Abstract
Introduction:ETV6::JAK2 is a fusion known to drive acute lymphoblastic leukaemia (ALL) in the presence of other genomic lesions which define the JAK/STAT class of Philadelphia chromosome-like acute lymphoblastic leukaemia (Ph-like ALL). Ph-like ALL comprises approximately 15% of ALL. Patients with mutations or gene fusions signalling through the JAK/STAT pathway have particularly poor prognosis. Emerging treatments targeting JAK2 fusions and mutations are promising, and phase 3 clinical trials are in progress. However, with widespread use of JAK2 inhibitors, it is important to anticipate and manage resistance mechanisms. The JAK2 p.G993A mutation confers resistance in vitro, even to high-dose JAK2 inhibitors such as ruxolitinib. We postulated that direct inhibition of STAT3 and STAT5, downstream from JAK2, may overcome resistance. Methods: Murine-derived IL-3-dependent Ba/F3 cells were transfected with ETV6::JAK2 containing a p.G993A mutation for this study. These cells were confirmed to demonstrate IL-3 independence and ruxolitinib resistance prior to use in experiments. An inhibitor-response assay was conducted using differing concentrations of SH-4-54 and pimozide (STAT3/5 inhibitors) applied to ETV6::JAK2 p.G993A cells and two control cell lines. Result: SH-4-54 and pimozide were effective against ETV6::JAK2 p.G993A cells with median lethal doses (LD50) of 296 nM for SH-4-54 and 455 nM for pimozide. Both drugs demonstrated a lesser effect on empty vector Ba/F3 cells, with an LD50 of 371 nM for SH-4-54 and 596 nM for pimozide. Neither drug demonstrated significant effect on non-JAK/STAT-activated KG-1a myeloid cells at doses near the LD50. Conclusion: SH-4-54 and pimozide both overcame treatment resistance in our in vitro model of JAK/STAT-driven Ph-like ALL with a mutation conferring JAK2 inhibitor resistance. While SH-4-54 demonstrates greater potency than pimozide, pimozide may be a more promising option given its demonstrated safety profile in humans. Direct STAT3 and STAT5 inhibition may be an effective approach for overcoming inevitable JAK2 inhibitor resistance-conferring mutations in patients with the poor prognostic subtype of JAK/STAT class Ph-like ALL.
Plain Language Summary
Philadelphia chromosome-like acute lymphoblastic leukaemia (Ph-like ALL) is a blood cancer that holds a poor prognosis in both children and adults. There are 4 types of Ph-like ALL, and we have focused on a type that is caused by genomic problems in the JAK/STAT pathway. Combination-targeted therapies and broad-spectrum chemotherapy are used to treat JAK/STAT pathway Ph-like ALL and do not narrowly target the disease. Consequently, therapy may confer many significant side effects and outcomes remain poor. We have tested two new treatments against murine cells that are representative of a form of JAK/STAT pathway genomic problems. The genomic problem in our model is resistant to current treatment options. We have shown that the new treatments help destroy cancer cells in our model. One of these treatments is experimental and has not been used in humans before, but the other is a tablet that has been used for other human diseases. The next step will be to establish the safety and efficacy of these therapies in full murine models.
Introduction
The JAK/STAT class of Philadelphia chromosome-like acute lymphoblastic leukaemia (Ph-like ALL) is a subgroup with particularly poor prognosis and high risk of relapse [1, 2]. ETV6::JAK2 is an example of a fusion that can drive disease in JAK/STAT class Ph-like ALL in the presence of other genomic lesions [2]. JAK2r alone is suggested to drive JAK2r B-ALL, and the frequency of ETV6::JAK2 in JAK2r B-ALL is cited as 9.6% [3]. The true frequency of JAK2 fusions in Ph-like ALL may be underestimated in some highly populated geographic areas [2]. There are a number of emerging therapies targeting JAK/STAT class genomic lesions undergoing phase I–III clinical trials [4]. Many inhibitors are already in mainstream clinical use for other diseases and may be successfully repurposed and combined with existing treatments. However, as the number of patients being treated with targeted inhibitors increases, resistance-conferring mutations will inevitably emerge. A historical example is the description of treatment resistance conferred by kinase domain mutations in relation to BCR::ABL1 inhibitors in ABL class Ph-like ALL [5]. Recent literature describing treatment of myeloproliferative neoplasms has already suggested that bispecific approaches are required to overcome treatment resistance in JAK/STAT-activated malignancies [6].
