Cell state-dependent chromatin targeting in NUT carcinoma

Abstract

Aberrant transcriptional programming and chromatin dysregulation are common to most cancers. Whether by deranged cell signaling or environmental insult, the resulting oncogenic phenotype is typically manifested in transcriptional changes characteristic of undifferentiated cell growth. Here we analyze targeting of an oncogenic fusion protein, BRD4-NUT, composed of 2 normally independent chromatin regulators. The fusion causes the formation of large hyperacetylated genomic regions or megadomains, mis-regulation of c-MYC, and an aggressive carcinoma of squamous cell origin. Our previous work revealed largely distinct megadomain locations in different NUT carcinoma patient cell lines. To assess whether this was due to variations in individual genome sequences or epigenetic cell state, we expressed BRD4-NUT in a human stem cell model and found that megadomains formed in dissimilar patterns when comparing cells in the pluripotent state with the same cell line following induction along a mesodermal lineage. Thus, our work implicates initial cell state as the critical factor in the locations of BRD4-NUT megadomains. These results, together with our analysis of c-MYC protein-protein interactions in a patient cell line, are consistent with a cascade of chromatin misregulation underlying NUT carcinoma.

Introduction

NUT carcinoma (NC) arises from squamous cells typically lining internal organs. The defining feature of NC is rearrangement of the NUTM1 gene (French et al. 2003, French 2018). NUTM1 expression is normally restricted to germ cells, but as a result of chromosomal translocation, NUTM1 becomes the 3′ partner of a widely expressed fusion protein. Somatic expression of NUTM1 protein is problematic because its normal function is to hyperacetylate histones for their removal during spermatogenesis (Shiota et al. 2018, Rousseaux et al. 2022). Thus, it is a powerful attractor and facilitator of the p300/CBP histone acetyltransferases (Wang and You 2015, Ibrahim et al. 2022, Yu et al. 2023).

The 5′ translocation partner of NUTM1 fusions in NC provides broad expression and chromatin association (French 2018). The most common 5′ partners are bromodomain-containing proteins such as BRD4 that bind acetylated histones (Zeng and Zhou 2002). BRD4 is a coactivator important for expression of many key genes including c-MYC (Filippakopoulos et al. 2010, Kotekar et al. 2022). The BRD4-NUT fusion is a powerful oncoprotein which forms distinctive nuclear foci (French 2018). We previously found that at the genomic level, these foci correspond to ∼200, hyperacetylated expanses of chromatin that reach up to 2 Mb in size (Alekseyenko, Walsh, et al. 2015, Eagen and French 2021). Based on their unprecedented size, we named these BRD4-NUT-associated regions “megadomains.”

Megadomains can be induced in non-NC cells via transgenic expression of the BRD4-NUT fusion protein. Interestingly, within hours of induction in HEK293T cells, megadomain formation typically starts at a subset of enhancers and subsequently spreads acetylation long distances in cis, sometimes filling whole topologically associated domains (TADs) (Alekseyenko, Walsh, et al. 2015). Spreading is thought to occur when the BRD4 bromodomain binds acetylated nucleosomes and the NUTM1 portion of the fusion protein attracts the EP300 lysine acetyltransferase, resulting in a self-perpetuating, feed-forward loop (Fig. 1a).

Fig. 1.

Induction of megadomains in HUES64 human embryonic stem cells. a) Model for formation of BRD4-NUT megadomains. b) Strategy for comparison of megadomain induction following transfection of HUES64 cells maintained in a pluripotent state or after differentiation along a mesodermal lineage. c). Anti-NUT immunostaining after transient transfection of HUES64 cells with a BRD4-NUT expression plasmid or a pUC19 negative control (400 × magnification). Formation of NUT nuclear foci is dependent on expression of the BRD4-NUT transgene.

