Dual mTORC1/2 Inhibition Synergistically Enhances AML Cell Death in Combination with the BCL2 Antagonist Venetoclax

Acute myelogenous leukemia (AML) is a heterogenous disease driven by deregulation of multiple pathways, including genetic aberrations, mutations, and overexpression of antiapoptotic proteins (1). Despite this heterogeneity, treatment options for AML have been limited to anthracyclines, nucleoside analogs, and alkylating agents until the approval in 2017 of midostaurin targeting the FLT3 mutation. Since then, 8 agents have been approved or reapproved for this disease (1). Of these recent advances, the introduction of the anti-BCL2 BH3 mimetic, venetoclax, has had a major impact on AML treatment (2). Anti-apoptotic BCL2 family members such as BCL2, BCL-xL, BCLW, and MCL-1 are frequently overexpressed in hematologic malignancies, including AML (3). These proteins bind to and sequester pro-apoptotic BCL2 family members such as BAK and BAX that trigger mitochondrial outer membrane permeabilization (MOMP), an irreversible step in apoptotic cell death (4). ABT-737 (navitoclax) was the first-in-class BH3 mimetic to be developed. This agent binds to BCL2, BCL-xL, BCLW, as well as MCL-1 but with a considerably weaker affinity for the latter (5). An orally bioavailable derivative of ABT-737, navitoclax (ABT-263) was subsequently developed but its potential in AML was limited by the risk of severe thrombocytopenia arising from on-target effects of BCL-xL (6). It is not currently used in this disorder.

This limitation stimulated the clinical development of venetoclax that exhibited a pronounced affinity for BCL2 but a much lower affinity for BCL-xL compared with navitoclax (7). Venetoclax displayed modest monotherapy efficacy in AML (8), but was subsequently approved when administered in combination with hypomethylating agents or low-dose cytarabine, providing an additional 5 months overall survival based on the results of a phase III clinical trial for patients not fit for intensive induction chemotherapy (9). As in the case of other targeted agents, de novo or acquired resistance to venetoclax occurs through various mechanisms, including upregulation of other anti-apoptotic proteins (10) and downregulation of pro-apoptotic proteins (11), raising the possibility that combination strategies circumventing these events may be necessary to enhance the therapeutic efficacy of this agent in AML.

The PI3K/AKT/mTOR pathway is frequently upregulated in AML secondary to FLT3, c-KIT, and RAS mutations (12) and inhibitors of PI3K such as idelalisib, copanlisib, and duvelisib have been approved for other hematological malignancies, including chronic lymphocytic leukemia (CLL) and non–Hodgkin lymphoma. The mTOR regulates signals acting through the AKT pathway as well as other cascades involved in energy maintenance and oxidative stress among others (13). mTOR functions as a serine/threonine kinase and participates in two complexes: mTORC1, implicated in RNA translation via 4EBP1 phosphorylation, and mTORC2, whose downstream target is AKT (14). Although mTORC1 inhibitors such as rapamycin have been approved for the treatment of certain solid tumors, for example, renal cell carcinoma (15), they have shown minimal activity in AML, despite displaying activity against leukemia stem cells in preclinical studies (16). It has been proposed that the limited activity of rapalogs reflects the failure to inhibit 4EBP1 and/or feedback activation of PI3K/AKT and MEK/ERK (17, 18). These considerations led to the development of second-generation dual TORC1/2 inhibitors (e.g., AZD2014 and INK128) that may offer advantages over pure TORC1 inhibitors (19, 20). INK128, currently referred to as spanisertib or TAK-228, has shown preliminary efficacy in solid tumors (21) and is currently under investigation in combination approaches (22, 23).

Previously, our group reported that dual mTORC1/2 inhibitors interacted synergistically with the BCL2/BCL-xL inhibitor navitoclax to induce cell death in AML cells through a mechanism involving, at least in part, MCL-1 downregulation (24). However, it would be important to determine whether similar mechanisms are operative in the case of venetoclax, given the approval of this agent in AML and the inability to employ navitoclax in this disease. Currently, mechanistic insights into interactions between dual TORC1/2 inhibitors and venetoclax in human myeloid leukemia cells are essentially lacking. In particular, it is presently unknown whether such a strategy would be effective in cells with high basal AKT activity or with acquired venetoclax resistance. Here, we report that this strategy synergistically induces cell death in AML cells, including those resistant to venetoclax or expressing constitutively active AKT, and is active against primary AML (but not normal) cells as well as in an AML patient-derived xenograft (PDX) model.

