue as described above. cDNA libraries were prepared using the NEBNext® Ultra™ II RNA Library Prep Kit (NEB, Ipswich, Mass) with ribo-depletion and random priming. Libraries were sequenced on an Illumina NextSeq sequencer (Illumina, San Diego, CA, USA) using paired 75 bp reads. Adapter sequences, primers, and too-short-read sequences were removed using Cutadapt v1.9.1 [15]. Reads were subsequently aligned to human reference genome hg19 using the HISAT2 v2.0.4 read-mapping algorithm [16] and HTSEQ v0.11.2 [17] for gene expression quantification. Data normalization was performed using Voom-limma R package [18]. Reads were subsequently aligned to human reference genome hg19 using the HISAT2 v2.0.4 read-mapping algorithm and gene abundance was quantified using HTSEQ v0.11.2. Library size normalization on the samples was performed using the TMM method [18]. MXI1-NUTM1 and BRD3-NUTM1 fusions were detected using JAFFA [19] and Arriba [20].

Gene set enrichment analysis was performed using GSEA [21] preranked mode, using Hallmark MSigDB gene sets. Single sample enrichment analysis were run using ranking based on transcript abundance quantification calculated Salmon [22] which accounts for transcript length bias. Comparative enrichment analysis were run using ranking based on log fold-change of the compared samples.

In vitro cells expressing MXI1-NUTM1 or MYC

MXI1-NUTM1 nucleotide sequences were synthesized and ligated into pcDNA3.1(+)-N-HA by GenScript Ltd (Hong Kong). Retrovirus vectors were constructed for HA-MXI1-NUTM1, MYC and human oncogenic HRAS (encoding HRAS with a glycine to valine substitution at amino acid 12 [23]). NIH-3T3 cells were maintained in DMEM (Gibco) + 10% FBS (HyClone). For retroviral transductions 2.5 × 105 NIH-3T3 mouse fibroblasts were seeded into 6 cm dishes. The following day they were infected with either pMSCVneo empty vector (EV), pMSCVneo HA-MXI1-NUTM1 or pMSCVpuro MYC diluted 1:5 in NIH-3T3 media with 8 μg/mL polybrene (Sigma-Aldrich). Some cells were simultaneously infected with pMSCV HRAS-IRES-GFP. Cells were re-infected the following day then put under selection with 800 μg/mL geneticin (Thermo Fisher Scientific) or 1.5 μg/mL puromycin (Sigma-Aldrich) for 1 week. Oncogenic HRAS-positive cells were selected by isolation of GFP-positive cells with a BD FACS Aria III and BD FACS DIVA software.

Immunoblotting

Cells were lysed in RIPA buffer (50 mM Tris-HCl pH 7.4, 1% NP-40, 0.5% sodium deoxycholate, 0.1% SDS, 1% glycerol, 137.5 mM NaCl) and quantified using the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific) according to the manufacturer’s instructions. Protein lysates (20–30 μg) were prepared with NuPage loading buffer and reducing agent (Life Technologies) and denatured by heating at 85 °C for 5 min. Samples were loaded onto 4–12% Bis/Tris or 3-8% Tris-Acetate gels (Life Technologies) and run using MOPS or Tris-Acetate running buffers (Life Technologies). Protein was transferred to PVDF membrane (Merck Millipore) using BioRad transfer modules. Membranes were blocked in blocking buffer (LiCOR) for 1 h at room temperature. Primary antibody diluted in 5% BSA/TBS was added to the membranes over night at 4 °C. Antibodies used were HA-Tag (1:1000, CST #3724), MYC (1:1000, Epitomics #1472), HRAS (1:2000, Merck #05-516) and β-actin (1:1000, Sigma-Aldrich #A5441). Fluorescent secondary antibody conjugated to IRDye680 or IRDye800 (LiCOR) diluted in Odyssey blocking buffer (1:15,000–1:20,000) were used for detection. An Odyssey CLx Infrared Imaging System (LiCOR) was used to image the immunoblots.

