Targetability of STAT3-JAK2 Fusions: Implications for T-cell Lymphoproliferative Disorders of the Gastrointestinal Tract

To the Editor:

The growth and survival of T cells broadly depend on JAK-STAT signaling, and dysregulation of this pathway is common and potentially targetable in a variety of T-cell lymphoproliferative disorders and lymphomas [1, 2]. We previously identified novel and recurrent STAT3-JAK2 fusion genes associated with t(9;17)(p24.1;q21.2) in patients with T-cell lymphoproliferative disorder of the gastrointestinal tract (GI TLPD)[3], a new provisional entity in the 2017 revision of the World Health Organization (WHO) classification of lymphoid neoplasms [4]. GI TLPD is characterized by debilitating gastrointestinal symptoms and a chronic, relapsing course. It is often refractory to both symptomatic management and conventional chemotherapy regimens. While characterized as an indolent lymphoproliferative disorder, some cases transform into overt T-cell lymphomas, particularly those with a CD4-positive phenotype [4]. We reported STAT3-JAK2 fusions in 4 of 5 CD4-positive GI TLPDs but not in cases with other phenotypes, and 1 patient with STAT3-JAK2 underwent fatal large-cell transformation [3]. The fusion transcript encoded a fusion protein containing most of STAT3 and the JAK homology 1 (JH1) tyrosine kinase domain of JAK2, and was associated with immunohistochemical detection of phosphorylated (p) STAT5^Y694 but not pSTAT3^Y705. The recurrence of the fusion and the chronic, refractory nature of GI TLPDs prompted us to characterize the function and targetability of STAT3-JAK2.

Because lymphoma-associated translocations are frequently the initial events in transformation [5], and since no GI TLPD model exists, we first assessed the ability of the fusion to induce growth in normal human CD4-positive T cells (see Supplementary Material for Methods). Lentiviral STAT3-JAK2 transduction increased growth by a factor of 9.4±1.0 over control vector at 72 h (P=0.0001; Figure 1A). We then evaluated growth in the murine Ba/F3 model, in which the transforming properties of JAK2 fusions have been validated previously [6]. STAT3-JAK2 rescued Ba/F3 cells from IL3 withdrawal and promoted growth 690-fold by day 13, while control-transduced cells died (P=0.00002; Figure 1B). Furthermore, STAT3-JAK2-transduced Ba/F3 cells grew subcutaneously in NSG mice without exogenous IL3 supplementation, reaching a volume of 1958±480 mm³ by day 21; control-transduced Ba/F3 cells did not form tumors (P≤0.0001; Figure 1C). To confirm these effects in a model of human T-cell lymphoma, we evaluated the effect of STAT3-JAK2 on growth of TLBR-3 cells derived from breast implant-associated anaplastic large cell lymphoma, which require IL2 in vitro. STAT3-JAK2 expression rescued TLBR-3 cells from IL2 withdrawal, expanding 145-fold over 14 days, while control-transduced cells died (Supplementary Figure 1A). TLBR-3 cells transduced with STAT3-JAK2 grew subcutaneously in NSG mice without exogenous IL2, reaching a volume of 3174±703 mm² by week 5; control-transduced TLBR-3 cells did not form tumors (P≤0.0001; Supplementary Figure 1B).

Figure 1.

Function of STAT3-JAK2 fusion. (A) STAT3-JAK2 expression in normal human CD4-positive T cells induced cell growth. (B) STAT3-JAK2 expression in IL3-dependent Ba/F3 cells rescued cells from IL3 withdrawal and promoted cell growth. The inset shows the early death of control-transduced cells after IL3 withdrawal. (C) Ba/F3 cells expressing STAT3-JAK2 but not control-transduced cells formed subcutaneous tumors in NSG mice.

