The case began innocuously. The patient developed persistent hoarseness after an upper respiratory infection, and flexible laryngoscopy revealed a tumor arising from the left true vocal fold and extending toward the anterior commissure. Scans showed the mass confined to the glottic and supraglottic regions of the larynx, with no lymph node involvement or spread to distant organs. A biopsy was read as poorly differentiated squamous cell carcinoma, the most common and most familiar cancer of the head and neck. Because the histology looked conventional, and because the patient was initially staged with a potentially curable stage II disease, no test for NUT carcinoma was performed at that time. That decision, the report makes clear, would shape everything that followed.

Given his youth, the clinical team pursued a larynx-preserving strategy. He received induction chemotherapy with docetaxel, cisplatin, and cetuximab, followed by definitive chemoradiotherapy delivering seventy gray of intensity-modulated radiation with concurrent carboplatin and 5-fluorouracil. The primary tumor vanished completely on endoscopy and imaging, and maintenance therapy with the oral agent S-1 followed. For ten months the remission held. Then, in August 20X1, a recurrent lesion appeared at the original site, this time with paralysis of the vocal fold on the same side. Surgeons performed a total laryngectomy, removing the entire voice box, and pathology again showed moderately to poorly differentiated squamous cell carcinoma. The tumor, everyone assumed, had been dealt with.

It had not. Five months after surgery the patient developed painful swallowing, and imaging revealed an ulcerative mass on the posterior pharyngeal wall along with a solitary lesion in the iliac bone of the pelvis. Chemotherapy with cisplatin and docetaxel failed within a single cycle. The immune checkpoint inhibitor nivolumab, which works well in a subset of head and neck cancers, also failed within weeks. Comprehensive genomic profiling using the FoundationOne CDx assay detected no actionable alterations and, crucially, did not detect the BRD4::NUTM1 fusion—a false negative that is well documented for NUTM1 rearrangements, since DNA-based sequencing depends on where the breakpoint falls and how well that region is captured. The tumor’s biomarker profile, with a low tumor mutational burden of four mutations per megabase and a microsatellite-stable genome, offered no foothold for immunotherapy.

The definitive answer came only after a third salvage operation. Radical surgery at Aichi Cancer Center involved circumferential pharyngectomy, removal of the cervical esophagus, and reconstruction with a free segment of jejunum. When yet another recurrence appeared around the permanent tracheostoma five months later, additional immunohistochemistry on the resected tissue revealed diffuse nuclear staining for the NUT protein with a characteristic speckled pattern, alongside expression of the squamous markers p63 and p40. That single stain established the true diagnosis: NUT carcinoma. Retrospective testing of archived specimens using fluorescence in situ hybridization and RNA-based reverse transcription polymerase chain reaction confirmed a BRD4::NUTM1 fusion and proved the tumor had been NUT carcinoma from the very beginning, hidden behind a costume of ordinary squamous cancer.

The biology explains both the disguise and the ferocity. NUT carcinoma is driven by fusion proteins that typically join the tandem bromodomains of BRD4, which latch onto acetylated histones, to the NUT moiety, which recruits the p300/CBP acetyltransferases. The result is enormous hyperacetylated chromatin domains that force MYC expression and lock immature epithelial cells into a perpetual, proliferating, non-differentiating state. Because the cells still express squamous markers such as p63 and p40, pathologists see a plausible squamous cell carcinoma under the microscope. Only diffuse NUT immunostaining or molecular detection of the NUTM1 rearrangement separates the two diseases, and the report argues forcefully that young, non-smoking patients with aggressive, poorly differentiated tumors at midline or head and neck sites should trigger NUT-specific testing early.

With the diagnosis finally in hand, the team tried what mechanistic logic suggested should work. A bromodomain and extra-terminal, or BET, inhibitor—the class of drugs designed precisely to displace BRD4 from acetylated chromatin—was given through a clinical trial. The response was swift and dramatic: imaging at three months showed a marked partial regression. But after roughly six months, new lesions emerged and treatment stopped after eight cycles. Resistance to BET inhibition is thought to arise not through target mutations but through adaptive rewiring, in which tumor cells restore MYC-centered transcriptional programs using compensatory co-regulators such as p300/CBP and histone deacetylases, or bypass signaling through the MAPK/ERK pathway. A trial of the histone deacetylase inhibitor vorinostat, intended to attack that same acetylation circuitry from the opposite direction, produced no control whatsoever, consistent with functional redundancy between HDAC1 and HDAC2.

