NSD3–NUT Fusion Oncoprotein in NUT Midline Carcinoma: Implications for a Novel Oncogenic Mechanism

Hematopoietic and mesenchymal malignancies are often characterized by translocation-associated fusion oncoproteins that block differentiation, whereas many epithelial cancers are typified by multiple sequential mutations that progress in a multistep pathway to carcinogenesis. One exception of an epithelial carcinoma that is driven by a fusion oncogene is NUT midline carcinoma (NMC). NMC is defined by chromosomal rearrangement of the NUT gene (aka NUTM1), which is most commonly fused to the BET family genes BRD4 and BRD3 (1, 2), defined by the presence of dual bromodomains and an extraterminal (ET) domain. BRD–NUT oncoproteins' primary mechanism is to block differentiation and maintain cells in a highly proliferative, poorly differentiated state. This poorly differentiated cancer is far more aggressive than even small cell carcinoma of the lung, with a median survival of 6.7 months (3), and there exist no effective treatment options.

Recent excitement in small-molecule BET inhibitors arose from the demonstration of the therapeutic targeting of BRD–NUT oncoproteins in NMC in vitro and in preclinical models (4). This has led to clinical trials using the GSK BET inhibitor drug, GSK-525762A, and the Tensha Therapeutics BET inhibitor, TEN-010, now enrolling patients with NMC and other solid tumors (http://www.clinicaltrials.gov/ct2/show/NCT01587703?term=NMC&rank=1; http://www.clinicaltrials.gov/ct2/show/NCT01987362?term=tensha&rank=1). BET inhibitors are acetyl-histone mimetics that target the acetyl-histone binding pocket of BET protein chromatin-reading bromodomains, such as those of BRD2, 3, 4, and T (4, 5). BET inhibitors induce differentiation and proliferation arrest of NMC, and are a potential form of differentiation therapy in this disease. However, it is not known how interference with chromatin binding leads to inhibition of the blockade of differentiation by BRD–NUT oncoproteins, because the mechanism by which BRD–NUT blocks differentiation is unclear. Evidence suggests that deregulation of MYC expression by BRD–NUT may be key to the blockade of differentiation (6), but it remains to be determined whether BRD–NUT acts directly or indirectly.

Known functional domains of BRD4 that are present in BRD–NUT fusions may give clues to its function. Wild-type BRD4 binds to acetylated histones and the positive transcriptional elongation factor, P-TEFb with its bromodomains (7), and is associated with transcriptional activation of target genes (7, 8). Although the function of NUT, an entirely unstructured protein, is unknown, it binds to and activates the histone acetyltransferase (HAT) p300 (9). Both of the bromodomains and the p300-binding domain are present in BRD–NUT oncoproteins. This has led to the hypothesis that BRD–NUT fusion oncoproteins tether HATs and transcriptional cofactors to chromatin via their bromodomains, leading to a feed-forward process of acetylation and recruitment that results in sequestration of these factors away from prodifferentiation genes to progrowth genes, such as MYC (2, 9).

The role of the ET domain and its binding proteins has not been investigated in the context of BRD–NUT oncoproteins. Here, we describe a novel fusion in a NUT-variant NMC between the methyltransferase protein NSD3, which has been previously shown to associate with the ET domains of BET proteins (8), and NUT. The finding suggested that NSD3 may be a key component of the BRD–NUT oncogenic complex. Thus, we investigated the oncogenic role of NSD3 in this NUT-variant NMC as well as more typical BRD4–NUT NMCs.

A poorly differentiated squamous cell carcinoma of the mediastinum (Fig. 1A) metastatic to the femur of a 13-year-old girl was referred to us for molecular diagnostic testing for NMC. Immunohistochemical analysis revealed diffuse nuclear expression of the NUT protein, a feature that is diagnostic of NMC (Fig. 1B; ref. 10). FISH demonstrated rearrangement of the NUT gene locus on chromosome 15q14; however, neither BRD4 nor BRD3 rearrangement was detected. Discarded live tumor tissue from a metastatic focus in the patient's lung was collected under Institutional Review Board approval through the NMC registry (www.NMCRegistry. org). From this tissue, the first known NUT-variant cell line, 1221, was established. To determine the putative partner gene to NUT, we performed comprehensive RNA sequencing on RNA purified from 1221 cells. We identified an in-frame transcript fusing the 5′ coding sequence of NSD3 (exons 1–7) to exons 2–7 of NUT (Fig. 1C). Expression of the NSD3–NUT fusion oncoprotein was verified by immunobloting with an antibody to NUT, revealing an approximately 200-kDa band that is similar in size to BRD3–NUT, but smaller than BRD4–NUT (Fig. 1D). Knockdown using siRNAs targeting NSD3 led to the disappearance of the putative NSD3–NUT band, as did siRNAs targeting NUT, confirming the identity of the NSD3 and NUT portions of the NSD3–NUT fusion protein (Fig. 1E). Genomic fusion of the NSD3 and NUT genes was confirmed by FISH, demonstrating bring-together of NUT and NSD3 probes (Fig. 1F). Likewise, the expression of an NSD3–NUT mRNA was demonstrated by reverse transcriptase PCR (RT-PCR; Fig. 1G). Cytogenetic analysis of the 1221 cell line was consistent with an NSD3–NUT fusion, revealing a t(8;15)(p12;q15) translocation, and metaphase FISH demonstrated localization of the NUT probe near the NSD3 chromosomal region (8p11.23; Supplementary Figs. S1–S2). Several additional aberrations of unknown significance were also present.