Human AML cell lines, U937 and MV4–11 (CRL-9591) were purchased from the ATCC, MOLM-13 (DSMZ Cat# ACC-554) and OCI-AML3 (DSMZ Cat# ACC-582) cells were purchased from DSMZ (Brunswick, Lower Saxony, Germany) and maintained as described previously (25). All cell lines were tested for Mycoplasma contamination using the MycoAlert Mycoplasma Detection Kit (Lonza) routinely. All experiments used logarithmically growing cells (3–4×10⁵ cells/mL).

Venetoclax-resistant MV4–11 and MOLM-13 cells were obtained by culturing in the presence of increasing venetoclax (from 1 nmol/L to 1 μmol/L) concentrations over a period of 2 months.

Venetoclax was a gift from AbbVie. INK128, AZD2014 was purchased from ChemieTek. Copanlisib was purchased from AdooQ. MK2206 was purchased from MedChemExpress (Monmouth Junction). Rapamycin was purchased from Cell Signaling Technology. All drugs were dissolved in DMSO, aliquoted, and stored at −80°C. Final DMSO concentrations did not exceed 0.1%.

Lentiviruses were generated in 293T cells by transfecting cells with plasmids, pLX307 MCL-1 plasmid (Addgene #117726, RRID:Addgene_117726)/pCDH-puro-myr-HA-Akt1 (Addgene #46969, RRID:Addgene_46969; ref. 26), psPAX2 (Addgene #12260, RRID:Addgene_12260), and pMD2-VSVG (Addgene #12259, RRID:Addgene_12259). PEI transfection reagents were used. Viral supernatant was collected 2 and 3 days after transfection, filtered through 0.45 μmol/L membranes, and added to U937 and MV4–11 cells in the presence of polybrene (8 μg/mL, Millipore). Puromycin (1.5 μg/mL) was used to treat cells for two days for selection, which eliminated all cells in the uninfected control group. MV4–11 cells lacking BAX, BAK, or BAK/BAX were generated using CRISPR-Cas9 system as previously described (25).

Apoptosis was evaluated by flow cytometry using Annexin V-FITC/PI staining as before (10). Loss of mitochondrial membrane potential and cell death were assessed by double staining with 7-AAD as before (10). Drug concentrations used in cytotoxicity studies were selected on the basis of relatively modest (e.g., generally <20%) single-agent induction of cell death.

For CD34⁺/CD38⁻/CD123⁺ and CD34⁺/CD45dim/SS_low analysis, CD34⁺ mononuclear cells isolated from patient with AML bone marrows were blocked by TruStain FcX (BioLegend, Cat#422302) on ice for 10 minutes, stained with CD38-PE/Cy7 (BioLegend, Cat#303516), CD123-APC (BioLegend, Cat#306012) and/or CD45-APC-FireTM750 (BioLegend, Cat#304062) on ice for 30 minutes followed by staining with Annexin V-FITC at room temperature for 15 minutes. The percentage of apoptotic (Annexin V⁺) cells in the CD34⁺/CD38⁻/CD123⁺ and CD34⁺/CD45dim/SS_low population was then determined using a FACSCanto flow cytometer (BD Biosciences).

Immunoprecipitation and immunoblotting were performed as previously described (27). The following primary antibodies were used: β-Actin (Sigma-Aldrich, Cat#A2066), AKT (Cat#4691), BAK (Cat#12105S), BAX (Cat#5023S), BCL2 (Cat#15071), BCL-xL (Cat#2762S), cleaved PARP (Cat#9541), cleaved caspase-3 (Cat#9661S), GAPDH (Cat#97166S), γH2A.X (Cat#2577S), phospho-AKT (Ser473; Cat#4060), phospho-4E-BP1 (Thr37/46; Cat#2855), and 4E-BP1 (Cat#9644) from Cell Signaling Technology. BAK AB1 (Cat# AM03100UG) was purchased from Thermo Fisher Scientific. BAX 6A7 (Cat# sc-23959) and MCL-1 (Cat#, sc-74437) were purchased from Santa Cruz Biotechnology.