Cell Titer Glo assay

Cells were seeded into white-walled 96-well plates with 2000 cells per well in 100 μL media with either 1 or 10% FBS. Blank wells with no cells were included to determine background luminescence. Cells were allowed to attach for 4 h, and then baseline measurements were determined using the CellTiter-Glo luminescent assay (Promega) as per the manufacturer’s instructions. Briefly, 100 μL CellTiter-Glo buffer was added to each well, plates were placed on an orbital shaker for 2 min, then incubated for 10 min at room temperature protected from the light. Luminescent readings were obtained using a FLUOstar Omega microplate reader (BMG Labtech). Background luminescence values were subtracted from each well. Measurements were taken in the same way 2, 4, and 5 days later and results were expressed as a fold-change relative to day 0.

Incucyte confluence and colony formation assay

NIH-3T3 cells were seeded in 24-well plates with 12,000 cells per well in 1 mL media in duplicate or triplicate. Cells were imaged using an IncyCyte ZOOM live cell imager (zoom40061, Essen BioScience) over a period of 6 days. Images over nine fields of view per well were acquired every 4 h. Phase images were used for calculating cell confluency for EV, MXI1-NUTM1, and MYC cells. GFP images at end point were used for calculating the number of colonies in EV + HRAS, MXI1-NUTM1 + HRAS, and MYC + HRAS transfected cells over five replicate experiments. The average colony counts for each of the transformants from each replicate was used for statistical analysis.

Soft agar colony formation assay

The soft agar colony formation assay was performed as previously described [24]. Briefly, wells of a 12-well plate were coated with 1 mL 0.5% agarose (Lonza Bioscience, Basel) in NIH-3T3 media. Gels were allowed to solidify for 30 min at room temperature, following which 1 mL of a single cell suspension of NIH-3T3 cells at a density of 1000 cells/mL in media containing 0.3% agarose was added. Once set, 500 μL of NIH-3T3 media was added to the wells and plates were incubated at 37 °C, 5% CO2. Fresh media was added to the wells every 2–3 days for 12 days. Colonies were fixed and stained in 0.1% crystal violet solution (Sigma-Aldrich) containing 10% ethanol for 15 min at room temperature. Gels were de-stained by rinsing five times in distilled water for 5 min each. Colonies were imaged using a Leica MZ12 dissecting microscope connected to a Leica DFC420C camera using LAS software. Colonies were quantified using FIJI image analysis software [25]. The number of colonies, average colony area, and total colony area was determined for 19 (EV + HRAS) or 18 (MXI1-NUTM1 + HRAS and MYC + HRAS) wells over six (EV + HRAS and MXI1-NUT + HRAS) or 5 (MYC + HRAS) replicate experiments.

Results

Clinical course

A 51-year-old woman, a previous smoker, presented with progressive dysphagia and weight loss. On endoscopy she was found to have a stricture of the lower esophagus and abnormal mucosa extending to the distal stomach. Biopsies showed infiltration of the lamina propria of the esophagus by undifferentiated small cells that expressed NUTM1 by IHC. Staging computed tomography, FDG-PET, and laparoscopy demonstrated a bulky tumor of the distal esophagus, gastroesophageal junction, and lesser curvature of stomach. Nodal, peritoneal, or distant metastases were not identified. The patient underwent an Ivor Lewis esophagectomy with total gastrectomy, colonic interposition, and Roux-en-Y jejunal reconstruction. The patient had a prolonged hospital admission due to post-operative complications. The patient was not eligible for clinical trials and was too unwell to receive chemotherapy, radiotherapy, or any specific targeted therapies. She died 7 months post-surgery from disease progression.

Tumor morphology

Gross examination of the resected specimen showed a near-circumferential tumor centered in the stomach and extending directly into the distal esophagus, a total length of 197 mm (Fig. 1a, b). It had a fleshy, cream-tan cut surface, undermined the normal mucosa and infiltrated the full thickness of the esophageal and gastric walls. Sections (Fig. 1c–e) showed that the tumor formed nests, sheets and strands of discohesive, monotonous, small to intermediate sized cells associated with a small amount of myxoid stroma. Tumor cells had ovoid nuclei with vesicular chromatin, single small but prominent nucleoli, and a small to moderate amount of eosinophilic cytoplasm, some with a vague rhabdoid appearance. There were occasional mitotic figures (2/10 hpf) and apoptotic bodies, as well as extensive lymphovascular, perineural, and intraneural invasion. It somewhat resembled the morphology of a NUT carcinoma. However, no squamous or other epithelial differentiation was seen (Fig. 1f). Metastatic tumor was present in multiple para-esophageal, lesser curve, greater curve, and subcarinal lymph nodes.

