Non-JAK-family gene fusions in cutaneous T-cell lymphoma highlights genetic diversity and potential therapeutic targets

Chromosomal translocations resulting in gene fusions have important implications in human cancer, ranging from diagnostic input to treatment indications.¹ Several recurrent gene fusions have been described in mature T- and NK-cell neoplasms, which comprise a rare, heterogenous group of over 30 distinct clinicopathological entities. The hallmark example is recurrent ALK translocations in anaplastic lymphoma kinase (ALK)-positive anaplastic large cell lymphoma (ALCL), which genetically define the disease and offer therapeutic opportunity through the use of ALK tyrosine kinase inhibitors. Other examples include recurrent IRF4/DUSP22 and TP63 rearrangements in ALK-negative ALCL and recurrent JAK2-STAT3 fusions in indolent T-cell lymphoproliferative disorder of the gastrointestinal tract. In the cutaneous T-cell lymphomas (CTCL), gene fusions are less well-described. Notable exceptions include recurrent JAK2 fusions in cytotoxic CTCL, in particular primary cutaneous CD8⁺ aggressive epidermotropic cytotoxic T-cell lymphoma (AECTCL) and recurrent TYK2 fusions in primary cutaneous CD30⁺ T-cell lymphoproliferative disorders (LPD). These events are well described, including a contemporaneous report from our center and a recent report of pediatric cases.^2,3 However, in other CTCL, including mycosis fungoides (MF), the frequency, identity, and implications of non-JAK-family gene fusions are undefined. Recently, we identified a cohort of patients with CTCL harboring non-JAK-family fusions and herein describe their molecular features and clinical course. Through a retrospective analysis of 113 specimens from 99 unique patients from our institutional CTCL clinic, we identified patients with CTCL harboring predicted productive non-JAK-family gene fusions. Fusions were detected through the RNA-based Archer FusionPlex™ Custom Heme Panel, which uses anchored multiplex polymerase chain reaction⁴ to detect expressed fusion transcripts. This panel consists of 199 cancer-related genes known to be recurrently involved in chromosomal rearrangements in hematologic malignancies. When available, we also reviewed next-generation sequencing results from the MSK-IMPACT Heme Panel, a custom paired tumor and normal DNA hybridization capture-based next-generation sequencing platform (see reference for genes covered).⁵ For diagnostic confirmation and characterization of CTCL, clinical and pathology records and images underwent re-review by a dermatopathologist, a dermatologist, and an oncologist with specific expertise in the field. All research was conducted under an Institutional Review Board-approved retrospective research protocol.

In total, we identified 17 patients with 20 unique gene fusions (3 patients had multiple fusions). An overview of these patients is shown in Online Supplementary Table S1. Fusions were seen across histological types. Six patients were best characterized as having primary cutaneous CD30⁺ T-cell LPD, including two with clinicopathological presentations consistent with lymphomatoid papulosis and one with primary cutaneous ALCL. Five patients were best characterized as having MF with a typical CD4⁺ phenotype, whereas three were considered to have MF with an unusual phenotype, which included one ‘double-positive’ case (CD4⁺/CD8⁺), one ‘double-negative’ case with a γ δ phenotype, and one CD8⁺ MF with large cell transformation. Finally, two patients were best characterized as having primary cutaneous peripheral T-cell lymphoma (PTCL) not otherwise specified (NOS) and one as having CD8⁺ AECTCL. As for recurrent gene fusions, TP63 fusions were detected in three patients with CD30⁺ T-cell LPD (2 cases with TBL1XR1-TP63 and 1 with FOXK2-TP63, both of which have been previously reported^6,7). TBL1XR1, which has broad biological functions, was present in a separate fusion involving BCL6 in the patient with CD8⁺ MF with large cell transformation, and BCL6 was involved in a separate fusion with CXCL8 in a patient with lymphomatoid papulosis. Finally, two ROS1 fusions were detected, one in a patient with lymphomatoid papulosis (VIM1-ROS1, novel) and one in the case of double-negative MF with a γδ phenotype (LMNA-ROS1, previously reported⁸). All other cases were unique within this cohort.

