ULK1 inhibition promotes oxidative stress–induced differentiation and sensitizes leukemic stem cells to targeted therapy

Abstract

Inhibition of autophagy has been proposed as a potential therapy for individuals with cancer. However, current lysosomotropic autophagy inhibitors have demonstrated limited efficacy in clinical trials. Therefore, validation of novel specific autophagy inhibitors using robust preclinical models is critical. In chronic myeloid leukemia (CML), minimal residual disease is maintained by persistent leukemic stem cells (LSCs), which drive tyrosine kinase inhibitor (TKI) resistance and patient relapse. Here, we show that deletion of autophagy-inducing kinase ULK1 (unc-51–like autophagy activating kinase 1) reduces growth of cell line and patient-derived xenografted CML cells in mouse models. Using primitive cells, isolated from individuals with CML, we demonstrate that pharmacological inhibition of ULK1 selectively targets CML LSCs ex vivo and in vivo, when combined with TKI treatment. The enhanced TKI sensitivity after ULK1-mediated autophagy inhibition is driven by increased mitochondrial respiration and loss of quiescence and points to oxidative stress–induced differentiation of CML LSCs, proposing an alternative strategy for treating patients with CML.

Introduction

Macroautophagy (hereafter referred to as autophagy) is an evolutionarily conserved, lysosome-dependent, catabolic process that maintains cellular energy amounts during starvation or metabolic stress. Whereas nonselective autophagy involves uptake of cytoplasmic cargo into forming autophagosomes, selective autophagy is accountable for specifically removing redundant components from cells such as protein aggregates and organelles, including superfluous mitochondria (mitophagy) (1). Numerous preclinical studies have demonstrated that autophagy plays a predominantly cytoprotective role in the context of cancer, highlighting the potential benefit of combining autophagy inhibition with current anticancer therapies (2). Whereas the lysosomotropic agent hydroxychloroquine (HCQ) inhibits autophagy in preclinical cancer models, a major drawback in the field is the lack of potency and efficacy of HCQ as an autophagy inhibitor in clinical trials (3–6). These phase 1/2 studies have demonstrated that maximal achievable doses of HCQ (up to 1200 mg/day) do not consistently inhibit autophagy in patients with cancer, driving the development of more potent compounds, with similar chemical and lysosomotropic properties as HCQ (7), or more selective autophagy inhibitors that target proteins required for autophagy initiation (8–12).

The nutrient-sensing mechanistic target of rapamycin complex 1 (mTORC1) is a serine/threonine kinase complex that inhibits autophagy through phosphorylation of downstream autophagy-related (ATG) proteins (13–16). Accordingly, inhibition of mTORC1 leads to activation of the ULK1 (unc-51–like autophagy activating kinase 1) complex, consisting of ULK1 itself, ATG13, focal adhesion kinase family–interacting protein of 200 kDa (FIP200), and ATG101. Autophagy regulation is also mediated by energy-sensing adenosine monophosphate–activated protein kinase (AMPK), which is activated in response to reduced intracellular adenosine triphosphate (ATP) concentrations, caused by nutrient deprivation or inhibition of mitochondrial function. Activated AMPK mediates its effect by direct phosphorylation of ULK1 (15–17), leading to ULK1 activation and phosphorylation of its downstream targets, including ATG13 (18). This triggers the initiation of the autophagy cascade, leading to the recruitment of a ubiquitin-like conjugation system and the conjugation of microtubule-associated protein 1 light chain 3 (LC3) to phosphatidylethanolamine (PE) on the phagophore (the precursors to autophagosomes), followed by autophagosome lysosome fusion (19). Activation of the ULK1 complex is one of the first critical steps in the autophagy process, and ULK1 is a druggable serine/threonine kinase (10–12); therefore, ULK1 represents an attractive target for pharmacological inhibition of autophagy (2, 19).

Chronic myeloid leukemia (CML) is a stem cell–driven myeloproliferative disorder, caused by t(9;22)(q34;q11) chromosomal translocation in a single hematopoietic stem cell (HSC). The defective chromosome 22, also known as the Philadelphia chromosome, carries the chimeric BCR-ABL oncogene, which codes for a fusion oncoprotein with constitutive tyrosine kinase activity. The introduction of imatinib and other second- and third-generation tyrosine kinase inhibitors (TKIs) has transformed the clinical management of CML (20, 21) and represents a paradigm for targeted therapy in cancer with the potential for cure (22). However, up to half of patients with stable deep molecular response (DMR) that qualify for TKI discontinuation trials relapse within 12 months (23). Hence, only 10 to 15% of patients with CML achieve sustained treatment-free remission (TFR) (24). In addition, the vast majority of the patients require life-long TKI treatment, associated with chronic side effects, including chronic anemia, which in turn has been linked to less favorable treatment responses (25, 26). Furthermore, 20 to 30% of patients with CML develop resistance to TKIs or progress to lethal blast phase (BP) with very limited therapeutic options (24).

It has been demonstrated that leukemic stem cells (LSCs) are inherently insensitive to TKI treatment (27, 28). LSCs are detectable in virtually all TKI-treated patients with CML, including those who achieve DMR (29). We have previously demonstrated that TKI treatment induces protective autophagy in CML LSCs ex vivo (30) and in vivo (31). However, chloroquine and imatinib combination to eliminate stem cells (CHOICES), a phase 2 clinical trial where imatinib combined with HCQ-mediated autophagy inhibition was tested in patients with chronic phase (CP)–CML, showed limited potency of HCQ and only a modest reduction in minimal residual disease (MRD) as assessed by quantitative reverse transcription polymerase chain reaction (qRT-PCR) for BCR-ABL (6). These results highlight the need for developing and testing specific autophagy inhibitors to achieve frequent and deeper molecular responses for patients with CML, enabling more patients to attempt and maintain TFR.

Here, we present our findings regarding the mechanism through which ULK1 regulates TKI-induced autophagy, central carbon metabolism, differentiation, and drug sensitivity of primitive leukemia cells. Using patient-derived CML cells and autophagy-deficient cell lines, we demonstrate that TKI treatment leads to AMPK and ULK1 activation and, subsequently, an increase in ULK1-dependent autophagy flux. Genetic deletion of ULK1 or treatment with an ULK1 kinase inhibitor, MRT403, blocked TKI-induced autophagy and promoted a metabolic shift from glycolysis to mitochondrial respiration. This led to increased oxidative stress and reactive oxygen species (ROS)–dependent differentiation. Last, we demonstrated that combined BCR-ABL and ULK1 inhibition restored a normal ratio of immature myeloid and erythroid cells in the bone marrow (BM) and targeted therapy-resistant CML LSCs in both transgenic and patient-derived xenograft (PDX) CML models, further supporting the concept that specific autophagy inhibition may enhance anticancer therapy.

Results

ULK1 is a key mediator of TKI-induced autophagy

mTORC1 and AMPK integrate energy signals to maintain cellular homeostasis by balancing anabolic and catabolic processes, including autophagy. To assess whether TKI-mediated BCR-ABL inhibition affected AMPK and mTORC1 activity in primary CML cells, stem/progenitor (CD34⁺) cells were isolated from four individuals with CP-CML at diagnosis (table S1) and treated with the TKI nilotinib in serum-free medium (SFM) supplemented with low concentrations of human growth factors (32). As expected, nilotinib treatment inhibited BCR-ABL kinase activity, measured by decreased phosphorylation of CrkL (Fig. 1A and fig. S1A). Furthermore, nilotinib decreased ULK1 phosphorylation on serine-757 and increased phosphorylation on serine-555, suggesting inhibition of mTORC1 (15) and activation of AMPK (17), respectively (Fig. 1A and fig. S1, B and C). Time course experiment (n = 5) demonstrated that nilotinib treatment rapidly inhibited mTORC1 activity, measured by decreased phosphorylation of ribosomal protein S6 (RPS6), and led to gradual time-dependent AMPK activation, demonstrated by an increase in phosphorylation on the conserved threonine-172 (Fig. 1, B and C, and fig. S1, D to K). The increased phosphorylation of ATG13 confirmed activation of the autophagy-inducing ULK1 complex (13, 14, 18). Although patient variability in LC3B-II concentrations was observed, the time-dependent reduction in the autophagy cargo receptor sequestosome 1 (SQSTM1/p62) indicated consistent induction of autophagy flux after nilotinib-mediated BCR-ABL inhibition. In addition to the increase in TKI-mediated autophagy, we observed an increase in mRNA transcript expression of key ATG genes, including members of the ULK1 complex (fig. S1L). We therefore hypothesized that ULK1 is a central energy-sensing node and required for TKI-induced autophagy in CML cells (fig. S1M). We next used KCL22 cells, human BP-CML cell line, and CRISPR-Cas9 technology to test whether specific inhibition of ULK1 affected autophagy flux in CML cells. Consistent with CD34⁺ CML cells, ULK1 activity and autophagy flux were visible in nilotinib treatment control cells (Fig. 1D and fig. S1N). ULK1 knockout (KO) in KCL22 cells prevented ATG13 phosphorylation and TKI-induced autophagy in multiclonal and single-cell KO clones. Similar results were obtained with imatinib treatment of KCL22 cells, confirming a BCR-ABL–mediated effect (Fig. 1D). Colony-forming cell (CFC) assays demonstrated increased sensitivity of ULK1-deficient cells to BCR-ABL inhibition (fig. S1O).

