Molecular mechanisms by which splice modulator GEX1A inhibits leukaemia development and progression

Abstract

Introduction

Splice modulators have been assessed clinically in treating haematologic malignancies exhibiting splice factor mutations and acute myeloid leukaemia. However, the mechanisms by which such modulators repress leukaemia remain to be elucidated.

Objectives

The primary goal of this assessment was to assess the molecular mechanism by which the natural splice modulator GEX1A kills leukaemic cells in vitro and within in vivo mouse models.

Methods

Using human leukaemic cell lines, we assessed the overall sensitivity these cells have to GEX1A via EC₅₀ analysis. We subsequently analysed its effects using in vivo xenograft mouse models and examined whether cell sensitivities were correlated to genetic characteristics or protein expression levels. We also utilised RT-PCR and RNAseq analyses to determine splice change and RNA expression level differences between sensitive and resistant leukaemic cell lines.

Results

We found that, in vitro, GEX1A induced an MCL-1 isoform shift to pro-apoptotic MCL-1S in all leukaemic cell types, though sensitivity to GEX1A-induced apoptosis was negatively associated with BCL-xL expression. In BCL-2-expressing leukaemic cells, GEX1A induced BCL-2-dependent apoptosis by converting pro-survival BCL-2 into a cell killer. Thus, GEX1A + selective BCL-xL inhibition induced synergism in killing leukaemic cells, while GEX1A + BCL-2 inhibition showed antagonism in BCL-2-expressing leukaemic cells. In addition, GEX1A sensitised FLT3-ITD⁺ leukaemic cells to apoptosis by inducing aberrant splicing and repressing the expression of FLT3-ITD. Consistently, in in vivo xenografts, GEX1A killed the bulk of leukaemic cells via apoptosis when combined with BCL-xL inhibition. Furthermore, GEX1A repressed leukaemia development by targeting leukaemia stem cells through inhibiting FASTK mitochondrial isoform expression across sensitive and non-sensitive leukaemia types.

Conclusion

Our study suggests that GEX1A is a potent anti-leukaemic agent in combination with BCL-xL inhibitors, which targets leukaemic blasts and leukaemia stem cells through distinct mechanisms.

Introduction

Acute myeloid leukaemia (AML) is an aggressive type of hematopoietic malignancy. With the current standard of intensive chemotherapy, 60–75% of AML patients achieve complete remission, and the remaining patients suffer from treatment failures owing to primary drug resistance. The overall 5-year survival rate for AML patients is <27% [1]. For patients who fail treatment, novel therapies are urgently needed. For example, the addition of targeted therapies such as FLT3 kinase inhibitors and IDH inhibitors into standard chemotherapeutic regimens significantly helps to improve the complete remission rates of AML patients with FLT3-ITD and IDH1/2 mutations, respectively [2–9]. For patients who achieved complete remission, disease relapses and acquired drug resistance due to persistent minimal residual disease (MRD) are the major problems [10, 11]. Novel medications are required for such patients to eliminate MRD in order to prevent disease relapse. AML patients harbouring FLT3-ITD mutations account for ~25% of all AML cases which show a significantly increased risk of relapse and shorter overall survival [12–15]. Drug-resistant leukaemia stem cells (LSCs) have been speculated to be the major cause of MRD. Thus, therapies targeted to LSCs are urgently required in order to minimise AML relapses [16, 17].

Many studies have demonstrated that overexpression of anti-apoptotic proteins such as BCL-2, BCL-xL and MCL-1 are commonly detected in both ALL and AML cells and that these are correlated with a poor prognosis among such patients [18, 19]. Fortunately, in many ALL and AML cases, the expression of pro-apoptotic proteins such as BAX, BAK and BIM is also increased, likely due to a feedback mechanism leading to a particular vulnerability of ALL and AML cells targeted by anti-apoptotic protein therapies [20–22]. Many chemical inhibitors of pro-apoptotic proteins have been developed and tested clinically to treat ALL and AML [23–25]. For example, ABT-263 (Navitoclax), a small-molecule inhibitor of BCL-2, BCL-xL and BCL-w, has been shown to be a potent killer of AML cells in preclinical models. However, it is not used clinically because it causes haemorrhage [26, 27]. ABT-199 (Venetoclax), a specific inhibitor of BCL-2, showed promising anti-AML and anti-ALL activities in both preclinical studies and clinical trials when used as monotherapy. The synergistic anti-AML activity was observed for ABT-199 when combined with the hypomethylating agent 5-azacitidine [23, 28, 29]. Mechanistically, ABT-199 + 5-azacitidine not only synergistically kills AML blasts by promoting apoptosis but also eliminates LSCs by effectively targeting mitochondrial metabolism [24]. However, challenges arose as important pro-survival proteins were either poorly targeted by BH3-mimetics, as is the case with MCL-1, or when their use resulted in severe thrombocytopenia, as in the case when targeting BCL-xL [30, 31]. Therefore, the development of novel medications that target such molecules with high specificity and low toxicity is of critical importance. Studies have suggested that many BCL-2 family members could be targeted by splicing modulators which induce splice isoform switching [32].

Somatic mutations of multiple spliceosome factors including SF3B1, U2AF1 and SRSF2 have been reported in many types of hematopoietic malignancies such as MDS, AML, CLL and CMML [33–37]. In addition, many splicing factors such as SRSF1, SRSF3, Sam68 and hnRNPA1 are direct transcriptional targets of MYC and are dysregulated in many types of malignancies [38, 39]. It was reported that malignant cells harbouring spliceosome mutations or displaying MYC overexpression rely upon normal splicing machinery for their survival, making such malignant cells more sensitive to treatments using splicing modulators [40–43]. Currently, the use of splicing modulators has shown much potential in treating numerous malignancies characterised by splicing factor mutations [42–50]. Almost all currently available splicing modulators, such as E7107 (a semisynthetic analogue of pladienolide) and H3B-8800 (an orally available small molecule), function by directly targeting the SF3B1 subunit of the U2 snRNP complex [42, 49, 51–53]. GEX1A (also known as herboxidiene) is one such splicing modulator which was originally isolated from Streptomyces sp. cultures. It induces a large shift in the pattern of exon skipping and intron retention events by inhibiting the SF3B1-PHF5A complex of the U2 snRNP [54]. The SF3B complex plays a key role in regulating pre-mRNA splicing by recognising the branchpoint sequence (BPS) within the intron and facilitating spliceosome assembly and activation [55]. While several of these splicing modulators have shown preclinical efficacy in carcinoma cell lines containing various spliceosomal mutations [44], the in vitro and in vivo effects of these modulators in leukaemic cells lacking spliceosomal mutations were largely unknown. In addition, the specific mechanism underlying splicing modulator-induced cell death remained to be elucidated. In this study, we addressed such questions by comparing the responses of different types of leukaemic cells to GEX1A treatment in both in vitro culture and in vivo xenograft models. We also deciphered the potential molecular mechanism by which GEX1A inhibits FLT3-ITD, leukaemic cells and LSCs.

