Dual STAT3/STAT5 inhibition as a novel treatment strategy in T-prolymphocytic leukemia

Abstract

T-prolymphocytic leukemia (T-PLL) is a rare, aggressive T-cell malignancy with poor outcomes and an urgent need for new therapeutic approaches. Integrating genomic data and new transcriptomic profiling, we identified recurrent JAK/STAT mutations (predominantly in JAK3 and STAT5B) as hallmarks in a cohort of 335 T-PLL cases. In line, transcriptomic and protein analyses revealed constitutive JAK/STAT activation in virtually all samples. Consequently, we explored the anti-leukemic potential of dual STAT3/STAT5 non-PROTAC degraders in T-PLL, with JPX-1244 as our lead substance. JPX-1244 efficiently and selectively induced cell death in primary T-PLL samples, including those resistant to conventional therapies, by blocking STAT3 and STAT5 phosphorylation and by inducing their degradation. The extent of STAT3/STAT5 degradation directly correlated with cytotoxicity. RNA-sequencing confirmed the treatment-related downregulation of STAT5 target genes. While JAK/STAT mutations did not predict responses to pharmacologic STAT3/STAT5 degradation, elevated expression of TOX, PAK6, and SPINT1 were associated with drug sensitivity. In subsequent combination screenings, cladribine, venetoclax, and azacytidine emerged as most effective combination partners of STAT3/STAT5 degraders, even in low-responding T-PLL samples, all synergistically reducing STAT5 phosphorylation. These findings highlight dual STAT3/STAT5 inhibition, particularly in combination with hypomethylating and BCL2-targeting agents, as a promising interventional approach in T-PLL, warranting further investigation.

Fig. 1. Genomic, transcriptomic, and proteomic signatures of enhanced JAK/STAT signaling in T-PLL.

We expanded a previously published meta-analysis containing sequencing data on 275 T-PLL cases [20] with 60 additional cases, assessed for JAK/STAT mutations by RNA profiling, resulting in a cohort of 335 T-PLL patients with sequencing data on at least one JAK or STAT gene locus. A Frequency of JAK and STAT mutations in 147 T-PLL patients. Only patients with data on the mutational landscape of all JAK/STAT members were included. The inner pie chart displays the proportion of T-PLL patients carrying one (blue), two (claret), three (black), or no (gray) mutation in members of the JAK/STAT signaling pathway. The outer pie chart gives information on the affected JAK/STAT gene. B Basal phosphorylation of STAT5 in 8 T-PLL cases compared to healthy-donor derived CD3⁺ T cells (n = 5), as assessed via immunoblot. The JAK/STAT mutation status is presented for all T-PLL patients. Quantification of STAT5 phosphorylation was performed in comparison to total STAT5 expression, and β-Actin was used as a housekeeper. C Volcano plot of STAT5 target genes, adopted from the HALLMARK IL2-STAT5 targets gene list, in 32 T-PLL cases compared to healthy-donor derived T-cell controls (n = 6). Genes significantly upregulated in T-PLL cells are presented in red, downregulated genes are marked in blue (FDR < 0.05, |log₂FC|≧1.5). D Bar chart displaying the viability of primary T-PLL cells with and without continuous stimulation with interleukins (IL-2 5 ng/ml; IL-6 2 ng/ml; IL-7 5 ng/ml) or in co-culture with the feeder cell lines KuSa and NKtert. Cell death was measured after 48 h via AnnexinV/7AAD flow cytometry (mean with SD is presented, two-tailed paired Student’s t test, **p < 0.01).

Fig. 2. Dual STAT3/STAT5 inhibitors efficiently and selectively induce cell death in primary T-PLL cells ex-vivo.