Fig. 1. ULK1 is critical for TKI-induced autophagy.

(A) Western blots showing protein expression and phosphorylation in CD34⁺ CML cells after 24 hours of 2 μM nilotinib treatment (n = 4). Means ± SD. P values were calculated using paired Student’s t test. (B and C) Western blots showing protein expression and phosphorylation in CD34⁺ CML cells after 4 to 24 hours of 2 μM nilotinib treatment; representative of n = 5 independent patient samples. (D) Western blot showing protein expression and phosphorylation in control and ULK1 KO KCL22 cells after 24 hours of 2 μM nilotinib or imatinib treatment. (E) Fold change in bioluminescence imaging (photon per second). NRGW⁴¹ mice were transplanted with luciferase⁺ KCL22 control and ULK1 KO cells. After 13 days, mice received either vehicle or imatinib (50 mg/kg, BID) for 4 weeks via oral gavage. IVIS imaging was performed after 2 weeks of treatment. Means ± SEM. P values were calculated using one-way ANOVA and Dunnett’s multiple comparisons test on logarithmically transformed variables. (F) Number and representative images of extramedullary tumors harvested from each mouse at experimental end point. Means ± SD. P values were calculated using one-way ANOVA and Dunnett’s multiple comparisons test. (G) Overall survival was monitored by Kaplan-Meier analysis. P values were calculated using log-rank (Mantel-Cox) test and refer to control + imatinib versus ULK1 KO + imatinib.

KCL22 cells form extramedullary tumors when transplanted via tail vein injection into immunocompromised mice and are suitable for target assessment studies. To test whether ULK1 deletion affects tumor formation in vivo, KCL22 cells were labeled with luciferase and transplanted by tail vain injection into nonirradiated immunocompromised mice. Mice were then left untreated or treated for 4 weeks with imatinib (50 mg/kg, BID), frontline treatment for patients with CML, and monitored weekly by luciferase bioimaging. This revealed a decrease in bioluminescence in mice transplanted with ULK1-deficient cells, compared to mice receiving control cells (Fig. 1E), and a significant (P = 0.0025) reduction in the number of tumors at week 4 (Fig. 1F). Imatinib treatment significantly (P = 0.008) reduced tumor burden in mice transplanted with ULK1 KO cells, when compared with imatinib-treated mice xenografted with autophagy-competent cells (Fig. 1, E and F). In a separate transplant experiment, although imatinib treatment had no effect on the overall survival of mice transplanted with ULK1-expressing cells, consistent with the lack of efficacy of imatinib in the BP of the disease, it significantly (P = 0.0033) prolonged the survival of ULK1 KO xenografted mice (Fig. 1G). These results suggest a specific role for ULK1 in drug resistance and encouraged us to examine further the consequence of pharmacological or genetic ULK1 inhibition on autophagy and energy homeostasis in CML cells.

Inhibition of ULK1 induces oxidative stress by promoting a metabolic shift from glycolysis to oxidative mitochondrial metabolism

We previously developed MRT68921, which potently inhibits ULK1 in vitro and blocks mTORC1-mediated autophagy (10). However, similar to other available ULK1 inhibitors (11, 12), MRT68921 displayed only modest specificity, which results in undesired ULK1-independent cytotoxic effects. We have subsequently developed MRT403 with comparable potency [median inhibitory concentration (IC₅₀) of ULK1, 1.6 nM]. Improved kinase selectivity was first illustrated for MRT403 in a single-point screen against a panel of 80 kinases (fig. S2A). The screen was conducted at a high concentration (1 μM) at which ULK1 and ULK2 were inhibited greater than 90% for both compounds. No other kinases were inhibited to this level by MRT403. In contrast, MRT68921 inhibited 11 other kinases, at greater than 90% inhibition, including TANK-binding kinase 1, AMPK, and AMPK-related enzymes as previously reported (10), which could affect autophagy. In vitro selectivity greater than 30-fold for ULK1 versus these targets for MRT403 was confirmed by potency measurements (table S2) and would suggest that interfering off-target cellular activity would be unlikely. Western blotting confirmed ULK1 inhibition without affecting BCR-ABL activity (fig. S2B). To assess autophagy cargo degradation in CML cells, we generated a K562 cell line stably expressing yellow fluorescent protein (YFP)–Parkin, enabling precise measurements of the rate of autophagy in response to agents that modulate autophagy flux (33, 34). Using this system, autophagy-dependent removal of mitochondria, also known as mitophagy, is induced. Mitophagy was measured by reduction in electron transport chain (ETC) proteins, after treatment with the ETC complex III inhibitor antimycin A and the ATP synthase inhibitor oligomycin in K562 YPF-Parkin cells (fig. S2C). Basal and stress-induced mitophagy was not shown in clonal ULK1-deficient cells or when ATG7, an enzyme required for LC3-PE conjugation (35), was deleted (fig. S2D). The increase in mitochondria in ULK1 and ATG7 KO cells was confirmed in live cells using green fluorescent mitochondrial stain (fig. S2E). We next measured the effect of MRT403 treatment on autophagy inhibition. Treatment of YFP-Parkin–expressing CML cells with increasing concentrations of MRT403 inhibited autophagic degradation of ETC complex components in a concentration-dependent manner, with complete inhibition of autophagy achieved at 3 μM (fig. S2, F and G). Similar results were obtained using MRT68921 (fig. S2, H and I). HCQ-mediated autophagy inhibition required 10 μM HCQ for complete inhibition, although inhibition of basal mitophagy was evident at 3 μM HCQ (fig. S2, J and K). This prompted us to investigate whether pharmacological ULK1 inhibition affected basal autophagy in primary patient-derived cells. Treatment of CD34⁺ CML cells (n = 5) with 3 μM MRT403 affected LC3B-II concentrations, in line with previous studies suggesting the accumulation of stalled early autophagosomal structures (10, 36). In addition, ETC complex concentrations were consistently increased, suggesting impairment in autophagy-dependent mitochondrial clearance (Fig. 2, A and B). Similar results were obtained by flow cytometry after pharmacological autophagy inhibition in live cells (fig. S2L).

Fig. 2. Inhibition of ULK1 induces mitochondrial respiration and oxidative stress.

(A) Western blots demonstrating mitochondrial protein expression and quantification relative CV-ATP5A (B) in CD34⁺ CML cells after 24 hours of 3 μM MRT403 treatment (n = 5). Means ± SD. P values were calculated using paired Student’s t test. (C) Representative oxygen consumption rate (OCR) profile and relative basal OCR in CD34⁺ CML cells after 3 days of 3 μM MRT403 treatment (n = 4). Means ± SD. P values were calculated using paired Student’s t test. (D) Relative basal OCR in untreated (UT) KCL22 cells or after 3 days of 3 μM MRT403 treatment (n = 3 independent experiments). Means ± SEM. P values were calculated using unpaired Student’s t test. (E) Representative flow cytometry plots showing measurements of mitochondrial ROS and relative mean fluorescent intensity (MFI) in CD34⁺ CML cells after 3 days of 3 μM MRT403 ± 100 nM NAC or 1 hour of H₂O₂ treatment (n = 5). Means ± SEM. P values were calculated using paired Student’s t test. (F) MFI of mitochondrial ROS in KCL22 cells (n = 3 independent experiments). (G) Representative extracellular acidification rate (ECAR) profile and relative basal ECAR in CD34⁺ CML cells after 3 days of 3 μM MRT403 treatment (n = 5). Means ± SD. P values were calculated using paired Student’s t test. (H) Relative basal ECAR in KCL22 cells ± 3 days of 3 μM MRT403 treatment (n = 3 independent experiments). SEM P values were calculated using unpaired Student’s t test. (I) Relative NADPH/NADP⁺ ratio in CD34⁺ CML cells after 3 days of 3 μM MRT403 treatment (n = 3). For all experiments, means ± SD or SEM. P values were calculated using unpaired Student’s t test.

The acute increase of mitochondria in patient-derived CML cells resulted in an overall increase in basal mitochondrial respiration and maximal respiratory capacity, measured by an increase in basal cellular oxygen consumption rate (OCR) or after treatment with the mitochondrial uncoupler carbonyl cyanide p-trifluoromethoxyphenylhydrazone (FCCP) (n = 4; Fig. 2C). Similar results were obtained in MRT68921-treated KCL22 cells (fig. S2M). An increase in mitochondrial respiration was also apparent in ULK1- and ATG7-deficient cells (Fig. 2D). MRT403 treatment did not have any additional effect in autophagy-deficient cells, confirming that the increase in OCR after pharmacological ULK1 inhibition was dependent on inhibition of the autophagy process.