Megadomain locations, mapped by ChIP-seq of NUT or H3K27 acetylation, are strikingly distinct in different NC patient cell lines, with the notable exception of the extensive regulatory region of the c-MYC gene which harbors a megadomain in all patient cell lines examined to date (Alekseyenko, Walsh, et al. 2015, Alekseyenko et al. 2017). In support of a model in which megadomains drive NC, treatment of NC patient cells in culture with a small molecule, JQ1, that inhibits interaction of the BRD4 bromodomains with acetylated lysines, leads to loss of megadomains, downregulation of c-MYC, and differentiation of NC cells (Filippakopoulos et al. 2010, Alekseyenko, Walsh, et al. 2015). The critical importance of MYC was revealed in studies in which ectopic c-MYC transgene expression reversed the differentiation induction caused by BRD4-NUT knockdown (Grayson et al. 2014). Furthermore, non-BRD4-NUT-dependent pathways such as KLF4 expression help maintain c-MYC levels in cells resistant to JQ1 (Liao et al. 2018). Thus, understanding the targeting mechanism of megadomains, as well as the molecular interactions of c-MYC protein in NC cells, should provide valuable insights into this therapeutically intractable disease.

Here we report our investigation into 3 questions relevant to this unusual, chromatin-driven disease. First, are the differences in megadomain locations in different patients due to differences in genotypes or instead related to their tissues of origin or epigenetic state? Second, is c-MYC the common target in NC patients because there is a powerful growth selection for cells mis-expressing this oncogene, or is c-MYC targeting somehow intrinsic to BRD4-NUT across cell states? Third, since MYC proteins are generally considered undruggable (Dang et al. 2017), can identification of c-MYC protein–protein interactions in an NC patient cell line reveal critical cofactors for therapeutic intervention?

Materials and methods

Cell culture

We purchased HUES64 from the Harvard Stem Cell Institute iPS core facility and maintained them on Matrigel (Corning) in mTeSR1 media (STEMCELL Technologies). Cells were routinely passaged as clumps after dissociation with 10 µM EDTA for 3 min as needed. For mesoderm formation, single hESC cells were plated at a density of ∼5 × 10⁵ per each well of a 6-well plate in the presence of ROCK inhibitor and exposed to commercially available mesoderm induction media (STEMCELL Technologies 05221).

BRD4-NUT expression in hESC and mesodermal lineage cells

Transfection of hESC

7 × 10⁶ cells were dissociated into single cells with Accutase (Thermo Fisher) at 37°C for 8–10 min. Ten milliliters of mTeSR1 media with 10 µM ROCK inhibitor was added, and cells were centrifuged 200×g for 5 min. Cell pellets was resuspended with 10 mL of mTeSR1 media with 10 µM ROCK inhibitors. Lipofectamine 3000-DNA transfection complexes were prepared by diluting 10 µg of pHAGE-EF1α-BRD4-NUT DNA (Supplementary Data S1) or pUC19 (negative control) in 500 µl of DMEM/F12 media (Thermo Fisher, catalog no. 11330-032), mixing well and incubating for 1 min, followed by addition of 20 µl of the P3000 reagent, mixing well and incubating for 1 min at room temperature (RT). At the same time, 20 µl of the Lipofectamine 3000 was diluted in 500ul of DMEM/F12 media, followed by addition of the diluted DNA, mixing well and incubating transfection complexes at RT for 15 min. The resulting transfection complexes were added to 7 × 10⁶ hESC cells in 10-mL mTeSR1 media with 10 µM ROCK inhibitors, mixed well, and plated cells on a 100-mm Matrigel-coated culture plate. For each immunostaining experiment, Matrigel (Corning)-coated round glass coverslips were added to the bottom of the plate. At 24-h postinfection, cells were immunostained with antibodies recognizing the NUT protein as described previously.

Transfection of mesodermal lineage cells

For mesoderm formation, single hESC cells were plated at a density of approximately 5 × 10⁵ per well of a 6-well plate (24 wells in total) in the presence of ROCK inhibitor and exposed to commercially available mesoderm induction media as instructed by the supplier (STEMCELL Technologies 05221). After 4 days of treatment cells were dissociated into single cells with Accutase (Thermo Fisher) at 37°C for 8–10 min. 7 × 10⁶ cells were transiently transfected with pHAGE-EF1α-BRD4NUT DNA or pUC19 (negative control) by the same protocol as for hESC except that in all steps, mTeSR1 media (STEMCELL Technologies) was replaced with mesoderm induction media (STEMCELL Technologies 05221).

