Dasatinib overcomes AML cells resistant to BCL2 inhibition by degrading MCL1

Summary

Acute myeloid leukaemia (AML) is a severe disease occurring mainly in the elderly population. Venetoclax (VEN) combined with azacitidine has changed the paradigm of treatment of AML. Nevertheless, approximately 30% of patients are primary refractory to VEN (VEN‐R), with no current therapeutic option. To target VEN‐R AML, we collected primary blasts at AML diagnosis in a prospective biobanking trial (NCT02320656). We performed targeted Next Generation Sequencing and ex vivo drug testing in 108 AML samples. We noticed that 17 (15.7%) were navitoclax‐resistant (NAV‐R). We observed a strong anticorrelation between NAV and Dasatinib (DASA) ex vivo sensitivity, also found in the BEAT‐AML cohort. As NAV and ABT797 are both BCL2/BCLxL inhibitors, we hypothesized that blasts sensitive to DASA (DASA‐S) were dependent on MCL1. We performed BH3 profiling in 25 samples confirming MCL1 dependency. Immunoblots showed a higher MCL1 and BIM protein expression. We found a dose‐dependent decrease in MCL1 protein expression associated with caspase 3 activation upon DASA in a primary AML sample. Collectively, these results suggest that DASA degrades MCL1 and effectively kills AML cells. To prove this hypothesis, we designed a phase II clinical trial named VEN‐R DASA‐IPC 2022 067 (EUCT 2023‐505846‐24‐00), currently enrolling VEN‐R patients.

INTRODUCTION

Acute myeloid leukaemia (AML) is a heterogeneous group of severe diseases with various molecular alterations that mainly occur in patients greater than age 60. ¹ Despite recent advances in management, the relapse rate is still high, and the long‐term survival does not exceed 20%. BH3‐mimetics drugs such as the specific BCL2 inhibitor venetoclax (VEN) and the BCL2/BLCxL inhibitors (−inh) navitoclax (NAV) are small molecules inducing intrinsic apoptotic cell death by inhibiting antiapoptotic proteins. ² , ³ Combination of VEN plus azacitidine (VEN‐AZA) has become a new standard of care for unfit AML patients ineligible for intensive chemotherapy. Nevertheless, more than one‐third of AML patients are primary refractory to this treatment, with no standard option, and are frequently included in early‐phase clinical trials or receive best supportive cares. ⁴ , ⁵ In contrast with relapsed AML, VEN primary resistance (VEN‐R) has been associated with specific features including monocytic blast differentiation, MCL1 dependency and signalling mutations in leukaemia stem cells. ⁶ , ⁷ , ⁸ Drug repurposing involves identifying the drug, evaluating its efficiency using preclinical models and proceeding to phase II clinical trials. Our aim was to select already existing drugs to be repurposed in refractory AML by combining drug sensitivity profiling and BH3 profiling.

METHODS

We collected clinical bone marrow or peripheral blasts at AML at diagnosis in a prospective biobanking clinical trial named HEMATO‐BIO‐IPC 2013‐015 (NCT02320656), between 2014 and 2019, and performed ex vivo drug sensitivity profiling (DSRP ⁹ , ¹⁰ ) using a home‐made library of 80 FDA‐approved drugs including NAV. In parallel, we collected clinical data and performed targeted exome sequencing using a standard next‐generation sequencing panel of recurrently mutated AML genes as previously described. ⁹ We performed BH3 profiling and immunoblots on selected samples (Method S1).

RESULTS AND DISCUSSION

AML characteristics

Taking the first 107 AML samples collected at diagnosis, we identified that 17 (15.7%) were primary resistant to the BCL2/BCLxL inhibitor navitoclax (NAV) and considered as NAV‐resistant (NAV‐R) in ex vivo culture, taking a normalized IC50 value (Z‐score ⁹ ) >0.75 as a threshold (Figure 1A). The median age of the patients in the VEN‐R group was 62.5 (range: 20–82), median leucocyte count, platelet count and percentage of blasts were 29.7 G/L (range: 1.1–86.3), 67 G/L (range: 18–163) and 72% (range: 21–93) respectively. Eleven AML (64.7%) were classified as French–American–British (FAB)‐M4 (n = 5) and FAB‐M5 (n = 6) versus 32.2% (FAB‐M4, n = 22; FAB‐M5, n = 7) in the navitoclax‐sensitive (NAV‐S) group (p = 0.015). Cytogenetics was normal in nine cases (53%). Signalling mutations (FLT3, N/KRAS, PTPN11) were more frequent in the NAV‐R group contrary to IDH1/2 mutations, whereas TP53 mutations were equally distributed between the two groups. Clinical and biological characteristics of AML are displayed in Table 1.

FIGURE 1.