BAX and BAK conformational changes were assessed as previously described (27).

Bone marrow or peripheral blood samples from patients with AML were obtained with written informed consent from the patients (both male and female, over the age of 18). These studies were conducted in accordance with the Helsinki Declaration. Mononuclear cells were isolated as previously described (28). Normal hematopoietic CD34⁺ cells were isolated from human umbilical cord blood obtained from patients undergoing normal deliveries. All studies were sanctioned by the Investigational Review Board of Virginia Commonwealth University.

After 16-hour treatment, primary patient mononuclear cells were incubated with 7-AAD (Sigma-Aldrich) and anti-human CD34 antibody conjugated with Alexa-Fluor488 (BioLegend, Cat#343518) for 20 minutes. Slides were then mounted using DAPI (4′,6-diamidino-2-phenylindole) Fluoromount-G (Southern Biotech). PDX bone marrow samples were fixed in 4% formaldehyde for 2 minutes and then blocked with PBX containing 5% BSA. Anti-human CD45-PE/Cy7 (BD Pharmingen, Cat#557748) was used for immunofluorescent staining. Slides were then mounted using DAPI Fluoromount-G (Southern Biotech). Images were captured using an Olympus IX71 Inverted System Microscope with a DP73; 17MP Color Camera.

Femurs from PDX mice were excised for histological examination and fixed in neutral-buffered formaldehyde (10%) overnight at 4°C. Samples were washed with water and decalcified in 10% EDTA (pH 7.4) for 14 days, until they lost normal structural rigidity. The bones were then embedded in paraffin blocks and 5 μmol/L sections were cutoff. Sections were subsequently immunohistochemically processed using anti-CD45 antibodies, and evaluated by histopathology. Sections were visualized and images captured using an Olympus BX41 Fluorescence Microscope with a DP71 Digital Camera.

Animal studies were conducted under an approved protocol by the Virginia Commonwealth University Institutional Animal Care and Use Committee. For the orthotopic murine model, NOD/SCID-gamma mice (The Jackson Laboratory, RRID: IMSR_JAX:005557) were injected intravenously via tail vein with 5×10⁶ luciferase-expressing MV4–11 cells. Eleven days after injection of tumor cells, mice were monitored for AML engraftment using the IVIS 200 imaging system (Xenogen Corporation) as described before (24). Mice were exposed to either oral venetoclax 80 mg/kg, oral INK128 0.5 mg/kg, or combined treatment 3 days/week for 4 weeks. Control animals were administered equal volume of vehicle. Tumor growth was also monitored every other day with an IVIS 200 imaging system. For the flank murine model, NOD/SCID-gamma mice were inoculated subcutaneously with 1×10⁶ U937 cells. From 5 days after injection, mice were treated with oral venetoclax 80 mg/kg, oral INK128 0.5 mg/kg, or combined treatment 3 days/week for 4 weeks or until tumor size reached 17 mm or other humane endpoints occurred. For the PDX model, NOD/SCID-gamma Il3- GM-SF (NSG-SGM3; The Jackson Laboratory; RRID: IMSR_JAX:013062) were inoculated with 4×10⁶ patient cells (#101–03340) via tail vein. After confirming human CD45⁺ cell engraftment in peripheral blood 59 days after injection, mice were treated with either oral venetoclax 80 mg/kg, oral INK128 0.5 mg/kg, or combined treatment 3 days/week for 6 weeks, after which they were sacrificed and the percentage of human CD45⁺ cells in bone marrows was assessed.

Values represent the means ± SD for at least 3 independent experiments performed in triplicate. The significance of differences between experimental variables was determined using the Student t test or one-way ANOVA with the Tukey's–Kramer multiple comparisons test. The significance of P values is *, P <0.05; **, P <0.01; or ***, P <0.001, wherever indicated. Analysis of synergism was performed by median dose effect analysis using the software Calcusyn (Biosoft; ref. 29). Kaplan–Meier analysis of mouse survival was performed using GraphPad Prism 6 software (RRID:SCR_002798).

The datasets generated and/or analyzed during the current study are available from the corresponding author on reasonable request. All updated original source data (chiefly Western blot data) are available to the journal and readers via storage on the website OSFHOME with the following web address: https://osf.io/8ak4z/?view_only=254bd20337064f6896d31e7b6a958150.

