Immune biomarker landscape and fusion partner–phenotype associations in thoracic and head-and-neck NUT carcinoma

Abstract

Background:

Thoracic and head and neck NUT carcinoma is a rare, highly aggressive NUTM1-rearranged malignancy with substantial diagnostic difficulty, marked phenotypic heterogeneity, and no established biomarker framework for immunotherapy stratification. Although fusion partners may contribute to clinicopathologic diversity, their relationship to phenotype and immune biomarkers remains insufficiently defined.

Materials and methods:

We performed an integrated literature-based analysis of thoracic and head and neck NUT carcinoma using English-language publications indexed in the Web of Science Core Collection from 1991 to 2025. Eligible studies were screened for clinicopathologic, fusion, and immune-biomarker data. Case-level variables were harmonized into anatomic and lineage-based strata to evaluate associations between fusion partners, phenotype, and immune features. PD-L1 status was harmonized into three descriptive categories based on author interpretation and/or reported TPS/CPS thresholds, and a binary sensitivity framework, Negative versus Any expression, was additionally applied to reduce the impact of heterogeneous assays and scoring systems. MSI and TMB were also extracted when available. Associations were assessed using two-sided Fisher’s exact test with odds ratios and 95% confidence intervals. To strengthen the robustness and translational relevance of the findings, we additionally analyzed an independent retrospective institutional cohort of thoracic and head and neck NUT carcinoma.

Results:

The integrated clinicopathologic dataset included 229 thoracic/head and neck cases with definitive fusion annotations from 109 studies. Squamous lineage predominated (71.62%), but non-squamous and mixed phenotypes were also identified. BRD4::NUTM1 was significantly enriched in squamous lineage and thoracic squamous subsets, whereas YAP1::NUTM1 and MGA::NUTM1 were preferentially associated with non-squamous phenotypes and specific strata. In the immune-biomarker subset, PD-L1 was predominantly negative (76.92%), with high expression being uncommon (6.41%). Thoracic squamous tumors showed significant enrichment for PD-L1 negativity. Among cases with both fusion and PD-L1 data, a binary sensitivity analysis confirmed that PD-L1 negativity predominated across fusion-defined subsets, with no significant association between fusion partner and PD-L1 status. Across reported MSI/TMB cases, tumors were uniformly MSS and generally low in TMB. An independent institutional cohort of 59 patients partially corroborated the major literature-derived signals, particularly BRD4::NUTM1 predominance among fusion-annotated cases and PD-L1 negativity among tested tumors, with no significant difference from the literature-derived pooled estimates.

Conclusion:

This study defines the immune biomarker landscape of thoracic and head and neck NUT carcinoma within a clinicopathologic and fusion-partner framework. The predominance of PD-L1 negativity together with MSS/low-TMB features suggests a generally immunologically “cold” phenotype in most reported tumors, while exploratory fusion partner–specific enrichment patterns may help refine diagnostic suspicion and generate hypotheses for future biomarker-guided stratification. These findings provide a translational basis for rare-cancer immunobiology research and for the rational design of biomarker-integrated prospective studies.

1 Introduction

NUT carcinoma is a rare yet highly aggressive malignancy driven by NUTM1 gene rearrangements. The prototypical t(15;19)-associated aggressive carcinoma was first reported in 1991 (1). Early reports emphasized tumors arising in midline structures, leading to the historical term “NUT midline carcinoma.” The entity subsequently gained broader recognition in WHO classification frameworks around 2015 onward, including thoracic pathology contexts (2). Importantly, NUTM1 rearrangement is not exclusive to NUT carcinoma and can also occur in other neoplastic contexts, including round cell sarcomas, acute leukemia, and selected adnexal tumors, underscoring the need for careful clinicopathologic correlation and molecular confirmation (3). In routine practice, NUT carcinoma often presents as a rapidly progressive poorly differentiated or undifferentiated malignancy with limited radiologic and morphologic specificity, and it is frequently misclassified as poorly differentiated squamous cell carcinoma, undifferentiated carcinoma, or other small round cell/sarcomatoid tumors (4, 5). Definitive diagnosis typically relies on NUT immunohistochemistry (NUT-IHC) demonstrating nuclear positivity, followed by confirmatory testing with FISH or NGS to document NUTM1 rearrangement. Owing to disease rarity, limited awareness, and marked morphologic heterogeneity, NUT-IHC is not consistently incorporated into standard diagnostic workflows, contributing to under-recognition and delayed confirmation (6, 7). Therapeutically, no universally accepted standard-of-care exists; multimodality regimens combining surgery, radiotherapy, and chemotherapy are commonly used but remain suboptimal. Targeted strategies leveraging the epigenetic dependencies of NUT carcinoma (e.g., BET inhibition) are under active investigation, yet robust and generalizable clinical evidence is still evolving (8–10).

At the molecular level, NUT carcinoma is defined by NUTM1 fusions (11), but the spectrum of fusion partners is increasingly diverse. Beyond canonical partners such as BRD4, BRD3, and NSD3, multiple zinc-finger proteins have been reported, and emerging data continue to expand the list to include non-BET partners such as MGA, YAP1, and CIC (3). This expanding fusion landscape raises the possibility of fusion partner–specific predilections for histologic lineage and anatomical site, which may contribute to phenotypic heterogeneity and further complicate real-world recognition, classification, and risk stratification. However, evidence addressing these relationships remains fragmented across case reports, small clinicopathologic series, and mechanistic studies. Consequently, there is a persistent lack of (i) a systematic, quantitative synthesis of fusion partner–histology–anatomy associations and (ii) an integrative framework that connects research trends, mechanistic insights, and clinical diagnostic challenges, particularly for the thoracic and head and neck presentations that account for a substantial proportion of reported cases.

Bibliometric analysis provides a quantitative and reproducible approach to mapping knowledge structures, collaboration networks, and the temporal evolution of research hotspots within a scientific field (12, 13). Nevertheless, bibliometric evidence specifically characterizing the intellectual landscape and thematic evolution of thoracic and head and neck NUT carcinoma remains limited. Moreover, how bibliometrically identified “frontier themes” align with—and potentially inform—major clinical unmet needs such as misdiagnosis, diagnostic delay, and fusion-driven heterogeneity has not been systematically articulated. Therefore, we aimed to: (i) delineate the knowledge structure, collaboration networks, and temporal evolution of research hotspots in WoSCC-indexed thoracic and head and neck NUT carcinoma literature using bibliometric approaches; (ii) synthesize WoSCC-indexed clinicopathologic evidence to quantify anatomical distribution, high-frequency histologic features, and patterns of initial misdiagnosis; (iii) explore associations between NUTM1 fusion partners and tissue-origin/histologic lineage as well as anatomical site; (iv) interpret these associations in the context of current mechanistic models and emerging multimodal technologies; and (v) assess whether major literature-derived clinicopathologic and immune-biomarker signals, particularly BRD4::NUTM1 predominance and PD-L1 negativity, could be corroborated in an independent institutional cohort.

2 Methods

2.1 Data source, search strategy, and corpus definition

We used the Science Citation Index Expanded (SCI-EXPANDED) within the Web of Science Core Collection (WoSCC) as the sole source database and applied a unified search strategy (fields and terms in Supplementary Table 1) to retrieve English-language publications from January 1, 1991 to December 31, 2025(executed on January 4, 2026), yielding 919 records. WoSCC was selected a priori because the bibliometric component required standardized citation metadata, complete records, and cited-reference information compatible with CiteSpace, VOSviewer, and bibliometrix. This single-source strategy also ensured that the bibliometric mapping and the downstream case-level synthesis were derived from a unified and reproducible corpus. PubMed/MEDLINE, Embase, and Scopus were not additionally searched because combining multiple citation databases would require extensive cross-database harmonization, deduplication of records, and reconciliation of non-equivalent cited-reference metadata, which could introduce database-dependent inconsistencies into co-citation, collaboration, and thematic evolution analyses. We acknowledge, however, that this strategy may reduce case capture for the clinicopathologic synthesis and may affect the representativeness of pooled estimates, particularly for rare case reports not indexed in WoSCC. Briefly, the WoSCC topic search combined three concept blocks: (1) NUT carcinoma/NUTM1-related terms, including NMC, “NUT carcinoma,” “NUT midline carcinoma,” “NUTM1 fusion,” “NUTM1-rearranged,” “NUTM1-rearranged carcinoma,” “NUT rearrangement,” “Nuclear protein in testis,” “NUTM1 protein,” and recurrently reported fusion partners such as BRD4, BRD3, NSD3, ZNF532, and ZNF592; (2) malignancy-related terms, including cancer, carcinoma, and neoplasm; and (3) thoracic and head and neck anatomical terms, including lung, pulmonary, bronchial, tracheal, pleural, mediastinal, thymic, esophageal, chest-wall, “head and neck,” “head & neck,” “upper aerodigestive tract,” HNSCC, sinonasal, nasal/paranasal sinus, pharyngeal, laryngeal, oral cavity, salivary gland, cervical/neck, thyroid/parathyroid, external auditory canal, temporal bone, and skull-base terms. The full line-by-line WoSCC search strategy, including Boolean combinations and search counts, is provided in Supplementary Table 1. The study selection process is shown in the PRISMA 2020 flow diagram in Figure 1A, and the completed PRISMA 2020 checklist is provided as a separate supplementary checklist.

