NC SNU-2972-1, SNU-3178S, Ty-82 and HCC2429 cell lines were kindly provided by Prof. Kim at the Seoul National University Hospital [25] and synovial sarcoma HS-SY-II cells were provided Prof. Seo from the Department of Orthopedics in Samsung Medical Center. Four NC cells were maintained in RPMI-1640 (SH30027.01, Cytiva - HyClone Laboratories Inc., Logan, UT) and HS-SY-II cells were maintained in DMEM (SH30243.01, Cytiva - HyClone Laboratories Inc.) supplemented with 10% fetal bovine serum (16000-044, Gibco, Grand Island, NY) and 1% penicillin–streptomycin–amphotericin B (15240-062, Gibco) in an incubator maintained at 37°C with 5% CO₂. All cell lines were performed short tandem repeat analysis for human cell line authentication.

A total of 18 cases diagnosed with NC between 2010 and 2020 were selected on the basis of electronic medical records and a previous study [26] at the Samsung Medical Center. Of these, 17 samples were available, on which quantitative reverse transcription polymerase chain reaction (qRT-PCR) and/or NGS assays were performed, to identify the partner genes of NUTM1 fusion from formalin-fixed paraffin-embedded (FFPE) tissue blocks. Hematoxylin and eosin–stained slides were prepared on 4-μm-thick whole-slide tissue sections of FFPE samples. All samples were stained with hematoxylin and eosin.

RNA was extracted using the RNeasy FFPE Kit (catalog No. 73504, Qiagen, Hilden, Germany) from the collected NC FFPE samples with sufficient tissue remaining. The concentration and purity of the extracted RNA were checked using a Nanodrop and Qubit 3.0 Fluorometer (Q33216, Thermo Fisher Scientific, Waltham, MA). Samples were primarily subjected to qRT-PCR with BRD4 exon 11 and NUTM1 exon 3 primer pairs. The samples that were not detected in the qRT-PCR were subsequently subjected to targeted NGS assay. The primers used for qRT-PCR included BRD4 F 5′-GAAAGTTGATGTGATTGCCG-3′ and NUTM1 R 5′-CTGGTGGGTCAGAAGTTGGT-3′. For targeted NGS assay, in four cases, the libraries were prepared using the Archer library kit (SK0093, ArcherDX, Boulder, CO). We proceeded with PreSeq QC using 10× VCP primer mix (SA1026, ArcherDX) after synthesizing first-strand cDNA. The second-strand cDNA was synthesized by selecting samples with VCP CT values below 31. Unidirectional gene-specific primers were used to enrich the target regions, followed by NGS using the Illumina MiSeq platform (Illumina, San Diego, CA). In three cases, libraries were prepared using the Sureselect RNA Direct Library Kit (catalog No. G7564, Agilent Technologies, Santa Clara, CA). The libraries were sequenced using CancerSCAN (Seoul, Korea), a targeted sequencing platform developed at the Samsung Genomic Institution (http://kr-geninus.com).

To assess the effect of siRNAs targeting the junction region of the B4N fusion gene, a series of 15 siRNA candidates targeting the junction region formed by the fusion of BRD4 exon 11 and NUTM1 exon 3 were designed and modified with 3′-dTdT overhangs (S1 Table). ON-TARGETplus Human NUTM1 siRNA (L-022211-02-0005, Dharmacon, Horizon Discovery, Ltd., Lafayette, CO) was used as a positive control NUTM1 siRNA. For each sample, 50 nM siRNAs were transiently transfected using the Neon Transfection System into HCC2429 cells expressing the B4N fusion gene. After 72 hours, the representative images were acquired, and the cells were then harvested for qRT-PCR assays. To determine the effect of siRNAs targeting the junction region of the SS fusion gene, a series of 16 siRNA candidates targeting the junction region formed by the fusion of SS18 exon 10 and SSX1 exon 6 were designed and modified with 3′-dTdT overhanging sequences (S2 Table). For each sample, 50 nM siRNAs were transiently transfected into HY-SY-II cells expressing the SS fusion gene. Representative images were acquired after 72 hours of transfection, and the cells were then harvested for qRT-PCR assays.

Total RNA from cell pellets was extracted using the RNeasy Mini Kit (catalog No. 74004, Qiagen), and cDNA was synthesized from 1 μg total RNA by using SuperScript III Reverse Transcriptase (18080093, Invitrogen, Thermo Fisher Scientific), according to the manufacturer’s instructions. qRT-PCR was conducted using a PRISM 7900HT Fast Real-Time PCR System (Applied Biosystems, Foster City, CA). The sequences of the primers used are listed in S3 Table. The experiments were performed in triplicate for all samples and t test was performed for statistical analysis.