Results

GEX1A inhibits the growth of some subtypes of leukaemic cells both in vitro and in vivo

To examine the anti-leukaemic effects of GEX1A in vitro, we first conducted a series of colorimetric proliferation assays to determine the EC₅₀ values of 18 different leukaemic cell lines (Fig. 1a). As indicated by these values, cell lines such as Eol-1, KOPN-8, Molm-13, MV4;11, SKNO-1, SEM, Kocl-48 and THP-1 exhibited high sensitivity to GEX1A with EC₅₀ values below 0.1 µM; cell lines NB4, Kasumi-1, RS4;11, REH and U937 showed median sensitivities to GEX1A with EC₅₀ values between 0.1 and 1 µM; and other cell types such as Jurkat, MM6, HL-60, K562 and ML-2 demonstrated little to no GEX1A sensitivity with EC₅₀ values greater than 1.0 µM (Fig. 1a). These cell lines, most being of either AML or ALL origin, encompass a wide range of different mutations (Supplementary Table 1). For example, Eol-1, Molm-13, MV4;11, SKNO-1, THP-1, NB4, Kasumi-1, U937, MM6, HL-60 and ML-2 are AML cell lines, whereas KOPN-8, SEM, Kocl-48, RS4;11, REH and Jurkat are ALL cell lines; K562 is a CML cell line. Interestingly, two highly sensitive cell lines, Molm-13 and MV4;11, contain the FLT3-ITD mutation (Fig. 1a). However, GEX1A sensitivity in other leukaemic cells did not associate with either genetic abnormalities or cellular subtypes. For example, sensitivity is not associated with the state of the genes encoding either TP53 or PTEN, since TP53 and PTEN mutations are randomly distributed within the sensitive and less-sensitive cell lines (Supplementary Table 1).

Fig. 1. GEX1A demonstrated potent anti-leukaemic activity in certain cell types as a single-treatment agent in vitro and in vivo.

a The EC₅₀ values of 18 leukaemic cell types depict a broad range of sensitivities to GEX1A in vitro. b The GEX1A-sensitive leukaemic cell types KOPN-8, Molm-13 and MV4:11 were xenografted into NRGS mice. GEX1A treatment in vivo showed a profoundly-increased survival compared to the untreated controls in all three cell types. **P < 0.01, ***P < 0.001, ****P < 0.0001, log-rank (Mantel–Cox) test.

Based on this in vitro data, we next assessed the in vivo responses of leukaemic cells to GEX1A treatment using xenograft models. Unlike other commonly used splicing modulators such as E7107 and sudemycin D, GEX1A is a naturally occurring compound that has the advantage of being water-soluble; thus, it can safely be dissolved in saline or culture media to ensure diluent toxicity is not a factor in determining a safe dosage to study. Our previous study demonstrated that GEX1A showed good cellular permeability with no evidence of efflux. Pharmacokinetic analyses demonstrated that GEX1A has a favourable plasma protein-binding profile and is quite stable in liver microsomes. Intraperitoneal injection (i.p.) of 2 mg/kg GEX1A into mice displayed a long plasma half-life of 4.53 h. These promising pharmacokinetic properties and the potential ability of GEX1A to cross the blood–brain barrier merited its advancement into animal models [56]. Because the safest and most effective dose of GEX1A in mice was unknown to us, we performed a pilot experiment by treating a group of mice with GEX1A at concentrations ranging between 0.5 and 3 mg/kg i.p. every 1–4 days. We discovered that although almost all mice could tolerate up to 2–3 mg/kg i.p. of GEX1A when injected only once, most mice did not tolerate multiple injections every 1–3 days even at lower dosages. However, we demonstrated that a 1.25 mg/kg i.p. injection of GEX1A once every 4 days was always safe and tolerated by mice (data not shown). By transplanting the three cell lines that were most sensitive to GEX1A treatment (KOPN-8, Molm-13 and MV4;11) into immunodeficient NOD.Rag1^-/-;γc^null-SGM3 (NRG-SGM3) mice by tail-vein injection (5–10 × 10⁵ cells/mouse), we generated corresponding leukaemia xenograft models. On day 3 post-transplantation, the mice were randomly divided into experimental and control groups and were treated with GEX1A and saline, respectively, via i.p. injection. The mice were then observed for the development of leukaemia and assessed for leukaemia-related death (Fig. 1b). As expected, control group mice transplanted with KOPN-8, Molm-13 or MV4;11 leukaemic cells died of terminal leukaemia within 18, 20 and 38 days, respectively. However, the lifespan of the mice in the experimental groups was significantly extended. In particular, the MV4;11 group showed the most profound response to treatment, with some mice in the treated group never developing leukaemia (Fig. 1b). Overall, these results demonstrate that GEX1A has notable anti-leukaemic activity both in vitro and in vivo as a solo treatment for some types of leukaemia. It is worth noting that GEX1A can be toxic in vivo at high doses or when administered too frequently, as we observed apparent liver damage in mice from our pilot experiments (Supplementary Fig. 1). However, such liver cytotoxicity was not observed when mice were treated every 4 days at 1.25 mg/kg i.p. or lower (Supplementary Fig. 1). Hematopoietic stem cells and hematopoietic progenitor cells were unaffected by such treatment (Supplementary Fig. 2).

GEX1A induces alternative splicing of MCL-1 and apoptosis in leukaemic cells both in vitro and in vivo

To determine the cellular mechanism by which GEX1A inhibits leukaemic cell growth, we examined the proliferation, differentiation and apoptosis of leukaemic cells after GEX1A treatment. We first found that GEX1A treatment influences neither the proliferation nor the differentiation of leukaemic cells (data not shown). In addition, GEX1A inhibits leukaemic cell growth primarily by inducing apoptosis as demonstrated by Annexin-V staining and Caspase 3 cleavage (Supplementary Fig. 3). GEX1A inhibits the activity of the SF3B1 complex by directly binding to the SAP155 subunit [54]. Previous studies have indicated that knocking down the SF3B1 spliceosomal complex via shRNA effectively induced the expression of pro-apoptotic isoforms and repressed the pro-survival isoforms of both BCL-xL and MCL-1 [57]. Therefore, we wished to examine whether GEX1A stimulated apoptosis by means of inducing the alternative splicing of MCL-1 and BCL-xL. To do so, we first treated four highly sensitive, three median sensitive and four resistant leukaemic cell lines with various dosages of GEX1A as indicated in Fig. 2a, b. Cells were then collected 6 h post-treatment for RNA extraction, and the spliceosomal isoforms of MCL-1, BCL-x and BIM genes were examined using RT-PCR (Fig. 2a). We found that GEX1A induced a shift in the MCL-1 gene from its pro-survival MCL-1L isoform to its pro-apoptotic MCL-1S isoform at as low a dose as 0.01 μM in all ten leukaemic cell lines regardless of their individual sensitivities to the drug. Additionally, the maximal splicing shift of MCL-1L to MCL-1S was observed with 0.1 μM GEX1A in all ten leukaemic cell types. Furthermore, ≥1.0 mg/kg GEX1A treatment in vivo was sufficient to induce a maximal shift of MCL-1L to MCL-1S in vivo (Supplementary Fig. 4). BIM is a pro-apoptotic gene that, with increasing concentrations of GEX1A, exhibited an isoform shift from its weaker pro-apoptotic isoform, BIM-L, towards its shortest and most potent pro-apoptotic isoform, BIM-ES, in all cells. However, contrary to what was observed in the previously reported SF3B1 shRNA knockdown [57], no cell type displayed a splicing isoform shift in the BCL-xL gene following treatment at any dose. GEX1A inhibits SF3B1 activity by interfering in the interaction between SF3B1 and PHF5A subunits of the U2 snRNP. The different results from SF3B1 knockdown and GEX1A treatment on BCL-xL splicing indicate that GEX1A selectively inhibits the alternative splicing of only some SF3B1 targets, but not all. Detailed mechanisms to explain this observation need to be determined in future studies.