Figure 1

All records were exported in the format of ‘Complete Records and Cited References’ to build the main dataset. Subsequently, publication-level deduplication was performed based on DOI, title, first author, and publication year, followed by title/abstract screening for relevance. For the clinicopathologic synthesis, patient-level deduplication was further conducted to minimize duplicate case inclusion across case reports and series by cross-checking institution, author group, publication year, age, sex, primary site, fusion partner, histology, treatment history, and outcome information. When overlapping reports were suspected, the most complete or most recent report was retained for case-level extraction, while earlier or less complete reports were used only for contextual information when necessary. This process resulted in a final set of 281 publications that met the analysis criteria. To focus on thoracic and head and neck disease, we included publications whose title/abstract/keywords explicitly addressed thoracic or head and neck NUT carcinoma. From the same eligible corpus, we predefined four analytic subsets (Figure 1B): (1) a bibliometric subset (original articles and reviews) for mapping knowledge structure, thematic evolution, and collaboration patterns; (2) a clinicopathologic synthesis subset including studies enabling extraction of case-level variables (fusion partner, primary anatomical site, and histologic descriptors for harmonized stratification); (3) an immune-biomarker subset to abstract PD-L1 expression and TMB/MSI status (and immunotherapy exposure/outcome descriptors where available); and (4) a narrative evidence-mapping component in the Discussion focusing on representative literature on pathogenesis, model construction, multimodal diagnostics, and evolving therapeutic modalities (with emphasis on immunotherapy).

We assembled an independent retrospective institutional cohort of thoracic and head and neck NUT carcinoma cases to provide clinical validation beyond published reports. This cohort was analyzed separately from the bibliometric and literature-synthesis datasets and was used specifically to assess the reproducibility of the major clinicopathologic and immune-biomarker signals identified in the pooled literature, with emphasis on fusion-partner distribution, PD-L1 landscape, and clinical aggressiveness.

2.2 Data extraction

For data extraction, bibliometric records were converted into standardized formats (plain text and CSV), and key fields were extracted, including title, authorship, affiliations, country/region, journal, publication year, keywords (author keywords and database-generated keywords), cited references, and citation metrics for descriptive statistics and network analyses. For clinicopathologic synthesis, two experienced pathologists independently abstracted case-level variables using a prespecified extraction form, with cross-checking and adjudication. Variables included fusion status (fusion partner and/or confirmatory evidence of NUTM1 rearrangement), primary anatomical site, histopathological descriptors enabling classification, diagnostic pathway variables (initial diagnosis category), and major treatment/outcome descriptors when available. To improve comparability across heterogeneous reports, cases were further mapped into harmonized anatomical regions and histologic groupings (squamous vs non-squamous), with disagreements resolved by consensus and third-pathologist adjudication when necessary. For immune-biomarker abstraction, we extracted PD-L1 expression results, including assay clone, scoring system, and cut-off when reported. PD-L1 results were harmonized into three descriptive categories: Negative, Low expression, and High expression. Author-reported categorical interpretations were prioritized. When only numeric results were available, cases were categorized as Negative if TPS was <1% or CPS was <1, Low expression if TPS was ≥1% to <30% or CPS was ≥1 to <30, and High expression if TPS was ≥30% or CPS was ≥30. Because TPS is reported as a percentage of viable tumor cells with membranous staining, whereas CPS is a unitless score incorporating staining in tumor cells and selected immune cells, these categories were used primarily for descriptive comparability rather than as directly interchangeable quantitative thresholds. Given the heterogeneity of PD-L1 assays, scoring systems, and reporting thresholds, we additionally used a binary sensitivity framework classifying cases as PD-L1 Negative versus Any expression. Author-reported negative cases, TPS <1%, or CPS <1 were classified as Negative, whereas author-reported low/high expression, TPS ≥1%, or CPS ≥1 were classified as Any expression. When both TPS and CPS were available, TPS was prioritized for tumor-cell expression, whereas CPS was retained descriptively and used only in the binary sensitivity framework.

We additionally extracted MSI status (categorized as MSS vs MSI-H/MSI-L when reported) and tumor mutational burden (TMB, mut/Mb) as a continuous variable. Due to heterogeneous reporting and substantial missingness across case reports, immune-biomarker analyses were performed using endpoint-specific complete-case datasets.

2.3 Bibliometric and clinicopathologic analyses, quality assessment

For analysis, bibliometric mapping and visualization were performed using CiteSpace,VOSviewer and R-bibliometrix, including keyword co-occurrence, clustering, and burst detection to identify thematic structure, evolving hotspots, and emerging fronts, as well as collaboration networks across authors, institutions, and countries/regions. In the clinicopathologic synthesis, categorical variables are reported as counts (%), and continuous variables as median (IQR) and/or mean (95% CI), as appropriate. Associations between fusion partners and clinicopathologic strata, including histologic grouping, anatomical region, anatomical-pathologic strata, and immune-biomarker categories, were assessed using Fisher’s exact test and reported as odds ratios (ORs) with 95% confidence intervals (CIs). Missing data were handled using endpoint-specific complete-case analyses. For PD-L1, both the prespecified three-category harmonized framework and a binary negative-versus-any-expression framework were summarized. Because analyses across fusion partners, anatomical strata, histologic categories, and immune-biomarker groups were exploratory and hypothesis-generating in the setting of a very rare disease, we did not apply formal multiplicity correction. All reported p values are two-sided and unadjusted; statistically significant associations, particularly those involving rare fusions, should therefore be interpreted as exploratory enrichment signals rather than confirmatory evidence. Methodological quality was appraised using the Agency for Healthcare Research and Quality (AHRQ) approach where applicable (Supplementary Table 2) (14). Because the available evidence consisted predominantly of case reports and small case series, the AHRQ assessment was used descriptively to characterize reporting completeness and inherent risk of bias rather than as an exclusion criterion. No predefined AHRQ score threshold was applied for study exclusion. Studies were retained if they met the eligibility criteria and provided extractable case-level clinicopathologic, fusion, or immune-biomarker data.

2.4 Independent clinical validation cohort

We retrospectively enrolled pathologically confirmed NUT carcinoma cases from our institution for independent cohort validation. Eligible cases were required to have at least one of the following available: fusion-partner data, PD-L1 testing data, or immunotherapy exposure history. Extracted variables included age, sex, smoking history, primary site, clinical stage, histologic category, fusion partner, PD-L1 status, Ki-67 proliferation index, and immunotherapy exposure history. Fusion-partner annotation was determined, when available, based on NGS RNA fusion panel results. PD-L1 expression in the institutional cohort was assessed uniformly using the 22C3 pharmDx assay and tumor proportion score (TPS). For consistency with the literature-derived three-tier harmonization, institutional PD-L1 results were descriptively mapped as Negative (<1%), Low expression (1% to <30%), and High expression (≥30%). Raw TPS values were retained for interpretation, and the binary framework, Negative versus Any expression, was additionally applied as a sensitivity analysis. Exact binomial tests were used to compare observed proportions in the institutional cohort with the corresponding WoSCC-indexed literature-derived estimates, and Fisher’s exact test was used for subgroup comparisons. All statistical analyses were performed using R version 4.3.x, with two-sided p < 0.05 considered statistically significant.

3 Results

3.1 Bibliometric analysis of NUT carcinoma research

3.1.1 Dataset characteristics and annual publication trends

This study retrieved 281 thoracic and head and neck NUT carcinoma-related publications from the Web of Science Core Collection (SCI-EXPANDED), distributed across 143 journals. The document types were predominantly original research articles (224 articles) and review articles (57 reviews). The dataset comprised 1,863 authors with 2,264 author appearances, averaging 8.06 authors per publication. Single-authored publications accounted for only 6 papers (2.14%), while internationally collaborative papers represented 19.93%, reflecting the highly multidisciplinary collaborative nature of this field. The publications contained 3,539 references, 533 author keywords (DE), and 384 KeyWords Plus (ID). The total citation count reached 11,589 times, with an average of 41.24 citations per publication, an annual citation intensity of 3.981 citations per publication per year, and a median citation count of 10. Zero-citation publications accounted for 16.37% (n=46), indicating a typical long-tail distribution where highly cited core publications coexist with a substantial proportion of low- or zero-citation research.

Annual publication volume analysis revealed a sustained growth trend in NUT carcinoma research, with an average annual growth rate of 16.49%. During the early phase (1991–2014), publication output was low and fluctuating; publications were sporadic before 2001, followed by gradual accumulation with annual outputs typically ranging from 1 to 8 papers. Research entered a growth phase in 2015 (13 papers) and accelerated markedly after 2020, remaining high during 2022–2025. A quadratic polynomial fitting model based on cumulative publication volume (y = 0.75X² − 9.44X + 28.25) demonstrated excellent goodness-of-fit (R² = 0.9868), supporting a sustained upward trajectory in research output (Figure 1C).