HCC2429 or HS-SY-II cells were seeded into 6-well plates, and 50 nM siRNAs were transiently transfected into these cells. Twenty-four hours later, 5,000 cells were plated into 96-well plates. After 48 hours, the medium was replaced with 10 μL of CellTiter 96 AQ One Solution reagent (G3582, Promega, Madison, WI) with 90 μL of DMEM per well. The plate was incubated at 37°C for 4 hours in a humidified atmosphere with 5% CO₂, and the absorbance was recorded at 490 nm using a SpectraMax 96-well plate reader (Molecular Devices, San Jose, CA).

NUT (1:50, NUT [C52B1] rabbit mAb, catalog No. 3625, Cell Signaling, Danvers, MA) and MYC (1:200, anti-c-MYC antibody [Y69], catalog No. ab32072, Abcam, Cambridge, UK) antibodies were used for IHC analysis, using a previously described procedure [26]. Immunoblotting was performed using antibodies against NUT (1:500, NUT [C52B1] rabbit mAb, catalog No. 3625), BRD4 (1:1,000, [E2A7X] rabbit mAb, catalog No. 13440), and MYC (1:1,000, c-MYC antibody, catalog No. 9402) (all from Cell Signaling Technology); SS18 (SYT) (1:1,000, [H-80], catalog No. sc-28698), β-actin (1:1,000, [C4], catalog No. sc-47778], and GAPDH (FL-335, catalog No. sc-25778) (all from Santa Cruz Biotechnology, Inc.), were used as reference, all under standardized procedures.

The experimental protocols were approved by the Animal Care and Use Committee of Samsung Biomedical Research Institute (Seoul, Korea). The procedures followed the institutional and National Institutes of Health guidelines for laboratory animal care, and animals were housed in an Assessment and Accreditation of Laboratory Animal Care International (AAALAC International)-accredited facility. The study was carried out in compliance with the ARRIVE guidelines. Tumors were generated in 6-week-old BALB/c nu/nu female mice. A total of 4×10⁶ HCC2429 cells were inoculated subcutaneously into the left flank of the mice in 50 μL of a mixture of Matrigel and RPMI media (1:1 dilution). For in vivo experiments, Accell siRNA B4N #9 was customized in Dharmacon, Accell non-targeting pool (catalog No. D-001910-10-05, Horizon Discovery, Ltd.) and Accell cyclophilin B pool (catalog No. D-001920-10-05, Horizon Discovery, Ltd.) were purchased from Dharmacon as controls. Cells were transfected with siRNAs by using the Neon Transfection System (Thermo Scientific). When tumors reached an average volume of approximately 50 mm³, the mice were randomly divided into control, B4N siRNA (2 mg/kg), and B4N siRNA (6 mg/kg) groups, with five animals in each group. Mice were intratumorally injected with the siRNA every 3 or 4 days for 20 days in 20 μL serum-free media (B-005000-500, Horizon Discovery, Ltd.) per mouse. The tumor size was measured using calipers and recorded daily. We also injected cyclophilin B siRNA into three mice. The fluorescence was measured at 24- and 48-hour post-injection. Tumor volumes were calculated by taking the length to be the longest diameter across the tumor and width to be the corresponding perpendicular diameter, and then using the following formula: (length× width²) mm³×0.5. None of the animals died during the experimental period. The animals were sacrificed by means of CO₂ inhalation or cervical dislocation at 48 hours after the last dose of treatment. The tumors were resected, fixed with 4% paraformaldehyde, and paraffin-embedded for sectioning and immunohistochemical staining with NUT antibody.

For cell-free RNA (cfRNA) extraction, blood was collected in 8 mL EDTA blood tubes and centrifuged at 120 ×g for 20 minutes at RT. Then, 4 mL of the upper phase was transferred into new tubes and centrifuged at 360 ×g for 20 minutes at RT. The supernatant was removed to collect platelet pellets with 200 μL plasma at the bottom of the tube. The platelet-rich plasma (PRP) was stored at −80°C. cfRNA was isolated from 200 μL of PRP by using the miRNeasy serum/plasma kit (cat. No. 217184, Qiagen) following the manufacturer’s instructions. cfRNA was converted to cDNA by using Superscript III (Life Technologies, Darmstadt, Germany), and 8 μL of cDNA was used for droplet generation in a QX200 ddPCR system (Bio-Rad Laboratories, Inc., Hercules, CA). PCR amplification was performed using the following primer pairs (S2 Table), with GUSB as an internal control. The concentration of transcript copies in the 20 μL ddPCR reaction was reported automatically by QuantaSoft software, which was then used to determine the transcript copies per plasma using the formula: (copies per well×RNA elution volume)/(PCR volume×plasma volume).