A Box plot showing inhibitory concentrations 50 (IC50s) of 28 dual Non-PROTAC STAT3/STAT5 degrader in primary T-PLL cases, in comparison to the clinically used cytostatics bendamustine and cytarabine, and the JAK inhibitor ruxolitinib. The number of T-PLL cases tested is given for each compound (¹ n = 5, ² n = 15, ³ n = 12). Mean with whiskers minimum to maximum are presented and substances were ordered from lowest to highest IC50. Cell viability was determined via cell titer glow (CTG) assay after 72 h treatment. JPX-1244 is marked as the selected lead substance, based on stability and safety data given in Supplementary Fig. 2A–C. B Violin plot of JPX-1244 lethal dosages 50 (LD50) values of 36 T-PLL cases, compared to CD3⁺ T cells and PBMC derived from age-matched healthy donors (n = 4). Cell death was determined by AnnexinV/7AAD flow cytometry after 48 h treatment with increasing concentrations of JPX-1244. The median is presented as an interrupted line and the quartiles as a dotted line. A two-tailed unpaired Student’s t test was performed (****p < 0.0001). C Dose-viability curves of primary T-PLL cells in monoculture or cocultures with stromal bone marrow feeder cell lines NKtert and KuSa (n = 4) upon treatment with increasing concentrations (0.1 µM to 10 µM) of JPX-1244 for 48 h (mean with SEM, two-way ANOVA with Geisser-Greenhouse correction and Bonferroni’s multiple comparisons test, *p < 0.05). D Bar chart displaying median LD50s of bendamustine, fludarabine, and JPX-1244 in 21 therapy naïve T-PLL cases compared to 12 relapsed (rel.) T-PLL patients. Median LD50s were calculated based on the median dose-viability curves after 48 h treatment with increasing concentrations of each compound, viability was assessed via AnnexinV/7AAD flow cytometry (see Supplementary Fig. 2E for individual dose-viability curves). Not all patients individually reached an LD50 (not reached with fludarabine: naïve 5/21 patients, relapsed 5/12 patients; not reached with bendamustine: naïve 13/21 patients, relapsed 5/12 patients with higher concentrations tested). The LD50 of bendamustine exhibited a 1.98-fold increase in relapsed patients, of fludarabine a 3.35 increase, and of JPX-1244 a 1.14-fold elevation, compared to treatment-naïve T-PLL cases.

Fig. 3. Underlying its marked anti-leukemic activity, JPX-1244 reduces STAT3/STAT5 phosphorylation paralleled by induced STAT3/STAT5 degradation.

The effects of JPX-1244 treatment on STAT3 and STAT5 phosphorylation, as well as protein stability, were investigated in immunoblots of 11 T-PLL patients, treated with 2.4 µM JPX-1244 for 8 h or 24 h. To enhance the phosphorylation of STAT3 and STAT5 in patients with low basal phosphorylation levels, each condition was doubled and stimulation with IL-6 (2 ng/ml) was performed in one sample of each condition (see Supplementary Fig. 3A for pSTAT3 and pSTAT5 quantification upon IL-6 stimulation). A Quantification of densitometry of pSTAT3, pSTAT5, STAT3, and STAT5 signals, untreated or upon treatment with JPX-1244 for 24 h, assessed by immunoblots and normalized to β-Actin (mean with SD, two-tailed paired Student’s t test, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001). B Bar chart showing the distribution of LD50s upon JPX-1244 treatment in 36 T-PLL patients (0.12 µM to 1.28 µM), as determined in viability assays shown in Fig. 2B. Patients were divided in JPX-1244 High-responder (0.12–0.45 µM, n = 12) and Low-responder (0.75–1.3 µM, n = 11). T-PLL patients who have been assessed for STAT3/STAT5 phosphorylation and protein degradation upon JPX-1244 treatment are labeled in green (n = 11). C Exemplary immunoblots of one JPX-1244 High-responder (P11) and one Low-responder (P26) upon JPX-1244 treatment. Signals of pSTAT5^Tyr694, STAT5, pSTAT3^Tyr705, and STAT3 are shown, densitometric quantification of each was calculated relative to β-Actin. Cleavage of Caspase 3 signals is presented as apoptotic marker, and β-Actin was used as a housekeeper. D Quantification of pSTAT3, pSTAT5, STAT3, and STAT5 signals upon JPX-1244 treatment divided between low- (n = 4) and high-responding cases (n = 7). Densitometric assessments were calculated relative to β-Actin and normalized to the untreated control (mean with SEM, two-tailed unpaired Student’s t test, *p < 0.05, **p < 0.01, ***p < 0.001). Quantification of pSTAT3 signals was only performed upon IL-6 stimulation, as most of the T-PLL patients did not show basal STAT3 phosphorylation.

Fig. 4. Marked transcriptomic alterations of JAK/STAT targets upon JPX-1244 exposure.