In both venetoclax-sensitive (MV4–11 and MOLM-13) or intrinsically resistant (U937 and AML-3) cell lines, combination treatment with venetoclax and INK128 sharply increased cell lethality at minimally effective single-agent dose levels (Fig. 1A and B; Supplementary Fig. S1A and S1B). In intrinsically sensitive cells, venetoclax concentrations of 5 nmol/L were used, whereas in less sensitive cells, higher but pharmacologically achievable venetoclax concentrations were used (e.g., 0.5–1.0 μmol/L). In each case, combination indices substantially less than 1.0 were obtained by median dose effect analysis (Fig. 1C and D; Supplementary Fig. S1C and S1D), indicating highly synergistic interactions. Similar synergism was observed with an alternative dual mTORC1/2 inhibitor, AZD2014 (Supplementary Fig. S2A–S2D). Interestingly, synergistic interactions were not observed with the selective mTORC1 inhibitor rapamycin (Supplementary Fig. S2E and S2F), and in fact, rapamycin failed to significantly enhance venetoclax-induced apoptosis in U937 and MV4–11 cells, even when administered at considerably higher concentrations compared with those of INK128 (e.g., 5 μmol/L vs. 100 nmol/L; Supplementary Fig. S2G and S2H). Consistent with these findings, co-treatment with venetoclax/INK128 induced pronounced caspase-3/PARP cleavage and increases in the DNA damage response marker, γH2A.X, whereas single agents had minimal effects (Fig. 1E; Supplementary Fig. S1E and S1F).

INK128 administered (24 hours) alone or in combination with venetoclax in MV4–11 or U937 cells diminished phosphorylation of AKT (downstream of mTORC2) at Ser473 and to a varying extent, phospho-4EBP1 (downstream of both AKT and mTORC1; ref. 30; Fig. 2A). Compatible results were observed at earlier intervals, for example, 6 and 12 hours (Supplementary Fig. S3A and S3B), and in other lines, for example, AML-3 and MOLM-13 (Supplementary Fig. S3C and S3D). For Western blot studies, slightly lower drug concentrations were used to minimize the confounding effects of extensive apoptosis. Corresponding quantitative analysis of cell death is shown in Supplementary Fig. S4.

To investigate the functional capacity of this strategy to circumvent the effects of AKT overexpression, U937 cells ectopically expressing constitutively active (myristolated) AKT were generated (Fig. 2B). Notably, constitutive AKT activation failed to reduce the lethal effects of the venetoclax/INK128 regimen (P > 0.05), manifested by Annexin V-FITC/PI staining and caspase-3 cleavage (Fig. 2C and D). Consistent with these results, both phosphorylated AKT and 4E-BP1 levels were similarly downregulated in control cells and in two subclones expressing constitutively active AKT.

In view of evidence by our group and others that PI3K inhibitors potentiate the activity of venetoclax in AML and lymphoma cells (31), parallel studies were performed with the PI3K inhibitor copanlisib. In marked contrast with results obtained with INK128, a regimen combining copanlisib and venetoclax was significantly less effective in inducing apoptosis (P < 0.01) in constitutively active AKT cells than in empty-vector (EV) controls (Fig. 2E). Consistent with these findings, CA-AKT cells largely failed to display dephosphorylation of AKT and 4EBP1 or cleavage of caspase-3 following venetoclax/copanlisib exposure, whereas pronounced reductions in p-AKT and p-4EBP1 as well as marked caspase-3 cleavage were observed in EV controls (Fig. 2F).

Finally, parallel results were obtained with the AKT inhibitor MK2206. Specifically, the lethal effects of the MK2206/venetoclax regimen were dramatically attenuated in AKT-CA cells compared with EV controls (Supplementary Fig. S5A). Consistent with these findings, AKT-CA cells exposed to MK2206/venetoclax exhibited reduced dephosphorylation of pAKT and p4EBP1, as well as minimal cleavage of caspase-3 compared with control cells (Supplementary Fig. S5B). Together, these findings raise the possibility that dual mTORC1/2 inhibition may offer advantages over PI3K or AKT inhibition in potentiating venetoclax antileukemic effects, at least in cells displaying basal AKT activation.