A specific group of JAK/STAT class Ph-like ALL patients with particularly inferior outcomes harbour rearrangements of JAK2, which may be targetable with existing JAK2 inhibitors such as ruxolitinib. Phase III trials of ruxolitinib are showing promising preliminary results in improving prognosis [7, 8]. Ruxolitinib binds at the ATP binding site on JAK2 [9], and mutations at this site can be predicted to lead to therapeutic resistance. A case of such resistance has already been demonstrated in a child [10]. Multiple mutations conferring resistance have been identified in vitro, one of which is the p.G993A mutation. This mutation has been demonstrated to confer resistance to multiple type I and type II JAK inhibitors, including to escalating doses of ruxolitinib [7].
JAK/STAT-activated Ph-like ALL demonstrates proliferation driven by constitutive activation of the JAK/STAT pathway including phosphorylation of STAT3 and STAT5. In JAK inhibitor-resistant models of JAK/STAT class Ph-like ALL, constitutive activation of JAK/STAT signalling continues in the presence of ruxolitinib [7]. In an in vitro murine-derived model of ruxolitinib-resistant ETV6::JAK2 containing a p.G993A mutation, we explored targeting STAT3 and STAT5 directly. We used, individually, two direct STAT3 and STAT5 inhibitors, SH-4-54 and pimozide. This model of overcoming treatment resistance bypassed the need to inhibit JAK fusions at the protein site.
Methods
Murine-derived IL-3-dependent Ba/F3 cells transfected with ETV6::JAK2 containing a p.G993A mutation which have demonstrated resistance to escalating doses of ruxolitinib (SellekChem™ CAT#S1378) were cultured in RPMI 1640 media (Sigma-Aldrich™ CAT#058M4856V) with the addition of foetal calf serum (CellSera™ Cat# AU-FBS-PG), penicillin-streptomycin (Sigma-Aldrich™, CAT#P4333), and l-glutamine (Sigma-Aldrich™ CAT# G7513). The cells demonstrated IL-3 independence. To the culture media,10 µL of 1 mm ruxolitinib (diluted in DMSO from 10 mm stock) was added to achieve a final ruxolitinib concentration of 1 µM. See supplement 1 (for all online suppl. material, see https://doi.org/10.1159/000543428) for detailed cell culture methods.
In two separate experiments, cells were treated with increasing concentrations of the STAT3 and STAT5 inhibitors SH-4-54 (SellekChem™ CAT#S7337) and pimozide (SellekChem™ CAT#S4358) as single agents. IL-3-dependent empty vector Ba/F3 cells and IL-3-independent KG-1a myeloid cells were treated with the same concentrations as controls. KG-1a cells were included because they do not demonstrate JAK/STAT activation and demonstrate less variability in growth compared to Ba/F3 cells [8, 11]. Three biological replicates of cells at the same passage were analysed. DMSO (drug solvent) and untreated cells were also analysed as controls. Inhibitors were applied to cells for 72 h. Percentage of live cells at each concentration was ascertained by flow cytometry on a BD FACSCanto™ Flow Cytometer using Annexin-V and LIVE/DEAD fixable aqua dead cell stain (Invitrogen™ CAT#L34957), and median lethal doses (LD50) were determined. Results were analysed using FlowJo™ software.