The fusion sequence is predicted to encode a 1694 amino acid protein containing amino acids 1–569 of NSD3, and 8–1132 of NUT. Interestingly, the NSD3 portion of the fusion protein lacks the Su(var)3–9, Enhancer-of-zeste and Trithorax (SET) domain and contains only its Pro–Trp–Trp–Pro motif (PWWP) domain, whereas nearly all of NUT is included in the fusion, as is typical in NMC (Fig. 1H; refs. 1, 11, 12). The NSD3–NUT fusion bears no resemblance to NUP98–NSD3/NSD1 fusion oncogenes that have been previously described in leukemia (13, 14), all of which fuse NUP98 to the 3′ end of NSD3/NSD1 containing their SET, PHD (Plant Homeo Domain), and C/H (Cys–His)-rich domains.

We next sought to determine whether NSD3 is a recurrent NUT fusion partner in NMCs; thus, we performed a dual-color NSD3 split-apart FISH assay on several NUT-variant cases. Four of eight non-BRD3/BRD4–NUT NMC cases (including the index case) demonstrated rearrangement of NSD3 and NUT, suggesting a frequent incidence of NSD3–NUT among NUT-variant cases (Fig. 1I).

The recurrent existence of NSD3–NUT in NMCs suggested that it may function similarly to BRD–NUT by blocking differentiation and maintaining proliferation of NMC cells (2). We thus knocked down endogenous expression of NSD3–NUT in 1221 cells to determine its effect on growth and differentiation. Seventy-two hours following knockdown, 1221 cells exhibited differentiation as evidenced by increased keratin expression, an epithelial differentiation marker, by immunofluorescence (Fig. 2A and B), and decreased proliferation as measured by Ki-67 fraction (Fig. 2C) and cell number (Fig. 2D). Notably, knockdown of wild-type NSD3 using siRNAs directed toward the 3′ end of NSD3 that is not included in the NSD3–NUT fusion gene had no effect on differentiation (Fig. 2B). These findings indicate that NSD3–NUT serves to block differentiation and maintain proliferation of 1221 cells.

NSD3 is one of several proteins that have been shown to bind to the ET domain of BET proteins (8). Thus, we hypothesized that NSD3 may have an oncogenic role in NMC through its interaction with BRD4–NUT's retained ET domain. It is noted that the BRD3 and BRD4 fusions with NUT in the BRD–NUT NMCs occur 3′ to the ET domain; thus, the ET domain is always included as part of the fusion protein (1, 2). We therefore tested whether NSD3 is required for the blockade of differentiation in BRD4–NUT-expressing NMC cells. In three different patient-derived BRD4–NUT⁺ NMC cell lines, TC-797 (15), PER-403 (16), and 8645 (17), siRNA knockdown of NSD3 resulted in differentiation, as measured by increased expression of the terminal squamous differentiation marker involucrin (Fig. 3A–C). This was accompanied by morphologic differentiation, as evidenced by flattening and enlargement of cells (Fig. 3B and Supplementary Fig. S3), as well as mild-to-moderate decreased proliferation quantified by Ki-67 staining (Fig. 3D) in all cell lines. Moreover, induced expression of the ET domain fused to a nuclear localization sequence (NLS) in a tet-inducible NMC derivative cell line, 797TRex, exhibited a dominant negative effect on BRD4–NUT function, inducing differentiation morphologically and immunophenotypically (Fig. 3E). In addition, induction of ET domain expression also negatively affected the proliferation rate of TC-797 cells, whereas the growth of heterologous, non-NMC cells, U2OS, or 293T, was unaffected (Fig. 3F) The findings indicate that expression of wild-type NSD3 protein and the ET domain of BRD4–NUT are required for the blockade of differentiation in BRD4–NUT⁺ NMC. The specific requirement of the ET domain for the oncogenic function of BRD4–NUT is evidenced by its conservation in all characterized BRD–NUT fusions (1, 2, 12, 18, 19), including uncommon splice variants (16, 18), and by the lack of growth inhibition induced by ET domain expression in non-BRD–NUT-expressing cell lines (Fig. 3F).

