Primary cytotoxic T-cell lymphomas harbor recurrent targetable alterations in the JAK-STAT pathway

TO THE EDITOR:

Cytotoxic cutaneous T-cell lymphomas (CTCLs) are a collection of uncommon but highly aggressive cancers of skin-homing cytotoxic T cells. These include the established World Health Organization (WHO) entity primary cutaneous γδ T-cell lymphomas (PCGDTLs), the provisional WHO entity primary cutaneous CD8⁺ aggressive epidermotropic T-cell lymphoma (PCAETCL), and samples that do not currently fit a specific WHO category, which we refer to as cytotoxic CTCL-NOS (not otherwise specified).¹ 

All 3 entities present with ulcerated skin lesions and express at least 1 cytotoxic enzyme by histopathology (supplemental Figure 1, available on the Blood Web site). These cytotoxic CTCL entities are clinically similar.^2,3 There are no statistically significant differences in clinical presentation, survival, or response to therapy.^2,3 Median survival across all 3 cytotoxic CTCL subtypes has been reported to be from 12 to 15 months.^3,4 Because of their dismal clinical outcomes, greater insights into disease pathogenesis are needed.

We recently defined the landscape of PCGDTLs.⁵ However, little is known about the other cytotoxic CTCLs. A recent whole-genome sequencing (WGS) study of 12 PCAETCLs found recurrent JAK2 fusions in 3 PCAETCLs.⁶ Otherwise, there is little known about these aggressive cancers.

To bridge this knowledge gap, we obtained 35 cases of cytotoxic CTCLs. All cases expressed a pan T-cell marker and at least 1 cytotoxic enzyme (TIA-1, perforin, or granzyme B). Cases were obtained from 10 academic institutions (Massachusetts General Hospital, Brigham and Women’s Hospital, Northwestern, Stanford, University of Utah, University of Pennsylvania, University of Wisconsin–Madison, Moffit Cancer Center, University of California–Irvine, City of Hope). All slides were reviewed in a blinded manner by 2 expert dermatopathologists (L.M.D. and J.G.).

All PCGDTLs expressed the γ or δ T-cell receptor. All 9 PCAETCLs expressed CD8 and had full-thickness epidermotropism. The 20 CTCL-NOSs did not meet criteria for PCAETCL because they were either CD8⁻, did not have full-thickness epidermal involvement, or had mixed histological features characteristic of other CTCLs (Table 1). Consistent with recent reports,³ there were no statistically significant differences in sites of disease, age, or sex between PCAETCLs and CTCL-NOSs in our cohort (Fisher’s exact test). Both PCAETCL and CTCL-NOS had aggressive clinical courses. Median survivals were 12.2 months and 11 months, respectively (Figure 1A).

To identify oncogenic gene fusions, we performed a clinically validated targeted RNA sequencing (RNA-Seq) assay using anchored multiplexed polymerase chain reaction (PCR) technology,⁷ the Archer Heme Fusion assay, on 6 PCGDTLs, 6 PCAETCLs, and 12 CTCL-NOSs formalin-fixed paraffin-embedded (FFPE) samples. Surprisingly, all 6 PCAETCLs harbored a kinase gene fusion; 5 had a JAK2 fusion, and one had a novel ABL1 fusion. In contrast, none of the CTCL-NOSs or PCGDTLs expressed a kinase fusion (Figure 1B).

To investigate this further, we performed WGS and/or RNA-Seq on 3 additional PCAETCLs (1 fresh frozen and 2 FFPE) and 7 CTCL-NOSs (4 fresh frozen and 3 FFPE). The WGS and RNA-Seq were prepared and analyzed as previously described.⁵ We also added custom probes spanning JAK2 to enrich our RNA-Seq for JAK2 fusions. Recently, we performed a comprehensive genomic analysis of PCGDTLs that identified mutations in JAK/STAT, MYC, and MAPK pathways.⁵ Thus, we did not pursue additional molecular studies on this cohort.

This analysis identified JAK2 fusions in all additional 3 PCAETCLs (Figure 1B). None of the additional CTCL-NOSs harbored an oncogenic fusion. Collectively, PCAETCLs in our cohort harbored JAK2 fusions reported in other cancers (PCM1-JAK2 and STAT3-JAK2),^8-10,JAK2 fusions recently reported for PCAETCL (KHDRBS1-JAK2),⁶ and 3 fusions not previously described in any cancer (PICALM-JAK2, CAPRIN1-JAK2, and SELENOI-ABL1) (Figure 1B).

All JAK2 translocations involved introns 15 to 18, preserved the catalytic domain, but lost the upstream FERM, SH2, and the regulatory pseudokinase domains. The ABL1 fusion also retained its catalytic tyrosine kinase domain but lost its N-terminal regulatory SH2 and SH3 domains. This fusion is analogous to NPM1-ABL1, but not BCR-ABL1, fusions, which also lose their SH2/SH3 domains.