Between 2017 and 2020, 18 patients with NC were identified through a retrospective collection of cases. The pathological characteristics of the patients are shown in Table 1. The median age at diagnosis was 46.5 years (range, 18 to 75 years) and 65% were male. Thirteen cases (72%) arose from thoracic origin such as the lung, trachea, pleura, and bronchus, and five cases occurred in the liver, nasal cavity, glottis, and metastatic lymph node. To determine the partner gene fused with NUTM1, qRT-PCR or NGS assays were performed using FFPE tumor samples. As expected, the majority of NUTM1 fusion partners were BRD4 (12 of 18, 66.7%), of which BRD4 exon 11 was found in 10 and BRD4 exon 14 was found in two cases, with BRD4 fused to NUTM1 exon 3 in both. NSD3-NUTM1 fusion was identified in two cases (of 18, 11%) (Fig. 1B). In four cases, the fusion partner genes fused to NUTM1 were not identified because one was not available for the test and two cases were not detected in the qRT-PCR assay. Unfortunately, in case No. 2, NUTM1 fusion gene was not detected in either the NGS or qRT-PCR assays. This is probably because the adequate amounts of tumoral DNA could not be extracted from FFPE samples for evaluation of the fusion gene.

Considering that the NUTM1 rearrangement can activate oncogenes such as MYC, SOX2, and TP63, and that MYC has been reported as a downstream target of the BRD4-NUTM1 fusion protein [10], the NUTM1 fusion-positive samples were further explored using MYC IHC assays (Fig. 1A, S4 Fig.). 10 cases were stained with anti-MYC and the representative cases with the different partners of NUTM1 fusion (case No. 7: NSD3, case No. 15: BRD4) are presented in Fig. 1A. The expressed MYC was found to be localized in the nuclear compartment and was colocalized with NUT in all cases, regardless of the partner genes (S4 Fig.). As previously reported, our results showed that the expression of NUT fusion protein is strongly correlated with MYC expression in NC.

NUTM1 fusion in NC is a major oncogenic driver gene, and we hypothesized that if the fusion oncogene was specifically eliminated from NC, the tumor growth and metastasis could be suppressed or delayed. To regulate an oncogenic function of the fusion by targeting the fusion-specific sequence, we preferentially identified the partner genes in cells with NUTM1 fusion using Sanger sequencing (Fig. 1C). Similar results obtained from clinical samples were observed in cell lines. BRD4-NUTM1 (B4N) fusion gene has been detected in three NC cell lines (SNU-2972-1, Ty82, and HCC2429), in which BRD4 exon 11 (NM_058243.2) and NUTM1 exon 3 (NM_001284292.2) were fused, and BRD3-NUTM1 fusion has been detected in SNU-3178S cell line (Fig. 1C). To explore whether targeting the junction region of a fusion oncogene is therapeutically useful, we designed a series of 15 siRNA candidates targeting the junction region of B4N (Fig. 2A, S1 Table). The candidate siRNAs were transiently transfected into the HCC2429 cells expressing B4N and 72 hours later, the relative mRNA level of parental genes and fusion gene was calculated using qRT-PCR (S5 Fig.). Commercial NUTM1 and scrambled control siRNAs were used as positive control and negative control. Next, B4N siRNA Nos. 6, 7, and 9 were treated to 293FT cells to test the effect of the siRNAs in normal cells, and the wild-type BRD4 mRNA level was measured with qRT-PCR. As expected, the junction target siRNA (B4N siRNA) did not have an effect on the BRD4 expression level (data not shown). As shown in Fig. 2A, we found that the relative mRNA level of B4N fusion decreased upon treatment with siRNAs Nos. 7 (p=0.01), 9 (p < 0.001), and 12 (p < 0.01), similar to that observed in treatment with NUTM1 siRNA, without a significant decrease in endogenous parent genes. siRNAs Nos. 12 and 15 also effectively inhibited B4N; however, they induced a concomitant decrease in the mRNA levels of B4N fusion and endogenous parent BRD4 as well. To further investigate whether the junction-specific siRNAs could affect the oncogenic function of B4N fusion, the three siRNA (Nos. 7, 9, and 12) was treated in the cell and the relative growth rate was measured using cell viability assay, and representative images were acquired (Fig. 2B and C). siRNA No. 6 was used as the negative control. As shown in Fig. 2B and C, when NUTM1 expression was downregulated by B4N-specific siRNA Nos. 7 and 9 (Fig. 2B), the relative cell growth was also delayed compared to the negative control (p < 0.001 in PC, Nos. 7, 9, 12) (Fig. 2C). These results were further investigated in synovial sarcoma. Synovial sarcoma is an aggressive neoplasm driven by the SS18-SSX fusion [27] and to determine whether siRNA targeting the junction region of a fusion gene has an inhibitory effect on cancer proliferation, 16 candidate siRNAs targeting the junction region of SS18-SSX (SS) were designed and experimented with HS-SY-II cells expressing SS fusion gene in the same manner as described above (S6A Fig.). As expected, a similar result was observed upon SS fusion gene-specific siRNA treatment. siRNA Nos. 3, 6, 7, 11, and 12 targeting the junction region of SS fusion gene induced a decrease in the relative level of SS fusion transcript and did not affect the parent mRNA expression (p < 0.001) (S6A Fig.). We showed that when SS fusion protein expression was downregulated by siRNAs Nos. 3, 6, and 11, the relative cell growth was delayed compared to that of the negative control (p < 0.001) (S6C Fig.). Similar to B4N fusion gene-specific siRNA, SS fusion gene-specific siRNAs Nos. 3, 6, and 11 selectively downregulated the expression of SS fusion protein, without affecting parent SS18 protein expression (S6B Fig.). SS siRNA No. 12 caused a decrease in SS mRNA level, but decreased the parent SS18 protein, rather than the SS fusion protein (S5B Fig.). Taken together, we showed that siRNAs targeting the junction region of a fusion gene could reduce oncoprotein expression and negatively regulate cellular growth without affecting parental protein expression. Therefore, we suggest that targeting the junction sequence of a fusion gene using siRNA could be a useful strategy to treat incurable diseases at the genetic level.