Because the ET domain is retained in BRD–NUT oncoproteins, we predicted that the interaction of NSD3 with BRD4 would be preserved when coexpressed with BRD4–NUT. BRD4–NUT normally localizes to discrete nuclear foci by immunofluorescence and IHC. We found that the HA-tagged portion of NSD3 present in NSD3–NUT (NSD3Tr, corresponding to amino acids 1–569 of NSD3) colocalized with BRD4–NUT foci (Fig. 4A). Moreover, HA-tagged NSD3, NSD3–NUT, and NSD3Tr (Fig. 4B) coimmunoprecipitated BRD4 in C33A cervical carcinoma cells. In reciprocal experiments, HA-tagged constructs of BRD4 and BRD4–NUT, but not NUT, were able to coimmunoprecipitate NSD3 (Fig. 4C). Of note, all the multiple NSD3 isoforms seen in this blot contain the N-terminal domain of NSD3 (NSD3Tr) that is present in the NSD3–NUT fusion protein that interacts with BRD4 (Fig. 4B). The findings indicate that NSD3 does associate with BRD4–NUT. To determine the role of the association of NSD3 with BRD4–NUT in the blockade of differentiation, NSD3Tr was expressed in 797TRex cells, and was found to induce differentiation (Fig. 4D). The findings, coupled with the dominant-negative effects of ET domain expression (Fig. 3E and F), suggest that the interaction of NSD3 with BRD4–NUT may be required for the blockade of differentiation. In support of this, other known interactors of the ET domain, including CHD4, ATAD5, GLTSCR1, and JMJD6 (MCB), were knocked down in TC-797 cells, but failed to induce differentiation (Supplementary Fig. S4A and S4B). Although the findings suggest that NSD3–ET domain interaction is required for the blockade of differentiation in NMC, they are not conclusive, because other unknown interactors with either of these domains may be critical.

The function of BRD4–NUT foci is unknown; however, it has been demonstrated that they are intensely enriched with BRD4–NUT and factors associated with transcription, including RNA polymerase II, the HAT, p300, the transcriptional elongation complex P-TEFb, and active histone marks (9, 20). On the basis of these observations, it is believed that BRD4–NUT foci are important to the function of BRD4–NUT in transcriptional regulation (9, 20), though this has not been tested directly. To determine what role, if any, NSD3 has in the formation of BRD4–NUT foci, we knocked it down in TC-797 cells and quantified foci number as compared with control siRNA-transfected cells. We found that the number of BRD4–NUT foci was significantly reduced within 24 hours following siRNA knockdown of NSD3 (Fig. 5A and B). Reduction of BRD4–NUT foci after NSD3 knockdown was not accompanied by a reduction of total BRD4–NUT protein levels (Fig. 5C), indicating that the effect is not due to loss of BRD4–NUT. Thus, the findings indicate that NSD3 is required for BRD4–NUT focus formation, suggesting a key role in BRD4–NUT complex or aggregate formation.

Because NSD3 is critical for BRD4–NUT function, we investigated whether NSD3–NUT was functionally equivalent to BRD4–NUT and as such replaces BRD4–NUT's function in blocking differentiation. To test this, we induced expression of a Bio-TAP tandem tagged NSD3–NUT construct with tetracycline in 797TRex cells subjected to BRD4–NUT knockdown using an siRNA targeting the 3′-untranslated region (UTR) of NUT not present in the NSD3–NUT introduced gene. Indeed, induction of expression of NSD3–NUT in 797TRex cells almost completely abrogated differentiation and proliferation arrest in cells subjected to knockdown of endogenous BRD4–NUT. The ability of NSD3–NUT to block differentiation was demonstrated morphologically, showing a lack of flattening (Fig. 6A), and immunophenotypically, showing markedly reduced involucrin expression (Fig. 6A and B), as compared with cells not induced to express NSD3–NUT (ethanol vehicle control–treated cells). The ability of NSD3–NUT to maintain proliferation was demonstrated by a Ki-67 fraction similar to scrambled siRNA-transfected cells, and markedly greater than ethanol vehicle control–treated cells subjected to BRD4–NUT knockdown (Fig. 6C).