The TP63 fusions we identified have been previously shown to act as bona fide oncogenes.⁶ The patients with these fusions (patients 1, 2, and 6) had generally aggressive clinical courses with multiply refractory disease requiring several sequential treatments, including two allogeneic hematopoietic stem cell transplants in patient 2. Together with prior literature,^6,7 these cases support that structural abnormalities in TP63 or events that alter TP63 function may correlate with aggressive clinical behavior and resistance to standard therapies. For the other unique non-TP63 fusions, we did not detect any clear unifying features to categorize these events. Online Supplementary Table S2 shows additional details of each fusion, including whether a kinase or transcription factor was involved, prior reports in the literature, and hypothesized function. Of the 20 total fusions in our cohort, 14 have not been reported to our knowledge, cross-referencing cases against the Mitelman Database of Gene Fusions in Cancer, a comprehensive online database of over 34,000 gene fusions previously reported in human cancers (https://mitelmandatabase.isb-cgc.org). Of the six previously reported fusions (FOXK2-TP63, TBL1XR1-TP63, NEK6-PBLX1, IKZF2-ERBB4, FIP1L1-PDGFRA, LMNA-ROS1), two had not been previously reported in T-cell lymphoma (NEK6-PBX1 has been reported in one instance in mantle cell lymphoma⁹ and FIP1L1-PDGFRA is a hallmark of various hypereosinophilic syndromes, though this patient had no peripheral eosinophilia nor evidence of organ system eosinophilia¹⁰). Six cases involved kinases (NEK6, PDGFRA, GUCY2C, ERBB4, ROS1) and 11 cases involved transcription factors (FOXK2, TP63, NF-KB, BCL6, PBX1, ETV, RUNX1, ZFHX3, MKL1, IKZF2, RORA). Aside from three patients with stage IA disease and one patient with stage IB disease responsive to phototherapy, all others had advanced disease (≥ stage IB) requiring multiple therapies, with three patients undergoing allogeneic hematopoietic stem cell transplantation and five patients dying of disease and/or infection. The tumor mutational burden varied across cases (Figure 1, Online Supplementary Table S3). Some patients had a generally high burden, including patients 1 and 2, each with CD30⁺ T-cell LPD, TP63 fusions, and >10 mutations. Other patients with comparably high burden among the cohort included patient 4 (CD30⁺ T-cell LPD, RORA-PDCD1LG2, 18 mutations, multiply relapsed), patient 9 (MF, KMT2A-IFT46, 11 mutations, died of disease), patient 11 (MF, IKZF2-ERBB4 and PML-FBXO25, 11 mutations, died of disease), patient 15 (CD8⁺ AECTCL, ETV-GUCY2C, 15 mutations, multiply relapsed), and patient 16 (primary cutaneous PTCL-NOS, SETD2-LMCD1, 15 mutations, died of disease). In contrast, other patients had a generally low burden, including the two patients with disease behavior most consistent with lymphomatoid papulosis (patients 3 and 5) and the one patient with limited-stage CD4⁺ MF responsive to topical steroids (patient 7), each having fewer than three mutations (patient 3 with lymphomatoid papulosis had a WT1 mutation, patient 5 with lymphomatoid papulosis had FLT3, EPHA7, ERBB4 mutations, and patient 7 with CD4⁺ MF had no detected mutations). Finally, across histological types, seven patients had TP53 and/or CDKN2A mutations (patients 1, 2, 4, 10, 11, 15, 17), generally uncommon events in CTCL although previously shown to associate with increased tumor burden and poor prognosis.¹¹ All of these patients had multiply relapsed courses and two died of disease. To gain insight into the potential clonality of the identified fusions (which could provide insight on driver function), we integrated the quantitative read support for each fusion (expressed as the percentage of unique breakpoint-spanning reads) with the estimated tumor purity of the sequenced sample (shown in Online Supplementary Table S3 as %Reads). Doing so revealed a spectrum of findings. In several cases, high fusion %Reads were observed in samples with substantial tumor purity, suggesting a clonal, potentially driver event. For instance, in patient 2 (TBL1XR1TP63, 80% tumor purity), the fusion was supported by over 50% of reads. Conversely, other fusions were detected at lower %Reads even in samples with moderate tumor purity. For example, the FIP1L1-PDGFRA fusion in patient 12 (30% tumor purity) was supported by only 3.55% of reads. Herein we describe 20 gene fusions across CTCL, 14 of which are previously unreported. CTCL have complex mutational landscapes, with few recurrent mutations and a high proportion of somatic copy number variants comprising driver mutations.¹¹ Our study contributes to this complexity, showing that multiple gene fusions are detectable in CTCL, even within the same patient. Gene fusions have been described in PTCL, but their contribution to the pathogenesis and behavior of CTCL remains poorly characterized. Many gene fusions create constitutively active kinase or transcription factor fusion proteins,¹ hence our attempt to categorize events based on the suspected fusion product. Most of the cases here did indeed involve either kinases or transcription factors, although despite our attempt to hypothesize function, the true consequences of most of these events and their contribution to pathogenesis are unclear and would require functional studies. The noteworthy exceptions are the three TP63 fusions, which have been shown through intricate pre-clinical experiments to coordinate the recruitment of epigenetic modifying complexes and drive upregulation of MYC and EZH2, thereby contributing to the development of aggressive B- and T-cell lymphomas.⁶ The patients with these fusions in our cohort had aggressive, multiply refractory disease. Knowing that TP63 rearrangements are associated with inferior survival in PTCL and have been previously demonstrated in aggressive cases of MF,¹² these findings support the notion that TP63 rearrangements correlate with aggressive disease biology. As for the other events, with the exception of the NEK6PBX1 fusion, all are in-frame, which at least suggests that these fusions could produce functional proteins, although the oncogenic activity and contribution to disease of these proteins would require further study. While we attempted to gain insight into such through integration of read support and tumor purity, the interpretation of these metrics requires caution due to technical and biological factors. First, the %Reads metric is derived from RNA sequencing, thereby reflecting transcript abundance, not direct DNA allele frequency. As such, a low %Reads could indicate multiple scenarios, including a genetically subclonal event present in only a fraction of tumor cells, low expression of the fusion transcript relative to wild-type alleles, or contribution from non-tumor cells not fully accounted for in the hematoxylin & eosin tumor purity estimate. For example, low read support for the KMT2A-IFT46 fusion in patient 9 (2.36 %Reads, <5% tumor purity) highlights a case in which low tumor content and potential subclonality complicate interpretation, whereas high support for the VIM1-ROS1 fusion in patient 5 (66.78 %Reads) despite a modest tumor purity (15%) seems to suggest high expression and clonal representation within the predominant tumor cell population. Therefore, while a high fusion fraction in a high-purity sample strongly supports a truncal driver role, a low fraction cannot be dismissed as non-pathogenic, as it could be a subclonal driver, an event with low transcriptional output, or an artifact of sampling heterogeneity. This quantitative framework highlights the candidate fusions for which orthogonal DNA-level validation or single-cell analysis would be most informative.