Even intensive cytotoxic chemotherapy delivered only borrowed time. Alternating cycles of vincristine, doxorubicin, and cyclophosphamide with ifosfamide and etoposide—a regimen borrowed from pediatric sarcoma practice—shrank the tumor markedly after two cycles, a reminder that this highly proliferative cancer retains some sensitivity to DNA-damaging agents. Yet regrowth appeared during the third cycle. Paclitaxel with cetuximab failed twice, and pembrolizumab combined with cisplatin and 5-fluorouracil achieved nothing. By June 20X4 the disease had replaced much of the anterior neck, and care shifted to palliation. Weekly applications of Mohs paste, a zinc chloride-based preparation rarely discussed in modern oncology, were used to harden the tumor surface and control malodorous exudate and bleeding, while morphine managed pain. The patient died in late July with his family present, having lost more than thirty kilograms from his baseline weight.

The autopsy delivered the report’s most consequential insight. Although the lungs were studded with innumerable metastatic nodules, each only a few millimeters across, the immediate cause of death was not disseminated cancer. It was overwhelming infection driven by uncontrolled locoregional disease—a massive twenty-three by twenty centimeter infiltrative tumor occupying the neck, invading the reconstructed jejunum, and compromising the airway, complicated by bilateral cavitary lung abscesses. This distribution matters therapeutically: fatal outcomes in NUT carcinoma may hinge less on distant metastasis than on relentless local progression along the aerodigestive tract, reinforcing evidence from prior series that durable locoregional control with surgery and radiotherapy underpins the only long-term survivals recorded.

The authors place their experience against the eighteen previously documented laryngeal cases, most of which arose in the supraglottis and claimed their victims within a year. Their patient’s glottic origin, four-year course, and autopsy characterization are all firsts for this site, and his survival of more than two years beyond the detection of an iliac bone metastasis stands against registry data showing zero percent two-year survival for metastatic head and neck NUT carcinoma. The broader lesson is uncomfortable but clear: despite its microscopic resemblance to squamous cell carcinoma, this is a transcription-addicted malignancy that standard head and neck regimens—induction chemotherapy, platinum-taxane combinations, and checkpoint inhibitors—all failed to hold in check. Some researchers now argue NUT carcinoma should be regarded as a molecularly defined subtype of squamous carcinoma rather than a separate entity, a framing that would push NUT testing into routine practice. Whether future patients benefit will likely depend on diagnosing the fusion early and combining epigenetic drugs with cytotoxic debulking and definitive local therapy, rather than adding agents one at a time to a network evolution too easily circumvents.

Subject of Research: Diagnosis and treatment resistance in BRD4::NUTM1 fusion laryngeal NUT carcinoma

Article Title: Laryngeal NUT Carcinoma With BRD4::NUTM1 Fusion: A 4‐Year Clinical Course Highlighting the Limitations of Current Multimodal Therapy

Article References: Kanno, M., Sasaki, C., Kato, E., Terada, H., Hanai, N., Sasaki, E., Fukushima, M., Masuishi, T., Takanari, K., Fukada, Y., Miyazaki, Y., Sonoda, Y., Kato, Y., Morikawa, T., Takabayashi, T., & Fujieda, S. (2026). Laryngeal NUT Carcinoma With BRD4 :: NUTM1 Fusion: A 4‐Year Clinical Course Highlighting the Limitations of Current Multimodal Therapy. Cancer Reports, 9(9), Article e70679. https://doi.org/10.1002/cnr2.70679

Image Credits: AI Generated

DOI: 10.1002/cnr2.70679

Keywords: NUT carcinoma, BRD4::NUTM1 fusion, laryngeal cancer, BET inhibitor, vorinostat, squamous cell carcinoma, NUTM1 rearrangement, chemoradiotherapy, immune checkpoint inhibitor, autopsy findings, epigenetic therapy, rare cancer

The scale of methylation disruption in cancer is staggering. Tumor genomes characteristically display widespread global hypomethylation, in which vast stretches of normally methylated DNA lose their marks, accompanied by focal hypermethylation concentrated at key regulatory regions, particularly the CpG-rich promoters of tumor suppressor genes. The result is an extensive rewiring of gene regulatory networks: genes that restrain cell division are switched off, repetitive elements that are normally locked down become active, and the carefully orchestrated patterns that define cell identity dissolve into chaos. These alterations are so consistent that methylation signatures now underpin diagnostic and prognostic assays used in clinics worldwide, helping pathologists classify tumors, predict patient outcomes, and even trace the tissue of origin of metastatic cancers.