The existence of NSD3 as a NUT fusion oncogene partner, whose encoded protein is also an important functional member of BRD4 and BRD4–NUT complexes, is reminiscent of the oncogenic mechanism of MLL fusion–associated leukemia (21). Thus, we surmised that the oncogenic function of NSD3–NUT may depend on its interaction with BRD4 as a component of a chromatin-modifying complex with similar function to BRD4–NUT. Indeed, siRNA knockdown of BRD4, both long and short isoforms, induced differentiation of 1221 cells (Fig. 7A), and treatment of 1221 cells with the BET inhibitor JQ1 resulted in differentiation and arrested proliferation, in a dose-dependent manner (Fig. 7B–D). These findings, together with the functional interchangeability of NSD3–NUT and BRD4–NUT, provide evidence that NSD3–NUT uses the chromatin-reading function of BRD4. Moreover, these data provide a sound rationale for treatment of patients with NSD3–NUT-positive NMCs using BET inhibitors.

A number of recent variant translocations have been described in NMC, illustrating the heterogenous nature of this disease (18, 22–24). In all of these previously described variants, both NUT and BET genes are fused. In this study, we describe a novel fusion gene in NMC that does not include a BET protein, but rather a BET-binding protein, NSD3. We find that the NSD3–NUT fusion oncogene encodes a protein that is both necessary and sufficient for the blockade of differentiation in NMC. We also find that wild-type NSD3, which binds to BRD4 in non-neoplastic cells, also binds to BRD4–NUT and is required for the blockade of differentiation by more common BRD4–NUT-expressing NMCs. The presence of a fusion oncoprotein involving constituents of a single oncogenic complex is well documented in cancer, and the existence of NSD3–NUT speaks to the importance of NSD3 and its association with BRD4–NUT. NSD3, also known as WHSC1L1, is a histone methyltransferase that belongs to the mammalian nuclear SET domain-containing (NSD) protein family of SET domain-containing methyltransferases, which also includes NSD1 and NSD2 (WHSC1/MMSET). Both NSD3 and NSD2 are known to bind the ET domain of BRD4; thus, the dominant-negative phenotype of ET and NSD3Tr expression in NMC cells is evidence that this interaction of BRD4 with NSD3 may be critical to BRD4–NUT function. However, it is not clear that the methyltransferase activity of NSD3 is needed for BRD4–NUT function as it is for NUP98–NSD3 fusions (14), because the NSD3 portion of NSD3–NUT lacks the SET domain. Moreover, knockdown of full-length NSD3 in the 1221 cells (Fig. 2B) does not induce differentiation, again suggesting that NSD3′s SET domain is not required for the differentiation blockade. Thus, it appears that the critical portion of NSD3 is its N-terminal, ET-binding domain in NSD3–NUT-expressing NMC. Combining these data with the fact that BRD4 expression and interaction with chromatin is required for NSD3–NUT function (Fig. 7), we have devised a model whereby the NSD3 portion links NSD3–NUT to BRD4, which tethers NSD3–NUT to chromatin, forming a complex that functions similarly to BRD4–NUT. In future studies, we plan to test this model by examining whether the interaction of NSD3–NUT with BRD4 is required for the oncogenic function of NSD3–NUT. In the context of BRD4–NUT-expressing NMC, we have shown that NSD3 is required for BRD4–NUT foci formation (Fig. 5A), suggesting a role in the aggregation of large BRD4–NUT-containing complexes. However, it remains to be determined what function NSD3 serves in the formation of these foci.

Although its importance is not known in NSD3–NUT⁺ NMCs, NSD3 methyltransferase activity may be important to the function of BRD4–NUT NMC. The SET domain of NSD proteins is homologous to the Saccharomyces cerevisae histone 3 lysine 36 (H3K36)–specific methyltransferase SET2 and is specific for H3K36 dimethylation (H3K36me2; ref. 25). NSD3 has been reported to regulate H3K36 methylation and thereby active gene expression (8, 26). It is possible that akin to the aberrant activation of HoxA1 expression by NUP98–NSD1-mediated methylation of H3K36 in acute leukemia (14), NSD3 may contribute to the transcriptional activation of key targets of BRD4–NUT. Although the NUP98–NSD3 fusion oncogene has been described in acute leukemia as well (13), it has not been further characterized mechanistically. Proteomic analysis of NSD3 suggested that the protein also interacts with the histone protein variant macroH2A1. MacroH2A1 replaces conventional H2A histones in a subset of nucleosomes, where it represses or activates transcription and participates in stable X chromosome inactivation (27–29). Moreover, macroH2A1 regulates cell growth and differentiation and is differentially expressed in cancer cells (30–32). Thus, NSD3 may regulate differentiation in BRD4–NUT-expressing NMC cells by affecting differentiation-specific genes via alterations of H3K36 or macroH2A1 levels.