We then evaluated the intracellular effects of STAT3-JAK2 in HEK-293 cells, which lack baseline activation of JAK2, STAT3, and STAT5 (Supplementary Figure 2A). STAT3-JAK2 expression resulted in phosphorylation of the residues within the fusion corresponding to JAK2^Y1007/1008 and STAT3^Y705 and phosphorylation of native STAT3^Y705 and STAT5^Y694. STAT3-JAK2 also resulted in phosphorylation of native STAT3 and STAT5 in TLBR-3 cells, with preferential phosphorylation of STAT5 (Supplementary Figure 2B). Using a JAK2 kinase activity assay, we then demonstrated that purified HA-tagged STAT3-JAK2, but not kinase-dead STAT3-JAK2^K882E, had intrinsic kinase activity (Supplementary Figure 3A). Furthermore, purified STAT3-JAK2 but not STAT3-JAK2^K882E directly phosphorylated recombinant STAT3 and STAT5 (Supplementary Figure 3B). Finally, native STAT3 and STAT5 were co-immunoprecipitated with HA-STAT3-JAK2 (Supplementary Figure 3C). These data suggest STAT3-JAK2 directly interacts with STAT3 and STAT5 and activates these proteins through its JAK2 kinase domain.

Since clinical cases of GI TLPD with STAT3-JAK2 expressed pSTAT5 but not pSTAT3, we examined the relative contributions of native STAT3 and STAT5 to the function of the fusion by silencing STAT3 or STAT5 in STAT3-JAK2-transduced cells. STAT5B siRNA significantly inhibited fusion-induced growth (46.2% inhibition; P<0.0001) while STAT3 siRNA showed only minimal effect that was not statistically significant (Supplementary Figure 4A,B). Luciferase reporter assays demonstrated a 31-fold increase in STAT5 transcriptional activity after STAT3-JAK2 expression (Nano-Glo/firefly ratio, 10.5±0.9 vs. 0.3±0.0 for vector-only control; P=0.004), compared to a 15-fold increase in STAT3 activity (5.0±0.3 vs. 0.3±0.0; P=0.002; Supplementary Figure 4C). RNA sequencing and gene set enrichment analysis demonstrated that STAT3-JAK2 transduction of TLBR-3 cells significantly enriched expression of STAT5 target genes (normalized enrichment score, 1.89; false discovery rate q-value, 0.0024) but not STAT3 target genes (Supplementary Figure 5). Together, these findings suggest that STAT5 activation is a predominant function of the STAT3-JAK2 fusion protein.

Since canonical STAT activation occurs in the cytoplasm, we performed immunofluorescence studies, which showed the fusion protein localized in the cytoplasm (Supplementary Figure 6A). Subcellular Western blots also showed the fusion predominantly in the cytoplasmic fraction (Supplementary Figure 6B). A very weak signal was seen in the nuclear fraction; although a nuclear role for the fusion cannot be entirely excluded, the lack of enriched expression of STAT3 targets suggests the fusion does not exhibit significant STAT3 transcriptional activity. Known STAT3 nuclear localization signal elements are retained in the fusion protein; possibly, unique tertiary structure of the fusion impedes nuclear entry. Consistent with its function in activating STAT proteins, reciprocal co-immunoprecipitation of HA-STAT3-JAK2 and FLAG-STAT3-JAK2 demonstrated dimerization (Supplementary Figure 6B). We developed mutated STAT3^R609Q-JAK2 versions of the fusion or treated STAT3-JAK2-transduced cells with the STAT3 dimerization inhibitor S3I-1757. However, these approaches did not eliminate STAT3-JAK2 homodimer formation (data not shown); therefore, we could not evaluate whether homodimerization was required for the kinase activity of the fusion.

To identify therapeutic opportunities that target STAT3-JAK2, we examined the ability of 5 JAK inhibitors either clinically available or in advanced phase clinical trials to inhibit STAT3-JAK2-induced growth in the previously validated Ba/F3 model [6]. Ruxolitinib and AZD1480 are selective JAK1/2 inhibitors; tofacitinib, gandotinib, and momelotinib have less JAK2 selectivity. All 5 inhibitors demonstrated dose-dependent inhibition of fusion-induced growth, with the most efficacy observed with ruxolitinib (IC₅₀=0.007 μM) and AZD1480 (0.02 μM; Figure 2A, Supplementary Figure 7A). Both ruxolitinib and AZD1480 also inhibited the growth of STAT3-JAK2-transduced TLBR-3 cells (IC₅₀=0.05 and 0.09 μM, respectively; Supplementary Figure 7B). Ruxolitinib inhibited STAT3-JAK2-induced phosphorylation of native STAT3 and STAT5 in Ba/F3 cells (Figure 2B). Reporter assays confirmed diminished STAT5 transcriptional activity (92.2±0.2% reduction in luciferase activity at 0.1 μM) and to a lesser degree STAT3 transcriptional activity (24.9±2.2% reduction; Supplementary Figure 8). We then established subcutaneous tumors of STAT3-JAK2-transduced Ba/F3 cells and treated them with twice-daily ruxolitinib at 90 mg/kg by oral gavage for 9 days. Ruxolitinib significantly inhibited tumor growth, with final tumor volumes of 74.3±48.6 mm³ in ruxolitinib-treated mice and 572.1±186.5 mm³ in control-treated mice (P≤0.0001; Figure 2C) and tumor weights of 0.082±0.030 g and 0.275±0.064 g, respectively (P=0.0001; Supplementary Figure 9). We validated these findings using TLBR-3 tumors to confirm in vivo efficacy of ruxolitinib against human T-cell lymphoma xenografts (final tumor volumes, 207.1±95.0 mm³ in ruxolitinib-treated mice and 1483.1±394.4 mm³ in control-treated mice; P<0.0001; Supplementary Figure 10).