Because electron leakage in the ETC is the principal source of ROS, we hypothesized that increased mitochondrial respiration in autophagy-inhibited CML cells would lead to increased oxidative stress. To address this, patient-derived CD34⁺ CML cells (n = 5) were treated with MRT403, and mitochondrial superoxide was measured. This revealed an increase in mitochondrial ROS after ULK1 inhibition, which was completely abrogated by the addition of the antioxidant N-acetyl-l-cysteine (NAC) (Fig. 2E). Similar results were obtained in ULK1-deficient cells (Fig. 2F) and after treatment with MRT68921, whereas the effect of 3 μM HCQ on mitochondrial ROS was not significantly different compared with untreated cells (P > 0.05; fig. S2N).

Consistently, ULK1 has been shown to maintain redox homeostasis after nutritional stress by direct phosphorylation of key glycolytic enzymes (37). This autophagy-independent effect sustains glycolysis and the pentose phosphate pathway (PPP), leading to increased NADPH (reduced form of nicotinamide adenine dinucleotide phosphate) production and protection from oxidative stress. We therefore tested whether ULK1 inhibition affects glycolysis in vitro. Treatment with MRT403 significantly (P = 0.010) reduced glycolysis in CD34⁺ CML cells (n = 5; Fig. 2G), measured by extracellular acidification rate (ECAR), which was similar to the effect observed after treatment with MRT68921 in these cells (n = 3; fig. S3A). Whereas decreased glycolysis was also observed in ULK1-deficient cells, this was not the case in ATG7-deficient cells, suggesting an ULK1-dependent, autophagy-independent effect (Fig. 2H). Concurrently, treatment with MRT403 reduced ECAR in ATG7-deficient cells but had no additive effect in ULK1 KO cells, confirming an ULK1-dependent effect. Furthermore, together with reduced glycolysis, CD34⁺ CML cells exhibited a reduced NADPH/NADP⁺ ratio after MRT403 treatment, suggesting a decrease in carbon flux to the PPP (Fig. 2I). An increase in total cellular ROS amounts after ULK1 inhibition in CD34⁺ cells was also observed, indicating increased oxidative stress (fig. S3B), although without a significant increase in apoptosis or DNA damage (fig. S3, C and D). Similarly, no increase in DNA damage was observed after pharmacological or genetic ULK1 inhibition in CML cells (fig. S3, E and F).

Inhibition of ULK1 promotes loss of quiescence and ROS-dependent differentiation ex vivo

Increased mitochondrial activity and oxidative stress have been associated with loss of quiescence and enhanced differentiation of HSCs (38, 39). We therefore hypothesized that ULK1 inhibition would drive loss of quiescence and reduce the number of phenotypic CML LSCs ex vivo. To address this, we treated patient-derived CD34⁺ CML cells (n = 4) with MRT403 and performed cell cycle analysis using the proliferation marker Ki67 and propidium iodide (PI)–mediated DNA staining. This revealed a significant (P = 0.021) reduction in the G₀ fraction (Ki67^lowPI^low) in treated cells and an increase in the S-G2-M phase (Fig. 3A). Furthermore, MRT403 treatment resulted in reduction in absolute numbers of CD34⁺ and CD34⁺CD133⁺ CML cells (n = 8), suggesting differentiation toward more mature immuno-phenotype (Fig. 3, B and C).

Fig. 3. MRT403-mediated ULK1 inhibition promotes loss of quiescence and ROS-dependent differentiation.

(A) Representative flow cytometry plots and G₀-G₁-SG₂-M cell cycle phase ratio determined using Ki67 and PI staining after 3 days of 3 μM MRT403 treatment (n = 4). Means ± SD. P values were calculated using paired Student’s t test. (B and C) Representative flow cytometry plots showing the expression of CD34, absolute numbers of CD34⁺, and CD34⁺CD133⁺ cells relative to untreated after 3 days of 3 μM MRT403 treatment (n = 8). Means ± SD. P values were calculated using paired Student’s t test. (D) CD34⁺ CML samples were exposed to EPO (3 U/ml), and expression of surface markers CD45RA and CD123 after 3, 6, and 9 days of 3 μM MRT403 treatment was evaluated; positive control of MEP commitment was performed using γEPO (25 U/ml). Representative flow cytometry plots showing the expression of CD45RA and CD123 at day 9. CMP/GMP/MEP ratio measured by flow cytometry at days 3, 6, and 9. Means ± SEM on logarithmically transformed variables. (E to G) Erythroid maturation was measured, monitoring the expression of GlyA/CD71/CD36/CD44 after 3 to 9 days of 3 μM MRT403 treatment ± NAC (100 mM); positive control of erythroid maturation was performed using γEPO (25 U/ml). Representative flow cytometry plots showing the expression of GlyA/CD44, CD71/GlyA, and CD36/GlyA at day 9. Percentage of GlyA⁺CD44⁻, CD71⁺GlyA⁺, and CD36⁺GlyA⁺ relative to untreated after 3 to 9 days of MRT403 treatment. Means ± SEM. P values were calculated using paired Student’s t test on logarithmically transformed variables. (H) Representative picture of cell pellets after 9 days of 3 μM MRT403 treatment or γEPO ± NAC.

To further investigate the role of ULK1 in differentiation of CML cells, we cultured primary CD34⁺ CML cells (n = 3) with additional low concentrations (3 U/ml) of erythropoietin (EPO), which sensitizes primitive blood cells to erythroid lineage commitment. Using the expression of surface markers CD45RA and CD123, we next assessed the ratio of common myeloid progenitor (CMP), granulocyte-macrophage progenitor (GMP), and megakaryocyte/erythroid progenitor (MEP) cell populations after treatment with MRT403, using a high concentration of EPO (γEPO; 25 U/ml) as a positive control for potent erythroid differentiation. This demonstrated that prolonged treatment with a high concentration of EPO resulted in an expansion of the MEP population, measured by the gradual loss of CD45RA and CD123 expression (Fig. 3D). However, using these markers, time-dependent differentiation after MRT403 treatment was less apparent. Therefore, we next asked whether MRT403 treatment increases erythropoiesis. CD34⁺ CML cells (n = 4) were treated with MRT403 and stained for markers of erythroid maturation: GlycophorinA (GlyA), found in the erythrocyte membrane; CD71 (transferrin receptor 1), mediates the uptake of transferrin-iron complexes; CD36, a glycoprotein expressed on the surface of erythrocytes; and CD44, which expression is lost on the surface of erythrocytes (40). After continuous MRT403 treatment, erythroid maturation was noticeable with increased expression of GlyA, CD71, and CD36 markers and loss of CD44 expression (Fig. 3, E to G). This effect was reduced with the addition of NAC, suggesting a ROS-dependent erythroid differentiation. After macroscopic examination, it was evident that a proportion of MRT403-treated cells had turned red, indicating increased hemoglobin production, in keeping with increased erythroid maturation (Fig. 3H). Similar results were obtained after ULK1 deletion in K562 cells (fig. S3, G and H), which are known to persist in culture in an undifferentiated state but can spontaneously differentiate toward early-stage erythrocytes, granulocytes, or monocytes (41). Together, these data demonstrate that ULK1 inhibition alters the metabolic state and increases oxidative stress in primitive CML cells, driving loss of quiescence and increased myeloid differentiation.

MRT403 treatment inhibits TKI-induced autophagy and selectively targets CML LSCs ex vivo

The notion that ULK1 activity was important for maintaining primary CML cells in a quiescent and less-differentiated stage prompted us to examine whether MRT403 treatment sensitizes persistent CML LSCs to TKI treatment. The TKI nilotinib induced autophagy in primary CML cells, indicated by reduction in p62 amounts and changes in LC3B-II amounts (Fig. 4, A and B). MRT403 treatment blocked nilotinib-induced autophagy, evident by the accumulation of p62. Similar results were obtained when either MRT403- or MRT68921-mediated ULK1 inhibition was combined with nilotinib treatment in KCL22 cells (fig. S4, A and B). For comparison, similarly to fig. S2 (J and K), HCQ partially inhibited autophagy at 3 μM but required 10 μM concentration to fully inhibit TKI-induced autophagy, evidenced by the accumulation of LC3B-II (fig. S4C). We next observed the fate of primitive CML cells after combined BCR-ABL and ULK1 inhibition. Although MRT403 alone had no effect on CFC numbers, it significantly (P ≤ 0.001, n = 7) enhanced the effect of nilotinib on CML progenitor cells (Fig. 4C). We next performed long-term culture-initiating cell (LTC-IC) assays, using CD34⁺ cells (n = 4). This most stringent ex vivo stem cell assay demonstrated that combined MRT403 and nilotinib treatment significantly (P = 0.002) reduced the number of colonies when compared with nilotinib alone (Fig. 4D), indicating that ULK1-mediated autophagy inhibition promoted the eradication of primitive CML cells when combined with TKI treatment. Similar results were obtained using MRT68921-mediated ULK1 inhibition in primary CML cells (fig. S4, D and E).