NSD family proteins have been associated with other cancers (reviewed in ref. 33). Chromosome translocation resulting in NSD2 (aka MMSET) overexpression leads to multiple myeloma, whereas reduction in NSD2 levels suppresses cancer growth (34–36). Moreover, NSD3 is amplified in breast cancer cell lines and primary breast carcinomas (37, 38). NSD3 has been reported to contribute to the transformed phenotype and invasiveness of these breast cancer cells (39, 40). Most recently, mutations in NSD3 have been identified in pancreatic adenocarcinoma (41). Despite these associations, the mechanism by which NSD3 contributes to oncogenesis in these cancers remains poorly understood.

A key to understanding the role of NSD3 in cancer may be through its association with BRD4. The indispensability of the BRD4 chromatin-binding bromodomains in NMC (4, 6) and its ET domain's role in recruiting NSD proteins demonstrates that BRD4 is a key player in BRD4–NUT chromatin-associated oncogenic complex formation. Moreover, recent studies with BET inhibitors have shown that BRD4 plays a role in other human cancers such as acute myeloid leukemia, multiple myeloma, and Burkitt lymphoma (42–44). In these cancers, as well as in NMC, BRD4 and BRD4–NUT, respectively, are required for the maintenance of MYC expression, and BET inhibitors repress expression of MYC, presumably through interference with BRD4–chromatin interaction (6, 43, 44). Our findings that siRNA knockdown or BET inhibitor blockade of BRD4 function induce differentiation in NSD3–NUT-expressing NMC cells indicate that NSD3 function depends on BRD4 and its interaction with chromatin (Fig. 7). If NSD function is dependent upon BRD4 in other cancers, these NSD-associated cancers may be responsive to BET inhibitor therapy. Thus, oncogenic NSD may be a biomarker of response to BET inhibitor therapy.

Apart from BET inhibitors, this study highlights the importance of the therapeutic potential of targeting the NSD family of proteins. Histone-modifying enzymes, including the NSD family, are often deregulated in cancer, and aberrant histone modification profiles are intimately linked to carcinogenesis (45). Histone deacetylase (HDAC) inhibitors have been found to be effective in inhibiting NMC (17) and are already approved by the FDA for certain leukemias, and inhibitors to histone methyltransferases, including the NSD family, are under development (33, 45–47). Such therapies hold promise for NMC as well as other cancers.

Dual-color FISH assays for NSD3 and NUT breakpoints were performed on formalin-fixed, paraffin-embedded, 4-mm tissue sections as described previously (20). Probes used for the 15q14 breakpoint flanking NUT included the telomeric bacterial artificial chromosome (BAC) clones RP11-1H8 and RP11-64o3, and the centromeric clones RP11-1084A12 and RP11-368L15. Probes used for the chromosome 8p11.23 NSD3 breakpoint were the flanking 5′ centromeric BAC clones CTD-2538P2 and RP11-957P17 and the 3′ telomeric BAC clones CTB-497A2 and RP11-90P5. Cytogenetic analysis and metaphase FISH was performed using standard methods (48, 49).

RNA was extracted from live cultured 1221 cells using the RNeasy Mini Kit (Qiagen). Elim Biopharmaceuticals performed the library preparation and sequencing. rRNA removal was performed using the Ribo-Zero Kit (Epicentre) following the manufacturer's instructions. The library was prepared using standard Illumina protocols with proprietary modifications and sequenced using HiSeq2500 (Illumina). TopHat-Fusion (v2.0.8b bundled with TopHat2) was run with default parameters (as described at http://tophat.cbcb.umd.edu/fusion_tutorial.html#toph but with −r 50 and −max-intron-length 1000000) to identify novel fusion transcripts from paired-end 50 base reads (50, 51).

RT-PCR was as described (6) using the following primers: NSD3f1382 AAGAGCCACCGCCTGTTAAA, NUTr388 GCTGTCA CAAATGGAGGTGC, GAPDH254f TCAAGTGGGGCGATGCTGGC GCT, and GAPDH788r AGGGGGCCCTCCGACGCCTGCT.