3.1.2 Core authors, institutions, and national contribution patterns

Author contribution analysis identified FRENCH CA as the most prolific researcher with 53 publications, the highest average citations per paper, and the most extensive collaborative network (Figures 1D, E), followed by SHAPIRO GI (12 papers), SHOLL LM (8 papers), and LAUER UM (7 papers).Research was concentrated in top-tier U.S. medical centers (Figure 1F), with Harvard-affiliated institutions (Harvard Medical School, Brigham & Women’s Hospital, Massachusetts General Hospital, Boston Children’s Hospital) leading globally through 60 publications (21.35%), 8,428 citations (140.47 per paper vs. field average 41.24), and an H-index of 35. The University of Texas MD Anderson Cancer Center was another major contributor. Co-organization networks revealed dense collaborations centered around Harvard institutions (Figure 2A). Figure 2B further visualizes the comparative impact of the top-10 institutions by plotting total citations versus H-index (bubble size = average citations per paper), underscoring the concentration of high-impact output in leading U.S. centers. Department-stratified analysis shows that publications from the top-10 institutions are primarily contributed by pathology and oncology departments (Figure 2C), and department-level impact—summarized by total citations vs H-index with bubble size denoting average citations—clusters in a subset of high-performing units (Supplementary Figure 1).

Figure 2

Regarding journal distribution, major publishing journals (Figure 2D) were concentrated in the fields of oncology and pathology (Figure 2E), and the subject-category overlay further indicates that NUT carcinoma research is positioned at the interface between clinical oncology/pathology and molecular biology disciplines (Figure 2F). Frontiers in Oncology (15 papers) and American Journal of Surgical Pathology (11 papers) being the primary sources, followed by Head & Neck Pathology, head and neck, and International Journal of Surgical Pathology (8 papers each). Bradford’s Law (15) analysis identified 14 core zone journals, constituting the primary vehicles for disseminating literature on this topic and demonstrating a pronounced journal clustering effect. The dual-graph overlay analysis of the journal shows that cited literature mainly comes from two major disciplinary clusters: MEDICINE, MEDICAL, CLINICAL and MOLECULAR, BIOLOGY, IMMUNOLOGY. The flow of knowledge presents a unidirectional pattern from the right MOLECULAR, BIOLOGY, IMMUNOLOGY cluster to the left MEDICINE, MEDICAL, CLINICAL cluster, reflecting the knowledge transfer pathway from basic research to clinical translation (Figure 3A).

Figure 3

National contribution analysis showed the United States dominated with 99 publications(35%), 8,678 total citations, and 87.66 citations per publication, leading in both output and impact (Figure 3B). China ranked second with 59 papers (accounting for 21%), but each paper was cited only 7.80 times on average, indicating that while its output is substantial, its influence is limited. However, in recent years, the growth rate of its literature has been significantly higher than that of other countries(Figure 3C).Other major contributors included Japan (22), Germany (17), and France (14).

International collaboration patterns revealed significant disparities: France (MCP ratio 0.50) and Germany (0.3529) demonstrated high transnational collaboration, while China showed minimal international engagement (MCP 3.39%, SCP 96.61%). Notably, France achieved 45.64 citations per publication despite moderate output, reflecting its significant contribution to foundational knowledge.

3.1.3 Citation patterns and knowledge foundation

High-citation literature analysis and bibliographic coupling identified milestone studies and core knowledge frameworks (Figure 3D). The temporal heatmap of the top 10 most-cited papers (Supplementary Figure 2) further illustrates sustained citation accumulation after 2015, with several seminal works showing pronounced citation bursts during 2018–2025, consistent with the consolidation of a core knowledge base in the field. By total citations (Figure 3E), among the cited references in our corpus, the most globally cited work was TRAVIS WD et al.’s 2015 lung cancer classification guidelines (Journal of Thoracic Oncology, 3,613 citations) (2), followed by FRENCH CA et al.’s 2008 Oncogene paper (349) (16), HAACK H et al.’s 2009 American Journal of Surgical Pathology paper (325) (6), and FRENCH CA et al.’s 2004 Journal of Clinical Oncology paper (323) (17). Co-citation analysis identified core publications repeatedly cited within the dataset (Supplementary Figure 3): BAUER DE 2012 (Clinical Cancer Research, Freq=164) (18), HAACK H 2009 (Freq=160) (6), FRENCH CA 2004 (Freq=142) (17), and FRENCH CA 2008 (Freq=111) (16), constituting the field’s core knowledge framework in establishing NUT carcinoma molecular characteristics, identifying chromosomal translocations/fusion genes, establishing diagnostic paradigms, and exploring therapeutic approaches. These highly cited publications primarily focused on the oncogenic mechanisms of BRD4::NUTM1 fusion proteins, epigenetic alterations involving histone hyperacetylation, establishment of diagnostic criteria, and exploration of targeted therapeutic strategies such as BET inhibitors, laying a solid theoretical foundation for subsequent research.

3.1.4 Research hotspots, thematic clustering, and temporal evolution

Time-sliced keyword word clouds revealed highly concentrated research themes and a clear temporal shift in thoracic and head and neck NUT carcinoma research (Figure 4A). The keyword co-occurrence network highlights ‘NUT carcinoma’ and fusion-related terms (e.g., BRD4::NUTM1, rearrangement) as central hubs with dense connectivity, reflecting the fusion-centric knowledge structure of the field (Figure 4B). Author keywords (DE) centered on “NUT carcinoma” (95 occurrences) and “NUT midline carcinoma” (71 occurrences), with high-frequency accompanying terms including “case report” (26 occurrences), “NUTM1” (23 occurrences), “BRD4-NUT” (17 occurrences), “immunohistochemistry” (15 occurrences), “immunotherapy” (11 occurrences), and “next-generation sequencing” (8 occurrences). Among KeyWords Plus (ID), “midline carcinoma” (107 occurrences), “differentiation” (73 occurrences), “outcomes” (58 occurrences), “rearrangement” (43 occurrences), and “BRD-NUT” (33 occurrences) occupied dominant positions, indicating that the research framework primarily revolves around disease entity, differentiation blockade/epigenetic aberrations, chromosomal rearrangements, and clinical outcomes.

Figure 4

Citation burst analysis further revealed temporal shifts in research hotspots (Supplementary Table 3). Early bursts concentrated on disease entity and translocation identification-related terms, including “thymic carcinoma” (2003–2013, Strength=4.82), “aggressive carcinoma” (2004–2013, Strength=6.83), and “translocation” (2004–2013, Strength=5.49). Mid-phase bursts featured fusion event and detection-related terminology, with “rearrangement” (2008–2017, Strength=7.61) and “BRD4::NUTM1 fusion” (2008–2012, Strength=4.29) indicating deepened fusion gene detection and molecular characterization. Recent bursts shifted toward precision diagnostics and targeted strategies, with “bromodomain inhibitor” (2020–2023, Strength=2.41) and “next-generation sequencing” (2021–2022, Strength=2.32) reflecting advances in therapeutic exploration and molecular diagnostic technologies. Notably, “NUT carcinoma” exhibited the strongest burst during 2022–2025 (Strength=11.96), suggesting further reinforcement of disease nomenclature standardization and research attention in recent years.

Thematic clustering analysis identified 11 major research directions spanning entity definition/translocation and fusion discovery, molecular diagnostics, site-stratified pathology, and therapeutic themes including BET/HDAC inhibition (Figure 4C: Keyword clustering map; details in Supplementary Table 4). Temporal evolution analysis revealed a four-phase development pattern. The Foundational Phase (pre-2010) was dominated by NUT carcinoma and Translocation themes, with a pivotal node around 2005 marking disease entity establishment and identification of t(15;19) translocation/BRD4-NUT fusion gene, laying the groundwork for subsequent molecular characterization. The Expansion Phase (2010-2015) witnessed thematic diversification, with the prominent emergence of BET inhibitor during 2012–2015 signaling a transition from mechanistic exploration to therapeutic strategies, while sustained growth in Case report, Nuclear protein, and RNA sequencing reflected parallel advancement in case accumulation, mechanistic research, and diagnostic technologies. The Diversification Phase (2015-2020) demonstrated multi-directional expansion, with the nodal appearance of RNA sequencing around 2015 indicating widespread adoption of next-generation sequencing, the emergence of Lung cancer and Oral squamous cell carcinoma representing deepened site-specific investigations, and Clinical imaging (2018-2020) reflecting enhanced emphasis on radiological features. The Contemporary Phase (2020-2025) emphasized precision medicine translation, with NUT carcinoma, BET inhibitor, Case report, and RNA sequencing maintaining continuous activity and increasingly strengthened cross-cluster connections, forming an integrated research paradigm encompassing molecular diagnostics, targeted therapies, and clinical management. The field evolved from disease definition and molecular characterization (2001-2010) through therapeutic exploration (2010-2015) to precision medicine implementation (2015-2025), with case documentation and molecular diagnostics as persistent themes. This trajectory is corroborated by keyword timeline analysis showing both emerging clusters and enduring core themes across 2001-2025 (Supplementary Figure 4).