Fig. 4. MRT403 treatment inhibits TKI-induced autophagy and targets CML LSCs.

(A) Western blots showing autophagy protein expression, and quantification (B), in CD34⁺ CML cells exposed to 2 μM nilotinib, 3 μM MRT403, and the combination for 24 hours (n = 4). (C) Representative picture of CFC-derived colonies and relative CFC number after exposure of CD34⁺ CML cells to 2 μM nilotinib, 3 μM MRT403, and combination for 3 days (n = 7). (D) Relative LTC-IC–derived colonies after exposure to 2 μM nilotinib, 3 μM MRT403, and the combination for 6 days (n = 4). (E) Representative picture of CFC-derived colonies and relative CFC number after exposure of normal CD34⁺ cells to 2 μM nilotinib, 3 μM MRT403, 100 nM omacetaxine, and the combination of nilotinib and MRT403 for 3 days (n = 5). (F) Relative LTC-IC–derived colonies after exposure of normal CD34⁺ cells to 2 μM nilotinib, 3 μM MRT403, and the combination for 3 days (n = 4). (G and H) Relative CFC number after 1 to 5 μM MRT403 and MRT68921 treatment in KCL22 cells. For all experiments, means ± SEM or SD. P values were calculated using paired Student’s t test.

To anticipate possible myelosuppression in humans, we assessed the effect of MRT403 and nilotinib on primitive nonleukemic hematopoietic human cells. Guided by previous results, normal CD34⁺ cells were treated with 3 μM MRT403, alone or in combination with nilotinib. This revealed no effect of single MRT403 treatment on primary CD34⁺ cells (n = 5) in contrast to cells treated with omacetaxine mepesuccinate (OMA), a Food and Drug Administration–approved inhibitor of protein biosynthesis, used on occasion for advanced phases of CML (Fig. 4, E and F). In contrast, MRT68921 significantly (P ≤ 0.0001) reduced the CFC and LTC-IC potential of normal cells when used at a concentration above 3 μM (fig. S4, F and G). Furthermore, whereas increasing the concentration of MRT403 alone did not affect the CFC potential of ULK1-competent or ULK1 KO cells, MRT68921 treatment reduced colony formation independently of the presence of ULK1, further highlighting the increased specificity of MRT403 (Fig. 4, G and H).

MRT403 targets human CML LSC xenografts when combined with TKI treatment

We next tested the target engagement and antileukemic effect of MRT403 in vivo. A dose-finding study demonstrated that a daily dose of ≥10 mg/kg (QD, oral gavage) resulted in >1 μg/ml concentration (>2 μM) of MRT403 in circulating plasma of mice (table S3). A 40 mg/kg dose was used in subsequent experiments. To evaluate MRT403 specificity in vivo, control and ULK1 KO cells were transplanted into nonirradiated immunocompromised mice. Mice were then left untreated or treated for 4 weeks with MRT403. MRT403 treatment significantly (P = 0.0029) reduced the number of tumors in mice transplanted with control cells (Fig. 5A). However, MRT403 treatment had no effect on the number of tumors in mice transplanted with ULK1 KO cells, indicating that the antitumor effects of MRT403 were mediated through ULK1 inhibition. To assess autophagy inhibition in BM-located human CML cells, we moved to a robust CML xenotransplantation model. Initially, primary CD34⁺ CML cells were transplanted into four sublethally irradiated mice. Four weeks after transplantation, mice were treated for 5 days with vehicle or MRT403 (Fig. 5B). p62 and ETC complex IV in isolated human CD45⁺ leukocytes were used as pharmacodynamic biomarkers to indicate autophagy/mitophagy inhibition and on-target effect of MRT403 (Fig. 5C and fig. S5, A and B). To allow for intra-experimental and intrapatient variability of human xenografts, we next performed two separate PDX studies using two different immunocompromised mouse models and independent primary CML samples. As previously demonstrated, increased human chimerism was achieved by using female NRGW⁴¹ (NOD.Cg-Rag1^tm1MomKit^W-41JIl2rg^tm1Wjl ) mice (42), when compared with male NRGW⁴¹ or conventionally used NSG (NOD-Prkdc^scid/scid-IL2Rγc^−/−) mice (fig. S5, C and D). After ensuring long-term BM engraftment of human cells, we treated the mice daily starting 9 to 12 weeks after transplantation for a period of 4 weeks with vehicle only, MRT403 (40 mg/kg, QD), imatinib (50 mg/kg, BID), or both drugs in combination. In all experimental arms, no signs of toxicity were observed during treatment as indicated by mouse weight and BM cellularity, confirming excellent tolerability (Fig. 5, D and E). In line with our previous results using the PDX model (43), imatinib had no effect on survival of primitive hCD45⁺CD34⁺CD133⁺ cells. Although MRT403 did not have a significant effect as a single agent, it significantly (P = 0.0018 and P = 0.0043) reduced the number of CML LSCs when combined with imatinib when compared with untreated or imatinib-only arms, respectively (Fig. 5F).

Fig. 5. MRT403 targets human CML LSC xenografts when combined with TKI treatment.

(A) NRGW⁴¹ (Exp#1) or NSG (Exp#2) mice were transplanted with luciferase⁺ KCL22 control and ULK1 KO cells. After 10 to 13 days, mice received either vehicle or MRT403 (40 mg/kg, BID) for 4 weeks via oral gavage. IVIS imaging was performed after 2 weeks of treatment. Number of extramedullary tumors harvested from each mouse at experimental end point. Means ± SD. P values were calculated using one-way ANOVA and Dunnett’s multiple comparisons test. (B) Diagram of PDX experimental design: Female NRGW⁴¹ mice were transplanted with human CD34⁺ CML cells, and engraftment was allowed to take place for 4 weeks. Mice were then treated in vivo for 5 days with vehicle or MRT403 (40 mg/kg) daily. (C) Western blot of autophagy and mitochondrial protein expression in purified human CD45⁺ cells after in vivo treatment. (D) Change in mice body weights upon drug treatment. (E) Absolute number of mouse BM cells engrafted with human CD34⁺ cells 9 to 12 weeks after transplant and after 4 weeks of in vivo treatment with the indicated drugs. (F) CD34⁺ CML cells (CML#5 and CML#6) were transplanted into female NSG or female/male NRGW⁴¹ mice. After engraftment of LT-LSCs (9 to 12 weeks), mice were treated by oral gavage with vehicle, MRT403 (40 mg/kg, QD), imatinib (50 mg/kg, BID), or the combination for 4 weeks. Absolute number of human CD45⁺CD34⁺CD133⁺ after treatment. Means ± SD. P values were calculated using one-way ANOVA and Dunnett’s multiple comparisons test on logarithmically transformed variables to meet the assumption of normal distribution. Representative flow cytometry plots showing expression of CD34/CD133 surface markers.

Combination of MRT403 and TKI treatment selectively targets Bcr-Abl ⁺ stem cells in a murine CML model

To complement the PDX work and to further investigate the effect of MRT403-mediated autophagy inhibition on leukemogenesis, we turned to an inducible CML mouse model. Using this transplantable model, mice developed leukemia after Bcr-Abl induction in HSCs, indicated by an increase in myeloid cells, splenomegaly, and anemia, resembling human CML (44). BM cells were isolated from Bcr-Abl ⁺ CD45.2 donor mice and transplanted into sublethally irradiated CD45.1 recipient mice (n = 27). After engraftment and irradiation recovery, Bcr-Abl was switched on by removing tetracycline from mice drinking water for 10 days, and separate mouse cohorts were treated with vehicle control, imatinib, equimolar concentration of MRT403, and HCQ, alone or in combination with imatinib for 4 weeks (Fig. 6A). Splenomegaly was evident in leukemic mice, which was reverted in mice receiving imatinib alone and when combined with MRT403 or HCQ (fig. S6A). Whereas a significant (P = 0.0327) reduction in the BM-located leukemic (CD45.2) donor cells was visible in mice receiving a combination of MRT403 and imatinib, the HCQ and imatinib combination was ineffective at reducing BM-located leukemic donor cells (Fig. 6B). The high percentage of engraftment of Bcr-Abl ⁺ cells using this model (>95% chimerism; fig. S6B) permits measurements of the most primitive leukemic cell population after single and combination drug treatments. Analysis of CD45.2 Lin⁻Sca1⁺c-Kit⁺ (LSK) cell subpopulations revealed that imatinib treatment enriched for BCR-ABL⁺ long-term (LT)–LSCs, in agreement with previously published work (31) (Fig. 6C). The enrichment in LT-LSCs after imatinib therapy was not visible when MRT403 was combined with imatinib treatment, suggesting that MRT403 may target BM-located LT-LSCs in vivo. To confirm the detrimental effect on the leukemia potential of the primitive leukemia-initiating stem cells, we performed a secondary transplantation experiment where CD45.2 BM cells from primary recipients were harvested and transplanted into sublethally irradiated CD45.1 mice (n = 26), together with 5 × 10⁵ CD45.1 carrier cells to avoid BM failure in the secondary recipient mice (Fig. 6A). As expected, limited long-term engraftment of CD45.2 cells was evident in mice receiving BM from vehicle- and MTR403-treated donor mice due to the reduced number of LT-LSCs in the BM of mice with high leukemia burden (Fig. 6D) (45). In contrast, leukemia, defined as >40% long-term CD45.2 engraftment and splenomegaly (>0.005 spleen/body weight ratio), was apparent in 70% of mice receiving BM cells from imatinib-treated mice (Fig. 6, D to F), in line with the increased number of LT-LSCs in these mice (Fig. 6C). Long-term engraftment of CD45.2 cells, as well as leukemia development, was completely reverted in mice transplanted with BM cells from MRT403 + imatinib–treated mice, confirming that the combination of ULK1 inhibition and imatinib treatment targets CML LSCs.