Fig. 1: Tumor morphology.
figure 1

Macroscopic photographs of the esophagectomy specimen from the MXI1-NUTM1 tumor with the mucosal (a opened longitudinally, proximal at the top of image) and cut (b) surfaces showing diffuse wall thickening. The serosa is inked black. H&E stained sections from the MXI1-NUTM1 tumor showing tumor cells invading the esophageal (c squamous mucosa at bottom of image) and gastric (d, e) walls. Note the tumor infiltrating around benign gastric glands (d) and invading the muscularis propria (e) f H&E stained section from a BRD3-NUTM1 carcinoma. Note the similar morphology to the MXI1-NUTM1 tumor (compare with d, e).

Full size image

RNA and DNA sequencing of the MXI1-NUTM1 tumor

Anchored Multiplex PCR detected chimeric transcripts involving MXI1 exon 5 (NM_005962.4) fused to NUTM1 exon 3 (NM_175741.2). These were in frame, and were confirmed by RT-PCR/Sanger sequencing (Fig. 2a) and RNAseq (Fig. 2b, see below). The fusion was subsequently also identified by Illumina TruSight RNA Fusion and RNA Pan-Cancer Panel analysis (not shown). The MXI1-NUTM1 fusion protein was predicted to possess the basic region/helix-loop-helix/leucine zipper domain of MXI1, which mediates its heterodimerization with MAX and sequence specific DNA-binding, as well as the acidic domain (AD1) of NUTM1 that is capable of binding and activating the histone acetyltransferase (HAT) p300 [26]. Other variants detected are shown in Supplementary materials.

Fig. 2: Identification of the MXI1-NUTM1 fusion.
figure 2

MXI1-NUTM1 fusion between MXI1 exon 5 (NM_005962.4) and NUTM1 exon 3 (NM_175741.2) confirmed by a Sanger sequencing of the RT-PCR product, and b RNA whole transcriptome analysis.

Full size image

Protein expression comparison of the MXI1-NUTM1 tumor and a BRD3-NUTM1 NC

Characterization of the MXI1-NUTM1 tumor by IHC showed nuclear expression of NUTM1 similar to that in a NC with a BRD3-NUTM1 fusion (Fig. 3). A “speckled” intra-nuclear pattern noted by others in NCs [5, 27] and thought to represent hyperacetylated chromatin “mega-domains” was not observed in either tumor by IHC or immunofluorescence (not shown). The MXI1-NUTM1 tumor also showed FLI-1 expression in most cells, focally CD99 membrane staining, and weak patchy NSE expression (not shown). It was negative for almost all cytokeratins tested (AE1/AE3, CK5/6, 34betaE12, and CK7), as well as EMA and BerEP4, all of which were robustly expressed in the BRD3-NUTM1 NC. The low molecular weight cytokeratin marker CAM5.2 also strongly stained the BRD3-NUTM1 NC but only showed focal faint expression in the MXI1-NUTM1 tumor. p40 (∆Np63) is a marker of squamous differentiation, and both MYC and TP63 are key transcriptional targets of BRD-NUTM1 fusion proteins [28, 29]. Notably, the MXI1-NUTM1 tumor cells showed no staining for MYC or p40. These were both readily detectable in the squamous cells in overlying esophageal epithelium, and in the BRD3-NUTM1 NC. SOX2 has also been shown to play a major role in BRD4-NUTM1-associated oncogenesis [30]. SOX2 protein was expressed in the BRD3-NUTM1 NC, but not in the MXI1-NUTM1 tumor. The results of other immunohistochemistry stains are discussed in Supplementary materials.