Treatment of MV4–11 cells with venetoclax modestly increased MCL-1 expression as shown by Western blot analysis (Fig. 3A), a phenomenon that has previously been observed with this and other BH3-mimetics, for example, ABT-737 (32). However, co-administration of INK128 abrogated this increase and in fact resulted in a net reduction in MCL-1 levels. Similar phenomena were observed with multiple other AML lines, for example, U937, AML-3, and MOLM-13 (Supplementary Fig. S6A–S6C). To gain further insights into the functional role of MCL-1 in venetoclax/INK128 interactions, MV4–11 cells ectopically expressing MCL-1 were used (Fig. 3B). Compared with control EV cells, MCL-1–overexpressing cells displayed significant resistance to combined treatment with the venetoclax/INK128 regimen (Annexin V-FITC/PI; P < 0.01) accompanied by increased expression of MCL-1 (Fig. 3B and C). Very similar results were observed in MCL-1–overexpressing U937 cells, for example, enforced expression of MCL-1 protected cells from venetoclax/INK128-induced lethality in association with diminished MCL-1 downregulation and PARP cleavage (Supplementary Fig. S6D and S6E). These findings argue that MCL-1 downregulation by INK128 contributes functionally to INK128/venetoclax antileukemic synergism.

Immunoprecipitation studies were performed in MV4–11 and U937 cells to monitor activation of BAX and BAK following venetoclax/INK128 exposure. Combined treatment of MV4–11 cells with venetoclax/INK128 triggered a modest increase in BAK conformational change, an event associated with diminished co-immunoprecipitation with MCL-1 (Fig. 3D). Combined exposure also induced a more pronounced conformational change in BAX compared with untreated controls or to single-agent treatment, although the reduction in MCL-1 co-immunoprecipitating with BAX was less marked than that observed with BAK (Fig. 3D). A virtually identical response pattern was observed in U937 cells (Supplementary Fig. S6F).

To assess the functional significance of these events, MV4–11 cells were engineered using CRISPR technology to knockout BAK, BAX, and both BAK and BAX. BAX individual knockout and BAK/BAX dual knockout displayed dramatically reduced lethality following venetoclax/INK128 exposure, whereas cell death was reduced to a lesser but still significant (P < 0.05) extent in BAK individual knockout cells (Fig. 3E and F). Reductions in caspase-3 cleavage were concordant with these findings (Fig. 3F). Collectively, these findings argue that the antileukemic activity of the venetoclax/INK128 regimen involves activation of BAK and particularly BAX.

Venetoclax-resistant MV4–11 cells were generated by continuously exposing cells to progressively higher concentrations of venetoclax. Two clones, designated MV4–11 R1 and R2, were generated that displayed minimal cell death at venetoclax concentrations of 500 or 1,000 nmol/L, concentrations that were lethal to the majority of control cells (Fig. 4A). Evidence of diminished caspase-3 and PARP cleavage confirmed these findings (Fig. 4B). Of note, resistant cells displayed increased basal activation of AKT (S473) and expression of MCL-1 accompanied by a pronounced reduction in BCL2 expression compared with controls (Fig. 4C). Significantly, in both resistant cell lines, INK128 and venetoclax interacted synergistically to induce cell death (Fig. 4D; actual flow data are shown in Supplementary Fig. S8A and S8B). These events were accompanied by marked AKT inactivation and downregulation of MCL-1 (Fig. 4E). Of note, when the R1 and R2-resistant lines were treated at the same (lower) concentrations used for sensitive MV4–11 cells, the cytotoxicity of the combination treatment was reduced compared with controls but was still significantly greater than that observed for single-drug treatment (Supplementary Fig. S9A and S9C).

Venetoclax-resistant MOLM-13 cells (Supplementary Fig. S7A and S7B) exhibited modest increases in p-AKT (S473), BCL-xL, and MCL-1 expression compared with controls, but no decline in BCL2 levels (Supplementary Fig. S7C). Very marked synergism between INK128 and venetoclax was observed in both resistant clones (Supplementary Fig. S7D; actual flow data are shown in Supplementary Fig. S8C and S8D), accompanied by inactivation of AKT and downregulation of MCL-1 (Supplementary Fig. S7E). Similar to MV4–11 R1 and R2, synergistic cell killing was reduced in MOLM-13 R1 and R2 compared with control MOLM-13 (Supplementary Fig. S9B and S9D). These findings demonstrate that INK128 and venetoclax interact synergistically in cells resistant to venetoclax via disparate mechanisms, and suggest that inactivation of AKT and downregulation of MCL-1 may be implicated in this phenomenon.