Statistical analysis was performed in GraphPad Prism v10 using the program’s Inhibitor vs. Normalised Response methods standard for dose-response experiments. Percentages of viable cells were normalised to the first and last means of each dataset. LD50 values and confidence intervals were generated using non-linear regression. Error bars provided standard error of the mean for 3 biological replicates. For the KG-1a control line in the pimozide graph only, linear regression was used to generate the final graph line given that all error bars were smaller than graph symbols. p values were generated using a two-way ANOVA test selected to perform correction for all comparisons.
Results
SH-4-54 and pimozide both demonstrated efficacy against the ETV6::JAK2 p.G993A-resistant Ba/F3 cell line. The LD50 of SH-4-54 was significantly (p < 0.0001) lower (75 nM lower, relative difference 20%) in the resistant as compared to control empty vector cell lines (296 nM [CI: 289–303 nM] compared to 371 nM [CI: 357–385 nM]). SH-4-54 demonstrated no cytotoxic/cytostatic effect on KG-1a control cells except at its highest concentrations indicating possible toxicity at high doses (Fig. 1a). The LD50 of pimozide was also significantly (p < 0.0001) lower (141 nM lower, relative difference 24%) in the resistant cell lines compared to the control empty vector cell lines (455 nM [CI: 448–462 nM] compared to 596 nM [CI: 579–613 nM]). Pimozide showed no cytotoxic/cytostatic effect on KG-1a control cells at lower concentrations and minimal effect at its highest concentrations (Fig. 1b).
a SH-4-54 induced greater cell death in ETV6::JAK2 p.G993A-transfected cells as compared to controls with minimal off-target toxicity only at high concentrations. b Pimozide induced greater cell death in ETV6::JAK2 p.G993A-transfected cells as compared to controls with miniscule off-target toxicity only at high concentrations.
a SH-4-54 induced greater cell death in ETV6::JAK2 p.G993A-transfected cells as compared to controls with minimal off-target toxicity only at high concentrations. b Pimozide induced greater cell death in ETV6::JAK2 p.G993A-transfected cells as compared to controls with miniscule off-target toxicity only at high concentrations.
Discussion
As phase 3 clinical trial results become available, the use of JAK/STAT pathway inhibitors in patients with JAK/STAT-activated Ph-like ALL is likely to be of increasing clinical relevance. High uptake of new therapy is likely given that JAK/STAT class disease confers particularly poor, often treatment-refractory, clinical outcomes [2]. Based on in vitro observations, clinical case reports, and prior clinical experience with ABL-kinase inhibitors, development of resistant mutations is expected to be a significant route of treatment resistance for a proportion of patients who receive JAK inhibitors. Anticipating these mechanisms, and devising strategies to combat them, is critical to maximising utility of agents available for treatment. This may include either novel combinations of inhibitors acting on multiple targets in the pathway or salvage therapy where relevant. We demonstrated the effectiveness of inhibition of the JAK/STAT pathway downstream from JAK using two STAT3 and STAT5 inhibitors against a JAK2 inhibitor-resistant cell line in vitro. The applicability of our results is limited by our model which has explored a single fusion known to drive JAK2r B-ALL. While it can be postulated that STAT3/5 inhibitors will display similar results in most instances where STAT3/5 are upregulated, this remains to be confirmed. Further studies are required to explore efficacy in the setting of the cooperative and common CRLF2r lesion in JAK/STAT B-ALL. The applicability of our results to B-ALL models demonstrating CRLF2r lesions without a Ph-like ALL gene expression profile is unclear. Ba/F3 may not be an ideal cell line for modelling JAK/STAT class Ph-like ALL, including in the current study [11]. Further studies are required to investigate other fusions known to drive JAK/STAT class Ph-like ALL. These fusions may demonstrate greater unreliability during Ba/F3 modelling [11] due to observed high proliferation rates. The use of human Ph-like ALL cell lines MHH-CALL-4 and MUTZ5 may be successful [12]. Further studies are indicated to demonstrate applicability of our results to the MHH-CALL-4 and MUTZ5 cell lines.