Fig. 2. GEX1A induces a pro-apoptotic switch in splicing of MCL-1, but not BCL-xL.

a, b Ten leukaemic cell types were treated with the indicated dosages of GEX1A for 6 h, followed by collection to extract mRNA (a) and protein lysates (b). The splice isoforms of MCL-1, BIM and BCL-X genes were examined by RT-PCR (a) and western blotting (b). The protein isoforms of other BCL-2 family members and SF3B1 were also examined by western blotting (b). c KOPN-8 and MV4;11 cells were transduced with Mcl-1L. The responses of the transduced cells to GEX1A were examined and compared to vector (VC)-transduced corresponding cells. C = untreated control; * = EC₅₀ of cell type. Cell types listed in descending order of GEX1A sensitivity based on EC₅₀. ** indicate P < 0.01 compared to VC.

Since the splicing isoform shift of the MCL-1 and BIM genes at the RNA level failed to predict the response of leukaemic cells to GEX1A treatment, we then examined the expression and isoform shift of the MCL-1, BIM and BCL-x genes at the protein level by western blotting using the same ten leukaemic cell types. Consistent with the RT-PCR data, we discovered that, while all cell types exhibited a shift from MCL-1L to MCL-1S post-treatment at the protein level, no isoform shift was observed for BCL-x (Fig. 2b). In contrast to the PCR data, we failed to detect any isoform shift of BIM at the protein level, possibly due to the longer natural lifespan of BIM proteins. Nevertheless, the splicing isoform shift of MCL-1 protein also failed to predict the GEX1A response. In addition, overexpression of MCL-1L protein only partially protected MV4;11 and KOPN-8 cells from the cytotoxic effects of GEX1A (Fig. 2c), indicating that, in addition to inducing an MCL-1 isoform shift, GEX1A might also induce apoptosis through an MCL-1-independent mechanism.

GEX1A + BCL-xL selective inhibitor combination blocks leukaemic cell growth in vitro in an additive-to-synergistic manner

Previous studies have suggested that cell survival is normally regulated by the balancing of cellular concentrations of pro-apoptotic and pro-survival proteins. Thus, we aimed to determine whether the levels of survival and apoptotic proteins of the BCL-2 family were altered in leukaemic cells upon GEX1A treatment. We first found that the levels of the pro-apoptotic proteins BAX and BAK, as well as the pro-survival protein BCL-2, were not affected by GEX1A treatment in most cell types (the exception being U937 cells). In addition, GEX1A induced SF3B1 expression in Molm-13 and Jurkat cells, repressed its expression in KOPN-8 cells, and induced an isoform shift in U937 cells. Nevertheless, none of these was correlated with the response of these leukaemic cells to GEX1A treatment (Fig. 2b). After comparing the basal levels of the key pro-apoptotic and pro-survival proteins in untreated leukaemic cells, we then found that only BCL-xL protein levels were negatively associated with the sensitivity of leukaemic cells to GEX1A (Fig. 3a). In addition, overexpression of BCL-xL partially protected MV4;11 and KOPN-8 cells (both sensitive) from GEX1A-induced cytotoxicity (Fig. 3b), while knockdown of BCL-xL sensitised HL-60 and MM6 cells (both resistant) to GEX1A-induced death, suggesting that BCL-xL protein level predicts the response of leukaemic cells to GEX1A-induced apoptosis (Fig. 3c, d).

Fig. 3. Bcl-xL predicts GEX1A resistance in leukaemic cells.

a Basal levels of Bcl-2 family members and NR4A family members were examined in leukaemic cells. b KOPN-8 and MV4;11, two Bcl-xL low-expressing cells, were transduced with the Bcl-xL gene. The sensitivities of the transduced leukaemic cells to GEX1A were examined and compared with their corresponding vector-only transduced leukaemic cells. c, d HL-60 (c) and MM6 (d), two Bcl-xL high-expressing cells, were transduced with Bcl-xL-specific shRNA (shBcl-xL). The sensitivities of the transduced leukaemic cells to GEX1A were examined and compared with their corresponding scrambled shRNA (Scr)-transduced leukaemic cells. *P < 0.05 and **P < 0.01, respectively, compared to VC or Scr controls.

Several inhibitors of pro-survival factors within the BCL-2 family have been developed and tested in the clinic as anti-leukaemic therapy, including the BCL-2-specific inhibitor ABT-199, BCL-2/BCL-xL/BCL-w pan-inhibitor ABT-263, and BCL-xL inhibitor A-1155463 (A-115). To examine whether GEX1A induces any synergistic effects when combined with any of these three inhibitors, we treated numerous leukaemic cell types in vitro with increasing dosage of GEX1A and ABT-199, ABT-263 or A-115, alongside individual treatments using each drug (Fig. 4a–c); cell growth was then assessed utilising a colorimetric assay. The anti-leukaemic effects of these two drug combinations were subsequently evaluated using the synergy calculation programme SynergyFinder [58]. We discovered that the BCL-xL specific inhibitor A-115 displayed significantly greater synergistic activity with GEX1A in almost all cell types. The strongest synergy was observed with GEX1A + A-115 in the BCL-xL high-expressing Jurkat, HL-60, U937, K562, THP-1 and MM6 cells (Fig. 4a). Interestingly, the GEX1A + ABT-199 combination displayed little to no synergistic effects in any of the cell types analysed, with the exception of REH (Fig. 4c). In fact, in Kasumi-1, Jurkat, RS4;11, SEM and Kocl-48 cells, which all express high levels of Bcl-2 protein (Fig. 3a), the GEX1A + ABT-199 combination appeared to generate notable antagonism (Fig. 4c). Furthermore, the Bcl-2 family pan-inhibitor ABT-263 + GEX1A combination resulted in antagonism in several types of cells that express high levels of BCL-B (BCL-L10) and BCL-2-A1 (also known as A1/Bfl-1) but notable synergistic activity in many other types of cells that express low levels of BCL-B and BCL-2-A1 (Figs. 3a and 4b). These data indicate that BCL-xL promotes leukaemic cell survival and is responsible for GEX1A resistance in leukaemic cells. The antagonistic anti-leukaemic effects of GEX1A + ABT-199 in BCL-2-expressing cells suggest that inhibition of BCL-2 might impair the anti-leukaemic activity of GEX1A. Further studies demonstrated that the knockdown of BCL-2 can partially prevent GEX1A-induced apoptosis in Kasumi-1 and THP-1 cells (Fig. 4d, e), while BCL-2 overexpression promoted GEX1A-induced apoptosis in HL-60 cells (Fig. 4f). The antagonistic anti-leukaemic effects of the GEX1A + ABT-263 combination in BCL-B/BCL-2-A1-expressing cells indicate that ABT-263 might also repress BCL-B and BCL-2-A1, since GEX1A also induces a splicing isoform switch of BCL-2-A1. Thus, we predict that ABT-263 also inhibits BCL-B and that the antagonistic effect of combining GEX1A + ABT-263 might be mediated by BCL-B. These data also suggest that BCL-2/BCL-B and BCL-xL play opposite roles in leukaemic cell survival when such cells are treated with GEX1A.