Figure 2.

Targetability of STAT3-JAK2 fusion. (A) Among a panel of 5 JAK inhibitors chosen for their potential clinical availability, all showed dose-dependent growth inhibition of Ba/F3 cells expressing STAT3-JAK2 in the absence of IL3, with the selective JAK1/2 inhibitors ruxolitinib and AZD1480 showing the most effective inhibition. (B) Ruxolitinib blocked phosphorylation of native STAT3^Y705 and STAT5^Y694 in Ba/F3 cells expressing STAT3-JAK2. The paradoxical hyperphosphorylation of JAK2^Y1007/1008 at the highest dose is a known effect of JAK2 inhibitors, possibly attributable to conformation-induced protection of these tyrosine residues from the action of phosphatases [15]. (C) Ruxolitinib (90 mg/kg twice daily by oral gavage) inhibited the growth of subcutaneous tumors derived from Ba/F3 cells expressing STAT3-JAK2 in NSG mice; tumor photographs are shown.

Taken together, these findings demonstrate an oncogenic role for the STAT3-JAK2 fusion recently identified in GI TLPD and identify differential impact of various JAK inhibitors on the activity of the fusion protein. This extends the spectrum of targetable genetic alterations of 9p24 that also includes amplifications of JAK2 and adjacent loci in Hodgkin and other lymphomas [7]. The predominant role of the fusion is to activate STAT5, resulting in increased STAT5 transcriptional activity. Native JAKs require dimerization resulting from interactions between extracellular receptors and their ligands to activate STATs. STAT3-JAK2 also forms homodimers; while this may be attributable to the STAT3 dimerization domain as suggested for other JAK2 fusions [8, 9], the mechanism and functional role of STAT3-JAK2 dimerization require future study. We also observed phosphorylation of native STAT3 and increased STAT3 transcriptional activity, albeit to a lesser degree than STAT5. pSTAT3^Y705 was seen in cell lines expressing STAT3-JAK2 but was absent by immunohistochemistry in most GI TLPDs, including those with STAT3-JAK2 [3, 10]. This difference might be due to limited sensitivity of immunohistochemistry, the amount of STAT3 available for activation, and/or variable expression of relevant phosphatases.

JAK2 inhibition is a clinically available strategy for treating myeloproliferative neoplasms and other diseases [11], and might benefit some patients with GI TLPDs. Ruxolitinib is an oral JAK1/2 inhibitor approved by the Food and Drug Administration for chronic myeloproliferative disorders. Ruxolitinib and related agents inhibit in vitro growth of Mac1 and Mac2a cells with PCM1-JAK2 fusion [12, 13], and efficacy of ruxolitinib in T-cell neoplasms is being evaluated in clinical trials. JAK2 inhibition also might show efficacy in large cell transformation of GI TLPD. In the patient we reported with transformed GI TLPD bearing the STAT3-JAK2 fusion [3], additional copies of STAT3 were present at transformation, a phenomenon also reported by another group [14]. In pathology practice, JAK2 rearrangements can be practicably tested by fluorescence in situ hybridization (FISH) or sequencing strategies. These pre-clinical data provide evidence supporting the clinical investigation of JAK inhibitors in patients with GI TLPD with STAT3-JAK2.