Immunofluorescence

hESC cells were plated in 6-well plates containing Matrigel (Corning)-coated round glass coverslips. At 24-h post-transfection, cells were immunostained as follows: coverslips in each well were washed in 3 mL of 1× PBS at RT, followed by the addition of 2 mL of fix solution (for 40 mL: 4 mL of 10× PBS + 24.8 mL of water + 10 mL of 16% ultrapure formaldehyde +1.2 mL of 10% Triton) for 10 min. After fixation, coverslips were washed 3× with 3 mL of 1× PBS + 0.3% Triton X-100 for 5 min each, with rocking. One milliliter of blocking solution was added per well (for 10mL: 500µl NGS (normal goat serum) + 0.5 g BSA + 7.5 mL of water + 1 mL of 10× PBS + 1 mL of 10% Triton X-100) and rocked for 1 h at RT. Blocking solution was then replaced with fresh blocking solution containing anti-NUT antibodies (1:500; Cell Signaling, catalog no. 3625) and incubated overnight at +4C. The next day, the coverslips were washed 3 × in 3 mL of 1× PBS + 0.3% Triton X-100 for 5 min each, again rocking the 6-well plate. Secondary antibodies [1:1000 anti-Rabbit IgG (H + L) Highly Cross-Adsorbed Secondary Antibody, Alexa Fluor 594, Thermo Fisher, catalog no. A-21207] were then added in fresh blocking solution and incubated overnight at +4C. The next day, nuclei were counterstained with ProLong Gold antifade reagent with 4′,6-diami-dino-2-phenylindole (DAPI) (Life Technologies, catalog no. P36935). Images were viewed on a Zeiss Axioskop 2 using AxioVision Rel. 4.8 software.

ChIP-seq analysis

We performed ChIP-seq using a rabbit anti-NUT monoclonal antibody (Cell Signaling Technology C52B1). We scaled up the Lipofectamine 3000 transfection protocol to obtain 5 × 10⁷ hESC and mesodermal lineage cells for each ChIP experiment. BRD4-NUT was expressed for 24 h prior to harvesting. Two biologically independent experiments were completed for each cell state. The main steps of the ChIP-seq procedure were performed as described in Alekseyenko, Walsh, et al. (2015). ChIP-seq reads were aligned to the human reference genome (GRCh37 assembly) using Bowtie (version 2.3.4.3) (Langmead and Salzberg 2012), retaining only uniquely mapped reads. Smoothed enrichment profiles were generated with a bandwidth of 5 kb. Bigwig files were generated using deeptools (version 3.0.2) with the parameters–RPGC–smoothLength 5000. To visualize ChIP-seq signal at individual regions, we used the Integrative Genomics Viewer (IGV; https://software.broadinstitute.org/software/igv/). Peak enrichment for BRD4-NUT was identified using the HOMER package (Heinz et al. 2010). Briefly, BAM files were converted into tag directories, and peaks were called using the findpeaks program with the parameters -region -size 20000 -minDist 200000 -F 2 -L 2 -C 2. A minimum size cutoff of 96 kb was used for megadomain detection. bedtools intersect (version 1.9) (Quinlan and Hall 2010) was used to identify common and differential peaks, and 48 kb was the minimum size scored as an overlap between megadomains in different cell states. Heatmaps for BRD4-NUT occupancy were scaled to an average megadomain size of 180 kb summing RPGC-normalized counts from 10 bp bins spanning ± 100 kb from peak start and end site and visualized with the deeptools plotHeatmap function (Ramírez et al. 2014). Heatmaps were plotted to visualize the score distributions of the BRD4-NUT enrichment associated with genomic regions specified in the BED files with peaks from HUES64 or dME cells. Metagene profiles were depicted with the plotProfile function. Genomic coordinates and the identities of genes that map within or flanking each megadomain are listed in Supplementary Data S2.