Fig. 4. GEX1A exhibits strong synergistic anti-leukaemic activity with Bcl-x inhibitors, but antagonised the effect of Bcl-2 inhibitor, in vitro.

a–c Leukaemic cells were treated with different dosage combinations of GEX1A + A-1155463 (a), GEX1A + ABT-263 (b) or GEX1A + ABT-199 (c). The synergistic or antagonistic activities of the two chemical combinations indicated were accessed by delta synergy score. d, e THP-1 (d) and Kasumi-1 (e), two BCL-2 high-expressing cells, were transduced with BCL-2-specific shRNA (shBcl-2). The sensitivity of the transduced leukaemic cells to GEX1A was examined and compared to scrambled shRNA (Scr)-transduced corresponding leukaemic cells. f BCL-2 low-expressing HL-60 cells were transduced with the BCL-2 gene. The sensitivity of the transduced leukaemic cells to GEX1A were examined and compared with their corresponding vector-only (VC) transduced leukaemic cells. **P < 0.01, compared to VC or Scr controls.

GEX1A kills leukaemic cells in part by inducing BCL-2-dependent apoptosis

NR4A1 (also known as Nur77 or TR3), NR4A2 (Nurr1) and NR4A3 (NOR-1) are three orphan nuclear receptors which belong to the steroid thyroid receptor family [59, 60]. In response to apoptotic stimuli, NR4A1, NR4A2 and NR4A3 translocate from the nucleus to the cytoplasm where they selectively interact with BCL-2, BCL-B and/or BCL-2-A1 (the latter also known as A1/Bfl-1). NR4A1 and NR4A3 target mitochondria and induce apoptosis by converting the pro-survival molecules BCL-2, BCL-B and/or BCL-2-A1 into pro-apoptotic molecules [61–66]. However, NR4A1 and NR4A3 do not interact with BCL-xL, MCL-1 nor BCL-W [61–66]. Therefore, we speculated that GEX1A may induce apoptosis in leukaemic cells by inducing the interaction of NR4A1 or NR4A3 with BCL-2, BCL-B and/or BCL-2-A1, thus converting the latter three molecules into cellular killers. To test such an idea, we examined the expression of NR4A1 and NR4A3 proteins in leukaemic cells and found that both proteins were expressed in all cell types used in the present studies (Fig. 3a). We found that GEX1A treatment induces nuclear exportation of both NR4A1 and NR4A3 in THP-1 and MV4;11 cells as demonstrated by western blotting of nuclear and cytosolic fractions (Fig. 5a) and was verified by immunofluorescent staining (Fig. 5b). Selinexor is a nuclear export inhibitor that prevents the nuclear exportation of NR4A1 and NR4A3 (Fig. 5a, b). We found that in BCL-2-expressing AML cells, selinexor also antagonizes the anti-leukaemic effect of GEX1A (Fig. 5c), suggesting that GEX1A kills leukaemic cells at least in part by inducing the exportation of NR4A1 and NR4A3 from the nucleus.

Fig. 5. GEX1A induces XPO-1 mediated nuclear exportation of NR4A proteins which can be inhibited by XPO-1 inhibitor selinexor.

a, b THP-1 and MV4;11 AML cells were treated with either GEX1A only or GEX1A + selinexor. Nuclear and cytosolic localisation of NR4A1 and NR4A3 were assessed by western blotting of the nuclear and cytoplasm fractions (a) and immunofluorescent staining (b). c THP-1 and MV4;11 cells were treated with varying dosage combinations of GEX1A + selinexor. Combination activity was assessed via delta synergy score, and graphs were generated via SynergyFinder 2.0 software.

GEX1A preferentially inhibits LSCs

After discovering notable synergy between GEX1A and A-115, we next focused on translating these data into an in vivo model. HL-60 and K562 cells, as well as murine MLL-AF9 AML cells, were resistant to treatment with GEX1A alone in vitro, though strong synergistic activity was exhibited within these cell types when combined with A-115 in vitro (Fig. 4a). Thus, we transplanted HL-60 and K562 cells into individual NRSG mice to generate xenograft models with immunodeficiency; we also transplanted murine MLL-AF9 AML cells into C57B6/J mice to generate AML mice with intact immunity. Three days post-transplantation, mice from each model were randomly divided into four groups and treated with vehicle, 1.0 mg/kg GEX1A (i.p. once every 4 days), 5.0 mg/kg A-115 (i.p. once every other day), or GEX1A + A-115 (Fig. 6a). Consistent with the in vitro data, the combination of GEX1A + A-115 displayed additive-to-synergistic activity in vivo as demonstrated by an increased survival in the GEX1A + A-115 groups compared to the control and individual drug-treated groups (Fig. 6a). Surprisingly, in all three models, these mice also demonstrated a significantly increased survival when treated with GEX1A alone compared to vehicle control groups. However, there was no significant survival benefit observed among mice treated with A-115 only. Since we treated the mice at an early time point following transplantation before obvious leukaemia developed, the difference observed between these leukaemic cells given the GEX1A-only treatment in vitro and in vivo suggests that GEX1A may repress leukaemia development by selectively targeting LSCs. The synergistic effects of GEX1A + A-115 indicated that this combination treatment might also collaboratively kill leukaemic blasts, which make up the bulk of ALL and AML cell populations.

Fig. 6. GEX1A + Bcl-xL inhibitor combination additively represses leukaemia development in vivo.

a, b Human HL-60 and K562 leukaemic cells, as well as murine MLL-AF9 leukaemic cells, were transplanted into recipient mice to generate leukaemia animal models. The recipient mice were treated with vehicle (control), GEX1A, A-115 or GEX1A + A-115 starting 3 days (a) or 15 days (b) post-transplantation. Mice were subsequently monitored for leukaemia development. The survival of mice is depicted by Kaplan–Meier analysis. c Leukaemic cells were collected from (b) and transplanted into secondary recipient mice. Mice were subsequently monitored for leukaemia development. Overall survival of mice is depicted using Kaplan–Meier analysis. d Equal numbers of MLL-AF9 leukaemic cells were treated with indicted agents for 24 h in vitro and were transplanted into recipient mice. Mice were subsequently monitored for leukaemia development. Overall survival of mice is depicted via Kaplan–Meier analysis.* and ** indicate P < 0.05 and P < 0.01, respectively, compared to control or A-115 treatment.

To test this idea, we treated HL-60, K562 or murine MLL-AF9 cells with 300 nM GEX1A in vitro for 24 hours. The percentage of LSCs in these populations was then examined by CM-H2DCFDA (DCF) staining of the ROS-low population in human AML cells or by Gr1 and CD11b staining of the Gr1^lowCD11b^low population in murine AML cells. Treatment with the topoisomerase inhibitor Daunorubicin was studied in parallel as a control. We found that GEX1A treatment preferentially kills LSCs compared to Daunorubicin as demonstrated by a reduction of the percentage of DCF^low and Gr1^lowCD11b^low populations (Supplementary Fig. 5). To examine this idea further, we transplanted another batch of mice with HL-60, K562 or murine MLL-AF9 cells. The mice were randomly divided into four groups on day 15 post-transplantation (after obvious leukaemia had developed) and treated with vehicle, GEX1A, A-115 or GEX1A + A-115, respectively. We found that only the GEX1A + A-115 combination could extend the lifetime of the mice (Fig. 6b). Interestingly, when equal numbers of leukaemic cells from all groups were transplanted into secondary recipients, the lifespans of the mice that received leukaemic cells from either the GEX1A or GEX1A + A-115-treated groups were significantly prolonged when compared to the mice that received leukaemic cells from either the vehicle or A-115-treated groups (Fig. 6c). Furthermore, we treated an equal number of leukaemic cells in vitro with vehicle, GEX1A, A-115 or GEX1A + A-115 (respectively) for 24 hours, followed by transplantation into respective recipient groups of mice to observe the development of leukaemia (Fig. 6d). We found that the lifespans of mice that received either GEX1A or GEX1A + A-115-treated leukaemic cells were significantly prolonged compared to mice which received vehicle- or A-115-treated leukaemic cells. All these data suggest that GEX1A treatment selectively inhibits LSCs in both immunodeficient and immune-intact mice.