Figure 1.Integrated OncoPrint of genomic alterations. The OncoPrint summarizes alterations across patients. Each column represents an individual patient and each row corresponds to a particular gene. Genes are grouped by biological pathways (left sidebar). Colored bars within the matrix indicate specific mutation types. The top annotation bar shows the detected gene fusion, and the bar immediately below denotes the diagnosis. AETCL: aggressive epidermotropic cytotoxic T-cell lymphoma; CD30⁺T-LPD: CD30⁺ T-cell lymph o proliferative disorder; LYP: lymphomatoid papulosis; MF: mycosis fungoides; PCPTCL-NOS: primary cutaneous peripheral T-cell lymphoma not otherwise specified.

Finally, we note that when the MSK-IMPACT Heme Panel was used, many cases demonstrated multiple additional mutations and copy number variants, some of which are considered to be oncogenic according to OncoKB,¹³ the precision-oncology knowledge base developed and maintained at our institution (highlighted in Online Supplementary Table S3). While the presence of these other oncogenic events could suggest that the fusions are passengers and more indicative of broad genomic instability, without functional characterization it is not possible to know for sure.

Whether any of the detected fusions represent therapeutic vulnerabilities is intriguing. EZH inhibition in TP63-rearranged cell lines and transgenic mice impairs tumor growth, and at least one patient with TP63-rearranged PTCL treated with the EZH inhibitor valemetostat in a phase I/II trial (NCT04703192) showed an initial disease response in the peripheral blood (though subsequently experienced cytomegalovirus reactivation and was removed from study).⁶ EZH inhibition is being investigated in CTCL (NCT06733441, NCT05944562) and could represent a therapeutic option for these aggressive cases. Other potential targetable fusions in our series include imatinib against FIP1L1-PDGFRA,¹⁰ pan-HER inhibitors, such as lapatinib, against ERBB4 fusions,¹⁴ tyrosine kinase inhibitors against ROS1 fusions,¹⁵ and checkpoint blockade, such as pembrolizumab, against PDCD1LG2 fusions, which could amplify PD-L2 and promote immune evasion. Acknowledging the uncertain contribution of these fusions to disease progression and the anecdotal nature of these “N of 1” cases, we see promise in the identification of a potential vulnerability in many of these otherwise refractory cases through fusion detection.

In summary, we identified several novel gene fusions across various CTCL, highlighting the genomic complexity in these heterogeneous diseases. While not routine, evaluation for fusions in multiply relapsed disease may lend insight into disease biology and uncover unrealized potential therapeutic options.

Footnotes

• Received October 23, 2025
• Accepted March 2, 2026

Correspondence

*RS and HT contributed equally as first authors

R. STUVER - stuverr@mskcc.org

Disclosures

No conflicts of interest to disclose.

Contributions

Funding

The study was funded in part by NIH/NCI Cancer Center Support grant P30 CA008748.