Fig. 6. MRT403 and TKI combination selectively targets LSCs in an inducible CML model.

(A) Diagram of experimental design: C57BL/6 mice were transplanted with 2 × 10⁶ tetracycline-inducible BCR-ABL⁺ [double transgenic (DTG)] cells after sublethal irradiation. After irradiation recovery (3 weeks), tetracycline was removed from the drinking water to induce BCR-ABL expression for 10 days. Then, mice received oral gavage vehicle, MRT403 (40 mg/kg, QD), imatinib (50 mg/kg, BID), or the combination for 4 weeks. After 4 weeks of in vivo treatment, mice were culled, and 3 × 10⁶ BMs were harvested and transplanted into irradiated recipient mice (2° recipient). The BM of the second recipient was analyzed after 8 weeks, or survival was measured after drug withdrawal. (B) Percentage of CD45.2 relative to the average of vehicle-treated mice. Means ± SD. P values were calculated using one-way ANOVA and Dunnett’s multiple comparisons test. (C) Flow cytometry gating strategy to measure amounts of primitive LSCs. Cells within the CD45.2⁺ LSK population were further selected on the basis of the expression of CD48/CD150 surface markers. Percentage of LT-LSC relative to the average of vehicle-treated mice. Means ± SD. P values were calculated using one-way ANOVA and Dunnett’s multiple comparisons test. (D) Percentage of CD45.2 relative to the average of vehicle-treated mice. Means ± SD. P values were calculated using one-way ANOVA and Dunnett’s multiple comparisons test. (E) Spleen weight relative to body weight from each treatment arm. Means ± SD. P values were calculated using one-way ANOVA and Dunnett’s multiple comparisons test. (F) Ratio of healthy and leukemic (>40% long-term CD45.2 engraftment; >0.005 spleen/body weight ratio) mice after secondary transplantation. (G) Overall survival monitored by Kaplan-Meier analysis. P values were calculated using log-rank (Mantel-Cox) test.

To further explore the antileukemic effect of MRT403 and imatinib combination, a separate cohort of mice was transplanted with CD45.2 BM cells and treated for 4 weeks as in the previous set of experiments. The rate of relapse and leukemia development was then evaluated by measuring survival after drug withdrawal. Imatinib monotherapy had no effect on overall survival, suggesting rapid relapse after drug withdrawal (Fig. 6G). However, the combination of MRT403 and imatinib significantly prolonged survival of recipient mice (P = 0.0055), providing further evidence that this combination approach targets the primitive LSC population that fuels the disease.

No changes were observed in progenitor and HSC populations when recipient mice were treated with MRT403 after transplantation with Bcr-Abl–negative CD45.2 BM cells, confirming MRT403’s limited effect on primitive nonleukemic cells (fig. S6, C to E). These results suggest a specific role for ULK1 in drug resistance and encouraged us to examine further the consequence of pharmacological or genetic ULK1 inhibition on autophagy and energy homeostasis in CML cells.

Combination of MRT403 and TKI treatment restores erythropoiesis in a murine CML model

Next, we wanted to further investigate the role of ULK1 in myeloid skewing in CML cells in vivo using the inducible CML mouse model. We observed an increase in MEPs after combined autophagy and BCR-ABL inhibition (Fig. 7A), with no detectable changes in CMPs or GMPs (fig. S7, A and B). To assess changes in erythroid cells, we measured the number of Ter119⁺ cells in the BM of mice. In agreement with Fig. 3 (D to H), this revealed a significant (P = 0.0006) increase in erythroid cells after MRT403 and imatinib combination treatment, reaching a similar amount to that seen in the BM of untreated healthy (Bcr-Abl OFF) mice (Fig. 7B). This was also demonstrated by an overall increase in the ratio of erythroid cells to more immature Gr-1⁺/Mac-1⁺ myeloid cells in leukemic mice treated with MRT403 and imatinib (Fig. 7C). Hematoxylin and eosin (H&E) staining of femur BM sections harvested from mice in each treatment arm revealed dense staining of vascular space or enucleated erythrocytes (cytoplasmic pink staining) in the BM of mice treated with MRT403 and imatinib combination (Fig. 7D). Gr-1 and Ter119 staining further confirmed that, whereas untreated mouse BM contained increased numbers of immature myeloid cells and decreased numbers of erythroid cells, mouse BM treated with MRT403 and imatinib presented with myeloid and erythroid cell ratios similar to noninduced wild-type mice (Fig. 7, D to F). To investigate erythroid maturation in more details, we determined distinct stages of in vivo erythropoiesis by using cell surface staining and flow cytometry. This strategy combines the expression of Ter119 and CD44 with cell size to quantify terminal erythroid maturation by defining distant erythroid developmental stages: proerythroblasts (I), basophilic erythroblasts (II), polychromatic erythroblasts (III), orthochromatic erythroblasts/immature reticulocytes (IV), and mature red blood cells (V) (Fig. 7G) (40, 46). We observed an increase in the number of cells in developmental stages II to V in leukemic mice treated with MRT403 and imatinib, confirming increased overall erythropoiesis in the BM (Fig. 7, H to L). This was also reflected by the proportion of mature red blood cells (region V), which reached 50% in mice treated with MRT403 and imatinib combination (Fig. 7M). Results obtained in MRT403-treated mice transplanted with Bcr-Abl– negative BM cells indicated no changes in CMP/GMP/MEP populations or in the ratio of erythroid/immature myeloid cells (fig. S7, C to F). However, a reduction was observed in the absolute number of Ter119⁺ cells (fig. S7G), reflecting the reduction in proerythroblasts, basophilic erythroblasts, polychromatic erythroblasts, and orthochromatic erythroblasts/immature reticulocytes (regions I to IV), although this did not significantly change the overall proportion of each developmental stage (fig. S7, H to M). Together, this suggests that the effect of pharmacological autophagy inhibition on erythrocyte development is distinct in Bcr-Abl–expressing cells compared to nontransformed cells.

Fig. 7. Combination of MRT403 and TKI treatments restore erythropoiesis.

(A and B) Absolute number of BM MEPs and Ter119⁺ cells relative to the average of vehicle-treated mice. Means ± SD. P values were calculated using one-way ANOVA and Dunnett’s multiple comparisons test. (C) Ratio of BM Gr-1⁺/Mac-1⁺ and Ter119⁺ cells from each treatment arm. Means ± SD. P values were calculated using one-way ANOVA and Dunnett’s multiple comparisons test. (D) H&E-, Gr-1–, and Ter119-stained femurs sections. Scale bars, 50 μM. WT, wild type. (E and F) Quantitation analysis of Gr-1 and Ter119. FoV, field of view. Means ± SD. P values were calculated using one-way ANOVA and Dunnett’s multiple comparisons test. (G) Flow cytometry plots demonstrating gating strategy to measure erythroid maturation. Cells were first gated on CD45⁻ cells. From GR1- and CD11b-excluded cells (Gr-1⁻CD11b⁻), lineage⁻, Ter119⁺ cells were gated, and expression of CD44 surface marker was used to visualize different stages of erythropoiesis. (H to L) Absolute number of BM proerythroblasts (region I), basophilic erythroblasts (region II), polychromatic erythroblasts (region III), orthochromatic erythroblasts/immature reticulocytes (region IV), and mature red blood cells (region V). Means ± SD. P values were calculated using one-way ANOVA and Dunnett’s multiple comparisons test. (M) Ratio of separate erythroid populations (regions I to V) from each treatment arm.