GEX1A represses LSCs by inhibiting the expression of mitochondrial FASTK

To discover the mechanisms by which leukaemic blasts are killed and LSCs are inhibited by GEX1A, we compared gene expression profiles and alternative mRNA splicing patterns among two types of GEX1A-sensitive cells, Molm-13 and MV4;11, and between HL-60 and K562, two types of GEX1A-resistant cells, before and after GEX1A treatment. Differential expression analysis revealed that 5391, 5788, 5956 and 5605 genes were upregulated by GEX1A in Molm-13, MV4;11, HL-60 and K562 cells, respectively, while 5270, 5145, 5645 and 5014 genes were downregulated in Molm-13, MV4;11, HL-60 and K562 cells, respectively (Supplementary Fig. 6). Although differential gene expression patterns were detected between sensitive cells and resistant cells, a large number of genes were co-expressed in all four cells (Supplementary Fig. 7). Gene Ontology and KEGG enrichment analyses demonstrated that the most differentially expressed genes are related to mRNA catabolic processes, ribonucleoprotein complex biogenesis, RNA splicing, and translational initiation in all four cell types, and there were no significant differences between sensitive cells and resistant cells (Supplementary Fig. 8a, b). GEX1A also induced a significant upregulation of FAS, DR-5, p21 and TP53INP1 genes in sensitive cells compared to resistant cells; this may be correlated to GEX1A-induced apoptosis (Supplementary Fig. 8c). Alternative splicing analysis demonstrated that GEX1A treatment-induced global splicing changes in a significant number of transcripts in all four types of cells. These splicing changes include exon skipping, intron retention, mutually-exclusive exons, alternative 5’-splice sites and alternative 3’-splice sites. Many intron retention events are associated with RNA downregulation, suggesting nonsense-mediated mRNA decay (Supplementary Fig. 9). For example, GEX1A-induced splicing changes in the Fas-activated serine/threonine kinase (FASTK) gene were present in both sensitive and resistant cells (Fig. 7a and Supplementary Fig. 10a) and were associated with the downregulation of FASTK mRNA and protein levels (Fig. 7b, c and Supplementary Figs. 10b and 11a).

Fig. 7. GEX1A inhibits LSCs by inducing intron retention of the FASTK gene and repressing m-FASTK expression.

a–c GEX1A induced FASTK intron retention (a) and repressed mRNA expression (b) as demonstrated by RNAseq. GEX1A repressed FASTK protein expression as shown by western blotting (c). d GEX1A repressed MTND6 expression as demonstrated by qRT-PCR. e, f FASTK knockdown confirmed via PCR (e) repressed leukaemia development as shown by xenograft assay (f). Scr: Scrambled; ShF1-3: 3 shRNAs specifics for FASTK gene. g, h Overexpression of m-FASTK but not c-FASTK as shown by Western blot (g) promotes leukaemia development and GEX1A resistance as demonstrated by xenograft models (h). ** indicates P < 0.01, compared to Scr groups in (f) and compared to VC and c-FASTK groups in (h).

FASTK has been described as a kinase that is activated by Fas signalling. Two isoforms exist: mitochondrial FASTK (m-FASTK, ∼50 kDa) and a nuclear/cytosolic-FASTK (c-FASTK, ∼60 kDa). The difference is attributable to an alternative translation initiation site (Fig. 7c). It was also proposed that m-FASTK serves as a sensor of mitochondrial stress and consequentially regulates mitochondrial RNA biology and function [67]. m-FASTK directly interacts with BCL-xL in the outer mitochondrial membrane to prevent Fas- or UV-induced apoptosis [68, 69] or to regulate the specific expression of ND6 mRNA (MTND6) in mitochondrial RNA granules. MTND6 encodes subunit 6 of NADH complex I of mitochondrial respiration, which is important for complex I activity. Such activity is critical for the self-renewal of LSCs [24, 70]. We found that GEX1A treatment repressed the expression of MTND6 in all leukaemic cell lines tested (Fig. 7d), as well as murine MLL-AF9 AML cells (Supplementary Fig. 11b). Compared to bulk leukaemic cells, LSCs have relatively lower MitoTracker Green staining (MTG). Such lower MTG in LSCs is not due to lower mitochondrial mass but rather is explainable by the existence of xenobiotic efflux pumps on mitochondrial membranes since it can be restored by calcium channel blocker Verapamil (Supplementary Fig. 11c) [71]. In addition, as distinct from bulk leukaemic cells which preferentially use glycolysis for their energy production, LSCs primarily use oxidative phosphorylation for their energy generation. Thus, LSCs have a higher oxygen consumption rate (OCR) than bulk leukaemic cells [24, 72]. We found that GEX1A treatment reduced xenobiotic efflux pump activity as demonstrated by increased MTG (Supplementary Fig. 11c) and inhibited oxidative phosphorylation as shown by decreased OCR (Supplementary Fig. 11d), suggesting impaired mitochondrial activity. However, the topoisomerase inhibitor daunorubicin did none of these (Supplementary Fig. 11c, d). Thus, we predict that GEX1A inhibits LSCs by impairing mitochondrial activity through its repression of m-FASTK and MTND6 expression.

To test this idea, we knocked down FASTK in murine MLL-AF9 AML cells as well as human MV4;11 and KOPN-8 leukaemic cells via shRNA and further assessed its role in LSCs by flow cytometry, in vitro colony assay and transplantation experiments. We found that knockdown of FASTK significantly reduced LSCs% and colony-forming ability of murine MLL-AF9 AML cells (Supplementary Fig. 11e–g). FASTK knockdown also significantly reduced the leukemogenic capacity of human leukaemic cells, suggesting that FASTK knockdown is associated with reduced LSC activity (Fig. 7e, f). In addition, we overexpressed m-FASTK and c-FASTK, respectively, in both MV4;11 and KOPN-8. The effect of m-FASTK and c-FASTK on GEX1A response in leukaemic cells was then assessed by in vivo transplantation. We found that overexpression of m-FASTK (but not c-FASTK) promoted GEX1A resistance in both MV4;11 and KOPN-8 leukaemic cells, suggesting that m-FASTK protects LSCs from the effects of GEX1A treatment (Fig. 7g, h).