Fig. 3: Protein expression in the MXI1-NUTM1 tumor and a BRD3-NUTM1 carcinoma.
figure 3

Representative paired images of protein expression by IHC in the MXI1-NUTM1 tumor (a–f) and a BRD3-NUTM1 carcinoma (a′–f′). Both tumors express NUTM1. The BRD3-NUTM1 carcinoma, but not the MXI1-NUTM1 tumor, expresses pan-cytokeratin AE1/AE3, p40, MYC, and SOX2. The MXI1-NUTM1 tumor, but not the BRD3-NUTM1 carcinoma, weakly expresses ER. Arrows show benign esophageal squamous epithelium in the MXI1-NUMT1 tumor expressing AE1/AE3, p40, MYC, and SOX2, but not ER.

Full size image

Transcriptome comparison of the MXI1-NUTM1 tumor with a BRD3-NUTM1 NC

RNAseq confirmed expression of genes directly involved in the oncogenic fusions (BRD3, MXI1, and NUTM1) (Fig. 4a) as well as chimeric MXI1-NUTM1 and BRD3-NUTM1 transcripts in the respective tumors. Expression of multiple high and low molecular weight keratins were detected in the BRD3-NUTM1 NC (Fig. 4a) and in benign esophageal squamous epithelium adjacent to the MXI1-NUTM1 tumor (not shown). In contrast, only expression of low molecular weight cytokeratins KRT8 and KRT18 were detected in a sample of the MXI1-NUTM1 tumor that was macrodissected away from esophagogastric epithelium, with low levels of the high molecular weight keratin KRT10. Whilst we cannot dismiss that some of these transcripts were from mesothelium, this was consistent with the weak staining for CAM5.2 by IHC in MXI1-NUTM1 tumor cells, as this clone reacts with KRT8. RNAseq confirmed that the key NC drivers MYC, TP63, and SOX2 were highly expressed in the BRD3-NUTM1 NC, but showed negligible expression in the MXI1-NUTM1 tumor, consistent with the IHC results (Fig. 4a). Other transcriptional changes are documented in Supplementary materials.

Fig. 4: Transcriptome analysis of the MXI1-NUTM1 tumor and a BRD3-NUTM1 carcinoma.
figure 4

a Scatterplot of normalized RNA expression. Genes known to be critical for oncogenesis in NC (MYC, TP63 and SOX2) are labeled green, keratins blue, genes involved in the fusions orange, and other genes discussed in the main text and supplementary material purple. Hallmark MYC Targets v1 analysis (b) and normalized enrichment scores (c). d MYC transcript expression (normalized data, converted into base 10 from log2 data in Fig. 4a).

Full size image

The MXI1-NUTM1 tumor has a “MYC-like” transcription signature in the absence of MYC

For BRD-NUTM1 fusion proteins the oncogenic mechanism is thought to involve a “feed-forward” process that results from the histone binding activity of BRD and the histone acetyltransferase activity of the NUTM1-binding protein p300. Binding of BRD-NUTM1 to acetylated histones localized at pro-proliferation genes leads to a cycle of further histone acetylation and BRD-NUTM1 binding. This process sequesters p300 away from pro-differentiation genes, resulting in their hypoacetylation [31]. The net effect is histone hyperacetylation and increased transcription of pro-proliferation genes, and silencing of pro-differentiation genes. However, since MAD-NUTM1 fusions have no known interaction with BRD complexes we have previously hypothesized that they are unlikely to exert their oncogenic effects through this feed-forward mechanism. Instead we have suggested that MAD-NUTM1 fusions will dimerize with MAX through the MAX-interacting domain of the MAD partner, and that these heterodimers would bind to E-boxes near MYC target genes and upregulate gene expression. This model predicts that increased expression of MYC target genes could occur in the absence of MYC itself [2].