3.1.5 Clinical problem-oriented thematic distribution

To enhance clinical interpretability of results, automatic thematic annotation statistics were performed on titles-abstracts-keywords (categories allowed to overlap). Fusion gene-related research comprised 173 papers (61.57%), with BRD4::NUTM1-related studies dominating at 82 papers (29.18%), followed by t(15;19) chromosomal translocation at 53 papers (18.86%), and BRD3::NUTM1 and NSD3::NUTM1 each at 12 papers (4.27%), reflecting the central role of fusion gene detection in diagnosis and molecular subtyping. Regarding treatment strategies, chemotherapy (99 papers, 35.23%) and radiotherapy (82 papers, 29.18%) showed the highest frequencies, indicating that conventional treatments remain the clinical mainstream; immunotherapy (63 papers, 22.42%) demonstrated rapid growth; BET inhibitors (31 papers, 11.03%) and HDAC inhibitors (9 papers, 3.20%) represented frontier explorations in targeted and epigenetic therapies. Research types were predominantly case reports/case series (93 papers, 33.10%), reflecting disease rarity and the importance of clinical experience accumulation; cohort studies (32 papers, 11.39%) and basic research (33 papers, 11.74%) were comparable; clinical trials (21 papers, 7.47%) were limited in number, indicating scarcity of high-quality prospective studies.

3.2 Systematic clinicopathologic synthesis and fusion partner–phenotype associations

Guided by Section 3.1 showing persistent emphasis on rearrangements/fusions and molecular diagnostics, we conducted a systematic clinicopathologic synthesis using the prespecified clinicopathologic subset from the same WoSCC corpus (Methods 2.1, Subset 2). Overall, 109 studies were included, yielding 229 thoracic and head and neck NUT carcinoma cases with extractable core clinicopathologic annotations and definitive fusion types. The anatomical distribution was nearly balanced between the head and neck and thorax (50.22%, 115/229 vs. 49.78%, 114/229): thoracic cases most commonly involved the lung (34.06%, 78/229) and mediastinum (10.92%, 25/229), whereas head and neck cases were enriched in the salivary glands (13.97%, 32/229), nasal cavity/sinuses (13.10%, 30/229), and thyroid gland (8.73%, 20/229). Histogenetic review indicated predominance of squamous epithelial origin (71.62%, 164/229), followed by glandular epithelium (17.03%, 39/229), mesenchymal tissue (4.80%, 11/229), biphenotypic squamous-and-glandular differentiation (3.93%, 9/229), and mixed origin (2.62%, 6/229).The overall mapping from tissue origin and sites to anatomical–pathologic strata and fusion types is summarized in Figure 4D.

The fusion landscape was dominated by BRD4::NUTM1 (64.19%, 147/229), followed by NSD3::NUTM1 (14.85%, 34/229) and BRD3::NUTM1 (10.92%, 25/229), with additional cases mostly harboring YAP1::NUTM1 (5.68%, 13/229) and MGA::NUTM1 (3.06%, 7/229).Under the prespecified Fisher’s exact and odds-ratio framework (Methods 2.3), fusion partners exhibited non-random distributions across histogenetic and anatomical–pathologic strata (Figure 4F; Supplementary Figure S5, S6). BRD4::NUTM1 was significantly associated with squamous epithelial origin (OR = 3.49,p=0.0001) and enriched in the thoracic squamous subgroup (OR = 2.33,p=0.0063), while being underrepresented in thoracic non-squamous and head and neck non-squamous subgroups (OR = 0.40, p=0.0256; OR = 0.33, p=0.0034). In contrast, YAP1::NUTM1 was strongly associated with glandular epithelial origin (OR = 36.93,p<0.001) and markedly enriched in the head and neck non-squamous subgroup (p<0.001), with no cases observed in squamous-related strata; MGA::NUTM1 showed an extremely strong association with mesenchymal origin (p<0.001) and was concentrated in the thoracic non-squamous subgroup (p<0.001). Collectively, these results suggest fusion partner–associated enrichment patterns across lineage and anatomical–pathologic stratification. However, because several non-BRD4 fusion groups were represented by small case numbers and wide confidence intervals, these findings should be interpreted as exploratory and hypothesis-generating rather than deterministic or mechanistically validated associations. In addition, recurrent microscopic patterns included sheet-like and nested architectures in both thoracic and head and neck tumors, with prominent nucleoli frequently reported (Supplementary Table 5). Abrupt keratinization was documented more often in head and neck tumors than in thoracic tumors (Supplementary Table 5). Finally, the most common initial diagnoses were squamous cell carcinoma (36.7%), undifferentiated tumor (16.2%), and poorly differentiated carcinoma (14.8%) (Supplementary Table 6).

3.3 Immune-biomarker landscape

Motivated by the bibliometric signal of increasing emphasis on immunotherapy and tumor microenvironment–related themes (Section 3.1), we further synthesized immunotherapy-relevant biomarkers from the clinicopathologic corpus.MSI status and numeric TMB values were available in 16 cases. All 16 were MSS. TMB values were generally low (median 1.35 mut/Mb; range 0–7.6 mut/Mb), with no cases reaching commonly used “high-TMB” thresholds.

PD-L1 status was extractable in 78 thoracic/head and neck cases. Overall, PD-L1 was Negative in 60/78 (76.92%), Low expression in 13/78 (16.67%), and High expression in 5/78 (6.41%). Across anatomical–pathologic strata, PD-L1 negativity was particularly prevalent in the thorax squamous subgroup (33/37, 89.19%), which showed increased odds of PD-L1 negativity compared with other strata (OR = 4.28, 95%CI 1.26–14.52; p=0.017) (Figure 4E; Figure 4G). Tissue-origin groupings and broad anatomical region (thorax vs head and neck) were not significantly associated with PD-L1 category in this dataset. To test whether the predominance of PD-L1 negativity was robust to simplified PD-L1 harmonization, we performed a binary sensitivity analysis. In the literature-derived PD-L1 subset, 60 of 78 cases were classified as PD-L1 Negative and 18 of 78 as Any expression, confirming the predominance of PD-L1 negativity. Across anatomical–pathologic strata, thoracic squamous tumors included 33 PD-L1-negative and 4 PD-L1-expressing cases, whereas the remaining strata included 27 PD-L1-negative and 14 PD-L1-expressing cases. These binary sensitivity results were consistent with the three-category analysis and are summarized in Supplementary Table 7.

Among 39 cases with both definitive fusion-partner annotation and PD-L1 status, PD-L1 was predominantly Negative (32/39, 82.1%), with Low expression in 5/39 (12.8%) and High expression in 2/39 (5.1%) (Figure 4H). No statistically significant association between the descriptive three-tier PD-L1 category and fusion partner was detected, although interpretability was limited by sparse counts for non-BRD4 and rare fusion groups. To address the heterogeneity of PD-L1 reporting across studies, we further performed a binary sensitivity analysis in which PD-L1 status was classified as Negative versus Any expression. Among the 39 literature-derived cases with both definitive fusion-partner annotation and PD-L1 data, 32 cases were PD-L1 negative and 7 showed any PD-L1 expression. By fusion partner, BRD4::NUTM1 accounted for 28 cases, including 23 PD-L1-negative and 5 PD-L1-expressing tumors; BRD3::NUTM1 accounted for 5 cases, all of which were PD-L1 negative; and NSD3::NUTM1 accounted for 4 cases, including 3 PD-L1-negative and 1 PD-L1-expressing tumor. The remaining ZNF532::NUTM1 and ZNF592::NUTM1 cases were represented by one case each. Exploratory Fisher’s exact testing did not identify a significant association between fusion partner and binary PD-L1 status (Supplementary Table 8). These findings support the predominance of PD-L1 negativity across fusion-defined subsets, while emphasizing that inference for non-BRD4 and rare fusion partners remains limited by sparse case numbers.

3.4 Independent cohort validation

We analyzed an independent institutional cohort of 59 pathologically confirmed NUT carcinoma cases. The baseline clinicopathologic characteristics of this cohort are summarized in Table 1. The cohort comprised 35 males and 24 females, with a median age of 34 years (IQR, 23.5–46.5; range, 7–76 years). Most patients presented with advanced-stage disease, including 44/59 (74.6%) with stage IV disease and 15/59 (25.4%) with stage III disease. Primary tumor sites were mainly thoracic (39/59, 66.1%) and head and neck (20/59, 33.9%). Eleven patients (18.6%) had a smoking history, and 20 (33.9%) had previously received immunotherapy.

Missingness was explicitly quantified in the institutional cohort. Histologic information was available for 39 of 59 patients and missing in 20 of 59 patients (33.9%). Fusion-partner annotation was available in 34 of 59 patients (57.6%), and PD-L1 testing was available in 33 of 59 patients (55.9%). Among fusion-annotated cases, BRD4::NUTM1 remained the predominant fusion partner (25/34, 73.5%), followed by BRD3::NUTM1 (8/34, 23.5%) and NSD3::NUTM1 (1/34, 2.9%). This proportion was not significantly different from the pooled literature-derived estimate (147/229, 64.2%; exact binomial test, p = 0.288).

Among PD-L1-tested cases, 25 of 33 tumors were PD-L1 negative, 5 showed low expression, and 3 showed study-level harmonized high expression under the three-tier framework; no TPS-scored institutional case reached the conventional TPS ≥50% high-expression threshold. This was consistent with the literature-derived estimate of PD-L1 negativity in the overall PD-L1-assessable subset (60/78, 76.9%; exact binomial test, p = 0.838; Supplementary Table 7) and supported the robustness of the immune-cold pattern under a simplified binary framework. However, because histologic information was missing in one-third of the institutional cohort and rare non-BET fusion partners such as YAP1::NUTM1 and MGA::NUTM1 were not represented, the institutional cohort was not sufficiently powered to independently validate fine fusion partner–phenotype enrichment patterns observed in the literature-derived analysis.