Drug sensitivity/resistance profiling identifies a correlation between NAV resistance and DASA sensitivity. (A) Heat map showing French–American–British (FAB), genomic distribution the mutations and drug sensitivity/resistance profiling of 108 primary AML samples collected at diagnostic. Z‐score (defined as: Patient EC50–meanEC50 of a patient reference matrix/standard deviation, in which the reference matrix was previously defined on a panel of 25 different samples treated identically). This Z‐score permits objective identification of a patient cell sensitivity using a quantitative threshold. In practice, a lower Z‐score indicates greater sensitivity of the patient's cells as compared to those of the other patients. (B, C) Graphs showing anticorrelation between navitoclax and DASA (B) or erlotinib drug sensitivity. IC50 is the estimated dose killing half of the cells at 48 h (C). (D) Graph showing normalized IC50 values of various tyrosine kinase inhibitors efficient in the 17 AML samples resistant to NAV. (E) Graphs showing anticorrelation between ABT‐737 (left panel) and VEN (right panel) and DASA sensitivity. Ex vivo drug response values are shown comparing the average area under the curve (AUC) for each patient samples.

TABLE 1.

Clinical and molecular characteristics of the 107 samples selected for drug sensitivity profiling.

	All (n = 107)	NAV‐R (n = 17)	NAV‐S (n = 90)
Patient's age	61 (20–82)	62 (33–89)	60 (20–82)
FAB classification
FAB‐0	11 (10%)	0	11 (12%)
FAB‐1	19 (18%)	3 (17.6%)	16 (17.7%)
FAB‐2	27 (25%)	1 (5.8%)	26 (28.8%)
FAB‐3	1 (<1%)	0	1 (1%)
FAB‐4	27 (25%)	5 (29.4%)	22 (24.4%)
FAB‐5	13 (12%)	6 (35.3%)	7 (7.8%)
FAB‐6	4 (3%)	1 (5.9%)	3 (3.3%)
Unknown	5 (5%)	1 (5.9%)	4 (4.4%)
De novo AML	97 (90%)	16 (94%)	76 (84%)
Secondary AML	10 (9%)	1 (6%)	14 (16%)
Prior MDS	4 (4%)	0	4 (4%)
Prior MPN	6 (6%)	1 (6%)	5 (6%)
WBC at inclusion (range)	15.1 (0.7–313)	29.7 (4–86)	13.5 (0.7–313)
Platelet count (range)	61 (3–252)	67 (18–175)	59.5 (3–252)
% of medullary blast (range)	66 (42–82.5)	72 (64–82.5)	64 (42–84)
Normal karyotype	38 (35.5)	9 (52.9)	29 (32.2)
Molecular alterations
NPM1	25 (23%)	7 (42%)	18 (20%)
FLT3‐ITD	12 (11%)	4 (24%)	8 (9%)
NRAS	20 (19%)	4 (24%)	16 (17%)
KRAS	10 (9%)	2 (12%)	8 (9%)
PTPN11	3 (3%)	3 (18%)	0
IDH1	7 (7%)	0	7 (8%)
IDH2	14 (13%)	1 (6%)	13 (14%)
KIT	4 (4%)	0	4 (4%)
TP53	11 (10%)	2 (12%)	9 (10%)

Abbreviation: AML, acute myeloid leukaemia; FAB, French–American–British; MDS, myelodysplastic syndroms; MPN, myelproliferative neoplasms.

Drug sensitivity profiling shows an inverse correlation between NAV and dasatinib sensitivity

We found a strong correlation between ex vivo resistance to NAV and sensitivity to most of the studied tyrosine kinase inhibitors (TKIs), including the Pi3K‐inh BKM120 (Buparlisib) and Idelalisib, the JAK2‐inh Ruxolitinib, the MEK‐inh Trametinib, the mTOR‐inh Temsirolimus, the FLT3‐inh Quizartinib and the BCR‐ABL‐inh Imatinib and Dasatinib (DASA, Figure 1A). Indeed, DASA sensitivity (DASA‐S) was statistically anti‐correlated with NAV‐R (correlation score was 0.492, p < 0.001, Figure 1B). On the other hand, we did not observe any statistical correlation with sensitivity to the EGFR‐inh erlotinib, indicating an absence of drug class effect in ex vivo testing (Figure 1A,C). Among the most active TKIs in the 17 AML samples considered as NAV‐R, DASA was found to be the most active with the lowest median normalized IC50 (Z‐score = 0.015, Figure 1D). As DASA has already been used in clinical settings in AML, ¹¹ , ¹² we chose it for further tests to be able to translate from preclinical data into a clinical trial. We first confirmed the anticorrelation between DASA and the BCL2/BCLxL‐inh ABT‐737 and with VEN in the BEAT‐AML published cohort (Figure 1E). ¹³ As NAV and ABT797 are both known BCL2/BCLxL‐inh, and given the fact that the most represented antiapoptotic proteins in AML blasts belonging to the BCL2‐family of proteins are limited to BCL2, BCLxL and MCL1, ¹⁴ we hypothesized that blasts resistant to NAV (and DASA‐S) were in fact independent from BCL2/BCLxL but dependent on MCL1 for maintaining their proliferation.