Generation of n-BioTAP-cMYC NC797 cells for proteomic analyses

To generate N-BioTAP cMYC fusions in NC797 cells (Toretsky et al. 2003), the following DNA fragment: pAVV-MCS-NheI-5′cMYC_BamHI-ATG-loxP-Blasti-P2A-loxP-PrA-Bio-3′cMyc was synthesized as a gBlock (Integrated DNA Technologies, Supplementary Data S1) and introduced into the NheI/BamHI restriction sites of pAAV-MCS2 by Gibson assembly (New England Biolabs). The N-BioTAP cassette was targeted to the first ATG of the MYC gene. The pAAV-nEFCas9 (Addgene, plasmid 87115) was used to express Cas9, and the pAAV-tagBFP U6-gRNA expression vector was used to express a gRNA targeting the start codon of cMYC: (5′ GACGTTGAGGGGCATCGTCGC 3′). Recombinant AAV2 was packaged in HEK293T cells using pHelper and pRC2-mi342 plasmids (Takara, Catalog #632608). Three days after transfection, cells were harvested, and AAV2 was isolated using AAVpro Extraction Solution (Takara, Catalog #6235). Two million NC797 cells were infected with 30 µl of each adeno-associated virus (AAV2). Following 1 week of culture post-AAV infection, cells were selected with 7.5 µg/mL Blasticidin (Invitrogen). Surviving cells were replated, single-cell clones were isolated and expanded, and genotyping PCR and sequencing were performed to check for proper integration of the BioTAP cassette in the c-MYC locus. We recovered both heterozygous and homozygous clones. Homozygosity of the cassette insertion was determined by absence of the wild-type PCR fragment (1033 bp) and presence of the larger PCR fragment (2083bp) with the following PCR primers: Genom-Myc(1S) 5′ GAGTGGGAACAGCCGCAG 3′ and Genom-Myc(1A) 5′ CAGGTACAAGCTGGAGGTGG 3′. The presence of the correctly tagged MYC protein was validated by western blot probed with a c-MYC antibody (Rabbit mAb, Cell Signaling Technology, Cat# 5605). The recombinant pAAV-CRE plasmid was packaged in HEK293T cells using pHelper and pRC2-mi342 plasmids (Takara, Catalog #632608). Cells were harvested 3 days after transfection and AAV2 was isolated using AAVpro Extraction Solution (Takara, Catalog #6235). Two million 797 cells with homozygous insertions were infected with 30 µl of each adeno-associated CRE virus (AAV2). Following 1 week of culture post-AAV infection, cells were replated, single-cell clones were isolated and expanded, and genotyping PCR and sequencing were performed to check for absence of the drug resistance cassette in the c-MYC locus.

Generation of n-BioTAP-cMYC expressing HEK293T cell for proteomic analyses

We used lentiviral transduction to generate stable HEK293T cell lines (Alekseyenko et al. 2017) carrying an inducible NBioTAP-cMYC transgene. The transgene was constructed using the Gateway recombination system to introduce the Myc ORF clone HsCD00039771 in pDONR221from the CCSB Human ORFeome Collection into the pHAGE-TRE-DEST-NBioTAP (Addgene #53568) lentiviral vector. Stable 293T-Trex-cell lines containing the transgene were generated by lentiviral transduction in the presence of 8 µg/mL polybrene, followed by selection with 2 µg/mL puromycin dihydrochloride (Sigma #P8833). To induce transcription of the cDNA clones from the CMV/2xtetO promoter, we added doxycycline (1 µg/mL) to the medium for 48 h.

Identification of cMYC interactions using BioTAP-XL

All cell lines were cultured as monolayers in DMEM (Invitrogen) supplemented with 1 × Penicillin–Streptomycin (Hyclone, South Logan, UT), 1× Glutamax (Gibco), and 10% (v/v) fetal bovine serum (FBS) (Hyclone). The main steps of the BioTAP-XL procedure were as previously described (Alekseyenko et al. 2017). SDS was removed with ethanol and acetonitrile washes, and proteins were digested with overnight trypsin incubation, yielding peptides for desalting and liquid chromatography-mass spectrometry (LC–MS) analysis. LC–MS acquisition and analyses were performed as previously described (Alekseyenko et al. 2017). The number of total spectra number for each protein was used to determine relative abundance.