4 Discussion

4.1 Key findings

This study integrates bibliometrics with systematic case-level synthesis to map thoracic and head and neck NUT carcinoma. The field has shifted since 2015 toward molecular diagnostics, epigenetic targeting, and immunotherapy. Clinically, cases are mainly squamous but include non-squamous and mixed phenotypes, with exploratory fusion partner–associated enrichment patterns observed across selected lineage and anatomical strata. Most reported tumors are PD-L1–negative and MSS/low-TMB, underscoring the need for prospective, biomarker-integrated registries.

4.2 Diagnostic challenges

Building on these findings, we further dissect the persistent causes of misdiagnosis and delayed diagnosis from three perspectives: morphologic clues, lineage/phenotypic spectrum, and testing workflows. Despite growing awareness, misdiagnosis and diagnostic delay remain common (4, 5, 19): (1) Non-specific morphology: NUT carcinoma often presents as poorly differentiated/undifferentiated tumors and substantially overlaps with poorly differentiated squamous cell carcinoma, undifferentiated carcinoma, and some small round cell or sarcomatoid neoplasms; moreover, thoracic and head and neck biopsy specimens are frequently limited by scant tissue, necrosis, and crush artifact, further reducing diagnostic clues. (2) Broad phenotypic spectrum: While focal “abrupt keratinization” can be suggestive, it is not consistently present or may be only focal; in addition, non-squamous or mixed phenotypes broaden the differential diagnosis, making NUT-IHC less likely to be incorporated as a “routine test.” (3) Heterogeneity in testing workflows and thresholds: Institutions vary in the threshold for “when to add NUT-IHC” and “when to escalate to fusion testing for confirmation,” resulting in under-testing and delayed testing.

4.3 Mechanistic context

Importantly, delineating the key molecular circuitry driven by fusions not only helps define more actionable triggers for diagnostic testing, but also provides a rationale for subsequent exploration of epigenetic dependency–based therapies. In canonical BRD4/BRD3/NSD3 fusions and a subset of non-canonical fusions, available evidence broadly supports a convergent epigenetic oncogenic logic: NUTM1 fusion proteins assemble an aberrant “reader–writer platform” that enhances histone acetylation and transcriptional amplification while blocking differentiation programs (11, 20–22), thereby maintaining an undifferentiated, highly proliferative state. As shown in Figure 5, bromodomains tether the fusion complex to acetylated chromatin while the NUT moiety recruits/activates p300/CBP to establish an acetylation feed-forward loop and expand active chromatin domains (23), providing a unified rationale for BET and HDAC inhibitor strategies. At a higher organizational level, liquid–liquid phase separation (LLPS) condensates (24), megadomain formation (25, 26), and 3D genome reorganization collectively shape a stable transcriptional niche (22), reinforcing core oncogenic transcriptional programs and suppressing differentiation-related pathways (21, 22, 27).Consequently, a poorly differentiated/undifferentiated morphology can be viewed as an expected epigenetic outcome rather than requiring the assumption of a single tissue of origin. The enrichment of specific fusion partners across lineage- and site-stratified subsets should be interpreted as a probabilistic and hypothesis-generating tendency related to cellular context and tissue microenvironment rather than as evidence of deterministic lineage specification. Structural differences among fusion partners may alter chromatin engagement (16), cofactor recruitment, or transcriptional circuit modulation, while the cell of origin and local environment may influence which fusion events are selected and how they manifest histologically. However, for rare non-BET fusions such as YAP1::NUTM1 and MGA::NUTM1, available evidence remains insufficient to establish whether the fusion directly drives the observed non-squamous differentiation states or instead marks a correlation shaped by sampling and publication bias.

Figure 5

4.4 Models and modalities

The mechanistic framework outlined above provides an interpretable coordinate system; however, its testability and translational utility still depend on appropriate model systems and robust phenotypic readouts. Consistent with the bibliometric signal of increasing emphasis on molecular diagnosis and translation, key advances in NUT carcinoma research have been driven to a large extent by complementary model platforms and rapidly evolving multimodal technologies (Figure 6). Patient/mice-derived cell lines (Supplementary Table 9) (11, 16, 28–31)offer a foundational platform for dissecting fusion-driven chromatin remodeling and transcriptional amplification and support drug screening for epigenetic dependencies such as BET and HDAC inhibition. Inducible-expression systems and genetically engineered tagging further improve the resolution of interaction networks, chromatin states, and dynamic processes. Meanwhile, in vivo animal model systems—including GEMMs, knockout models, and xenografts (Supplementary Table 10)—provide indispensable evidence to test the oncogenic sufficiency of a single fusion driver, capture early lesions, and interrogate tumor–microenvironment interactions (20, 23, 27, 29–33).

Figure 6

In parallel with controllable validation enabled by model systems, multi-omics and spatial multimodal technologies further enhance resolution for tumor heterogeneity and microenvironmental remodeling. These approaches can delineate intratumoral differentiation states, invasion-associated subpopulations, and microenvironmental changes, thereby offering finer stratification cues and potential biomarkers; they can also provide trackable molecular readouts of treatment response and tolerance/resistance mechanisms to inform rational combination strategies (9, 34, 35). Proteomics and immunopeptidomics further expand the discovery space for circulating/tissue biomarkers and immunologic targets, whereas computational approaches such as machine learning may help distill key determinants from multi-omic or epigenetic features (36–39). Therefore, promoting multimodal strategies and aligning these technologies with clinical workflows may reduce under-detection and shorten time to diagnosis in the future.

Future mechanistic validation will be particularly important for variant non-BET fusions. Patient-derived cell lines, organoid or xenograft models, and genetically engineered mouse models carrying defined fusion partners could be used to test whether rare fusions directly reshape chromatin engagement, epigenetic reader–writer complex assembly, and downstream lineage programs. Such models would help distinguish fusion-driven lineage specification from statistical enrichment caused by case selection, tissue context, or publication bias.

4.5 Treatment strategies for NUT carcinoma

Outside immune-based strategies, management of NUT carcinoma remains anchored in aggressive local control and empiric systemic therapy. For resectable disease, early radical surgery with an R0 margin offers the best chance of durable control, with adjuvant radiotherapy and/or systemic therapy considered according to margin status, nodal involvement, and site-specific staging principles (5, 40–43).For unresectable disease or non-R0 resection, definitive high-dose IMRT-based radiotherapy combined with systemic therapy is commonly adopted, and tumor burden metrics (e.g., GTV size) may further stratify prognosis and guide intensification (10). Cytotoxic chemotherapy can induce responses but is typically limited by rapid resistance; regimen selection (e.g., platinum- vs ifosfamide-based) may influence short-term response patterns without consistently translating into long-term survival benefit (8).Epigenetic and transcriptional targeting—most notably BET inhibition, as well as HDAC inhibition and other emerging agents—provides an additional systemic option and a rational partner for multimodal combinations, albeit with practical constraints such as hematologic toxicity and limited CNS penetration in some settings (9, 34, 38, 44–49).

Consistent with the recent bibliometric shift toward immunotherapy and tumor microenvironment–related topics, immunotherapy has emerged as a practical frontier in thoracic and head and neck NUT carcinoma. To contextualize the heterogeneous clinical reports, we integrate our biomarker synthesis (PD-L1, MSI, and TMB) with a focused evidence map of ICI regimens and outcomes. As shown in Table 2, evidence is largely limited to case reports/series and retrospective cohorts, with no randomized data (50–69).

The immune checkpoint inhibitors (ICIs) reported to date are predominantly PD-1/PD-L1 inhibitors, including pembrolizumab, nivolumab, sintilimab, camrelizumab, durvalumab, and atezolizumab (52, 54, 55, 57–68).In terms of treatment strategy, ICIs are most often used in combination with platinum-based chemotherapy and/or radiotherapy (e.g., maintenance after concurrent/sequential chemoradiotherapy or combined with chemotherapy), suggesting that in real-world practice ICIs are typically embedded within multimodal treatment rather than used alone (52, 54, 55, 57, 59, 60, 63, 65).Beyond conventional chemoradiotherapy combinations, ICIs have also been combined (or sequenced) with BET inhibitors, anti-angiogenic/multi-target TKIs, or the oncolytic virus T-VEC in an attempt to enhance efficacy (54, 56, 59, 60, 62).

Overall efficacy is markedly heterogeneous: in primary pulmonary NUT carcinoma, there is a reported case of nivolumab monotherapy achieving sustained stable disease (SD) for approximately 29 months, with continued treatment beyond radiographic progression and ongoing clinical benefit, indicating a potential long-term benefit window in a minority of patients (50).Meanwhile, there are reports of CR/pCR achieved after perioperative or chemoradiotherapy combined with immunotherapy, suggesting that deep responses may be attainable with specific sequencing and combination patterns (54, 55).However, most cases still show primary resistance or rapid progression after only transient benefit (including early progression on first-line platinum-based chemotherapy plus PD-1 inhibitors), indicating that ICI efficacy in NUT carcinoma is unstable and may be substantially influenced by the tumor’s aggressive biology and confounding from concomitant therapies (52, 61, 64–66, 69).