GEX1A represses FLT3-ITD-STAT5 signalling by inducing aberrant splicing of the FLT3-ITD gene

Two FLT3-ITD⁺ AML cell lines, Molm-13 and MV4;11, were shown to be highly sensitive to GEX1A treatment. We found that GEX1A treatment induces the retention of introns 12 and 13 within FLT3 RNAs in both Molm-13 and MV4;11 cells as demonstrated by both RNAseq (Fig. 7a) and qRT-PCR (Fig. 8b). As a consequence, FLT3-ITD expression (including the wild-type FLT3 allele) is significantly repressed in these cell lines by GEX1A treatment as demonstrated by RNAseq (Fig. 8c), qRT-PCR (Fig. 8d), and flow cytometry (Fig. 8e). Consequently, GEX1A also repressed Stat5α/ß activity in MV4;11 cells (Fig. 7f) as demonstrated by a reduction of p-Stat5α/ß levels, though we were unable to detect p-Stat5α/ß in Molm-13 cells. We also found that GEX1A treatment-induced intron retention in FLT3-ITD RNA transcripts (Fig. 8g), repressed FLT3-ITD protein expression (Fig. 8h), and decreased p-Stat5 levels (Fig. 8i) in primary FLT3-ITD⁺ AML samples; however, daunorubicin treatment did none of the above. This suggests that FLT3-ITD⁺ might predict the sensitivity of AML cells to GEX1A treatment.

Fig. 8. GEX1A represses FLT3-ITD signalling by inducing intron retention in the FLT3-ITD gene.

a–e The FLT3-ITD⁺ AML cell lines Molm-13 and MV4;11 were treated with 100 nM GEX1A. Cells were collected 6 hours (for a–d) and 12 h (for e) post-treatment. Intron 12 retention (iR) was examined by RNAseq + IgV software (a) and RT-PCR (b). FLT3-ITD expression was examined by RNAseq (c), qRT-PCR (d) and flow cytometry (e) assays. f MV4;11 cells were treated with indicated concentrations of GEX1A. Levels of Stat5α/ß and p-Stat5α/ß proteins were examined by western blotting. g–i FLT3-ITD⁺ AML blasts from two patients were treated with indicated concentrations of GEX1A. Cells were collected 12 h post-treatment. Intron 12 retention (iR) was examined by RT-PCR (g). FLT3-ITD expression (h) and p-Stat5a levels (i) were examined by flow cytometric assay. Vehicle and daunorubicin (d, 500 nM) treatments were studied in parallel as controls. ** and *** indicate P < 0.01 and P«0.001, respectively, compared to the untreated group. Data in (e, h, and i) are representative of three biological replicates.

Discussion

The spliceosome has been identified as a potential and promising anti-cancer target for many types of cancer, especially for those patients whose tumours involve spliceosome mutations or MYC overexpression. Several splicing modulators have been developed and tested in preclinical models, and two of them have been tested in Phase I/II clinical trials as cancer therapies [50, 73]. Recent studies have suggested a promising but limited efficacy when splice modulators were used as a monotherapy for cancer treatment. Thus, a better understanding of the molecular mechanism by which splicing modulators inhibit the growth of malignant cells would be of significant help in the development of combination therapies using currently available splice modulators together with standard chemotherapeutic agents or other targeted therapies, as well as novel and more effective anti-cancer splicing modulators. Most currently available modulators are derivatives of FR901464, pladienolides and GEX1A, which all happen to be natural products [74]. Such molecules specifically target the SF3B complex and attenuate the fidelity and efficiency of SF3B complex-mediated alternative splicing of pre-RNAs, which cause intron retention and exon skipping. However, the transcript targets of these modulators are not always the same due to the difference in their binding sites and strength of binding to the SF3B complex [45, 75]. In this study, we used GEX1A in both in vitro culturing and in vivo xenograft experiments to study the how SF3B complex modulators kill leukaemic cells and LSCs. We found that GEX1A kills leukaemic cells by inducing both an MCL-1 isoform switch and BCL-2-NR4As-mediated apoptosis. BCL-2 and BCL-xL play opposing roles in GEX1A-induced apoptosis in leukaemic cells. We also demonstrated that GEX1A kills LSCs by inhibiting m-FASTK-mediated mitochondrial metabolism. Most importantly, we determined that leukaemic cells harbouring a FLT3-ITD mutation are highly sensitive to GEX1A treatment because the latter also represses FLT3-ITD expression. Our study suggests that GEX1A + a BCL-xL-specific inhibitor is a more efficacious combination to treat leukaemia, specifically FLT3-ITD⁺ AML.

AML patients harbouring FLT3-ITD mutations typically have a poor prognosis. Although several specific inhibitors of FLT3-ITD have been evaluated in large clinical trials and showed promising results, such inhibitors improve disease-free survival largely by eliminating AML cells but fail to improve overall survival in the setting of additional acquired mutations and/or drug resistance in a sub-clone of leukaemic cells [4, 12]. Our study suggests that GEX1A and other splice modulators represent a new type of treatment for FLT3-ITD⁺ AML patients, even those whose FLT3-ITD diseases are typically resistant to other inhibitors. Our future studies will determine whether a splice modulator + FLT3-inhibitor combinations have synergistic anti-leukaemic effects in FLT3-ITD⁺ AML.

Therapeutically targeting BCL-2 pro-survival proteins shows great promise for the treatment of cancer. Many small-molecule inhibitors of such proteins have been developed and tested in clinical trials. Thus far, only ABT-199 has been successfully used clinically for the treatment of all types of leukaemia. However, ABT-199 monotherapy is only moderately effective due to compensation by other pro-survival BCL-2 family members, including MCL-1 [76]. Thus, the success of ABT-199 in the clinic is observed only when it is used in combination with standard chemotherapies or other targeted therapies. There are splice isoforms for many members of the BCL-2 family. Importantly, the splice isoforms of most BCL-2 proteins have opposing activities in the regulation of mitochondrial-related apoptosis. For example, the long splice isoforms of MCL-1, BCL-x, BCL-W and BCL2-A1 are pro-survival factors, whereas their short splice isoforms are pro-apoptotic. Thus, inducing splice isoform switching of BCL-2 family members has been proposed as a potential strategy to treat cancer [32]. Consistent with what was already known for other SF3B splicing modulators, we found that GEX1A induces splice isoform switching of MCL-1 and BCL-2-A1 without affecting the splicing of BCL-xL [48, 77]. In addition, we demonstrated that GEX1A also induces BCL-2-dependent apoptosis, which explains the antagonistic effects observed when GEX1A and a BCL-2-specific inhibitor are used in combination.

Inhibitors of BCL-xL are limited clinically by their adverse off-target effects, especially platelet depletion [78]. This hindrance can be resolved by developing cancer cell-specific BCL-xL inhibitors, such as selective BCL-xL PROTEC [79]. It has been claimed that MCL-1 is an oncoprotein that is difficult to target through chemotherapy [80]. The anti-leukaemic effects of MCL-1 inhibitors have been tested in several preclinical models and are still in early clinical evaluation [81–85]. We predict that GEX1A + a BCL-xL inhibitor will emerge as a more effective combinatorial treatment in inhibiting leukaemia compared to an MCL-1 inhibitor + a BCL-xL inhibitor, because GEX1A also represses the oncogene FLT3-ITD and induces BCL-2-dependent apoptosis in leukaemic cells while also inhibiting m-FASTK-mediated survival in LSCs. However, the potential dosage window for the use of GEX1A is relatively small but could be expanded with typical medicinal chemistry [86]. The major side effect of GEX1A in mice is hepatic toxicity (Supplementary Fig. 1), which could be related to tissue specificity of the parent structure. Future studies need to determine the mechanism by which GEX1A induces such damage to the liver in order to develop newer GEX1A derivatives that reduce liver toxicity without affecting their anti-leukaemic capacity.