BRD4 ET domain containing fragment (BRD4 444–722) was cloned into pcDNA5 FRT/TO-FLAG (Invitrogen) with an N-terminal SV40 NLS sequence to generate pcDNA5 FRT/TO-Flag-NLS-BRD4-ET (p6894). MSCV-CMV-Flag-HA-NSD3 (p6351) has been described previously (8). To make tetracycline-inducible N- and C-terminal BioTAP-tagged constructs, we transferred the gateway destination cassette from pHAGE-TRE (gift of Steven Elledge, Harvard Medical School, Boston, MA) to the pcDNA5 frt/to mammalian expression vector with N-terminal BioTAP tandem tag to make pcDNA5 frt/to-DEST-NBioTAP. NSD3–NUTwas constructed by fusion PCR into pDONR223, then transferred into the pcDNA5 frt/to-DEST-NBioTAP vector by gateway cloning. Full-length NSD3, NSD3Tr encoding amino acids 1–569 of NSD3, full-length NUT, and BRD4–NUT (derived from pcDNA5 frt/to FLAG-BRD4-NUT; ref. 17) were PCR-cloned into pDONR223, then transferred by gateway cloning into the pHAGE-P CMVt N-HA GAW expression vector derived from the PHAGE lentiviral vector (52).

A tetracycline-inducible HA-tagged NSD3Tr was gateway cloned from pDONR223 into the tetracycline-inducible pHAGE-P CMVt N-HA GAW expression vector and into the tetracycline-inducible pHAGE-TRE-HA (gift of Steven Elledge).

NMC cell lines, TC-797 (15), 10–15 (6), 8645 (17), 293T, U2OS, and C33A cells were maintained as monolayer cultures in DMEM (Invitrogen) supplemented with 10% (v/v) FBS (SH3008803; Hyclone) and 1% pen–strep (GIBCO/Invitrogen). The 797TRex cell line was created using Flip-In technology as described previously (Invitrogen; ref. 6) and maintained as above, but with the addition of Hygromycin (150 μg/mL; Sigma-Aldrich) and Blasticidin (7.5 μg/mL; Life Technologies) to maintain selection of cDNA insert and tet repressor genes, respectively. 797Trex/Flag-NLS-ET and N-BioTAP–NSD3–NUT cell lines were generated by recombination with the plasmid pcDNA5 FRT/TO-FLAG-NLS-ET and N-BioTAP–NSD3–NUT (above) using Flp-In technology (Invitrogen). The resulting cell line was maintained in DMEM (Invitrogen) supplemented with 10% (v/v) FBS (SH3008803; Hyclone), 1% pen–strep (GIBCO/Invitrogen), 7.5 μg Blasticidin/mL and 150 μg Hygromycin/mL. The 1221 cell line is derived from a lung metastasis from the index case of a 13-year-old female with NSD3–NUT-positive NMC. The 1221 cells were grown and maintained in WIT media as described previously (17, 53). Tet-inducible C33A cells (C33A-6TR) were established by transfecting C33A cells with pcDNA6/tet repressor (TR) vector (Invitrogen). Cells were selected and maintained with 5 μg/mL Blasticidin, and single-cell clones were obtained. Single-cell clones were then tested for leakiness using a tet-inducible enhanced green fluorescent protein (EGFP) construct, and clone 13 was chosen for further experiments due to a tight regulation of EGFP expression by the TR. The 1221, TC-797, 797TRex, 8645, and 10–15 cell lines were authenticated by FISH (above) demonstrating rearrangement of the NUT, BRD4, and/or NSD3 genes. The C33A cell line has been authenticated by documentation of p53 and pRb mutations (54). Neither the 293T nor U2OS cell lines have been authenticated.

Cells were plated at a density of 3,000 per well in a 96-well plate, and CellTiter-Glo (Promega) was used to determine cell viability as a measure of ATP content according to the manufacturer's instructions.