From a biomarker perspective, PD-L1 expression spans a wide range in NUT carcinoma: some cases show high TPS (e.g., 70–80% and 100%), whereas others have TPS of 0% or near 0% (55, 62, 63, 66, 69).Among cases with available data, NUT carcinoma generally exhibits low TMB and is predominantly MSS; for example, a BRD4::NUTM1 cohort reported a mean TMB of ~1.7 mut/Mb (range 0–4) with all cases MSS, suggesting that the classic “high TMB” paradigm may not apply to most NUT carcinomas (63).Notably, objective responses or substantial metabolic responses have also been observed in the setting of PD-L1 negativity and low TMB: for example, an index case with a BRD4::NUTM1 fusion achieved near-complete response to nivolumab despite PD-L1 negativity and TMB 0.8 mut/Mb; another PD-L1–negative patient achieved pCR after adding a PD-L1 inhibitor to a chemoradiotherapy backbone (55, 63).In addition, in the context of very low PD-L1 expression (TPS <1%), multimodal therapy involving sequential T-VEC plus pembrolizumab has also shown signals such as PR/SD, suggesting that “immune-cold” tumors may be reshaped through combination strategies (54, 56).Mechanistically, a study using HLA typing and pMHC affinity prediction proposed that junctional peptides generated at the BRD4::NUTM1 fusion breakpoint may bind with high affinity in specific HLA-C subtypes, potentially triggering effective T-cell responses even in a low-TMB context, providing a direction for future biomarker development (63).From a practical trial-design perspective, these observations suggest that PD-L1 and TMB should not be used as the only biomarker axes for future NUT carcinoma immunotherapy studies. A feasible biomarker module could include: (i) fusion breakpoint characterization by RNA-based or hybrid-capture sequencing; (ii) HLA typing from tumor or germline DNA; (iii) in silico prediction of fusion-junction peptide binding affinity to patient-specific HLA alleles; and, when tissue quantity permits, (iv) immunopeptidomic validation of presented peptides (63). This approach remains investigational and should not be interpreted as a current replacement for standard PD-L1 or MSI/TMB assessment. However, it may be particularly relevant for rare fusion-driven tumors in which conventional immunotherapy biomarkers are frequently negative or low. Recent work also suggests that PRAME may represent an additional HLA-restricted T-cell target in NUT carcinoma, supporting the broader concept that antigen-presentation-based biomarkers may complement PD-L1 and TMB in future trials (39).

Regarding safety, reported immune-related adverse events (irAEs) are typified by endocrine toxicities, including hypothyroidism/thyroiditis and suspected immune-related type 1 diabetes (which may be complicated by DKA), and immune-related pneumonitis leading to treatment discontinuation has also been described (50, 55, 66). In summary, current evidence remains limited by small sample sizes, heterogeneous study designs, confounding from combination therapies, and inconsistent endpoints/response criteria; prospective registries and integrated studies incorporating fusion breakpoints, HLA typing, pMHC affinity, and tumor microenvironment features are urgently needed to define optimal ICI combinations and the populations most likely to benefit (54, 56, 63).

4.6 Limitations and outlook

This study has several limitations. First, both the bibliometric analysis and case-level synthesis were based on WoSCC-indexed English-language publications. This unified source improved compatibility with citation-based bibliometric tools and ensured that bibliometric mapping and clinicopathologic synthesis were derived from the same reproducible corpus; however, it may have reduced capture of region-specific reports and rare case reports. Accordingly, pooled estimates should be interpreted as literature-derived estimates from a WoSCC-indexed English-language corpus rather than exhaustive global incidence estimates.

Second, the clinicopathologic synthesis relied predominantly on case reports and small case series, in which diagnostic timelines, immunophenotypic details, treatments, and outcomes were inconsistently reported. Endpoint-specific complete-case analyses were therefore necessary, limiting causal inference and increasing susceptibility to reporting, referral, and publication biases. Similarly, fusion partner–phenotype associations should be interpreted as statistical enrichment and hypothesis-generating signals rather than deterministic rules. This caveat is particularly important for rare fusions, for which small sample sizes and wide confidence intervals limit the stability of effect-size estimates.

Third, PD-L1 and MSI/TMB data were available only in subsets of cases and were heterogeneous with respect to assays, scoring systems, and cutoffs. Although we harmonized PD-L1 into broad categories to enable synthesis, TPS and CPS are not directly interchangeable, because TPS reflects membranous staining in tumor cells whereas CPS incorporates staining in tumor cells and selected immune cells. Therefore, three-tier PD-L1 categorization across heterogeneous literature reports may introduce misclassification and residual heterogeneity. To reduce dependence on assay-specific and cutoff-specific distinctions, we performed a simplified Negative versus Any expression sensitivity analysis, which confirmed the central finding that PD-L1 negativity predominates in reported thoracic and head and neck NUT carcinoma. Nevertheless, the small number of PD-L1-expressing tumors and the scarcity of rare fusion partners preclude definitive conclusions regarding fusion-specific immune-biomarker patterns. Although institutional PD-L1 testing was performed uniformly using the 22C3 pharmDx assay and TPS scoring, the literature-derived cases used heterogeneous assays, clones, and scoring frameworks; therefore, cross-cohort comparisons should be interpreted as broad pattern-level comparisons rather than assay-equivalent biomarker estimates.

Finally, missingness in the institutional cohort limited subgroup validation. Histologic information was unavailable in 20 of 59 patients, and rare non-BET fusions such as YAP1::NUTM1 and MGA::NUTM1 were not represented among fusion-annotated cases. Accordingly, the institutional cohort supports the broader signals of BRD4::NUTM1 predominance and PD-L1 negativity, but cannot independently validate rare fusion-specific phenotype enrichment.

Future research should focus on: (1) establishing more standardized case reporting and data elements (fusion type, primary site, histologic description, immunophenotype, diagnostic timeline, etc.); (2) leveraging registries and multicenter datasets to reliably link molecular subtypes with treatment and outcomes; (3) testing, within defined cellular origin/lineage contexts, the interaction between “fusion-partner domains × cellular background” in shaping chromatin organization and transcriptional circuits; (4) prospectively evaluating workflow-embedded multimodal diagnostic strategies and mechanism-informed diagnostic and combination-therapy approaches; and (5) developing better biomarkers and rational combination strategies for predominantly immune-cold tumors. Aligning bibliometrically identified frontiers with unmet real-world clinical needs may help accelerate the translation of NUT carcinoma from “recognition” to “reproducible diagnosis and effective therapy”.

5 Conclusion

This study integrates bibliometric analysis with systematic clinicopathologic synthesis to characterize thoracic and head-and-neck NUT carcinoma. We identified non-random fusion-partner distributions and a predominantly PD-L1-negative immune-biomarker profile. The independent institutional cohort supported the broader literature-derived signals of BRD4::NUTM1 predominance and PD-L1 negativity, but was not powered to validate rare fusion-specific enrichment patterns. Exploratory fusion partner–phenotype enrichment patterns may refine diagnostic suspicion and guide future mechanistic research, but require validation in larger multicenter cohorts and dedicated model systems. Overall, these findings highlight the clinical value of integrating fusion architecture with immune profiling and support prospective biomarker-integrated studies in NUT carcinoma.

References

• 1

  KeesURMulcahyMTWilloughbyML. Intrathoracic carcinoma in an 11-year-old girl showing a translocation t(15;19). Am J Pediatr Hematol Oncol. (1991) 13:459–464. doi: 10.1097/00043426-199124000-00011. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 2

  TravisWDBrambillaENicholsonAGYatabeYAustinJHMBeasleyMBet al. The 2015 World Health Organization classification of lung tumors: Impact of genetic, clinical and radiologic advances since the 2004 classification. J Thorac Oncol. (2015) 10:1243–1260. doi: 10.1097/JTO.0000000000000630. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 3

  LuoWStevensTMStaffordPMiettinenMGatalicaZVranicS. NUTM1-rearranged neoplasms-a heterogeneous group of primitive tumors with expanding spectrum of histology and molecular alterations-an updated review. Curr Oncol. (2021) 28:4485–4503. doi: 10.3390/curroncol28060381. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 4

  LuoJBishopJADuBoisSGHannaGJShollLMStelowEBet al. Hiding in plain sight: NUT carcinoma is an unrecognized subtype of squamous cell carcinoma of the lungs and head and neck. Nat Rev Clin Oncol. (2025) 22:292–306. doi: 10.1038/s41571-025-00986-3. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 5

  YeZLiXXieFSunJYangDDengCet al. Expert consensus and current landscape of NUT carcinoma: A comprehensive strategy from diagnosis to treatment. Tumor Discov. (2024) 3:4904. doi: 10.36922/td.4904

  • CrossRef
  • Google Scholar
  • View reference in article

• 6

  HaackHJohnsonLAFryCJCrosbyKPolakiewiczRDStelowEBet al. Diagnosis of NUT midline carcinoma using a NUT-specific monoclonal antibody. Am J Surg Pathol. (2009) 33:984–991. doi: 10.1097/PAS.0b013e318198d666. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 7