NR4A1 and NR4A3 are two well-documented repressors of AML [59, 87]. BCL-2-dependent apoptosis is triggered by NR1A1 and NR4A3. These two are highly homologous nuclear receptor transcription factors which regulate targeted gene expression in the nucleus. However, upon apoptotic signal stimulation, NR1A1 and NR4A3 translocate to the cytoplasm where they promote mitochondrial-mediated apoptosis through their interaction with BCL-2, BCL-B and/or BCL2-A1 by converting these pro-survival factors to pro-apoptotic molecules [61–64, 88–90]. We found that both NR1A1 and NR4A3 can be detected in all types of leukaemic cells studied. GEX1A treatment did not increase the expression of any of the NR4As (data not shown), suggesting that the basal levels of NR4As seem sufficient to convert BCL-2 to a cytocidal form. NR4A-BCL-2-mediated apoptosis is regulated by several signalling pathways. For example, AKT signalling represses the nuclear export and mitochondrial translocation of NR4A1 by phosphorylating it on Ser351, while the MEK-ERK-RSK cascade promotes such export and translocation by phosphorylating NR4A1 on serine 354 [91, 92]. p38α MAPK signalling promotes the interaction of BCL-2 and NR4A1 by phosphorylating BCL-2 on its Ser87, Ser70 and Thr56 residues [60, 93–97]. Importantly, NR4A1/3-mediated apoptosis is independent of p53 [92]. Future studies will need to determine the signalling pathway GEX1A influences that results in the induction of NR4A-BCL-2-mediated apoptosis in leukaemic cells. It will be also important to test whether activators of NR4A-BCL-2 apoptotic signalling can be used to treat leukaemia using a combination that includes GEX1A [66, 98].

Both FASTK molecules have been identified as RNA-binding proteins, but c-FASTK and m-FASTK have distinct biological functions. For example, c-FASTK regulates pre-RNA processing and splicing as well as its degradation and translation, in concert with other RNA-binding proteins such as eukaryotic initiation factor 4E (eIF4E), T-cell intracellular antigen-1 (TIA-1) and its related protein TIAR/TIAL1 [99–103]. Likewise, it induces apoptosis by promoting the expression of c-IAP-1 and XIAP, which are inhibitors of apoptosis. m-FASTK has also been reported to be required for the biogenesis of the mitochondria; it interacts with mitochondrial ND6 mRNA in RNA granules, protecting MTND6 mRNA from degradation by the degradosome. ND6 mRNA encodes an essential subunit of mitochondrial respiratory complex I (CI, NADH:ubiquinone oxidoreductase), regulating mitochondrial CI activity [67, 104]. Our study suggests that GEX1A inhibits LSCs by repressing m-FASTK. It was reported that m-FASTK also interacts with BCL-xL on the outer mitochondrial membrane [68]. We predict that the LSC inhibitory effect of GEX1A is independent of BCL-xL, since the BCL-xL-specific inhibitor alone failed to inhibit LSCs. Future studies must focus on elucidating the molecular mechanisms by which m-FASTK-MTND6 regulate the survival and self-renewal of LSCs.

Materials and methods

Production, isolation, and purification of GEX1A 1 from Streptomyces chromofuscus

Streptomyces chromofuscus (ATCC 49982) was acquired from the American Tissue Culture Collection (Manassas, VA) and maintained in ISP2 (0.4% yeast extract, 1.0% malt extract, 0.4% glucose) liquid medium. Production of GEX1A from S. chromofuscus was conducted on our developed production medium (0.4% yeast extract, 0.8% soy flour, 1.0% glucose, 0.1% CaCO₃, 2.0% agar). S. chromofuscus seed cultures were grown in ISP2 liquid medium (50 mL) for 48 h at 28 °C, 325 rpm. Agar production medium plates were prepared, and the seed culture was spread onto the plates (∼0.5 mL/plate), which were cultured at room temperature in the dark for 14 days. After incubation, the plates were homogenised into 1 L of ethyl acetate and sonicated for 48 h. The crude organic extract was filtered via vacuum filtration through the Whatman filter paper and concentrated under reduced pressure. All crude organic material obtained through these methods was purified via flash column chromatography (10–30% Acetone−Hexanes) to yield a pale-yellow oil, which was then recrystallised in hot 9:1 Et2O:Hexanes to obtain the desired compound as a powdery white solid. Isolated yields of GEX1A 1 were determined to be ∼50–70 mg/L.

Cell culture, treatments, EC₅₀ assessment and synergy analysis

Each cell type specified was grown in culture to adequate confluence and counted via Trypan blue exclusion assay using a BioRad Cell Counter (Model #TC10). Viable cells were extracted, centrifuged at 1600 rpm for 6 min, and supernatants were aspirated. Cells were re-suspended in fresh growth medium (respective to each cell type) and seeded at 5 × 10⁴ cells/well, 100 µL/well in replicates of 3–4 unless otherwise specified in figure legends. Cells remained in culture for 12 h prior to treatment(s); each well was seeded with 1 µL of the respective drug and/or vehicle (if diluted in DMSO) to minimise non-specific cell death. Viabilities were then assessed via Cayman WST-8 Proliferation Assay in accordance with the kit protocol (Product #10010199). Absorbances were read at least twice on an Omega Plate Reader at 450 nm wavelength. EC₅₀ curves were generated via GraphPad Prism 8 software, and synergy scores were calculated using SynergyFinder software: https://synergyfinder.fimm.fi/.

Cell lysate collection and western blot analysis

Cells were collected from culture after designated treatments as described above and were re-suspended into 1× SDS lysis buffer. Suspensions were briefly sonicated for 5 pulses and subsequently centrifuged at 14,000 rpm for 15 min at 4 °C, followed by supernatant extraction. Lysates were either used immediately or stored at −80 °C until further use. Between 20–30 µg of protein per sample was loaded onto freshly made 1.5-mm gels (8–14% acrylamide) alongside 2–3 µL of protein molecular weight standard, and electrophoresis was performed at 60–65 V for 10–15 min, then at 120–140 V for 80–90 min. Proteins were immediately transferred onto a nitrocellulose membrane using the Thermo Fisher TurboTransfer System in accordance with equipment use protocols, followed by a brief ddH₂O wash of each membrane. Membranes were blocked using 3% BSA in 1 × TBST for at least 60 min and incubated overnight at 4 °C on a shaker platform with the primary antibody(s) specified (1:500-1:1000 dilution in fresh blocking agent). Membranes were then washed with 1 × TBST 3–5 times for 5–10 min and then incubated with secondary antibody for 30–60 min (1:1250 dilution in fresh blocking agent). Membranes were washed again with TBST 3–5 times for 10–20 min and subsequently imaged using BioRad Imager at various exposure times. If stored, membranes were washed in TBST for 10 min and then briefly with 1 × TBS; each was wrapped individually with plastic wrap and stored at −20 °C For re-blotting purposes, membranes were thawed in TBST for 10 min and stripped (if applicable) with 1 × Thermo Fisher Stripping Buffer for 20–30 min, followed by 2–3 TBST washes for 10 min and re-blocking for at least 30 min. New primary antibody was used at the same dilution conditions specified above for at least 60 min. All subsequent steps for re-blotting were identical to those described above.