Results and discussion

BRD4-NUT megadomain locations are determined by cell state

Our primary goal was to ask why megadomain patterns were distinct in different cancer patient cell lines. From our previous work, we already knew that this was unlikely to be due to a random, stochastic process, as megadomain patterns were remarkably reproducible in replicates induced in cell lines with no prior exposure to BRD4-NUT after transfection of a BRD4-NUT transgene (Alekseyenko, Walsh, et al. 2015). Therefore, we proposed 2 hypotheses to account for this distinctive targeting: (i) Different patients' cells have critical DNA sequence variations, which could explain the divergence of megadomain patterns, or (ii) Different transcriptional or epigenetic states are the key to megadomain pattern formation. For example, NC can arise in squamous cells lining the lung or salivary gland, so the patterns might reflect these initial tissue-of-origin differences.

To distinguish between these hypotheses, we selected a human stem cell model to study how BRD4-NUT megadomain formation responds to changes in cell state (Fig. 1b). We purchased HUES64 cells (Chen et al. 2009) from the Harvard Stem Cell Institute iPS core (https://ipscore.hsci.harvard.edu/home) and maintained them in the pluripotent state on Matrigel in mTeSR1 media (see Materials and methods). We chose the mesodermal lineage as our comparison cell state because a parallel study confirmed that treatment of HUES64 cells with a commercially available kit (STEMCELL Technologies) produced conversion to ∼98% EpCAM-negative/NCAM-positive mesodermal cells by FACS analysis (Naxerova et al. 2021). To express the fusion oncoprotein in hESC and in cells differentiated along a mesodermal lineage (dME), we constructed pHAGE-EF1α -BRD4NUT, in which the fusion cDNA was driven by a strong constitutive promoter derived from the human elongation factor EF-1α gene (Supplementary Data S1).

Following transient transfection, we performed immunostaining to test whether expression of BRD4-NUT protein caused the formation of NUT nuclear foci in the 2 distinct conditions. In each case, we found that foci qualitatively formed in similar number and appearance to those of endogenous BRD4-NUT in cultured patient-derived cells (Fig. 1c). Nucleofection efficiency of the DNA into hESC was typically around 30%. Since the NUT epitope was uniquely expressed in successfully transfected cells, this allowed our experiments to proceed without drug selection or cell-sorting.

Encouraged by the appearance of nuclear foci after transfection, we performed ChIP-seq using anti-NUT antibodies (Cell Signaling). We scaled up the Lipofectamine 3000 transfection protocol to obtain 5 × 10⁷ hESC or dME for each ChIP experiment and performed 2 independent biological replicates for each cell state. In each experiment, BRD4-NUT was expressed for 24 h. Our results confirmed that BRD4-NUT produced megadomains in highly reproducible patterns in replicates of the same cell state (Fig. 2, a and b). In contrast, we found that megadomains were formed in highly divergent patterns when cells in the pluripotent state were compared to the same cell line following induction along a mesodermal lineage (Fig. 2a–d, Supplementary Data S2). We conclude that BRD4-NUT initiates megadomain formation in response to the preexisting transcriptional or epigenetic state of recipient cells.

Fig. 2.

Comparison of BRD4-NUT megadomains by ChIP-seq. a) Top: human chromosome 8 IGV browser view of log₂ IP/Input BRD4-NUT read counts from ChIP-seq experiments performed in pluripotent HUES64 (hESC) and derived mesoderm (dME). Middle: boxed regions are enlarged, showing the strong concordance between independent biological replicates. Bottom: examples of cell type-specific megadomains in hESC and dME are indicated along with their respective genomic sizes. Transcription units are displayed below the ChIP-seq profiles. b) Venn diagrams showing the strong reproducibility of independent biological replicates of the same cell state and the divergence when comparing the same cell line in different cell states. c) IGV browser view comparing differences between cell states in a 3 Mb region of chromosome 4. d) Metagene profiles and heatmaps of BRD4-NUT megadomains calculated as log2 ChIP/Input. The x-axis shows megadomains scaled to the average size of 180 Kb, flanked by linear 100 Kb segments of upstream and downstream sequences. Top: each row of the heatmap represents a megadomain sorted in descending order for BRD4-NUT occupancy in HUES64, revealing many hESC-specific megadomains (left) along with a few that overlap with dME megadomains (right). Bottom: heatmap of dME-specific megadomains sorted in descending order (right) with the corresponding regions of hESC (left).