In order to test this hypothesis we analyzed the MXI1-NUTM1 tumor for a MYC transcriptional signature [32]. We found a hallmark MYC target gene expression signature in both the MXI1-NUTM1 tumor (normalized enrichment score 2.43, p < 0.001) and the BRD3-NUTM1 NC (normalized enrichment score 2.64, p < 0.001) both with a false discovery rate (FDR) of <0.001 (Fig. 4b, c). MYC protein expression was undetectable by IHC in the MXI1-NUTM1 tumor (Fig. 3d) and MYC transcripts in the normalized samples were ~23-fold less than in the BRD3-NUTM1 NC, most consistent with a origin of the signature in the MXI1-NUTM1 tumor from a non-MYC source. However, we had observed some MYC expression in some endothelial cells and stromal cells within the MXI1-NUTM1 tumor by IHC, and in a sample comprised of benign esophageal wall adjacent to the tumor there was a MYC transcriptional signature (not shown). This raised the possibility that benign cells and not tumor cells were in part responsible for the MYC signature in the MXI1-NUTM1 tumor sample. To control for this potential effect we performed comparative enrichment analysis based on log fold-change of MXI1-NUTM1 tumor versus normal esophageal wall transcriptome data. We estimated tumor cell purity to be about 80% in the MXI1-NUTM1 sample, but nevertheless if the MYC signature was derived in part or even wholly from benign stromal cells then the FDR from a hallmark MYC target gene analysis of the comparison should not reach significance. In fact a MYC transcriptional signature was still readily detected in this comparison analysis (normalized enrichment score 3.79, p < 0.001, FDR < 0.001). This argues strongly that the MXI1-NUTM1 tumor cells were the source of the MYC target signature.

MXI1-NUTM1 partially phenocopies the effect of MYC on cell proliferation and anchorage-independent growth in vitro

The contribution of MYC to oncogenesis is complex, but it has important and well-described roles in promoting cell proliferation and anchorage-independent growth in vitro. In order to determine if MXI1-NUTM1 also shared these MYC-type roles we transfected NIH-3T3 mouse fibroblasts with vectors constitutively expressing MXI1-NUTM1 and performed cell growth assays. We used MYC-expressing cells as a positive control. Expression of MXI1-NUTM1 and MYC was confirmed with immunoblots (Fig. 5a; the MXI1-NUTM1 was HA-tagged in order to facilitate its detection). In low serum (1%) conditions cell numbers were modestly but significantly increased in MXI1-NUTM1-expressing cells compared to empty vector (EV) controls as assessed by both ATP abundance (Cell Titer Glo assays) and Incucyte cell confluency assays (Fig. 5b, c). We noted similar, but greater increases in both ATP abundance and cell confluency in MYC-expressing cells (Fig. 5b, c). ATP concentrations fell rapidly after day 2 in MYC-expressing cells, probably due to exhaustion of nutrients.

Fig. 5: MXI1-NUTM1 effect on cell proliferation in vitro.
figure 5

a Immunoblots with anti-hemagglutinin (HA), MYC, HRAS, and β-actin (loading control) antibodies on cell lysates from NIH-3T3 cells transduced with combinations of empty vector (EV), HA-MXI1-NUTM1, MYC, and oncogenic HRAS expressing retroviruses. A duplicate HA-MXI1-NUTM1-only lane has been cropped from the immunoblots for clarity. ATP abundance (b Cell Titer Glo assays) and cell confluency (c Incucyte assay) in NIH-3T3 cells transduced with either empty vector (EV), HA-MXI1-NUTM1 or MYC (1% serum conditions). All data are expressed as the mean + SD.

Full size image

MYC synergizes with activated RAS/MAPK pathway in rodent cells to produce anchorage-independent growth through suppression of senescence and apoptosis, key hallmarks of oncogenic transformation [33]. We performed soft agar colony formation assays in NIH-3T3 mouse fibroblasts cells to test if MXI1-NUTM1 could similarly cooperate with a constitutively active oncogenic HRAS (G12V) to promote anchorage independent growth. High levels of HRAS and pERK1/2 were confirmed in these cells (Fig. 5a and not shown). HRAS-expressing cells grew in aggregates that confounded the confluency assay, and were difficult to lyse, confounding the Cell Titer Glo assay. However, well-formed colonies grew in both adherent culture and semi-solid soft agar assays. Colony numbers were modestly but significantly increased in MXI1-NUTM1 + HRAS cells when compared with EV + HRAS control cells both on plastic (1.3-fold increase, p = 0.0047; Fig. 6a), and in the semi-solid soft agar assay (1.2-fold increase, p = 0.0029; Fig. 6b). Mean and total colony area were also significantly increased in the soft agar assays in MXI1-NUTM1 + HRAS cells (1.2-fold increase, p = 0.0002 and 1.4-fold p < 0.0001, respectively; Fig. 6b). As a positive control for these experiments we also overexpressed MYC together with oncogenic HRAS. This resulted in a 2.7-fold increase in the number of colonies grown on plastic (p < 0.0001, Fig. 6a). In the semi-solid soft agar assay MYC + HRAS increased colony numbers by 1.9-fold (p < 0.0001), mean colony area by 1.3-fold (p < 0.0001) and the total colony area by 2.6-fold (p < 0.0001) when compared with EV + HRAS control cells (Fig. 6b).