Discussion

Autophagy has been shown to be required for growth of rat sarcoma virus (RAS)–driven cancer cells (47, 48) and is frequently induced as a survival mechanism against anticancer treatments, such as chemotherapy, radiotherapy, and targeted therapy, contributing to drug resistance (2). For example, cancer cells with activating RAS mutations, such as pancreatic and colorectal cells, exhibit dependence on autophagy as a resistance mechanism after inhibition of the mitogen-activated protein kinase (MAPK) signaling pathway (49–51). This provided a rationale for testing the antimalarial HCQ, a nonspecific autophagy inhibitor used in the clinic, in combination with anticancer therapy. However, completed clinical trials indicate that treatment with antimalarial HCQ, a nonspecific autophagy inhibitor used in the clinic, does not achieve consistent autophagy inhibition in patients when used at maximum tolerated doses (3–6). This is increasingly being recognized by the academia and industry, driving the pursuit for druggable autophagy targets and the development of more specific autophagy inhibitors. These efforts have already led to the development of small molecular inhibitors against ULK1 (10–12). However, the lack of specificity of these inhibitors is a concern, and compounds suitable for preclinical in vivo studies have so far been lacking. Thus, the findings we present here regarding MRT403 report the first preclinical ULK1-specific inhibitor that is suitable for animal studies of cancer and other diseases. Moreover, a phase 1 clinical trial, designed to evaluate the safety, tolerability, and clinical activity of the investigational ULK1 inhibitor DCC-3116, as a single agent and in combination with MAPK inhibition in patients with RAS-driven cancers, has recently been initiated (ClinicalTrials.gov identifier: NCT04892017). This will further evaluate the concept of ULK1 inhibition as a new treatment option for specific tumor types.

Despite decades of research and increased knowledge about the hematopoietic system, MRD remains a clinical problem in myeloid malignancies. CML is driven by LSCs and represents an ideal tumor type to investigate cancer stem cell biology where findings may be applied to other tumor types. We have recently shown that CML LSCs rely on aberrant oxidative mitochondrial metabolism for survival (43). The core function of autophagy is related to metabolism. In this study, we show that TKI treatment activates ULK1-dependent autophagy after activation of energy-sensing AMPK, and the increase in autophagy flux supports the survival of therapy-resistant CML LSCs when BCR-ABL is inhibited. Furthermore, our study supports the concept that inhibition of ULK1 and autophagy-dependent mitochondrial clearance, also known as mitophagy, in CML models promotes mitochondrial ROS-dependent myeloid differentiation and maturation of erythrocytes.

One limitation in our study is that we did not test the amount of mitochondria clearance during erythropoiesis in our system. Studies have indicated that mitophagy plays an important role during erythropoiesis, whereby the ablation of ATG7 (52), the autophagy receptor NIX (53), or ULK1 (54) leads to defects in mitochondrial clearance during erythroid maturation. Therefore, it is likely that alternative systems for removing mitochondria are in place when ULK1 is inhibited, supported by the fact that mitochondrial clearance from reticulocytes is diminished, but not completely blocked, in autophagy-deficient erythroid cells (54, 55). Because MRT403 inhibits both ULK1 and ULK2, the potential clearance of mitochondria in our system is unlikely to be mediated by residual activity of ULK1-related proteins. This suggests that the way mitochondria are cleared may be context dependent and potentially regulated by both autophagy-dependent and autophagy-independent mechanisms (or noncanonical autophagy pathway), as previously suggested (55), or even via mitochondria-derived vesicles, which can be activated after ROS production (56). Another limitation in our study is that the clinical relevance of the increased erythropoiesis after ULK1 inhibition is not clear. Whether combined autophagy inhibition and TKI treatment promotes differentiation of BCR-ABL⁺ cells or reduces anemia in individuals with CML therefore remains to be elucidated.

Resistance of CML LSCs to TKIs has been linked to their undifferentiated and quiescent state. Of clinical relevance, we demonstrated that pharmacological ULK1 inhibition drives differentiation of phenotypically and functionally defined CML stem cells ex vivo and in vivo. This suggests that BCR-ABL expression in LSCs, which is known to drive increased mitochondrial activity (43), increased mitochondrial ROS amounts (57), and reduced long-term engraftment (58) when compared with normal HSCs, makes CML LSCs susceptible to stress-induced differentiation. This exposes a metabolic vulnerability, whereby “exaggeration” of the disease phenotype, in our case, through inhibition of mitochondrial clearance, reduces “stemness” and drives CML LSCs out of quiescence. The consequence is sensitization to TKI treatment, supporting the concept that testing ULK1 inhibitors such as MRT403 or DCC-3116 in combination with TKI treatment is an attractive approach to target LSCs and for patients with CML to achieve TFR. Future laboratory and clinical studies should aim to investigate whether pharmacological ULK1 inhibition may also have broader application to related leukemias and potentially provide a paradigm for other malignancies with a cancer stem cell hierarchy.

Materials and Methods

Study design

This study was designed to identify and validate effective drug combination to treat CML models that are driven by TKI-resistant CML LSCs. Functional work, including unbiased transcriptional analysis of patient-derived CML cells, highlighted ULK1 as a potential target for autophagy inhibition. To this end, we applied CML cell line xenograft model where gene KO is valuable for target validation/drug specificity studies, CML PDX model, which allows measurements of human CML LSCs after drug treatment, and transplantable double transgenic (DTG) CML model that is suitable for measurements of leukemia burden and primitive LT-HSCs. The design of the mouse treatment trials is described below. Mice were transplanted with CML or BM cells by tail vain injection and randomized to the various experimental cohorts at weeks 1 to 2 (cell line xenograft), weeks 9 to 12 (PDX), or weeks 4 to 5 (DTG model). Although no statistical methods were used to calculate the exact cohort sample size, the number of animals used per arm in each experiment was estimated on the basis of SD measurements in preliminary and previous experiments. To minimize variables for xenograft studies, same sex and similarly aged animals were transplanted with the same number of donor cells. For data analysis, investigators were blinded to the experimental conditions when assessing the outcomes. All mice were cared for equally in an unbiased fashion by animal technicians and investigators, and no animal was excluded from the analysis. P values were calculated by the method specified in the figure legends. In vitro experiments were all performed on at least three separate occasions, except where noted.

Primary samples

Primary CD34⁺ CML samples were obtained from patients with CP-CML (Ph⁺) at diagnosis, before TKI treatment. CD34⁺ normal primary samples were surplus cells collected from femoral head BM, surgically removed from patients undergoing hip replacement. CD34⁺ cells were isolated using the CD34 MicroBead Kit or CliniMACS (Miltenyi Biotec, #130-100-454). Informed consent for all human samples was obtained in accordance with the Declaration of Helsinki and with approval of the National Health Service Greater Glasgow Institutional Review Board and Clyde Institutional Review Board. Ethical approval for this work has already been granted by the West of Scotland Research Ethics Service (REC reference: 15/WS/0077 and 10/S0704/60).

Cell culture

Primary cells were cultured in SFM supplemented with stem cell factor (SCF; 0.2 ng/ml; BioLegend, #573902), granulocyte colony-stimulating factor (G-CSF; 1 ng/ml; BioLegend, #578602), granulocyte-macrophage CSF (GM-CSF; 0.2 ng/ml; BioLegend, #572902), interleukin-6 (IL6; 1 ng/ml; BioLegend, #570802), macrophage inflammatory protein α (MIPα; 0.2 ng/ml; PeproTech, #300-08), leukemia inhibitory factor (LIF; 0.05 ng/ml; PeproTech, #300-05), [bovine serum albumin (BSA), insulin, and transferrin (BIT)] (20%; STEMCELL Technologies, #09500), low-density lipoprotein (40 μg/ml; Sigma-Aldrich, #L4646), 2-mercaptoethanol (0.1 mM; Invitrogen, #31350-010), and 1% penicillin/streptomycin (LifeTech, #15140-122) and resuspended in Iscove’s modified Dulbecco’s medium (LifeTech, #12440-053). The CML cell lines K562 and KCL22 (DSMZ) were cultured in RPMI 1640 medium (LifeTech, #11875-093) supplemented with 1% penicillin/streptomycin, 1% l-glutamine, and 10% (v/v) fetal calf serum. Both primary cells and cell lines were passaged every 2 to 3 days, maintained at a concentration of 2 × 10⁵ cells/ml at 37°C with 5% CO₂, and routinely tested for mycoplasma. ULK1 KO single-cell cloning was performed with serial dilution in a 96-well plate (100 μl per well). Puromycin selection (3 μg/ml) was performed after cell expansion, and Western blotting was performed to validate ULK1 KO.

Ex vivo lineage commitment and erythroid maturation

CD34⁺ CML cells were cultured in SFM media with addition of EPO (3 U/ml; PeproTech, #100-64). Cells were treated for 3, 6, and 9 days with 3 μM MRT403 or with dimethyl sulfoxide (DMSO), and media was refreshed every 3 days. EPO (γEPO) (25 U/ml) was used as a positive control. Lineage commitment and erythroid maturation were assessed by BD FACSVerse (BD Biosciences) after staining with CD235a (BioLegend, #349106), CD71 (BioLegend, #334108), CD44 (BioLegend, #338804), CD36 (BioLegend, #336222), CD123 (BioLegend, #306010), CD45RA (BD, #347723), CD34 (BD, #555824), and CD133 (Miltenyi, #130-080-801).