Yet a profound gap separates correlation from causation. The overwhelming majority of cancer-associated methylation changes have never been proven to drive tumor behavior; they may instead be passive consequences of the transformed state. Distinguishing driver marks from passenger marks using observational methylome profiling alone has proved remarkably difficult, because the epigenome is dynamic and context-dependent. This ambiguity has real consequences for drug development. Conventional hypomethylating agents, such as nucleoside analogs that trap and deplete methyltransferase enzymes, can partially reverse cancer-associated methylation alterations and have demonstrated clinical efficacy in selected hematological malignancies. But these drugs act genome-wide, with no capacity to target a specific promoter or enhancer. Their blunt mechanism limits both mechanistic insight—since any observed effect could stem from thousands of altered loci—and therapeutic precision, since desirable and undesirable demethylation occur simultaneously across the genome.

CRISPR-based epigenome editing offers a fundamentally different approach. Rather than cutting DNA, these systems repurpose catalytically inactive Cas proteins, most famously dCas9, as programmable delivery vehicles. Guided to a chosen genomic address by a complementary guide RNA, dCas9 can be fused to methylation ‘writers’—enzymes such as DNMT3A and DNMT3L that install methyl marks—or methylation ‘erasers’, including TET1 catalytic domains and engineered demethylases that oxidize and remove them. Because the underlying DNA sequence is left untouched, the editing is, in principle, reversible, mirroring the reversible nature of the epigenetic alterations themselves. The result is a technology capable of quantitatively tuning methylation at individual promoters, enhancers, and CpG shores—transition zones flanking CpG islands that are frequent sites of cancer-associated methylation gain.

The experimental power of this approach lies in its precision. A single tumor suppressor promoter can be hypermethylated in a controlled setting to test whether silencing of that gene is sufficient to confer growth advantages, or demethylated to ask whether restoring its expression suppresses malignant behavior. Similar logic applies to enhancers, whose methylation status can govern the activity of oncogenes and immune-modulatory genes alike. By editing specific loci, researchers can interrogate mechanisms of chemoresistance, asking whether methylation changes at particular drug-response genes cause treatment failure or merely accompany it. They can also probe immune modulation, since methylation patterns influence the expression of antigen-presenting molecules and immune checkpoint pathways that determine whether tumors are recognized by the immune system. In this way, methylation editing transforms the epigenome from a readout of tumor state into an experimentally addressable variable.

The review also highlights the modular nature of modern CRISPR recruitment platforms, which extend the technology well beyond simple enzyme fusions. Systems such as SunTag, MS2-based scaffolds, and related multiplexing strategies allow multiple effector domains to be recruited to a single target site, amplifying the magnitude of methylation change or combining writers and erasers with transcriptional activators and repressors. Such platforms enable graduated, quantitative manipulation of methylation levels rather than all-or-nothing switching, which is critical because the relationship between methylation density and gene expression is often dose-dependent. Combined with single-cell and multi-omics readouts, these tools allow researchers to observe how a targeted methylation edit propagates through chromatin state, transcription, and ultimately cellular phenotype in diverse cancer models, from cell lines to patient-derived organoids.

None of this progress comes without caveats, and the authors are candid about the obstacles that stand between bench and bedside. Off-target activity remains a central concern: both the DNA-binding specificity of guide RNAs and the promiscuity of the tethered enzymes can produce unintended methylation changes at sites resembling the intended target, and even at sites tethered transiently by direct enzyme-to-dCas9 fusions. Because methylation changes can be heritable through cell division, a single off-target event could have lasting consequences. Context-dependency poses a further complication. The same edit can yield different outcomes depending on chromatin context, cell type, CpG density, and the developmental or disease state of the cell, complicating efforts to generalize results across models. Delivery and durability add clinical dimensions to these challenges: getting large editing constructs efficiently into tumor cells—whether in culture or in a patient—and ensuring that the methylation change persists long enough to be therapeutically meaningful remain unsolved engineering problems.