To functionally validate these fusions, we obtained cDNA for CAPRIN1, JAK2, and ABL1 (Horizon; Addgene). We cloned the fusion genes into pMSCVpuro-DEST vector (Addgene). We transduced Ba/F3 cells with retrovirus encoding CAPRIN1-JAK2, SELENO-ABL1, or wild-type JAK2 or ABL1. Like other oncogenes,¹¹ both gene fusions, but not wild-type kinases, conferred IL-3–independent growth (supplemental Figure 2A-B). Both CAPRIN1-JAK2 and SELENOI-ABL1 increased phosphorylation of targets downstream in their respective signaling cascades (Figure 1C). For CAPRIN1-JAK2, these include phospho-JAK2 (pJAK2), pSTAT3, and pSTAT5. For SELENOI-ABL1, these include pABL, pSTAT5, and pCRKL.

Next, we assessed their potential targetability. The JAK1/JAK2 inhibitor ruxolitinib decreased pJAK2, pSTAT3, and pSTAT5 and decreased viability of cells expressing CAPRIN1-JAK2 fusions (50% inhibitory concentration [IC₅₀] = 402.2 nM). Similarly, imatinib inhibited SELENOI-ABL1–dependent phosphorylation cascades and viability (IC₅₀ = 1641 nM) (Figure 1D). As previously shown for NPM1-ABL1, the allosteric inhibitor asciminib that binds a myristoyl site in ABL1¹² had no effect on SELENOI-ABL1 fusion expressing cells (data not shown).

We then analyzed the RNA-Seq data for transcriptional evidence of JAK-STAT activation. As expected, gene set enrichment analysis demonstrated a significant upregulation of a JAK-STAT gene signature in PCAETCLs (P value = .005) compared with unstimulated CD8⁺ T cells (Figure 1E). To confirm this observation at the protein level, we performed IHC for pSTAT3 (clone D3A7, Cell Signaling Technology) on 5 samples. These include 1 PCAETCL expressing the SELENOI-ABL1 fusion (predicted to phosphorylate pSTAT5 but not pSTAT3), 3 PCAETCL samples with JAK2 fusions (PCM1-JAK2 and STAT3-JAK2), and a CTCL-NOS without a known JAK-STAT mutation. As expected, we found pSTAT3 in the 3 samples with JAK2 fusions but not in the other samples (supplemental Figure 3).

Next, we analyzed our WGS data as previously described¹³ (supplemental Table 1). Five of 7 (71%) CTCL-NOSs had a somatic single nucleotide variant (SSNV) predicted to activate JAK-STAT signaling, including hotspot mutations in STAT3, p.D661Y (NU9) and p.P714L (NU11), JAK1, p.G1097D (NU14), JAK3, p.P906S (UT22), a damaging mutation in SOCS1, p.Q108* (NU7), and a SH2B3 splice site mutation (NU11). As previously reported, we found CDKN2A mutations in all PCAETCLs.⁶ These were also present in 43% of CTCL-NOSs (Figure 1F). Moreover, CTCL-NOSs and/or PCAETCLs harbored mutations in other genes recurrently mutated in CTCL,^11,14 including chromatin modification (ARID1A, KMT2D, CREBBP), cell cycle (CDKN1B), and DNA damage genes (ATM). Consistent with their genetics, CTCL-NOSs also had significant enrichment of the JAK-STAT gene set in their RNA-Seq data (P = .00001). Despite the negative pSTAT3 IHC in the 1 CTCL-NOS sample, these data suggest transcriptional activation of ≥1 STATs in this disease cohort.

Our observations have immediate clinical importance for both classification and treatment. Strikingly, all 9 PCAECTLs in our series harbored kinase fusions, suggesting that these fusions may be a disease-defining characteristic. At the same time, the majority of CTCL-NOSs harbor SSNVs that activate JAK/STAT signaling. Although both cancers may require oncogenic JAK-STAT signaling, the literature suggests variable genotype-driven responses to JAK inhibitors. JAK2 fusions predict exceptional response to small molecule inhibitors, while responses in cases with activating SSNVs are variable.¹⁵ A recent study of ruxolitinib in peripheral T-cell lymphoma and CTCLs reported responses in 23% of patients with activating JAK/STAT SSNVs.¹⁶ 

The reasons underlying the genetic differences between PCAETCLs and CTCL-NOSs remain unclear. It is possible that these kinase fusions and JAK/STAT SSNVs have subtle molecular differences. Nonetheless, our data add to the growing body of literature wherein molecular differences between tumors correlate with histological patterns.^17,18

We received informed consent for these patients in accordance with Northwestern Institutional Review Board standards.