  KimJJWaltonSAMahadevanNRHaradonJPaoloniFPaikPKet al. Molecular characterization of NUT carcinoma: A report from the NUT carcinoma registry. Clin Cancer Res. (2025) 31:3922–3931. doi: 10.1158/1078-0432.CCR-25-1071. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 8

  LuoJSanchezMLeeEHertzlerHLuongNMazzolaEet al. Initial chemotherapy for locally advanced and metastatic NUT carcinoma. J Thorac Oncol. (2024) 19:829–838. doi: 10.1016/j.jtho.2023.12.022. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 9

  YeZLiXZhangMXieFLuoXDengCet al. Complete response to BET inhibitor in primary pulmonary NUT carcinoma with single-cell sequencing-based analysis: A case report. JTO Clin Res Rep. (2025) 6:100885. doi: 10.1016/j.jtocrr.2025.100885. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 10

  YeZYanHLiXYangDXieFDengCet al. Radiotherapy outcomes and prognostic factors in head and neck NUT carcinoma: A single-center cohort analysis. Head Neck. (2026) 48:975–985. doi: 10.1002/hed.70119. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 11

  FrenchCAMiyoshiIAsterJCKubonishiIKrollTGDal CinPet al. BRD4 bromodomain gene rearrangement in aggressive carcinoma with translocation t(15;19). Am J Pathol. (2001) 159:1987–1992. doi: 10.1016/S0002-9440(10)63049-0. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 12

  HuangXJinJLiYZhangYZhangLSiWet al. Clinical treatment, research topic and global trends of Kaposiform hemangioendothelioma: a review and bibliometric analysis. Int J Surg. (2025). doi: 10.1097/JS9.0000000000004432. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 13

  PengLSongHLiTMaYYanCSunKet al. Epigenetic regulation of varicella-zoster virus infection: a systematic synthesis of mechanisms. Int J Surg. (2026) 112:4467–4485. doi: 10.1097/JS9.0000000000003912. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 14

  ViswanathanMAnsariMTBerkmanNDChangSHartlingLMcPheetersMet al. Assessing the risk of bias of individual studies in systematic reviews of health care interventions. In: Methods guide for effectiveness and comparative effectiveness reviews. Rockville, MD: Agency for Healthcare Research and Quality (2008). doi: 10.23970/ahrqepcmethguide2

  • CrossRef
  • Google Scholar
  • View reference in article

• 15

  BradfordSC. Sources of information on specific subjects. Engineering. (1934) 137:85–86. doi: 10.1300/j105v01n03_06. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 16

  FrenchCARamirezCLKolmakovaJHickmanTTCameronMJThyneMEet al. BRD-NUT oncoproteins: a family of closely related nuclear proteins that block epithelial differentiation and maintain the growth of carcinoma cells. Oncogene. (2008) 27:2237–2242. doi: 10.1038/sj.onc.1210852. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 17

  FrenchCAKutokJLFaquinWCToretskyJAAntonescuCRGriffinCAet al. Midline carcinoma of children and young adults with NUT rearrangement. J Clin Oncol. (2004) 22:4135–4139. doi: 10.1200/JCO.2004.02.107. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 18

  BauerDEMitchellCMStraitKMLathanCSStelowEBLuerSCet al. Clinicopathologic features and long-term outcomes of NUT midline carcinoma. Clin Cancer Res. (2012) 18:5773–5779. doi: 10.1158/1078-0432.CCR-12-1153. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 19

  YuanJXuZGuoY. Diagnosis, treatment and prognosis of primary pulmonary NUT carcinoma: A literature review. Curr Oncol. (2022) 29:6807–6815. doi: 10.3390/curroncol29100536. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 20

  ReynoirdNSchwartzBEDelvecchioMSadoulKMeyersDMukherjeeCet al. Oncogenesis by sequestration of CBP/p300 in transcriptionally inactive hyperacetylated chromatin domains. EMBO J. (2010) 29:2943–2952. doi: 10.1038/emboj.2010.176. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 21

  WangRLiuWHelferCMBradnerJEHornickJLJanickiSMet al. Activation of SOX2 expression by BRD4-NUT oncogenic fusion drives neoplastic transformation in NUT midline carcinoma. Cancer Res. (2014) 74:3332–3343. doi: 10.1158/0008-5472.CAN-13-2658. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 22

  EagenKPFrenchCA. Supercharging BRD4 with NUT in carcinoma. Oncogene. (2021) 40:1396–1408. doi: 10.1038/s41388-020-01625-0. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 23

  RousseauxSReynoirdNKhochbinS. NUT is a driver of p300-mediated histone hyperacetylation: From spermatogenesis to cancer. Cancers (Basel). (2022) 14:2234. doi: 10.3390/cancers14092234. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 24

  KosnoMCurrieSLKumarAXingCRosenMK. Molecular features driving condensate formation and gene expression by the BRD4-NUT fusion oncoprotein are overlapping but distinct. Sci Rep. (2023) 13:11907. doi: 10.1038/s41598-023-39102-9. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 25

  AlekseyenkoAAWalshEMWangXGraysonARHsiPTKharchenkoPVet al. The oncogenic BRD4-NUT chromatin regulator drives aberrant transcription within large topological domains. Genes Dev. (2015) 29:1507–1523. doi: 10.1101/gad.267583.115. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 26

  ShiotaHElyaJEAlekseyenkoAAChouPMGormanSABarbashOet al. "Z4" complex member fusions in NUT carcinoma: Implications for a novel oncogenic mechanism. Mol Cancer Res. (2018) 16:1826–1833. doi: 10.1158/1541-7786.MCR-18-0474. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 27

  ZhengDElnegiryAALuoCBendahouMAXieLBellDet al. Brd4::Nutm1 fusion gene initiates NUT carcinoma in vivo. Life Sci Alliance. (2024) 7:e202402602. doi: 10.26508/lsa.202402602. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 28

  HarukiNKawaguchiKSEichenbergerSMassionPPGonzalezAGazdarAFet al. Cloned fusion product from a rare t(15;19)(q13.2;p13.1) inhibit S phase in vitro. J Med Genet. (2005) 42:558–564. doi: 10.1136/jmg.2004.029686. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 29

  KuzumeTKubonishiITakeuchiSTakeuchiTIwataJSonobeHet al. Establishment and characterization of a thymic carcinoma cell line (Ty-82) carrying t(15;19)(q15;p13) chromosome abnormality. Int J Cancer. (1992) 50:259–264. doi: 10.1002/ijc.2910500216. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 30

  ToretskyJAJensonJSunCCEskenaziAECampbellAHungerSPet al. Translocation (11;15;19): a highly specific chromosome rearrangement associated with poorly differentiated thymic carcinoma in young patients. Am J Clin Oncol. (2003) 26:300–306. doi: 10.1097/01.COC.0000020960.98562.84. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 31

  DurallRTHuangJWojenskiLHuangYGokhalePCLeeperBAet al. The BRD4-NUT fusion alone drives malignant transformation of NUT carcinoma. Cancer Res. (2023) 83:3846–3860. doi: 10.1158/0008-5472.CAN-23-2545. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 32

  MoriniereJRousseauxSSteuerwaldUSoler-LopezMCurtetSVitteALet al. Cooperative binding of two acetylation marks on a histone tail by a single bromodomain. Nature. (2009) 461:664–668. doi: 10.1038/nature08397. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 33

  ShiotaHBarralSBuchouTTanMCouteYCharbonnierGet al. Nut directs p300-dependent, genome-wide H4 hyperacetylation in male germ cells. Cell Rep. (2018) 24:3477–3487.e6. doi: 10.1016/j.celrep.2018.08.069. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 34

  YeZLiXXieFSunJYangDDengCet al. A single-cell sequencing-based analysis of a 13-year-old with maxillary sinus NUT carcinoma. Oral Oncol. (2025) 162:107185. doi: 10.1016/j.oraloncology.2025.107185. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 35

  BellDChobyGAfkhamiMMaghamiEFerrarottoRSnydermanCet al. Magnifying glass on sinonasal NUT carcinoma heterogeneity via spatial transcriptomics. Head Neck. (2026) 48:E1–E9. doi: 10.1002/hed.28231. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 36

  HamiltonGARuizPDYeKGambleMJ. Acetylation of histone H2B on lysine 120 regulates BRD4 binding to intergenic enhancers. bioRxiv. (2025). doi: 10.1101/2025.02.07.637147. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 37

  WangQSadeghiSAFangFZhengDLuoCGaoGet al. High-throughput plasma/serum proteomics by coupling nanoparticle protein corona-based sample preparation and capillary zone electrophoresis-tandem mass spectrometry. Anal Methods. (2025) 17:6356–6365. doi: 10.1039/d5ay00721f. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 38

  BauerNBalourdasDISchneiderJRZhangXBergerLMBergerBTet al. Development of potent dual BET/HDAC inhibitors via pharmacophore merging and structure-guided optimization. ACS Chem Biol. (2024) 19:266–279. doi: 10.1021/acschembio.3c00427. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 39

  JensenJLPetersonSKSambadeMJAlleyJRYuSKinjoTet al. Shared PRAME epitopes are T-cell targets in NUT carcinoma. J Immunother Cancer. (2026) 14:e013539. doi: 10.1136/jitc-2025-013539. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 40