Effects of the venetoclax/INK128 regimen were examined in primary patient-derived AML cells in vitro. The clinical features of patients donating specimens are listed in Supplementary Tables S1 and S2. In a representative specimen, immunofluorescence staining with CD34 and 7-AAD showed a marked increase in 7-AAD staining (red) colocalizing with CD34 following combined compared with single-agent treatment (Fig. 5A). In both the bulk leukemia cell population (ref. 33; CD45_dim, side scatter low, CD34⁺; N = 11; Fig. 5B) and more primitive leukemia cell population [ref. 24; CD34⁺ CD38⁻ CD123⁺ (ref. 34) N = 5; Fig. 5C], cell viability was significantly decreased in the combined versus the single-treatment groups. In addition, studies were performed using a limited number of specimens treated with 10 or 50 nmol/L venetoclax. As shown in Supplementary Fig. S10A–S10D, results were generally similar to those employing 5 nmol/L venetoclax, for example, combined treatment induced significantly more cell death than individual drug exposure. In marked contrast, no significant difference was observed between treatment groups in the case of normal CD34⁺ cord blood hematopoietic cells (Fig. 5D). Finally, Western blot analysis for cleaved PARP and caspase 3, MCL-1, and phosphorylated AKT in two specimens for which sufficient cells were available displayed similar patterns as observed in leukemia cell lines (Fig. 5E).

To assess the in vivo anti-leukemic activity of the venetoclax/INK128 regimen, NOD/SCID-gamma mice systemically bearing luciferase-labeled MV4–11 cell–derived xenografts were used. Coadministration of venetoclax (80 mg/kg) and INK128 (0.5 mg/kg) significantly (P < 0.05 vs. single agents) reduced the leukemia cell burden in vivo whereas single agents had only modest activity (Supplementary Fig. S11A and S11B). Survival was also significantly prolonged in the combination treatment arm compared with the placebo and single-agent arms (Supplementary Fig. S11C) without significant difference in body weight (Supplementary Fig. S11D). Similar tumor growth reduction was also observed in a flank tumor model involving U937 cells (Supplementary Fig. S12A and S12B). Western blot analysis of flank model tumors showed sharp increases in caspase-3 cleavage and downregulation of MCL-1, p-AKT, and p-4EBP1 expression with combined treatment as observed in leukemia cell lines (Supplementary Fig. S12C). Again, we did not observe significant difference in body weight among treatment groups (Supplementary Fig. S12D).

This combination treatment strategy was evaluated in a PDX model. NSG-SGM3 (stem cell factor, GMCSF and IL03) mice were injected with primary AML cells and treated with venetoclax (80 mg/kg) ± INK128 (0.5 mg/kg) when human CD45⁺ cells appeared stably in the peripheral blood (Supplementary Fig. S14A and S14B). In vitro exposure (24 hours) of these cells with INK128 + venetoclax showed a significant reduction in cell viability compared with single-agent treatment (Supplementary Fig. S13A and S13B). Mice were treated for 6 weeks and sacrificed at day 50 of treatment. At the time of sacrifice, CD45-positive human-derived cells in bone marrow were significantly reduced in the combined treatment group compared with each single-agent and control cohorts (Fig. 6A and B) whereas body weight was not significantly different across groups before and during the entire treatment interval (Fig. 6C). Similar findings were obtained by IHC (Fig. 6D, stained with anti-human CD45) and immunofluorescence analysis of bone marrow cells (Supplementary Fig. S14C). Collectively, these results indicate that combined INK128/venetoclax treatment is both tolerable and effective in xenograft and PDX AML models.

BCL2 inhibitor monotherapy exerts only modest therapeutic benefit in AML, and current clinical approval in this disease involves combination with hypomethylating agents or low-dose cytarabine (8). Mechanisms of resistance to venetoclax include MCL-1 upregulation through activation of various pathways (35, 36), predominance of monocytic phenotype expansion (37), acquired BAX mutations (38), and TP53 mutations (intrinsic resistance; ref. 39). We and others have previously reported synergistic interactions between BCL2 ± BCL-xL and PI3K/AKT/mTORC pathway inhibitors in AML (24, 25) and other hematopoietic malignancies (40). However, the ability of dual mTORC1/2 inhibitors to potentiate the activity of venetoclax in AML cells, including those displaying intrinsic or acquired venetoclax resistance, has not yet been investigated. The present findings indicate that a strategy combining venetoclax with a dual mTORC1/2 inhibitor is effective in these settings as well as in multiple in vivo AML models.