Based on in vitro findings, we observed the effect of siRNAs targeting the junction region of the fusion gene in an in vivo experiment. For xenograft mouse experiments, B4N siRNA #9 was synthesized by Dharmacon Research with a self-delivery-modified siRNA system (Accell siRNA). HCC2429 cells were injected subcutaneously into the flanks of nude mice on day 0. When the average tumor volume reached approximately 50 mm³, the siRNAs were directly injected into the tumor areas at dosages of 2 mg/kg or 6 mg/kg, twice a week, for a total of 5 times (Fig. 3A). The localized delivery of the siRNA to a tumor site was validated using Accell cyclophilin B control siRNA, which is a highly reliable positive control for delivery and RNAi efficiency (Fig. 3B). As shown in Fig. 3C, tumor regression was observed in the B4N siRNA No. 9 treatment groups, regardless of the treatment concentration. In the 2 mg/kg treatment group, a statistically significant decrease was observed, as compared to that in the control group (p=0.024). Although not statistically significant, there also was a decrease in tumor size in the 6 mg/kg group. To determine B4N expression in xenograft tumors, the tumors were analyzed using qRT-PCR and Western blot (Fig. 3D and E). The tumors showed that B4N mRNA levels significantly decreased in the siRNA treatment groups, as compared to the control group (Fig. 3D); in addition, B4N protein expression was also reduced upon treatment of the tumors with siRNA (Fig. 3E, arrowhead).

We next evaluated whether the junction sequence could be used to detect fusion oncogenes in the circulating blood of patients. This hypothesis was tested in case No. 13, a 45-year-old man, who was diagnosed with NC in the lung in November 2019 as of pT4N2M0, stage IIIB (Table 1). He underwent pneumonectomy and concurrent chemoradiation therapy following chemotherapy. Three months later, in July 2020, multiple bone metastases were detected in his case. Blood sample was the first obtained from the patient in June 2020, before radiotherapy, and subsequent blood samples were collected every 3 weeks for 28 weeks (Fig. 4B). To detect the circulating fusion oncogene in the blood, PRP was prepared from whole blood (5 mL) by means of a two-step centrifugation method (Fig. 4A). His blood was analyzed a total of nine times and 1-D fluorescence amplitude plots presented the oncogenic B4N fusion gene (S7 Fig.). To explore the association between disease progression and the number of detected droplets, the number of drops per one mL of sample was counted and plotted in Fig. 4B. The droplets in PRP were first detected when multiple tumor metastases occurred and interestingly, when the patient received radiotherapy, the plots of the B4N fusion gene immediately decreased and then increased again after a certain period of time (Fig. 4B). This showed that radiotherapy had worked for his tumor; however, unfortunately, the effect was maintained for less than 6 weeks. In this experiment, we found that the junction sequence of a gene fusion could be used to monitor the therapeutic response of patients.