Primary cutaneous CD8⁺ aggressive epidermotropic cytotoxic T-cell lymphoma with novel FGFR1 fusion treated with pemigatinib

TO THE EDITOR:

A 70-year-old male presented with a 2-year history of cutaneous T-cell lymphoma (CTCL), reported as CD8⁺. During this time, the patient had been treated with cyclophosphamide, doxorubicin, vincristine, and prednisone (6 cycles); gemcitabine and oxaliplatin (3 cycles); and brentuximab vedotin; with no response. A subsequent specimen demonstrated CTCL with marked epidermotropism by CD3⁺, CD4/CD8–double negative, T-cell receptor (TCR) βF1-positive, and cytotoxic granule–associated RNA-binding protein (TIA-1)-positive cells. His lesions rapidly progressed. On physical examination, there were patches, plaques, and tumors with varying degrees of ulceration involving the back, chest, and extremities, including palms and soles (Figure 1A).

A subsequent skin biopsy demonstrated an epidermotropic and folliculotropic CTCL, positive for CD8, CD5, CD7, and TIA-1 (Figure 1C-H), focally expressing granzyme B and perforin. A diagnosis of primary cutaneous CD8⁺ aggressive epidermotropic cytotoxic T-cell lymphoma (PCAECTL) was made. Positron emission tomography–computed tomography scan studies revealed multiple cutaneous focal uptakes and nonspecific hypermetabolic adenopathy. Flow cytometry analysis of the peripheral blood was negative for aberrant T cells. Skin-directed radiation therapy was recommended to the patient for symptomatic control. He received 6 Gy in 2 fractions to the bilateral palms (delivered with superficial 6 MeV electrons) and 6 Gy in 2 fractions to the bilateral feet (delivered with photons and bolus to treat the entire foot). Clinical improvement of the radiation-treated lesions was noted but they recurred shortly after.

Next-generation sequencing testing for somatic mutations performed on formalin-fixed, paraffin-embedded (FFPE) sections of the tumor disclosed a DNMT3A p.E545G missense mutation (interpreted as a variant of unknown significance). A clinically validated targeted RNA-sequencing assay using anchored multiplex polymerase chain reaction technology (Archer Heme Fusion assay) was also performed on FFPE sections, as previously described.¹ Anchored multiplex polymerase chain reaction revealed an in-frame fusion transcript involving special AT-rich sequence-binding protein 1 (SATB1) and fibroblast growth factor receptor 1 (FGFR1; Figure 2A). Fluorescence in situ hybridization showed 5′FGFR1 deletion (Figure 2B). Cell viability assays were conducted. The SATB1::FGFR1 fusion transcript was demonstrated to be an oncogene (Figure 2C-E).

The patient was started on pemigatinib, 13.5 mg daily. He reported almost 90% improvement in all his ulcerated lesions after 2 weeks of therapy and remained in partial remission for ∼4 months (Figure 1B). Side effects included hyperphosphatemia, nail changes, and eye irritation/sensitivity. Because of ocular toxicities, the pemigatinib dose was decreased to 9.0 mg daily. The patient’s lesions recurred 1 month later as thick ulcerated erythematous scaly patches and plaques on his trunk and extremities. Skin biopsy showed CTCL, weakly positive for CD8, positive for TCR βF1, TIA-1, and granzyme B. Whole-genome sequencing (WGS) and RNA sequencing were performed on DNA and RNA extracted from FFPE samples, as previously described.⁴ Single nucleotide polymorphism and indel discovery was performed using MuTect2.0.⁵ An internal normal panel was used to denoise contaminating germ line and variants occurring frequently in healthy populations, along with FFPE artifacts, as previously described.⁶ 

Next-generation sequencing reidentified the missense mutation in DNMT3A (p.E545G). WGS also identified a nonsense mutation in XRCC1 (p.R31∗). The complete list of mutations detected by WGS in the relapse specimen can be found in supplemental Table 1. WGS and RNA sequencing detected persistent SATB1::FGFR1 fusion and, in addition, 2 point mutations, FGFR1 p.V561L and p.E462V, absent in the pretreatment specimen (Figure 2F-G). The patient passed away 2 months after relapse.

Although most patients with PCAECTL show a rapidly progressive and lethal clinical course, some cases with initial presentation mimicking mycosis fungoides (MF) and longer survival have been reported.⁷ Early diagnosis of PCAECTL can be challenging because it may mimic lymphomatoid papulosis type D^8,9 and pyoderma gangrenosum.¹⁰ PCAECTL tends to show preserved expression of CD7 and loss of CD5; expression of CD5 is however not rare.¹¹ CD4/CD8–double-negative phenotype is recognized in PCAECTL, along with variable intensity of CD8 and phenotype switches.^7,12 Expression of at least 1 cytotoxic molecule such as TIA-1, granzyme B, and perforin is usually detected.⁷ 

PCAECTL harbors recurrent deregulation of Janus kinase 2 (JAK2) signaling by the presence of gene fusions involving JAK2 and PTPRC (CD45),^13,14 rarely detected in other primary cutaneous cytotoxic lymphomas. Kinase fusion partners preserve the entire kinase domain and are sensitive to JAK inhibitors. By array comparative genomic hybridization, loss of 9p21.3 (CDKN2A and CDKN2B loci) appears to be frequent.¹⁵ In our case, a missense mutation in DNMT3A was present in both initial and recurrent lesions. Somatic mutations in chromatin-modifying genes such as DNMT3A with loss of function are relatively frequent in CTCL (42.5%) and peripheral T-cell lymphomas.⁶ We also detected a XRCC1 truncating mutation in the relapsed lymphoma. Although DNMT3A and XRCC1 abnormalities might be oncogenic, we must be cautious to not overinterpret the significance of these variants in a single case. A recent report using single-cell RNA with TCR sequencing demonstrated PCAECTL to display a cytotoxic effector phenotype, expressing GZMA, NKG7, CTSW, and CST7, in contrast to the central memory phenotype exhibited by MF. Interestingly, upregulation of transcripts associated with epidermal repair and innate immune response (such as KRT6 and KRT16), strongly detected in MF, was not present in PCAECTL, suggesting a molecular mechanism underlying its characteristic clinical presentation with ulcerated and necrotic lesions.¹⁶ 