Fig. 6: MXI1-NUTM1 effect on anchorage-independent growth in vitro.
figure 6

NIH-3T3 cells transduced with oncogenic HRAS (from a HRAS-IRES-GFP vector) and either empty vector (EV), HA-MXI1-NUTM1, or MYC. a Cells grown on plastic substrate in 10% serum, showing box and whisker plots of colony number and representative images of colonies (visualized by GFP fluorescence). b Cells grown in soft agar, 10% serum. Box and whisker plots for colony number, area per colony and total colony area. c Images of colonies from cells grown in soft agar, 10% serum (top row) and IHC for NUTM1 and MYC on paraffin blocks of cells grown on plastic in 10% serum. Statistical comparisons were made using unpaired two-tailed t-tests. Data were analyzed using GraphPad Prism 8.1.0. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Full size image

The apparently greater potency of MYC compared to MXI1-NUT did not appear to be due to increased cell proliferation or decreased apoptosis as assessed by Ki67 and cleaved caspase-3 labeling performed on cell blocks, respectively (not shown). In part, it was likely due to epigenetic silencing of the MXI1-NUTM1 construct over time, as by immunoblot we observed a marked reduction of MXI1-NUTM1 protein expression with increasing passage number (not shown). IHC of cell blocks showed that only ~40% MXI1-NUTM1 + HRAS cells expressed detectable NUTM1, which could also explain why the effects of MXI1-NUTM1 were modest when compared to MYC (Fig. 6c). Unfortunately, we were unable to restore MXI1-NUTM1 expression by re-selection with antibiotics, and progressive silencing rendered further experiments untenable. Therefore, expression of MXI1-NUTM1 partially recapitulated the MYC-induced increase in proliferation and anchorage-independent growth in concert with oncogenic HRAS.

Discussion

MAD-NUTM1 tumors are probably best classified as sarcomas. They have been reported in soft tissue as well as viscera [3] and can metastasize to lymph nodes and lung. The MXI1-NUTM1 tumor described here showed similar features to previously described MAD-NUTM1 tumors, with small round cells and rhabdoid cell morphology, and no specific differentiation lineage by IHC and transcriptome profiling. NCs also have a primitive small round cell component, but in contrast to MAD-NUTM1 tumors they typically show small foci of squamous differentiation. These foci may be infrequent and therefore not be identified in small biopsies. The MXI1-NUTM1 tumor did not express typical markers of NCs including cytokeratins and p40 (∆Np63). For example, a recent paper documented expression of cytokeratins or p40/p63 in 17 of 19 NCs [5]. Furthermore, oncogenic drivers of NCs, including MYC, TP63, and SOX2 [28,29,30], were not expressed in the MXI1-NUTM1 tumor. MYC, TP63, and SOX2 provide a direct link between the BRD-NUTM1 fusion and the oncogenic mechanism in NCs. They map to chromatin megadomains that are hyperacetylated and transcriptionally activated by BRD4-NUTM1. RNAi knockdown of any one of these genes blocks cell proliferation in NC cell lines, and in the case of MYC and SOX2, also induces differentiation [28,29,30].