RNA extraction and reverse transcription–polymerase chain reaction (RT-PCR)

RNA was extracted from indicted cells using Tri Reagent following the instructions provided by the vendor (Invitrogen). RNA extracts were quantified via NanoDrop 2000, and 2 µg of RNA from each sample was treated with DNAse I (1U/µL) in 1 × DNAse I buffer at 37 °C for 30 min, followed by 2.5 mM EDTA treatment at 65 °C for 10 min. Reverse transcription (RT) was then conducted to generate cDNAs using SuperScript^® III Reverse Transcriptase (Life Technologies) in accordance with the supplied protocol. The expression of the genes of interest in each cDNA sample was examined by PCR assay using gene-specific primers (see primer list in table 3) along with 0.67 µM MgCl₂ and Taq polymerase in 1×PCR buffer (Step 1: 94 °C for 5 min; Step 2, 10 cycles: 94 °C for 30 s, 60 °C for 45 s, 72 °C for 30 s; Step 3, 30 cycles: 94 °C for 30 s, 55 °C for 45 s, 72 °C for 30 s; Step 4: 72 °C for 5 min; 4 °C hold).

Animals, leukaemic cell transplantation and treatment

Both NOD.Cg-Rag1^tm1MomIl2rg^tm1WjlTg(CMV-IL3,CSF2,KITLG)1Eav/J,-NRGS mice (Stock No: 024099) and C57Bl/6J mice (Stock No: 000664) were purchased from the Jackson Laboratory and housed under a 12-h light/dark cycle in micro-isolator cages contained within a laminar flow ventilation system. All procedures were conducted in accordance with the National Institutes of Health guidelines for the care and use of laboratory animals for research purposes and were approved by Loyola University Chicago’s Institutional Animal Care and Use Committee (IACUC) (AU 518815). Both male and female mice were used in these studies and were randomly divided into different experimental groups. At the time of transplantation, all mice were 7–12 weeks old. To generate xenograft leukaemic models, human leukaemic cell lines KOPN-8, Molm-13, MV4;11, K562, or HL-60 were transplanted into NRGS mice via intravenous (i.v.) tail-vein injection individually, 1–2 × 10⁶ cells per mouse. To generate murine MLL-AF9 (MA9) leukaemic models, MA9 leukaemic cells collected from our previous studies [105] were transplanted into C57BL/6 J mice by tail-vein injection. All mice were treated with GEX1A, ABT-199, ABT-263 or A-115 by i.p. injection individually or in different combinations at the indicted dosages and timepoints. All mice were then monitored for leukaemia development and analysed for leukaemia-related death by Kaplan–Meier Survival graphing. After the mice were sacrificed, leukaemia was verified by the following techniques: (1) examining dissected corpses for splenomegaly and hepatomegaly; (2) histologic analysis for leukaemic cell infiltration into the liver, spleen and kidney; (3) microscopic observation for leukaemic cells in PB smears and BM cytospins after Wright’s Giemsa staining; and 4) flow cytometric analysis of human CD45⁺ leukaemic cells (for human leukaemic cells) and mCherry⁺ cells (for murine MA9 cells, since MA9 AML cells were labelled with mCherry) in PB, BM, and spleen. Some mice transplanted with MV4;11 and K562 cells developed individual tumour masses. Tumour masses were sectioned and stained with human anti-CD45 antibodies to confirm the leukaemic cell of origin.

RNA sequence analysis

Molm-13, MV4;11, HL-60 and K562 leukaemic cells were treated with vehicle or 100 nM GEX1A for 6 h and were collected for mRNA extraction. RNAseq was conducted by Novogene Corporation Inc. Raw data FASTQ files were aligned to the human genome (GRCh38) using Tophat (version 2.0) and Bowtie22. The gene expression profiles for the individual samples were calculated as FPKM (paired-end fragments per kilobase of exon model per million mapped reads) values. Gene Ontology analysis was carried out with the Database for Annotation, Visualisation and Integrated Discovery (DAVID) tool (https://david.ncifcrf.gov/, Version 6.8).3. The differential expression cluster for the heatmap was generated using JavaTreeview.4.

Plasmid generation and virus production/transduction

High-titre viruses expressing specific genes of interest were generated by co-transfection of HEK293T cells with packaging plasmids and retroviral/lentiviral plasmids containing the individual genes of interest using the Calphos mammalian transfection kit (Clontech). Retroviral/lentiviral supernatants were harvested 24 and 48 h after transfection. After filtration by passage through a 0.45 μM filter, retroviral supernatants were aliquoted and frozen at −80 °C. Retroviral/lentiviral titres were determined by infecting 3T3 cells. Human leukaemic cells were infected with viral supernatants by spinoculation in the presence of 4 μg/mL polybrene. Twenty‐four hours after spinoculation, the infected cells were purified by sorting for GFP⁺ cells using FACS. We had previously generated MSCV-BCL-2-YFP, MSCV-MCL-1-GFP and MSCV-BCL-xL-GFP. c-FASTK and m-FASTK DNA fragments were amplified using forward primers  TTAGATCTCACCatgaggaggccgcggggggaac for c-FASTK and TTAGATCTCACCatgcttcgagtcctgctctctg for m-FASTK with same reverse primer TTGAATTCtcagcccccttcaggcccccagcg and pDONR223-FASTK as templates. MSCV-c-FASTK-GFP and MSCV-m-FASTK-GFP plasmids were generated by subcloning c-FASTK and m-FASTK DNA fragments into the MSCV-GFP vector.

Annexin-V and 7-AAD staining to analyse for apoptosis

The treated leukaemic cells were collected at the indicated timepoints and then stained with allophycocyanin-conjugated Annexin V followed by 7-amino-actinomycin D (7-AAD) staining in binding buffer following the manufacturer’s instructions (BD Biosciences). The death of infected cells was examined by analysing the percentages of Annexin-V⁺ and Annexin-V⁺/7-AAD⁺ cells by flow cytometry.

Statistics

Data are expressed as means ± SD. Two-way ANOVA (multiple groups) and Student’s t tests (two groups) were performed to determine the statistical significance of differences among and between experimental groups. Mantel–Cox log-rank test was performed for Kaplan–Meier survival comparison between experimental groups. P < 0.05 was considered significant.

Reporting summary

Further information on experimental design is available in the Nature Research Reporting Summary linked to this paper.

Author contributions

MS, RM, MR and KJ, LZ, SL, SB, WW, PL, PB and JZ conducted the experiments and analysed the data. MS drafted the first version of the paper. PB, RT and JZ contributed to the writing and editing of this manuscript.

Funding

This work was supported by NIH grants R01 HL133560-01 and R01 CA223194-01 through Loyola University Chicago, as well as Loyola programme development funds to Jiwang Zhang. This work was also partially supported by a grant from the National Institutes of General Medical Sciences (R01-GM129465), as well as the Walther Cancer Foundation through the Harper Cancer Research Institute at the University of Notre Dame.

Data availability

RNA sequence data have been submitted to the National Center for Biotechnology Information Gene Expression Omnibus (NCBI GEO) and is available for public access (GEO Submission GSE166591).

Competing interests

The authors declare no competing interests.

Ethics approval and consent to participate

Not applicable.

Consent to publish

Not applicable.

Supplementary information

The online version contains supplementary material available at 10.1038/s41416-022-01796-5.