Cell cycle analysis

CD34⁺ CML cells were cultured at 2 × 10⁵ cells/ml and fixed in cold 70% ethanol while vortexed and incubated at −20°C for at least 2 hours. Cells were then stained with Ki67 (BD, #558026) antibody for 20 min. PI was used as a nucleic acid dye.

CFC assay

CD34⁺ cells (3 × 10³) were seeded in 1.5 ml of MethoCult H4034 Optimum (STEMCELL Technologies, #04034). For CML cell lines, 1 × 10³ cells were seeded in MethoCult 4230 (STEMCELL Technologies, #04230). The number of CFCs was counted after 14 days.

LTC-IC assay

Irradiated M2-10B4 and S1/S1 genetically engineered mouse cell lines, expressing human cytokines (feeder cells), were seeded (7 × 10⁴ cells each) in MyeloCult H5100 (STEMCELL Technologies, #05150) supplemented with hydrocortisone in collagen-coated plates. The following day, 5 × 10⁴ primary CML cells (pretreated for 6 days) were harvested, resuspended in 500 μl of MyeloCult H5100 and seeded on top of the feeder cell layer, and cultured for additional 5 weeks with a weekly refresh of half of the media. Cells were then harvested and transferred to 1.5 ml of MethoCult H4034 Optimum, and the number of CFC was counted after 14 days.

OCR and ECAR measurements

OCR and ECAR were measured using the Seahorse XF96 flux analyzer (Seahorse Bioscience). The XF96 well plate was coated with 25 μl per well of Cell-Tak solution (22.6 μg/ml; Corning, #354240) and left for at least 30 min at room temperature (RT). CD34⁺ CML cells or KCL22 cells were seeded in each well in 175 μl of XF Assay Medium (Seahorse Bioscience, #100965–000). For measurements of OCR in KCL22 cells, media was supplemented with 2 mM glutamine and 11 mM glucose. For CD34⁺ CML cells, media was supplemented with 2 mM glutamine and 25 mM glucose, SCF (0.2 ng/ml), G-CSF (1 ng/ml), GM-CSF (0.2 ng/ml), IL6 (1 ng/ml), MIPα (0.2 ng/ml), LIF (0.05 ng/ml), low-density lipoprotein (40 μg/ml), and 2-mercaptoethanol (0.1 mM). After 30 min in a CO₂-free incubator at 37°C to ensure adherence, the plate was transferred to the Seahorse XF96 analyzer. Measurement of OCR was done at baseline and after sequential injections of oligomycin (0.7 μM; an ATP synthase inhibitor), FCCP (1.6 μM; a mitochondrial uncoupler), antimycin A (1 μM; complex III inhibitor), and rotenone (1 μM; complex I inhibitor), respectively. Basal respiration relative to untreated was calculated as the average of last rate measurements before oligomycin injection minus the average of nonmitochondrial respiration rate. For measurements of ECAR, media was prepared as above without glucose. Measurement of ECAR in KCL22 was performed at baseline and after sequential injections of glucose (11 mM), oligomycin (0.7 μM), and 2-deoxyglucose (5 mM). For CD34⁺ CML cells, glucose (25 mM), oligomycin (0.7 μM), and 2-deoxyglucose (10 mM) were used. ECAR relative to untreated was calculated as the average of maximum measurements before oligomycin injection minus the average of last rate measurements before glucose injection.

NADPH measurements

Cells (1 × 10⁶) per condition were harvested and washed in phosphate-buffered saline (PBS); NADPH was measured and analyzed as indicated by the manufacturer (Abcam, #ab65349).

ROS measurements

Cells were washed in PBS and stained using 1 ml (10 μM) of MitoSOX (Thermo Fisher Scientific, #M36008) or CellROX (Thermo Fisher Scientific, #10422) and incubated for 30 min at 37°C. Before staining, cells were treated for 1 hour with 100 nM NAC and 1 μl of 3.7% H₂O₂. Cells were then washed twice in 2 ml of PBS and analyzed by BD FACSVerse.

Mitochondrial mass measurements

Cells were washed in PBS and stained using 1 ml (10 μM) of MitoTracker green (Thermo Fisher Scientific, #M7514) and incubated for 30 min at 37°C. Cells were then washed twice in 2 ml of PBS and analyzed by BD FACSVerse.

DNA damage measurements

CD34⁺ CML cells were fixed in cold 70% ethanol while vortexed and incubated at −20°C for at least 2 hours. Cells were then stained with anti-H2A.X Phospho (Ser¹³⁹) (BioLegend, #613411) antibody for 20 min. DAPI (4′,6-diamidino-2-phenylindole) staining was used as a nucleic acid dye.

Reverse transcription quantitative polymerase chain reaction

RNA was extracted using the PicoPure RNA Isolation Kit (Thermo Fisher Scientific). Reverse transcription was performed with the High-Capacity cDNA Reverse Transcription Kit (Thermo Fisher Scientific). TaqMan Array Fast Plates (10 μl of reaction) were designed to cover a panel of autophagy genes (Thermo Fisher Scientific, #4413255-57). Plates containing autophagy primers were used with TaqMan Advanced Fast Master Mix (Thermo Fisher Scientific). The PCR was performed with the following steps: 2 min at 50°C, 10 s at 95°C, followed by 40 cycles at 95°C for 15 s, and 60°C for 30 s on a C1000 Touch thermal cycler (Bio-Rad). The relative mRNA expression was calculated using the ΔΔCT method.

Western blotting

Cells were prepared using lysis buffer containing protease and phosphatase inhibitors. Protein lysates were quantified using a BCA protein assay kit (Pierce; Thermo Fisher Scientific, #23228) and equilibrated according to standard curve. Lysates were resolved in 8 to 12% SDS–polyacrylamide gel electrophoresis gels. For LC3B analysis, 15% gels were home-casted according to Bio-Rad recipe.

Western blotting antibodies

Antibodies from Cell Signaling Technology are as follows: AMPKα (1:1000; #2532), p-AMPK^Thr172 (1:1000; #2531), ATG13 (1:1000; #13468), Crkl (1:500; #3182), p-Crkl^Tyr207 (1:100; #3181), mTOR (1:500; #2983), p-mTOR^Ser2448 (1:500; #2971), RPS6 (1:1000; #2317), p-RPS6^Ser240/244 (1:1000; #5364), ULK1 (1:1000; #8054), p-ULK1^Ser757 (1:500; #6888), p-ULK1^Ser555 (1:500; #5869), p-ATG13S^er318/ATG^Ser355 (1:1000; #46329), LC3B (1:500; #2775), ATG7 (1:1000; #8558), glyceraldehyde phosphate dehydrogenase (GAPDH) (1:1000; #5174), β-tubulin (1:1000; #2146), mouse immunoglobulin G (IgG) (1:5000; #7076), and rabbit IgG (1:5000; #7074). OXPHOS cocktail (1:2000; Abcam, #110413), green fluorescent protein (1:1000; Roche, #118144600001), MT-CO2 (1:2000; Thermo Fisher Scientific, #A-6404), p62 (1:1000; BD, #610832), and p-ATG13^Ser318 (1:1000; Abnova, #NBP2-19127) were also used.

CRISPR-Cas9–mediated gene editing

To target human ULK1 and ATG7, guides were designed using the optimized tool https://zlab.bio/guide-design-resources. Guides within the first six coding exon and with the highest score and lower probability of off-target effect were selected. Synthesis of the oligonucleotides was performed by Integrated DNA Technologies, which were in vitro annealed and cloned in Bsmb I–digested lentiCRISPRv.2-puro (Addgene, #52961). After stable integration of lentiCRISPRv.2 using lentiviral transfection and 1-week selection using puromycin (3 μg/ml), guides were validated by performing Western blotting. Oligonucleotides from Addgene are as follows: ATG7 (5′-GAAGCT-GAACGAGTATCGGC-3′), ULK1-1 (5′-CGAAGGCGCCGTGG-CCGATC-3′), ULK1-2 (5′-GCAGCGTCTGAGACTTGGCG-3′), and ULK1-3 (5′-AGCAGATCGCGGGCGCCATG-3′).

Lentivirus production

Lentiviruses for pLentiCRISPRv.2 and pLenti CMV V5-LUC blasticidin (Addgene, #21474) were produced by calcium phosphate method using pCMV-VSV-G (envelope plasmid) and psPAX2 (packaging constructs) vectors and human embryonic kidney (HEK) 293FT for transfection. Retroviruses for YFP-Parkin-IRES-zeocin (plasmid donated by S. Tait) were produced by transfection of Phoenix-Ampho HEK.

Fluorescence-activated cell sorting analysis

BM lineage–specific markers were assessed using flow cytometry analysis. BM was resuspended in PBS/2% fetal bovine serum and stained with antibodies for surface-specific markers for 20 min in the dark. Cells were then washed in PBS and analyzed using BD FACSVerse.