Our patient’s tumor was extensively studied by WGS and RNA sequencing. FFPE tissue was used as source of DNA and RNA. In order to prevent bias due to artifacts introduced by formalin fixation, we used an established strand orientation bias filter to eliminate variants with high probability of FFPE artifacts.¹⁷ We identified a novel SATB1::FGFR1 fusion, not reported in PCAECTL or other T-cell lymphomas. FGFR1 fusions can be detected in several solid tumors (lung and breast carcinoma, rhabdomyosarcoma, glioblastoma, pilocytic astrocytoma, among others) and some hematopoietic disorders such as 8p11 myeloproliferative syndrome, acute myeloid leukemia, and B-cell lymphomas, in which different fusion partners have been identified (ZMYM2, BCR, and others, but, to our knowledge, no SATB1). SATB1 alterations appear to have a role in acute myeloid leukemia progression and lymphocyte differentiation.¹⁸ A SATB1 (intron 5)::ALK (exon 7) gene fusion has been detected in anaplastic lymphoma kinase–positive anaplastic large-cell lymphoma.¹⁹ SATB1 is a chromatin binding protein with dimerization domains, therefore presumed to lead to dimerization activation of the FGFR1 kinase. The SATB1::FGFR1 fusion links the nuclear localization signal, the ubiquitin-like domain, the CUT1-like DNA binding domain of SATB1 and the tyrosine kinase (TK) domains of the FGFR1 protein. The preserved domains in FGFR1 are similar to those preserved in other fusions described as type 1 FGFR fusions. To functionally validate this translocation, we transduced Ba/F3 cells with the novel SATB1::FGFR1 fusion. SATB1::FGFR1 but not controls showed interleukin-3–independent growth (Figure 2C), supporting its oncogenic nature. The US Food and Drug Administration–approved FGFR inhibitor pemigatinib inhibited growth in SATB1::FGFR1–transduced cells in vitro (50% inhibitory concentration = 5.3 nM; Figure 2D). SATB1::FGFR1 induced phosphorylation of known target genes of FGFR1, including itself and signal transducer and activator of transcription 5, again inhibited by pemigatinib (Figure 2E).

Therapies for PCAECTCL are extremely limited, with some partial responses with immune checkpoint inhibitors and infliximab²⁰ and isolated reports of long-term remission after aggressive multiagent chemotherapy,²¹ brentuximab vedotin,²² and stem cell transplantation.²³ In our patient, pemigatinib induced almost complete response but recurrence was observed right after a dose reduction because of side effects. WGS and RNA sequencing of the postrelapsed sample showed a gatekeeper mutation (V561L) and E462V, a novel FGFR1 mutation within the juxtamembrane fusion region of FGFR1. Using the aforementioned method for detection of bias due to FFPE artifacts, the calculated probability of artifact of the FGFR1 p.V561L variant was 0.004791, whereas the probability of the FGFR1 p.E462V mutation was 0.0006002. Furthermore, we confirmed the existence of these variants in both the genomic and RNA-sequencing data sets, further supporting their existence.

One common mechanism of resistance to TK inhibitors is the replacement of the gatekeeper residue, located deep in the active site of the tyrosine, which controls access of the TKIs to the adenosine triphosphate–binding pocket.²⁴ FGFR1 V561 site mutations such as V561M and V561I have been characterized to show resistance to FGFR inhibitors including pemigatinib.²⁵ To our knowledge, the E462 site has not been identified as a mutated site in FGFR1. To determine whether the V561L and E462V variants drive SATB1::FGFR1 resistance to pemigatinib, Ba/F3 cells were transduced with SATB1::FGFR1 expressing the FGFR1 V561L and E462V variants.; V561L alone conferred resistance to pemigatinib (Figure 2G). Further studies are needed to better define the mechanisms that drive FGFR1 resistance.

In conclusion, this is a unique case of PCAECTL harboring a SATB1::FGFR1 fusion and showing initial clinical response to pemigatinib. Our case expands the spectrum of activating fusions detected in PCAECTL and adds a potential therapeutic option for these aggressive lymphomas.

This study was approved under an institutional review board–approved protocol.

Contribution: C.A.T.-C., J.C., and V.N. designed the research; C.A.T.-C., J.A.Y., Y.Z., K.T., G.T., J.C., and V.N. performed the research; A.H., G.S., G.T., J.L.C., J.G., B.D., and S.I. contributed analytical tools; C.A.T.-C., J.A.Y., Y.Z., A.H., G.S., and S.I. collected data; C.A.T.-C., J.A.Y., Y.Z., R.N.M., J.C., and V.N. analyzed and interpreted the data; J.A.Y., Y.Z., and K.T. performed statistical analysis; and C.A.T.-C., J.A.Y., and Y.Z. wrote the manuscript.

Conflict-of-interest disclosure: J.C. is a cofounder and stakeholder in Moonlight Bio, which has no relevance to the findings in this manuscript. The remaining authors declare no competing financial interests.

Correspondence: Carlos A. Torres-Cabala, Department of Pathology, The University of Texas MD Anderson Cancer Center Pathology, 1515 Holcombe Blvd, Unit 85, Houston, TX 77030; email: ctcabala@mdanderson.org.