Megadomains form at N-MYC rather than c-MYC in pluripotent cells

We next asked whether the c-MYC regulatory region, harboring a common megadomain in all NC patients studied to date (Alekseyenko, Walsh, et al. 2015), was a target of BRD4-NUT in HUES64 stem cells and found that it was not (Fig. 3a). However, the MYC family of cellular oncogenes is comprised of 3 related genes, c-MYC, commonly known as MYC, as well as N-MYC and L-MYC (Henriksson and Luscher 1996, Facchini and Penn 1998). While the c-MYC regulatory region was not a target, we found that a prominent megadomain formed on the N-MYC regulatory region in HUES64 pluripotent cells (Fig. 3a). To test whether this might be common to other pluripotent cell lines, we transfected Ntera-2 cells, a human embryonal carcinoma cell line (Lee and Andrews 1986), and found similar results, i.e. the N-MYC regulatory region was a prominent target of BRD4-NUT megadomain formation (Fig. 3b). As these experiments were performed after 24 h of transgene expression, selection for BRD4-NUT-driven proliferation was unlikely to be a factor in this association. c-MYC and N-MYC have similar functions but distinct expression patterns during development (Malynn et al. 2000). Therefore, we speculate that BRD4-NUT may have intrinsic attraction to the relevant MYC gene in undifferentiated cell types, potentially because endogenous BRD4 is normally a key coactivator of MYC during proliferation.

Fig. 3.

Comparison of BRD4-NUT ChIP-seq of the N-MYC, c-MYC, and L-MYC loci. N-MYC, c-MYC, and L-MYC IGV browser views of log₂ IP/Input BRD4-NUT read counts from ChIP-seq experiments performed in a). HUES64 and derived mesoderm dME, and a) Ntera2 cells at 0, 4, and 16 h of retinoic acid induction, which causes differentiation toward a neural lineage. Both N- and c-MYC have large adjacent regulatory regions encompassing noncoding RNA transcripts (such as GACAT3 and PVT1) but lacking additional protein-coding genes, while L-MYC resides in a region with many protein-coding genes. Colored bars below the ChIP profiles indicate megadomains (>96 Kb contiguous occupancy of BRD4-NUT).

The devastating effect of MYC overexpression is well established in human cancer, but less is known regarding whether dysregulated MYC might acquire aberrant or unexpected protein interactions when driving malignancies. Therefore, we next turned to analysis of the downstream consequences of BRD4-NUT-driven c-MYC dysregulation in an NC patient cell line at the level of protein–protein interactions.

c-MYC protein interacts with the NuA4 acetyltransferase complex in NC797 cells

Given the strong evidence for dysregulation of c-MYC as the common feature of NC, we turned our focus to its protein interactions in a well-characterized NC cell line, NC797 (Toretsky et al. 2003). For many years, comprehensive proteomic analyses of MYC interactions were impeded by technical challenges. In addition to a relatively short half-life, it has been difficult to extract intact MYC complexes from chromatin using traditional biochemical methods. This has been overcome by the use of biotinylation as a high affinity tag (Kim et al. 2010) or by the implementation of BioID (Roux et al. 2018), in which proximity labeling alleviates the need for release of MYC protein from chromatin to identify interactors (Dingar et al. 2015, Kalkat et al. 2018). Using BioID, the Penn laboratory has mapped the interactions of both the NuA4 (KAT5/TIP60) and STAGA (KAT2A/GCN5) complexes to c-MYC homology box II (MBII). Furthermore, they discovered that MBII is linked to both complexes through interaction with TRRAP and confirmed the importance of MBII in stimulation of acetylation by c-MYC (Kalkat et al. 2018). Thus, comprehensive proteomics has strongly confirmed the initial discoveries of c-MYC association with acetyltransferases (McMahon et al. 2000, Bouchard et al. 2001, Frank et al. 2001, Frank et al. 2003, Liu et al. 2003).