The path forward, the review argues, lies in integration. Locus-specific editing is poised to converge with comprehensive multi-omics profiling, in which methylome, transcriptome, chromatin accessibility, and three-dimensional genome architecture are measured simultaneously before and after an edit. Machine-learning-guided target selection promises to accelerate this process, mining large epigenomic datasets to prioritize the methylation changes most likely to act as drivers, and to predict how editing a given locus will ripple through regulatory networks. Rational combination therapies represent the clinical endpoint of this vision: methylation editing might, for instance, be paired with immune checkpoint inhibitors by demethylating and reactivating antigen presentation genes, or with conventional chemotherapy by erasing methylation marks that drive drug resistance. Each of these strategies depends on the causal knowledge that only targeted editing can generate.

The significance of this synthesis extends beyond the technical details. For a generation, epigenetic therapy has been constrained by an uncomfortable irony: doctors could alter the cancer epigenome, but only in a coarse, genome-wide fashion that revealed little about which alterations mattered. CRISPR-based methylation editing inverts that logic, making the epigenome writable at single-locus resolution and thereby testable in a way no previous technology allowed. If the challenges of specificity, delivery, and context can be met, the result would be a new class of therapeutics that silence oncogenes or resurrect tumor suppressors without touching a single base of DNA sequence. For now, the technology’s greatest value is as an engine of discovery—a way to finally separate the drivers from the passengers in the cancer methylome. But the trajectory is clear, and the authors’ message is that rewriting the cancer epigenome has moved from metaphor to laboratory reality, with clinical translation as the next horizon.

One reason methylation is such an attractive target for editing is its intrinsic stability. Unlike histone modifications, which turn over rapidly and often require continuous reinforcement, DNA methylation at CpG sites can be faithfully copied through cell division by maintenance methyltransferases that recognize hemimethylated DNA after replication. This means a single, well-placed edit has the potential to persist across many generations of cells, a property that is both a strength for durable therapy and a liability if the wrong loci are modified. The review’s emphasis on precision reflects this double-edged character of the methylation mark.

The choice of effector enzymes also matters greatly. DNMT3A, one of the writer domains commonly fused to dCas9, is itself a gene frequently mutated in hematological malignancies, illustrating how the machinery of methylation regulation is intimately entangled with cancer biology. On the eraser side, catalytic domains derived from the TET enzymes, which normally initiate active demethylation through oxidation of 5-methylcytosine, can be repurposed for targeted removal of marks. That these same pathways are dysregulated in tumors underscores why locus-specific manipulation is such a valuable experimental tool: it allows individual components of an altered network to be tested in isolation.

The work emerges from a research group with a sustained focus on cancer epigenomics at the University of Otago’s Department of Pathology and Molecular Medicine, and the article is published open access under a Creative Commons Attribution license, supported by New Zealand funding bodies including the Health Research Council and the Marsden Fund. Open availability is significant for a methods-oriented review, since widespread adoption of these editing approaches depends on researchers across disciplines and resource settings being able to evaluate the technical trade-offs in detail.

It is also worth noting that the review appears at a moment when epigenome editing more broadly is maturing beyond proof-of-principle demonstrations. The same dCas9 recruitment logic underlies efforts to edit histone marks, chromatin architecture, and transcriptional states, suggesting that methylation editing will ultimately function as one module within a larger toolkit for writing and rewriting the regulatory layer of the genome.

Subject of Research: CRISPR-based DNA methylation editing technologies for interrogating and potentially treating aberrant cancer epigenomes

Article Title: Rewriting the cancer epigenome: CRISPR technologies for DNA methylation editing

Article References: Sadeghi, N., Rodger, E. J., & Chatterjee, A. (2026). Rewriting the cancer epigenome: CRISPR technologies for DNA methylation editing. Cellular and Molecular Life Sciences. https://doi.org/10.1007/s00018-026-06429-1

Image Credits: AI Generated

DOI: 10.1007/s00018-026-06429-1

Keywords: cancer, epigenetics, DNA methylation, CRISPR, DNA methylation editing, dCas9, epigenome editing, tumor suppressor genes, off-target effects, epigenetic therapy, gene regulation, precision medicine