INK128 is a dual mTORC1/2 inhibitor that potently inhibits Ser473 AKT phosphorylation, an event required for TORC2 activation (41). mTORC1/2 inhibitors exhibit theoretical advantages over clinically approved selective mTORC1 inhibitors such as rapamycin through multiple possible mechanisms, including more pronounced inhibition of protein translation (19, 41, 42), more effective downregulation of MCL-1 (24), and prevention of the compensatory rebound AKT activation observed with rapamycin (43). In the current study, combining venetoclax with rapamycin was not associated with synergistic induction of cell death, whereas combining venetoclax with INK128 dramatically induced apoptosis in multiple leukemia cell lines, including those resistant to venetoclax, as well as patient specimens in vitro. Such findings raise the possibility that a dual mTORC1/2 inhibitor such as INK128 may be more effective when combined with venetoclax than pure mTORC1 inhibitors in patients with AML.

It is noteworthy that INK128 alone or in combination with venetoclax effectively suppressed phosphorylation of AKT and 4EBP1 in multiple AML models. Importantly, INK128 significantly enhanced venetoclax-mediated cell killing in cells exhibiting constitutively active AKT. In sharp contrast, a PI3K inhibitor (copanlisib), which we and others have shown to increase venetoclax lethality in malignant hematopoietic cells (25, 44) failed to do so. Similar findings were obtained with the AKT inhibitor MK2206. Moreover, INK128 was more effective than either of these agents in downregulating phospho-AKT expression in cells expressing constitutively active AKT. Dual mTORC1/2 inhibition by INK128 also diminished 4EBP1 phosphorylation in constitutively active AKT cell lines, in contrast with copanlisib and MK2206. These findings most likely reflect the ability of dual mTORC1/2 inhibitors to act at sites both upstream and downstream of AKT (14). Inhibition of AKT results in MCL-1 degradation through a GSKβ-dependent mechanism as we and others have previously reported (45, 46). In addition, deactivation of p4EBP1 by blocking phosphorylation plays diverse roles in cellular survival, including regulation of autophagy, transcription, and selective translation (47). Importantly, as MCL-1 has a very short half-life (e.g., 2–3 hours), inhibition of translation represents a potent stimulus for MCL-1 downregulation (28). It is therefore plausible that the capacity of INK128, but not MK2206 or copanlisib to downregulate p4EBP1 may contribute to the ability of the dual mTORC1/2 inhibitor but not MK2206 or copanlisib to circumvent the protective actions of constitutive AKT activation. Given frequent upregulation/activation of AKT in AML (12), it is tempting to speculate that combining venetoclax with dual mTORC1/2 inhibitors may be particularly effective against a subset of leukemic cells in which AKT is constitutively activated. Efforts to test this concept are currently underway.

The ability of BCL2 inhibitors to increase expression levels of MCL-1, including in vivo, is well described previously (48), although the mechanism underlying this phenomenon remains to be elucidated. Such a compensatory process represents an important mechanism of venetoclax resistance (36). It has been shown that MCL-1 (± BCL-xL) downregulation/knockdown can re-sensitize venetoclax-resistant AML cell lines to this agent (49). Consistent with these findings, ectopic expression of MCL-1 markedly protected leukemic cells from the INK128/venetoclax regimen. Taken together, such findings argue that the ability of INK128 to prevent venetoclax-mediated MCL-1 upregulation, and instead downregulate expression of this protein, are likely to be implicated in the antileukemic activity of this regimen. Furthermore, induction of leukemic cell death by the INK128/venetoclax regimen was functionally dependent upon BAK and particularly BAX activation, based on the observation that BAK and BAX knock-out cells were highly resistant to the INK128/venetoclax regimen. In the case of BAK, this may reflect the observed reduction in binding of downregulated MCL-1 to this protein, culminating in MOMP (50).