  FlaadtTLemelleLAbeleMVirgoneCBen-AmiTKachanovDet al. NUT carcinoma in children and adolescents: An analysis of the European Cooperative Study Group on pediatric rare tumors (EXPeRT). Lung Cancer. (2025) 201:108449. doi: 10.1016/j.lungcan.2025.108449. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 41

  KlokerLDSidirasMFlaadtTBrechtIBDeinzerCKWGrossTet al. Clinical management of NUT carcinoma (NC) in Germany: Analysis of survival, therapy response, tumor markers and tumor genome sequencing in 35 adult patients. Lung Cancer. (2024) 189:107496. doi: 10.1016/j.lungcan.2024.107496. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 42

  ViswanathanKHahnEDoganSWeinrebIDicksonBCMacMillanCet al. The histological spectrum and immunoprofile of head and neck NUT carcinoma: A multicentre series of 30 cases. Histopathology. (2024) 85:317–326. doi: 10.1111/his.15204. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 43

  ChauNGMaCDangaKAl-SayeghHNardiVBarretteRet al. An anatomical site and genetic-based prognostic model for patients with nuclear protein in testis (NUT) midline carcinoma: Analysis of 124 patients. JNCI Cancer Spectr. (2020) 4:pkz094. doi: 10.1093/jncics/pkz094. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 44

  LauerUMAwadaAPostel-VinaySShapiroGIThieblemontCPiha-PaulSAet al. Final results from the phase Ia/Ib study of the novel bromodomain and extra-terminal domain inhibitor, BI 894999, in patients with advanced solid tumors or diffuse large B-cell lymphoma. ESMO Open. (2025) 10:104499. doi: 10.1016/j.esmoop.2025.104499. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 45

  Tontsch-GruntUTraexlerPEBaumAMusaHMarzinKWangSet al. Therapeutic impact of BET inhibitor BI 894999 treatment: backtranslation from the clinic. Br J Cancer. (2022) 127:577–586. doi: 10.1038/s41416-022-01815-5. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 46

  EstevesGFerreiraJAfonsoRMartinsCZagaloCFelixA. HDAC overexpression in a NUT midline carcinoma of the parotid gland with exceptional survival: A case report. Head Neck Pathol. (2020) 14:1117–1122. doi: 10.1007/s12105-020-01130-6. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 47

  Piha-PaulSAHannCLFrenchCACousinSBranaICassierPAet al. Phase 1 study of Molibresib (GSK525762), a bromodomain and extra-terminal domain protein inhibitor, in NUT carcinoma and other solid tumors. JNCI Cancer Spectr. (2020) 4:pkz093. doi: 10.1093/jncics/pkz093. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 48

  ShapiroGILoRussoPDowlatiADoKTJacobsonCAVaishampayanUet al. A phase 1 study of RO6870810, a novel bromodomain and extra-terminal protein inhibitor, in patients with NUT carcinoma, other solid tumours, or diffuse large B-cell lymphoma. Br J Cancer. (2021) 124:744–753. doi: 10.1038/s41416-020-01180-1. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 49

  LewinJSoriaJCStathisADelordJPPetersSAwadaAet al. Phase Ib trial with Birabresib, a small-molecule inhibitor of bromodomain and extraterminal proteins, in patients with selected advanced solid tumors. J Clin Oncol. (2018) 36:3007–3014. doi: 10.1200/JCO.2018.78.2292. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 50

  DavisAMaharAWongKBarnetMKaoS. Prolonged disease control on Nivolumab for primary pulmonary NUT carcinoma. Clin Lung Cancer. (2021) 22:e665–e667. doi: 10.1016/j.cllc.2020.10.016. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 51

  LeeMSAnSSongJYSungMJungKChangESet al. Cancer-specific sequences in the diagnosis and treatment of NUT carcinoma. Cancer Res Treat. (2023) 55:452–467. doi: 10.4143/crt.2022.910. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 52

  CostaBAMaraveyasAWilkoffMHCorreiaGSCTallon de LaraPRohsNCet al. Primary pulmonary NUT carcinoma: Case illustration and updated review of literature. Clin Lung Cancer. (2022) 23:e296–e300. doi: 10.1016/j.cllc.2022.03.001. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 53

  AryalSCZiaSRodgersSShenYPerryKYuanL. BRD3-NUTM1-expressing NUT carcinoma of lung on endobronchial ultrasound-guided transbronchial needle aspiration cytology, a diagnostic pitfall. Diagn Cytopathol. (2022) 50:E47–E53. doi: 10.1002/dc.24885. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 54

  KlokerLDCalukovicBBenzlerKGolfABohmSGuntherSet al. Case report: Immunovirotherapy as a novel add-on treatment in a patient with thoracic NUT carcinoma. Front Oncol. (2022) 12:995744. doi: 10.3389/fonc.2022.995744. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 55

  ChoYAChoiYLHwangILeeKChoJHHanJ. Clinicopathological characteristics of primary lung nuclear protein in testis carcinoma: A single-institute experience of 10 cases. Thorac Cancer. (2020) 11:3205–3212. doi: 10.1111/1759-7714.13648. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 56

  FlaadtTWildHAbeleMFruhwaldMDirksenUClassenCFet al. NUT carcinoma in pediatric patients: Characteristics, therapeutic regimens, and outcomes of 11 cases registered with the German Registry for Rare Pediatric Tumors (STEP). Pediatr Blood Cancer. (2024) 71:e30821. doi: 10.1002/pbc.30821. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 57

  UgurluerGDincerNDanyeliAECelikLGunerALCorapciogluFet al. Nuclear protein in testis (NUT) midline carcinoma of the larynx: A rare case report of a paediatric patient and literature review. Cancer Radiother. (2024) 28:385–389. doi: 10.1016/j.canrad.2024.02.004. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 58

  CadeskyASchulman-RosenbaumRCarterAPaulEJaggiS. A rare case of NUT carcinoma of the thyroid. JCEM Case Rep. (2024) 2:luae037. doi: 10.1210/jcemcr/luae037. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 59

  CaoJLiuYLuC. Primary thyroid nuclear protein in testis carcinoma: a case report and literature review. Gland Surg. (2024) 13:1116–1125. doi: 10.21037/gs-24-77. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 60

  HerbisonHDavisSNicklessDHaydonAAmeratungaM. Sustained clinical response to immunotherapy followed by BET inhibitor in a patient with unresectable sinonasal NUT carcinoma. J Immunother Precis Oncol. (2024) 7:67–72. doi: 10.36401/JIPO-23-19. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 61

  ZhouJDuanMJiaoQChenCXingASuPet al. Primary thyroid NUT carcinoma with high PD-L1 expression and novel massive IGKV gene fusions: A case report with treatment implications and literature review. Front Oncol. (2021) 11:778296. doi: 10.3389/fonc.2021.778296. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 62

  FuSWangZLiCLiYZhangKZhongZet al. The whole treatment process and thinking of a patient with NUT carcinoma of the parotid gland: a case report. Front Oncol. (2023) 13:1094770. doi: 10.3389/fonc.2023.1094770. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 63

  RiessJWRahmanSKianWEdgerlyCHeilmannAMMadisonRet al. Genomic profiling of solid tumors harboring BRD4-NUT and response to immune checkpoint inhibitors. Transl Oncol. (2021) 14:101184. doi: 10.1016/j.tranon.2021.101184. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 64

  YangGLiuRYangLYangXTangXMaoH. Pulmonary NUT carcinoma, an elusive and refractory entity, shows transient response to chemotherapeutics and PD-1 inhibitor: a case report and literature review. Front Immunol. (2025) 16:1497124. doi: 10.3389/fimmu.2025.1497124. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 65

  CanerBOrhanSODeligonulAEvrenselT. Immunotherapy experience in sinonasal NUT midline carcinoma, case report. J Cancer Res Ther. (2024) 20:479–481. doi: 10.4103/jcrt.jcrt_1083_22. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 66

  BadranAAliSSArabiTZAlaklabiAMAbdallaHMMohammedSet al. Suspected NUT carcinoma progressing on pembrolizumab, carboplatin, and paclitaxel as first-line treatment: a case report. Ann Med Surg (Lond). (2024) 86:1061–1065. doi: 10.1097/MS9.0000000000001501. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 67

  HaebeSSchuebbeGJurmeisterPvon Bergwelt-BaildonMWestphalenCBLauerUMet al. Adding checkpoint inhibitors to first-line chemotherapy for NUT carcinoma patients. NPJ Precis Oncol. (2025) 9:26. doi: 10.1038/s41698-024-00768-7. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 68

  ChenMChenXZhangYWangWJiangL. Clinical and molecular features of pulmonary NUT carcinoma characterizes diverse responses to immunotherapy, with a pathologic complete response case. J Cancer Res Clin Oncol. (2023) 149:6361–6370. doi: 10.1007/s00432-023-04621-5. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 69

  LaiJKPanasciLPateyNGWangH. Placental metastasis from maternal NUT carcinoma: diagnostic pitfalls and challenges. BMJ Case Rep. (2024) 17:e259538. doi: 10.1136/bcr-2023-259538. PMID:

  • Pubmed Abstract
  • CrossRef
  • Google Scholar