BH3 profiling shows a MCL1 dependency in dasatinib‐sensitive samples

To explore the link between DASA sensitivity and BCL2‐family protein dependency, we performed a BH3 profiling on 25 primary AML samples from the same cohort of patients (Figure 2A,B). BH3 profiling is a functional assay that measures mitochondrial apoptotic priming by using BH3 peptides to provoke a loss in mitochondrial membrane potential and cytochrome c release. ⁶ The test allows us to identify the dependency of cancer cell to each antiapoptotic protein. Unsupervised analysis showed three separate groups depending on the relative sensitivity to MS1 peptide (interacting with MCL1) and BAD peptide (interacting with BCL2). Group 1 had a global low sensitivity to MS1 and a higher sensitivity to BAD, indicating a relative BCL2 dependency; group 2 displayed an equal sensitivity to MS1 and BAD, indicating dependency on both BCL2 and MCL1, and group 3 had a higher sensitivity to MS1 than BAD in favour of MCL1 dependency only. We observed that BH3 profiles were highly correlated with ex vivo drug response to DASA and NAV (Figure 2C). Indeed, group 3 (MCL1‐dependent) had the lowest DASA IC50 and the highest NAV IC50. We confirmed these results by immunoblotting primary AML samples with an MCL1 antibody. We showed a higher expression of MCL1 protein and the proapoptotic peptide BIM, which specifically interacts with MCL1 in most of the DASA‐S samples (Figure 2D).

FIGURE 2.

DASA targets MCL1 dependency of navitoclax‐resistant cells by degrading MCL1. (A) Heat map showing the BH3 profiling of 25 AML samples. Unsupervised analyses allowed the discrimination of three groups depending on their relative mitochondrial susceptibility to MS1 (revealing MCL1 dependency) or BAD (revealing BCL2 dependency) peptides. (B) Examples of BH3 profiling in the three groups based on the heat map. (C) Ex vivo drug response to DASA and NAV in the three subgroups of patients identified by BH3 profiling. (D) Immunoblots showing MCL1 and BIM protein expression based on DASA sensitivity (Figure 1A). (E) Immunoblot showing MCL1, phospoT163 MCL1 and caspase 3 in a DASA sensitive patient sample after 8 and 24 h of treatment.

Dasatinib induces an indirect MCL1 degradation

Finally, we hypothesized that DASA restored apoptosis sensitivity in resistant AML blasts by targeting MCL1. We performed an immunoblot of MCL1 in a selected DASA‐S sample among the sensitive ones with enough available biological material and treated the cells with DASA. We observed a dose‐dependent decrease in MCL1 and phospho‐T163 MCL1 protein expression upon DASA treatment, associated with a MCL1 cleavage and a caspase 3 cleavage indicating a MCL1‐dependent apoptosis (Figure 2E). Collectively, these results suggest that AML cells resistant to BCL2/BCLxL inhibition can be cleared by targeting MCL1 dependency with DASA. It is known that DASA targets Src kinases (which stabilize MCL1), partially explaining its synergy with MCL1 inhibitors. ¹⁵ Preclinical data from the Beat AML cohort identified that AML samples sensitive to DASA were enriched in FLT3‐ITD and PTPN11 mutations and had a unique gene expression signature. ¹⁶ Another study showed that NAV and DASA are synergistic in NUP98‐NSD1/FLT3‐ITD AML. ¹⁷ To better understand the AML subgroups that would benefit from DASA, we designed a phase II clinical trial (VEN‐R DASA‐IPC 2022 067; EUCT 2023‐505846‐24‐00) using DASA in monotherapy for patients failing a minimum of two VEN‐AZA cycles.

CONCLUSION

DASA may overcome BCL2 inhibition‐resistant AML by triggering apoptosis through an MCL1‐dependent manner. Preclinical studies based on the DSRP in AML samples followed by orthogonal validation allowed us to ask clinical questions subsequently addressed in an early‐phase clinical trial. Ex vivo drug response companion assay can be used to identify the patient samples sensitive to DASA given the high correlation with ex vivo BH3 profiling.

AUTHOR CONTRIBUTIONS

SG performed the research, analysed the data and wrote the paper. CM, MB, AJ, RC, MC, FB, M‐AH and JA performed the research. PA, YC and NV designed the research study.

FUNDING INFORMATION

This work has been partially funded by A*Midex (AMX‐21‐PEP‐018).

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

PATIENT CONSENT STATEMENT

All patients consented to be enrolled in the biobanking protocol named HEMATO‐BIO‐IPC 2013‐015.

CLINICAL TRIAL REGISTRATION

NCT02320656.

Supporting information

Data S1.

Garciaz S, Montersino C, Bourgoin M, Jacquel A, Castellano R, Guille A, et al. Dasatinib overcomes AML cells resistant to BCL2 inhibition by degrading MCL1 . Br J Haematol. 2025;207(2):381–386. 10.1111/bjh.20195

DATA AVAILABILITY STATEMENT

Data are available on demand.