For TC-797, Per403, and 8645 cells, 7 × 10⁶ cells were transfected with 50 nmol/L siRNA using Nucleofector II (Lonza) and Amaxa solution R and plated in 100-mm cell culture dish. 1221 cells were transfected using RNAi-MAX (Invitrogen) using the manufacturer's protocol. Briefly, reverse transfection procedure was used to deliver 50 nmol/L siRNA to 33 × 10⁴ cells in a 6-well plate. Cells were analyzed for mRNA levels 24 hours after transfection. Sequences of siRNA used were: siControl, ON-TARGET plus siRNA #1 (Dharmacon; cat no. D-001810-01-20); siNUT-1 (targeting coding sequence), AAACUCAGAACUUUAUCCUUAUU; siNUT-2 (targeting the 3′-UTR), UUACCUUUGGAAGGAGCUA; siBRD4-1, 5′ siGENOME Human BRD4 (Dharmacon; cat no. D-004937-02); siBRD4-2, 3′ GGGAGAAAGAGGAGCGUGAUU; siNSD3-6, ON-TARGETplus Human WHSC1L1 (54904; Dharmacon; cat no. J-012875-06); siNSD3-7, ON-TARGETplus Human WHSC1L1 (54904) siRNA (Dharmacon; cat no. J-012875-07); siNSD3, 3′-1 CUGUAAACCUCUAAAGAAAUU; siNSD3, 3′-2 GAAAGGUGCCAGCGAGAUUUU; siJMJD6-12, GGUAUAGGAUUUUGAAGCA (Dharmacon; cat no. J-010363-12-0020); siJMJD6-13, GGAUAACGAUGGCUACUA (Dharmacon; cat no. J-010363-13-0020); siNSD3-06, GAACGUGCUCAGUGGGAUA (Dharmacon; cat no. J-012875-06-0020); siNSD3-07, GCUUGAGGUUCAUACUAAA (Dharmacon; cat no. J-012875-07-0020); siGLTSCRI-05, GUAAUGAUCGACCGAAUGU (Dharmacon; cat no. J-020751-05-0020); siGLTSCR1-08, CCACCACGUUCAAUGGGAA (Dharmacon; cat no. J-020751-08-0020); siATAD-05, UAGCAGAAAUGUACAACUA (Dharmacon; cat no. J-004738-05-0020); siATAD5-06, GCGCAAUAAUG UAUACUUU (Dharmacon; cat no. J-004738-06-0020); siCHD4-07, (Dharmacon; cat no. J-009774-07-0020); and siCHD4-08 (Dharmacon; cat no. J-009774-08-0020).

Immunofluorescence on TC-797 cells was performed as described previously (55) and nuclei were counterstained with ProLong Gold antifade reagent with 4′,6-diamidino-2-phenylindole (DAPI; Life Technologies). Primary antibodies used were anti-NUT (1:1,000, rabbit monoclonal clone C52; Cell Signaling Technology) and anti-HA (1:500, mouse monoclonal; Sigma-Aldrich). Secondary antibodies included goat anti-rabbit Alexa Fluor 594 and goat anti-mouse Alexa Fluor 488 (1:1,000; Life Technologies). Photographs were taken on a Nikon Eclipse E600 fluorescent microscope using a Spot RTSlider camera (Diagnostic Instruments, Inc.), and Spot Advanced software (Diagnostic Instruments, Inc.).

Quantitation of BRD4–NUT foci was performed by analyzing immunofluorescent images with ImageJ software. BRD4–NUT foci were identified by equal threshold adjustment of all images. The Analyze Particles function in ImageJ was used to determine foci number for each cell. Forty cells were counted for each experimental condition, and experiments were performed in triplicate. Statistics shown in all figures are from Student t tests (two-tailed).

Cells were transfected in 384-well format using 50 nmol/L control siRNA (above), NSD3 siRNA (above), BRD4 siRNA, and NUT siRNA (above), as described previously (6), or a dose range of JQ1 treatment. Cells were stained with AE1/AE3 antibody (1:4; Dako) to measure keratin intensity, Ki-67 antibody (1:500; Cell Signaling Technology), and nuclei were stained with Hoechst 33342 (4 μg/mL; Molecular Probes). Imaging for keratin expression was performed using the ImageXpress high-throughput microscope with MetaXpress software (Molecular Devices) as described previously (56). Shown are representative images taken at ×400 magnification. Each condition was performed in triplicate in 384-well plate format in three separate experiments, where wells were analyzed using MetaXpress with Multi Wavelength Cell Scoring for cell number, average keratin fluorescence pixel intensity, and % Ki-67 per well. Statistics shown are from Student t tests.

Total RNA was harvested at the indicated time points using TRizol (Invitrogen) and the RNeasy Mini Kit (Qiagen), and RNA (1 μg) was reversed transcribed into cDNA using the iScript cDNA Synthesis Kit (Bio-Rad). Quantitative real-time PCR (qRT-PCR) was performed as described previously (6) using the reference gene ribosomal protein L13a (RPL13a) mRNA levels as normalizing control. All qRT-PCR experiments are representative of triplicate qRT-PCRs from one of three independent experiments. qRT-PCR was performed in triplicate on a Bio-Rad iCycler in 96-well plate format with IQ SYBR Green supermix (Bio-Rad) and 1 μL of cDNA template per reaction. Amplification curves and C_t values were generated using MyiQ Single-Color Real-Time software (Bio-Rad). Primers used were: 3′ NSD3 fwd TTCTAGGAGTGCGGCCAAAG; 3′ NSD3r CAGCTCTCCACCATCTCCAC; 5′ NSD3f GCCCCAGTTCAGCCAATACT; 5′ NSD3r ACCATACAAGGCCACCAAGG; RPL13Af CCTGGAGGA GAAGAGGAAAGAGA; and RPL13Ar TTGAGGACCTCTGTGTATT TGTCAA, and using the TaqMan primers (Life Technologies), JMJD6 (Hs00397095_m1), NSD3 (Hs00256555_m1), GLTSCR1 (Hs00185249_m1), and ATAD5 (Hs00227495_m1).