Interestingly, leukemic cells displaying acquired venetoclax resistance exhibited activation of AKT accompanied by increased expression of MCL-1, and these events were reversed by INK128 administration. This may stem from the ability of AKT inhibition to enhance MCL-1 degradation through a GSK-dependent process (51) and by interfering with 4EBP1-dependent protein synthesis, leading to downregulation of short-lived proteins such as MCL-1 (52). Although the magnitude of specific perturbations in relevant proteins varied between resistant cell lines, common features included marked increases in the relative expression levels of MCL-1/BCL2, and increased expression of p-AKT. The former findings are consistent with observations that patients with multiple myeloma whose cells exhibit low BCL2/MCL-1 ratio are likely to be resistant to venetoclax (53). The higher baseline MCL-1 expression in acquired venetoclax-resistant cell lines raised the possibility that they might be more sensitive to INK128 due to their dependence upon this protein. However, the absolute cytotoxicity of dual treatment was reduced compared with non-resistant cell lines, arguing against this notion. Notably, synergistic interactions between INK128 and venetoclax accompanied by p-AKT and MCL-1 downregulation in leukemic cells displaying acquired venetoclax resistance were recapitulated in cell lines (e.g., U937) exhibiting intrinsic venetoclax resistance. Taken together, such findings raise the possibility that the INK128/venetoclax strategy may act through a mechanism-based manner to induce leukemic cell death in venetoclax-resistant cells.

Administration of INK128 significantly increased the lethality of venetoclax toward primary AML blasts in association with downregulation of MCL-1 and inactivation of AKT, but had little or no effect on normal hematopoietic cells (CD34⁺). The basis for this selectivity is unclear, but may reflect the increased dependence of leukemic cells, including primitive leukemic progenitors, on BCL2, MCL-1, and activated AKT for survival compared with their normal counterparts (54). Determining whether this strategy will be particularly effective against leukemic cells exhibiting basal AKT activation and potentially addicted to this pathway for survival will require analysis of a considerably larger number of specimens. However, the finding that leukemic cells genetically engineered to express constitutively active AKT were highly susceptible to the INK128/venetoclax regimen raises the possibility that this may be the case.

Notably, the INK128/venetoclax regimen was well tolerated in mice, and resulted in a significantly greater reduction in leukemic cell burden compared with single-agent treatment in flank and systemic xenograft as well as AML PDX models. Such findings are consistent with the previously observed in vitro selectivity of this strategy. Significantly, leukemic cells extracted from animals receiving both agents exhibited several of the perturbations observed in vitro, for example, downregulation of p-AKT, p-4EBP1, and MCL-1, arguing that mechanisms responsible for enhanced antileukemic effects observed in vitro can be recapitulated in the in vivo setting.

In summary, the present findings demonstrate that dual mTORC1/2 inhibition robustly increases venetoclax antileukemic activity through an MCL-1 and BAX/BAK-dependent mechanism, including in cells resistant to venetoclax or expressing constitutively active AKT. It is also active against primary leukemic but not normal cells, as well as in multiple AML model systems. The recent success of the HMA/venetoclax regimen in AML has prompted the search for new strategies, particularly those targeting MCL-1, to increase the efficacy of venetoclax, including co-administration of novel MCL-1 inhibitors (55). The present observations argue that combining dual mTORC1/2 inhibitors with venetoclax, if tolerable, may represent a promising addition to the therapeutic armamentarium for AML.

S. Grant reports grants from NCI during the conduct of the study. No disclosures were reported by the other authors.

T. Satta: Conceptualization, resources, data curation, software, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration. L. Li: Conceptualization, resources, data curation, software, formal analysis, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. S.L. Chalasani: Investigation, methodology. X. Hu: Investigation, methodology. J. Nkwocha: Investigation, methodology. K. Sharma: Investigation, methodology. M. Kmieciak: Resources, data curation, formal analysis. M. Rahmani: Methodology, writing–review and editing. L. Zhou: Investigation, visualization. S. Grant: Conceptualization, supervision, funding acquisition, validation, writing–original draft, project administration.

Supported by R01CA205607, R01CA167708, Leukemia and Lymphoma Society of America #6472–15, P30CA16059 (to S. Grant), and Rising Scholar program by Virginia Commonwealth University (to T. Satta).

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.