A Flow chart visualizing the experimental setup for subsequent RNA-sequencing. In total, 11 T-PLL patients were treated, with 2.4 µM JPX-1244, either for 8 or 24 hours, against DMSO controls. In addition, in each condition, one sample was stimulated with 2 ng/ml IL-6, resulting in 8 samples per patient. B Venn diagram showing the overlap of differentially (|log₂FC|≧3; FDR < 0.05) expressed genes (DEG) upon JPX-1244 treatment (without IL-6 stimulation) between 8 h and 24 h of treatment. While 91 genes were exclusively deregulated after 8 h (1.9%), 1,220 genes showed differential expression at both time points (24.9%), and 3581 genes (73.2%) were exclusively deregulated after 24 h of JPX-1244 treatment. C Volcano plot of DEGs upon JPX-1244 treatment (without IL-6 stimulation) after 8 h (left) and 24 h (right). Genes significantly downregulated upon JPX-1244 treatment are marked in blue, and genes upregulated upon JPX-1244 treatment are colored in red, with a cut-off at |log₂FC|≧3 and an FDR < 0.05. D Gen set enrichment analyses (GSEAs) of HALLMARK gene sets upon JPX-1244 treatment compared to DMSO controls, presented for 8 h (light gray) and 24 h of treatment (dark gray). All significantly altered gene sets (FDR < 0.05) are presented in the graph. Gene sets of particular interest, specifically those related to apoptotic signaling and the JAK/STAT signaling pathway, are highlighted with red boxes. E Heatmap showing the expression of expert knowledge-selected STAT5 target genes, including JPX-1244 treated and DMSO control samples after 24 h without IL-6 stimulation. Unsupervised hierarchical clustering revealed a striking separation between treated and untreated T-PLL cases. Results were confirmed in an unbiased approach, utilizing the KEGG JAK/STAT gene list (Supplementary Fig. 4A). The patient identification codes (as summarized in Supplementary Table 2) are displayed in the legend at the top.

Fig. 5. The sensitivity of T-PLL towards JPX-1244 is independent of JAK/STAT mutational status, but correlates with elevated TOX, PAK6, and SPINT1 expression.

A Volcano plot showing differentially expressed genes between JPX-1244 LR (n = 11) and HR (n = 12) T-PLL cases, as assessed by RNA-sequencing (without treatment or stimulation). Significantly upregulated genes in HR cases are marked in red and significantly downregulated genes in blue (cut-off at |log₂FC|≧3 and FDR < 0.05). B Box plot showing TOX, PAK6, and SPINT1 expression patterns, divided between HR and LR T-PLL cases (mean with minimum to maximum, two-tailed unpaired Student’s t test, *p < 0.05, FDR < 0.05). C Association of white blood cell counts (WBC) at diagnosis (left) and first-line therapy responses (right) with TOX mRNA expression, considering all T-PLL patients with data available on the respective clinical characteristic and TOX mRNA expression. T-PLL patients were divided in two groups based on the mean TOX mRNA expression. Left: WBC at diagnosis comparing low (n = 33, mean = 162.4 × 10⁹cells/l) and high (n = 5, mean = 61.1 × 10⁹cells/l) TOX expressing T-PLL cases (box plot with median as line, mean as rhombus, and whiskers as the interquartile range, p = 0.025, two-tailed unpaired Student’s t test). Right: Response to first-line therapy, divided in patients with low (n = 29) and high (n = 3) TOX expression (fisher’s exact test, p = 0.0121). Treatment responses were assessed as clinical complete remission (clinical CR), clinical complete remission with incomplete count recovery (clinical CRi), CR, CRi, progressive disease (PD), and partial response (PR), as previously defined [1]. D Association of JPX-1244 treatment responses with the JAK/STAT mutation status. A cohort of T-PLL patients (n = 32) was included, for whom cytotoxicity data from JPX-1244 treatment (see Fig. 2B) and corresponding JAK/STAT mutation profiles (see Fig. 1) were available. Box plots of JPX-1244 LD50 values, divided between T-PLL patients (left) with and without any mutation in a JAK/STAT family member (left: without ‘J/S wt’ n = 11, mean = 0.49 µM; with ‘J/S mt’ n = 21, mean = 0.69 µM), between T-PLL patients with and without JAK3 mutation (middle: without ‘JAK3 wt’ n = 25, mean = 0.60 µM; with ‘JAK3 mt’ n = 7, mean = 0.70 µM), and between T-PLL patients with and without STAT5B mutation (right: without ‘STAT5B wt’ n = 23, mean = 0.64 µM; with ‘STAT5B mt’ n = 9, mean = 0.62 µM). Mean with whiskers minimum to maximum is presented. No significant differences were observed in all three comparisons (p > 0.05, two-tailed unpaired Student’s t test). E Dose-viability curves of the TCL1A positive T-PLL-like cell line SUP-T11 upon transgenic JAK/STAT alterations. The strains SUP-T11^empty, SUP-T11 STAT5B^wildtype, SUP-T11 STAT5B^N642H, SUP-T11 JAK3^wildtype, SUP-T11 JAK3^M511I, were treated with increasing concentrations (0.1 µM to 10 µM) of JPX-1244 and viability was assessed via AnnexinV/7AAD flow cytometry after 48 h.

Fig. 6. Combination screening identifies cladribine, venetoclax, and azacytidine as effective partners for JPX-1244 towards promising dual inhibitory concepts.