The absence of MYC, TP63, and SOX2, together with the functional disparity between BRD and MAD proteins, argues for a different oncogenic mechanism for MAD-NUTM1 fusions compared with BRD-NUTM1 fusions. Previously we have suggested that the fusion of NUTM1 to the C-terminal region of a MAD family protein could potentially convert it from a repressor to an activator of MYC target genes [2]. MAD proteins compete with MYC for binding to MYC-Associated Factor X (MAX). MYC/MAX heterodimers bind to sequences in promoters known as E-boxes and increase gene expression through transcriptional elongation, whereas MAD/MAX heterodimers bind E-boxes but inhibit transcription through mSin3 and histone deacetylases [34, 35]. The set of MYC responsive genes can alter depending on the cellular context, but they have important roles in cell cycle, cell growth, apoptosis and cellular metabolism, amongst others. Therefore, competition between MYC and MAD for binding to MAX has a major impact on many “hallmarks of cancer” processes. Our hypothesis is that since the MAD portion of MAD-NUTM1 fusions retain their basic region/helix-loop-helix/leucine zipper domains they would be capable of heterodimerizing with MAX and binding E-boxes. However, unlike wild-type MAD, they would also recruit the histone acetyltransferase p300 through the AD1 domain of NUTM1 and cause transcriptional activation of MYC target genes [2]. Here we provide support for this model by demonstrating a similar transcriptome-wide enrichment of MYC target genes in the MXI1-NUTM1 tumor and a NC, despite the absence of MYC expression in the former. We also show that the MXI1-NUTM1 fusion can at least partly phenocopy the effect of MYC on cell proliferation and anchorage-independent growth in vitro, which to the best of our knowledge is the first experimental demonstration of an oncogenic mechanism for a MAD-NUTM1 protein.

We stress that our model for the oncogenic mechanism of MAD-NUTM1 fusions remains speculative. Analysis of other tumors with MAD-NUTM1 fusions and mechanistic studies on their chimeric proteins are required to confirm the key features including binding to MAX and E-box motifs in MYC target genes. Furthermore, it was evident that MXI1-NUTM1 was not as potent as MYC in our in vitro assays. This may be partially due to the use of a fibroblast cell line, but was clearly also due to silencing of the integrated MXI1-NUTM1 construct. At the time of our colony assays, only about 40% of the MXI1-NUTM1 transformed cells continued to express NUTM1. Stable cell lines expressing recombinant BRD4-NUTM1 have yet to be created [29], suggesting that it may be similarly difficult to create stable cell lines expressing MAD-NUTM1, in which case either cultured primary tumor cells or inducible cell lines may be required for thorough mechanistic analyses.

This work raises the likelihood that NUTM1-fusions from other NRNs also have different oncogenic mechanisms to BRD-NUTM1 fusions. NRNs are increasingly being viewed as several distinct subgroups, and we have recently argued that their classification, clinical behavior, and therapeutic options may be best defined by the NUTM1 fusion partner, rather than by tumor morphology or IHC profile [2]. Although there are some promising results from the treatment of NCs with BET inhibitors [36, 37], it would follow that BET inhibitors may be ineffective against NRNs that lack the bromodomain-histone acetylated chromatin feed forward mechanism seen in NCs. Consistent with this, some non-NC NRNs respond to standard therapies for their tumor group.

MAD-NUTM1 tumors and other NRNs illustrate that there are important diagnostic and potentially therapeutic implications if a diagnosis of NC is incorrectly made based on a positive NUTM1 IHC or FISH alone. We suggest that in addition to the liberal use of NUTM1 IHC as a relatively cheap and sensitive screen for NRNs, all NUTM1 IHC-positive cases should be rigorously assessed for evidence of epithelial and squamous differentiation by morphology and IHC prior to making a diagnosis of NC. Ideally, confirmatory molecular testing should also be performed in order to identify the NUTM1 fusion partner for diagnostic certainty. In this regard RNA sequencing appears to be more sensitive than DNA intron tiling sequencing. We readily detected the MXI1-NUTM1 fusion with multiple RNA sequencing technologies but failed with a DNA targeted panel approach, most likely because the NUTM1 intron where the breakpoint occurred (intron 2) contains several regions of low complexity DNA sequence. In summary, the novel MXI1-NUTM1 fusion that we describe here illustrates how a diagnostic approach that integrates morphological, immunohistochemical and molecular data can be critical to inform appropriate clinical management, as well as reveal novel biological mechanisms.