Cell lysis and immunoblotting was performed as described previously (17). Antibodies used were rabbit anti-NUT (1:500, AX.1 rabbit polyclonal antibody; ref. 57), anti-GAPDH (1:5,000, mouse monoclonal 6C5; Life Technologies), anti-BRD4 (1:1,000; Bethyl Laboratories Inc.), anti-FLAG (1:1,000, mouse monoclonal; Sigma-Aldrich), anti-involucrin (1:1,000, mouse monoclonal; Sigma-Aldrich), anti-HA (1:1,000, mouse monoclonal; Sigma-Aldrich), anti-NSD3 (1:500, mouse monoclonal clone 2E9; Lifespan Biosciences), antiperoxidase complex/PAP antibody (1:5,000, rabbit polyclonal; Sigma-Aldrich), anti-histone H3 (1:1,000, mouse monoclonal ab10799; Abacam), anti-actin (1:1,000, mouse monoclonal clone 4; Millipore), and anti-p300 (1:1,000, mouse monoclonal clone RW128; EMD Millipore).

For coimmunoprecipitation experiments of HA-tagged proteins, C33A-6TR cells stably expressing HA-tagged constructs under a tet-inducible promoter were used. Protein extracts were prepared 24 hours after transfection or induction with doxycycline (1 μg/mL). Cells were lysed in lysis buffer [50 mmol/L Tris–HCl (pH 7.5), 150 mmol/L NaCl, and 0.5% Nonidet P-40] with freshly added protease inhibitors (Roche Complete, EDTA-free protease inhibitor cocktail; Roche). Extracts were adjusted for protein concentration, and 10% of extracts were used as input. Immunoprecipitations were performed using 15 μL of HA-resin (Sigma A2095). Extracts were incubated overnight at 4°C, and precipitated proteins were detected by Western blot analysis.

Formalin-fixed, paraffin-embedded cell blocks of cultured cells were prepared using Histogel (Richard-Allan Scientific) as described previously (2). Sections were stained with hematoxylin and eosin (H&E) or by IHC, which was performed on 5-mm sections prepared from formalin-fixed, paraffin-embedded cell blocks. Immunohistochemical stains were performed using anti-NUT antibody (1:100, rabbit monoclonal clone C52; Cell Signaling Technology), anti-involucrin antibody (1:12,000; Sigma-Aldrich), and Ki-67 (MIB-1 clone; Dako; ref. 2). Three hundred cells were counted per sample for Ki-67 percentage. Standard deviations for triplicate counts are shown in figures.

No potential conflicts of interest were disclosed.

Conception and design: C.A. French, S. Rahman, E.M. Walsh, A.R. Grayson, P.M. Howley

Development of methodology: C.A. French, S. Rahman, E.M. Walsh, A.R. Grayson

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C.A. French, S. Rahman, E.M. Walsh, S. Kühnle, A.R. Grayson, B.P. Rubin, C.R. Antonescu, R. Venkatramani

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C.A. French, S. Rahman, E.M. Walsh, S. Kühnle, A.R. Grayson, M.E. Lemieux, N. Grunfeld, C.R. Antonescu, P. Dal Cin, P.M. Howley

Writing, review, and/or revision of the manuscript: C.A. French, S. Rahman, E.M. Walsh, S. Kühnle, B.P. Rubin, C.R. Antonescu, S. Zhang, R. Venkatramani, P.M. Howley

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C.A. French, E.M. Walsh, S. Zhang

Study supervision: C.A. French, P.M. Howley

Karyotype of the cell line, and FISH analysis to confirm the NUT rearrangement: P. Dal Cin

The authors thank James E. Bradner (Dana-Farber Cancer Institute, Boston, MA) for sharing the BET inhibitor JQ1. The authors are grateful to Jeffrey Galligan in the Howley laboratory who generated the C33A-6TR cell line used in these studies. The authors also thank Ann Thomas and Michelle Speranza for their technical expertise that was very helpful for the cytogenetic analysis.

This work was supported by grants from the NIH: R01CA116720 (to P.M. Howley), 2R01CA124633-06A1 (to C.A. French), T32CA009361 (to S. Rahman), T32HL007627-27 (to E.M. Walsh); as well the Samuel Waxman Cancer Research Foundation (to C.A. French).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.