A Combination screening in 20 T-PLL cases and 3 healthy controls (CD3⁺ T cells and PBMC). JPX-1244 was combined with 19 combination partners, comprising 8 compounds chosen based on (i) T-PLL’s pathogenesis and previous trials [30, 31, 47, 48]: KRT-232, belinostat, ruxolitinib, cladribine, bendamustine, trametinib, dinaciclib, and azacytidine, and (ii) a computational framework, predicting 11 compounds to transform the transcriptome of LR patients to resemble the one of HR patients [49]: elesclomol, danusertib, BAY872243, panobinostat, idarubicin, gemcitabine, sirolimus, pralatrexate, cobimetinib, venetoclax and NMS1286937. T-PLL samples and healthy controls were treated for 48 h, with 6 increasing concentrations of each compound (see Supplementary Table 4 for concentrations), a 1:1 combination was used in pairwise drug combination testing with JPX-1244, followed by predictions with the DECREASE machine learning model to fill the full (7 ×7) drug combination dose-response matrices [57]. Cell viability was assessed via CellTiter-Glo luminescent assay. Efficacy scores and synergy were calculated as previously described [31]. Left: Heatmap showing color-coded selective efficacy of 19 combination therapies, presented per patient (red: selective efficacy score >0, white: selective efficacy score =0, blue: selective efficacy score <0). Selective efficacy scores were calculated with the toxicity volume score and efficacy volume score using the SynToxProfiler [58]. The combinations were tested in 20 T-PLL patient samples and in CD3⁺ T cells derived from 3 age-matched healthy donors. For the selective efficacy in T-PLL samples compared to PBMC of healthy donors, see Supplementary Fig. 6B. Top legend displays the response status of each T-PLL patient towards JPX-1244 in previous single compound screenings (see Fig. 3C, gray: low-responder n = 11, black: high-responder n = 9). Unsupervised hierarchical clustering of both combination partners and T-PLL cases was performed. The respective patient ID is given in the bottom legend. The respective compound combined with JPX-1244 is given on the right. Middle: Selective efficacy values of 19 combination therapies in 20 T-PLL cases displayed as box plot (mean with minimum to maximum). Right: Box plot showing the most synergistic area of 19 combination therapies in 20 T-PLL cases (mean with minimum to maximum). For detailed information on the conditions of the combination screening and calculation of efficacy and synergy, see Supplementary Methods. B Immunoblots of one HR (P20) and one LR (P14) T-PLL case, upon 24 h treatment with JPX-1244 (1.2 µM), cladribine (10 nM), venetoclax (30 nM) and azacytidine (3 µM), or the combination of JPX-1244 with either cladribine, venetoclax or azacytidine, with the same sublethal dosages as in the mono-therapy condition, compared to DMSO control. Left: pP53^Ser15 and P53, Caspase-3 (Casp-3) and cleaved (cl.) Caspase-3, PARP, and cleaved PARP were stained as markers for apoptotic signaling. Densitometric quantification of pP53^Ser15, cleaved Caspase-3 and cleaved PARP was calculated normalized to the housekeeper β-Actin and relative to the DMSO control (rel. to ctrl). Right: pSTAT5^Y694, STAT5, pSTAT3^Y705, and STAT3 signals are shown, quantified relative to the DMSO control, and normalized to the housekeeper β-Actin. As two different gels were used, the housekeeper β-Actin is shown below the respective proteins (Blot 1: pP53^Ser15, P53, Caspase-3, pSTAT5^Y694 and STAT5; Blot 2: PARP, pSTAT3^Y705, and STAT3). C Schematic overview showing functional points of attack of dual STAT3/STAT5 inhibition and the combination partners cladribine, venetoclax, and azacytidine. JPX-1244 directly targets STAT3 and STAT5, affecting both protein phosphorylation and stability. Azacytidine carries demethylating functions by inhibiting the DNA methyltransferase [59], induces DNA damage by incorporation in DNA and RNA [60], and inhibits MCL-1 and BCL-XL expression [61]. Cladribine induces DNA strand breaks as an analog of the nucleoside deoxyadenosine [62], is able to evoke P53 activation [31], and harbors hypomethylating functions as well [63]. Venetoclax selectively binds and antagonizes BCL2, by mimicking the BH3 domain of pro-apoptotic proteins [64]. The effects on these known targets are shown as black, solid arrows, and the interrupted physiological signaling cascades as black, dotted arrows. We propose a synergism of these three compounds with JPX-1244 through indirect effects, exemplarily through hypomethylation of regulators or effects on BCL2 family members, or through direct effects on STAT3/STAT5 phosphorylation (red inhibitors), ultimately leading to decreased expression of STAT target genes. Graph was created with Biorender.