SH-4-54 has demonstrated significantly superior efficacy to pimozide in vitro. Importantly, there were minimal effects on KG-1a myeloid control cells, a line which does not demonstrate JAK/STAT activation, at higher doses. This observation may indicate that SH-4-54 could have off-target effects at high doses.
We chose SH-4-54 and pimozide as the prototypical models for our experiments to demonstrate the in vitro effect, and pre-clinical animal modelling is necessary as the next step. Pimozide, while showing slightly less in vitro efficacy as compared with SH-4-54, is a widely used antipsychotic with a known and manageable side effect profile. Pimozide levels in patients treated for neuropsychiatric disorders are generally lower than 1,000 nM [13], and it is encouraging that our results provided an LD50 below 1,000 nM. Pimozide can cross the blood-brain barrier [14] which may suggest it could be a future candidate therapy for providing important treatment for, or prophylaxis against, acute lymphoblastic leukaemia involving the central nervous system. Pimozide demonstrated minimal off-target effect on KG-1A cells, even at high concentrations, suggesting targeted inhibition of JAK2-activated cells. However, murine models are required to further quantify minimal off-target effects. Our findings may also spur development of specific STAT3 and STAT5 inhibitors for therapeutic use which can be combined with other targeted novel therapies. There could be potential to engineer a single bispecific or trispecific inhibitor targeting STAT3/STAT5/JAK2. Our findings may also be relevant in other cancers that depend on STAT3 and STAT5 signalling where therapies generating meaningful remission are limited and treatment resistance occurs, such as certain myeloproliferative neoplasms.
Conclusion
We have demonstrated efficacy of two STAT3 and STAT5 inhibitors, SH-4-54 and pimozide, in an in vitro murine-derived model of ruxolitinib-resistant ETV6::JAK2 containing a p.G993A mutation. This finding indicates potential for SH-4-54 and/or pimozide to provide important combination treatment options for Ph-like ALL patients with the poor prognostic JAK/STAT class of disease. Further in vitro and in vivo animal studies are required to explore STAT3 and STAT5 inhibition in the context of JAK inhibitor resistance in acute lymphoblastic leukaemia. Future findings may be applicable to other neoplasms driven by JAK/STAT genomic lesions.
Acknowledgments
The authors acknowledge the Myeloma Research Laboratory, University of Adelaide, SA, Australia, for the supply of the Ba/F3 cell line and the Chronic Myeloid Leukaemia Laboratory, SAHMRI, SA, Australia, for supply of the KG-1A cell line.
Statement of Ethics
Ethics committee approval and patient consent were not required given that no patient samples were used in this study. An ethics statement was not required for this study type since no human or animal subjects were used.
Conflict of Interest Statement
The authors declare no conflicts of interest. None of the funding sources were involved in conducting this research.
Funding Sources
NHMRC postgraduate scholarship 2020/GNT2005394 provided specific funding for the inhibitors used in this research. Jane Thompson is supported by a National Health and Medical Research Council (NHMRC) postgraduate scholarship with the applicable Grant No. being 2020/GNT2005394. Deborah L White is supported by LF, Cancer Council SA (CCSA), and NHMRC. Deborah White has received honoraria from Amgen. Randall Grose is supported by McMahon Family Fund. Randall Grose’s facility is funded by Australian Cancer Research Foundation, Zero Childhood Cancer, CCSA, and Detmold Group. David Yeung is supported by CCSA, has received research funding from BMS and Novartis, and has received honoraria from Novartis, BMS, Amgen, Pfizer, and Takeda. The funding sources had no involvement in carrying out this study.
Author Contributions
Jane Frances Thompson developed the concept, designed the research, performed the experiments and statistical analysis, and wrote the paper. Deborah L White assisted with data interpretation and writing the paper. Randall Grose assisted with flow cytometry, data analysis, and the construction of figures. David Yeung assisted with data interpretation and writing the paper.
Additional Information
David Yeung and Deborah L. White should be considered as joint senior authors.
Data Availability Statement
The data that support the findings of this study are not publicly available due to the nature of the study and form of data but are available from the corresponding author upon reasonable request.