Here, we applied an orthogonal approach, BioTAP-XL (Alekseyenko, McElroy, et al. 2015), to characterize c-MYC protein interactions in NC and non-NC cells without requiring initial release from chromatin. BioTAP-XL employs cross-linking, sonication, stringent 2-step affinity purification, and analysis of enrichment over input by mass spectrometry. Using BioTAP-XL, we found that c-MYC interacts primarily with subunits of the NuA4 acetyltransferase complex, but not STAGA in NC797 cells (Fig. 4a). To test whether the lack of STAGA enrichment might be related to our cross-linking method rather than the NC cell type, we performed the BioTAP-XL experiment in the non-NC cell line HEK293T, which was used in the previous BioID studies (Kalkat et al. 2018). Similar to NC797 cells, we found enrichment of NuA4 components with BioTAP-MYC in HEK293-TREx cells, but we also observed enrichment of STAGA components such as ATXN7 and KAT2A (Fig. 4b). Therefore, c-MYC interaction with NuA4 but not STAGA appears to be a characteristic of the NC797 patient cell line rather than somehow biased by our cross-linking approach, although it should be noted that BioTAP-tagged c-MYC was expressed from its endogenous locus in NC797 cells, and ectopically expressed in HEK293T cells.

Fig. 4.

MYC protein interactions in NC797 cells. a) Scatterplot depicting enrichment over input of proteins from N-terminally tagged c-MYC using BioTAP-XL affinity purification in NC797 cells. Each gray dot represents an individually identified protein, where its position is a measure of enrichment efficiency based on the number of total peptides identified in pulldown compared to input and normalized to the molecular weight of the protein. The dashed gray lines denote the 99th percentile of enrichment. The right plot is a zoomed in region of the most enriched proteins in the left plot from replicate experiments. b) Comparison of the total peptides mapping to NuA4 and STAGA components in HEK293T and NC797 cells following BioTAP-XL of c-MYC. Full data can be found in Supplementary Data S3.

As chromatin acetylation complexes were the most enriched MYC interactors (top 1%) using our stringent cross-linking, tandem affinity purification approach, our results support the extensive previous work cited above, in which acetyltransferase activity is strongly linked to the biological function of c-MYC. Previous nascent RNA-seq analyses revealed transcription of the STAGA acetyltransferase KAT2A gene at similar levels in HEK293T and NC797 cells (Supplementary Fig. 1). Thus, it remains to be determined why the NC cells we studied here show strong enrichment for MYC interaction with NuA4 but not STAGA. Demonstrations that c-MYC is dependent on acetyltransferases for its function have suggested potential therapeutic options to what is generally considered an undruggable oncoprotein (Mustachio et al. 2020). Our results suggest that development of specific inhibitors of KAT5/TIP60 for synergy with JQ1 could be particularly relevant to the treatment of NC.

In conclusion, while much remains to be discovered regarding BRD4-NUT chromatin-driven oncogenesis, it is striking that its targeting is clearly related to cell state rather than to genotype or DNA sequence. Although we did not directly assess acetylation in the current study, we speculate that differential targeting of BRD4-NUT is likely due to the underlying differences in preexisting genomic acetylation patterns in pluripotency vs differentiation, based on our prior work mapping the initiation of megadomain formation to a subset of enhancers in HEK293T cells (Alekseyenko, Walsh, et al. 2015). Thus, acetylation is implicated in both the initiation of megadomains and their aberrant transcriptional consequences. c-MYC upregulation and strong physical association with acetyltransferases confirm a third connection to a cascade of aberrant acetylation underlying NC.

Data availability

The BioTAP-XL mass spectrometry data are available on the ProteomeXchange Consortium via the PRIDE partner repository, accession number PXD041334. Chip-seq data sets are available on the NCBI GEO database, accession number GSE229558.

Supplemental material available at GENETICS online.

Funding

This research was funded by the National Institutes of Health, grant number R35 GM126944 to MIK.