Animal studies

Animal work was carried out with ethical approval from the University of Glasgow Animal Welfare and Ethical Review Board under Home Office License PPL 60/4492 and personal license IA68AFD03.

Luciferase bioimaging

KCL22 control or ULK1 KO cells were transfected with pLenti CMV V5-LUC blasticidin vector using lentiviral transfection and selected on the basis of antibiotic resistance. NSG or NRGW⁴¹ mice transplanted by tail vein injection with control or ULK1 KO cells (4 × 10⁶ cells per mouse) were weekly injected subcutaneously with 200 μl of d-luciferin (PerkinElmer, #122799-2) and analyzed by luciferase bioimaging, using the IVIS Spectrum In Vivo Imaging System, to monitor tumor burden. After engraftment of the cells (10 to 13 days after transplant), mice were treated for 4 weeks with either MRT403 (40 mg/kg, QD) or imatinib (50 mg/kg, BID).

Drug escalation study

Mice were treated daily for 2 weeks with MRT403, with an initial dose of 5 mg/kg and highest dose of 40 mg/kg (dose doubled every 5 days). Blood was harvested, performing cardiac puncture after 24 hours of treatment with the lowest dose (5 mg/kg), after 5 days of treatment with 10 mg/kg, and after 5 days of treatment using 40 mg/kg, and immediately poured in tubes containing EDTA. Tubes were inverted three times to avoid blood clotting. Samples were centrifuged at 2000g/20 min at RT to separate plasma from each sample. Plasma was then collected into a fresh tube and stored at −80°C.

PDX studies

CD34⁺ CML cells were tail vein transplanted (1 × 10⁶ cells per mouse) into 8- to 12-week-old sublethally irradiated [2.25 grays (Gy)] female NSG and female/male NRGW⁴¹ mice. After 9 to 12 weeks posttransplant, mice were treated by oral gavage with vehicle [water + N-methyl pyrrolidone (NMP), polyethylene glycol (PEG)], MRT403 (40 mg/kg, QD), imatinib (50 mg/kg, BID), or the combination of MRT403 and imatinib for 4 weeks. After treatment, BM was extracted from the hip, tibia, and femur of each mouse. CD45, CD34, and CD133 expression was measured by fluorescence-activated cell sorting (FACS) analysis.

In vivo autophagy measurements

CD34⁺ CML cells were tail vein transplanted (1 × 10⁶ cells per mouse) into 8-week-old sublethally irradiated (2.25 Gy) female NRGW⁴¹ mice. After 4 weeks of engraftment, mice were treated by oral gavage with either vehicle (water + NMP-PEG) or MRT403 (40 mg/kg, QD) for 5 days. Mice were then culled, and BM was extracted from the hip, tibia, and femur of each mouse. Human CD45⁺ cells were FACS sorted, and autophagy/mitophagy markers were evaluated by Western blotting.

Transgenic mouse model

CD45.2 BM cells harvested from the tibia, femurs, and hips of Scl-tTa–BCR-ABL mice were tail vein transplanted (2 × 10⁶ cells per mouse) into 8-week-old sublethally irradiated (2 × 4.25 Gy; 3 hours apart) female and male CD45.1 C57/Bl6 recipient mice. Mice were maintained on tetracycline (0.5 g/liter) for 3 weeks before removal of tetracycline from the drinking water for transgene induction. After 10 to 14 days off tetracycline, mice were treated with vehicle (water + NMP-PEG), MRT403 (40 mg/kg, QD), HCQ (36 mg/kg, QD), imatinib (50 mg/kg, BID), or the combinations of MRT403 and imatinib and HCQ and imatinib, for 4 weeks. After treatment, BM was extracted from the hips, tibias, and femurs of each mouse. For survival study, mice were kept off tetracycline after 4 weeks of treatment and culled after 20% loss of total body weight.

Secondary transplant study

CD45.2 BM cells were harvested from the tibia, femurs, and hips of Scl-tTa–BCR-ABL mice after 4 weeks of treatment with either vehicle (water + NMP-PEG), MRT403 (40 mg/kg, QD), imatinib (50 mg/kg, BID), or the combination of MRT403 and imatinib. BM cells (3 × 10⁶ CD45.2 cells per mouse, plus 5 × 10⁵ CD45.1 carrier cells per mouse) were tail vein injected into 8-week-old sublethally irradiated (2 × 4.25 Gy; 3 hours apart) female and male CD45.1 C57/Bl6 recipient mice. Mice were maintained off tetracycline (0.5 g/liter) for 8 weeks. At week 8, mice were culled, and BM was extracted from the hips, tibias and femurs of each mouse and analyzed for leukemia burden.

IHC analysis

Femurs were collected after 4 weeks of in vivo treatment as indicated above. Four-micrometer-thick sections were prepared and fixed in formalin for 16 hours, transferred to 70% ethanol, and paraffin-embedded. Bone sections were prepared and stained with H&E, Ter119 (1:400; BD, #550565), and Ly6G (1:60,000; 2B Scientific, #BE0075-1) immunohistochemistry (IHC). Antigen retrieval was performed for 20 min using ER2 buffer from Leica and performed on-board the Leica Bond Rx autostainer.

Kinase selectivity assay

Compounds were profiled against a panel of 80 kinases at Dundee [Division Signal Transduction Therapy (DSTT)] in single-point mode at 1 μM. ULK1 versus ULK2 kinase activities were compared against selected kinases using a radiometric assay to measure the incorporation of radiolabeled ³³P into kinase substrate using a glass fiber capture filter method. Reaction conditions for ULK1 were myelin basic protein (0.2 mg/ml), 20 μM ATP (0.25 μCi per well), 50 mM tris-HCl (pH 7.5), 10 mM MgCl₂, 0.1% β-mercaptoethanol, 0.1 mM EGTA, and 0.01% bovine serum albumin. For ULK2 (DSTT, Dundee), 5 μM ATP (0.25 μCi per well) was substituted. AMPK and Janus kinase 2 (JAK2) were measured using a mobility shift assay (Caliper EZ Reader II), which measures both substrate peptide and phospho-peptide product. AMPKα2/β1/γ1 (#02-114) and Jak2 (#08-045) were sourced from Carna Biosciences. Reaction conditions were set at 1 μM peptide (AMPK peptide, peptide #10 or Jak2 peptide, peptide 22; Caliper Life Sciences), Km ATP (50 and 5 μM AMPK and Jak2, respectively), 50 mM tris-HCl (pH 7.5), 0.1 mM EGTA, 0.1% beta-mercaptoethanol, and 10 mM MgCl2. Reactions were incubated at RT and terminated with the addition of 100 mM EDTA before more than 10% of the peptide had been phosphorylated. Compounds were tested in duplicate, and values were normalized to 3% DMSO-only controls. The data were fitted to a four-parameter fit equation, and the IC₅₀ values shown are averages of at least two independent experiments.

Statistical analysis

Raw data obtained from the Seahorse XF96 flux analyzer, RT-qPCR, and FACS analysis assays were copied into Excel (Microsoft) or Prism (GraphPad Software) spreadsheets. Graphs were generated, and statistical analysis was performed with Prism. Error bars indicate SD or SEM. The number of biological replicates and applied statistical analysis are indicated in the figure legends. All raw numerical data are presented in data files S1 to S14. All mean values, SD, and SEM values of technical or biological replicates were calculated using the calculator function. Graphical representation was produced in Prism. Statistical significance for data was determined by paired or unpaired Student’s t test. For multiple comparisons, analysis of variance (ANOVA) and Dunnett’s multiple comparisons test were used (logarithmically transformed variables were used in Figs. 1E, 3, D to G, 5F, and 7M) to meet the assumption of normal distribution. For in vivo work, overall survival was monitored by Kaplan-Meier analysis, and P values were calculated using log-rank (Mantel-Cox) test. For Kaplan-Meier plots, log-rank P values are presented.

Attacking autophagy in cancer.

Autophagy is known to play a cytoprotective role in cancer; however, autophagy inhibitors thus far lack efficacy in clinical trials. ULK1 regulates autophagy and represents a targetable intervention to inhibit autophagy. Here, Ianniciello and colleagues use patient-derived chronic myeloid leukemia (CML) cells to demonstrate that ULK1 is responsible for tyrosine kinase inhibitor resistance through the induction of autophagy. They also show that combined inhibition of ULK1 and BCR-ABL1 was able to overcome targeted therapy-resistant CML in patient-derived xenograft models. Although further studies are required, this suggests that inhibition of ULK1 represents a promising strategy to overcome therapeutic resistance in CML.

Data and materials availability

All data associated with this study are present in the paper or the Supplementary Materials. Materials in this study will be made available by contacting the corresponding author. MRT403 will be made available to academic researchers by LifeArc upon reasonable request and after completion of a material transfer agreement.