Signalling mechanisms that regulate metabolic profile and autophagy of acute myeloid leukaemia cells

Abstract Acute myeloid leukaemia (AML) comprises a heterogeneous group of hematologic neoplasms characterized by diverse combinations of genetic, phenotypic and clinical features representing a major challenge for the development of targeted therapies. Metabolic reprogramming, mainly driven by deregulation of the nutrient‐sensing pathways as AMPK, mTOR and PI3K/AKT, has been associated with cancer cells, including AML cells, survival and proliferation. Nevertheless, the role of these metabolic adaptations on the AML pathogenesis is still controversial. In this work, the metabolic status and the respective metabolic networks operating in different AML cells (NB‐4, HL‐60 and KG‐1) and their impact on autophagy and survival was characterized. Data show that whereas KG‐1 cells exhibited preferential mitochondrial oxidative phosphorylation metabolism with constitutive co‐activation of AMPK and mTORC1 associated with increased autophagy, NB‐4 and HL‐60 cells displayed a dependent glycolytic profile mainly associated with AKT/mTORC1 activation and low autophagy flux. Inhibition of AKT is disclosed as a promising therapeutical target in some scenarios while inhibition of AMPK and mTORC1 has no major impact on KG‐1 cells’ survival. The results highlight an exclusive metabolic profile for each tested AML cells and its impact on determination of the anti‐leukaemia efficacy and on personalized combinatory therapy with conventional and targeted agents.

bioenergetics needs. 9,10 This metabolic reprogramming provides tumour cells with advantages necessary for sustaining their high proliferation rates, such as the rapid generation of ATP and intermediates for the synthesis of fatty acids, nucleotides and amino acids. 11 Studies in AML cell lines and human primary AML blasts correlated metabolic reprogramming with chemo-resistance showing that enhanced glycolysis decreases the AML cells sensitivity to cytarabine, while the inhibition of glycolysis potentiates the cytotoxicity of this anti-leukaemia agent. 12 Furthermore, it was also proposed that the extent of myeloblast glycolysis may be an effective method to determine the pretreatment prognosis of AML. 13 Importantly, the metabolic reprogramming in cancer cells is mainly associated with the deregulation of the major nutrient-sensing pathways: the AMP-activated protein kinase (AMPK), the mammalian target of rapamycin complex 1 (mTORC1) and the phosphoinositide 3-kinase (PI3K)/serine/threonine protein kinase B (AKT). 14 Deregulation of these signalling pathways, which enhance cellular survival and proliferation, seems to cooperate with genetic abnormalities to the pathogenesis of AML. 15 In fact, while PI3K/AKT pathway is often found activated in AML, mTORC1 appears to be active in all reported AML cases. 16,17 Both mTORC1 and AKT seem to contribute for the glycolytic metabolism of some AML cells and human primary AML blasts. 18,19 Globally, it is still debatable and controversial if the deregulation of AMPK, mTORC1 and/or AKT in AML cells would function as a tumour suppressor or promoter. 15,17,[20][21][22][23][24][25][26] Nevertheless, once activated, AMPK [27][28][29] and AKT 29,30 may control macroautophagy in mTORC1-(in)dependent pathway(s). Macroautophagy, hereafter referred as to autophagy, is a multi-step self-degradative process by which cytoplasmic content, such as long-lived proteins and superfluous/damaged organelles, is delivered to lysosomes for degradation. 31 Deregulation of autophagy has been extensively described in AML acting both as tumour promoting and suppressing. 26,[32][33][34] Therefore, the elucidation of the interconnection between the nutrient-sensing players, autophagy and energetic metabolism is of major relevance to understand cellular homeostasis and survival of AML cells. Herein, results provide evidence that different AML cells present diverse metabolic profiles.
Indeed, whereas KG-1 cells exhibited preferential OXPHOS metabolism with co-activation of AMPK and mTORC1 associated with increased autophagy flux, NB-4 and HL-60 cells displayed high intracellular ATP levels and a glycolytic profile mainly associated with AKT/mTORC1 activation and low autophagy flux. Inhibition of AKT is disclosed as a promising target for therapeutic intervention in some scenarios while inhibition of AMPK and mTORC1 has no major impact on KG-1 cells survival.

| Determination of the extracellular glucose and lactate levels
NB-4, HL-60 and KG-1 cells were plated at 0.5 9 10 6 cells/mL/well, cultured for 24 hours with or without the respective treatment(s), collected and the supernatant reserved. Measurement of the extracellular glucose and lactate levels was performed using the glucose test kit from R-Biopharm â and the lactate test kit from Spinreact â according to the manufacturer's instructions. At least, 3 independent biological replicates were performed.

| Quantification of the intracellular ATP levels
NB-4, HL-60 and KG-1 cells were plated at 0.5 9 10 6 cells/mL/well, cultured for 24 hours, collected and the pellet reserved. Intracellular ATP levels were determined using the ENLITEN ATP Assay System from Promega â according to the manufacturer's instructions. At least, 3 independent biological replicates were performed.

| Measurement of cell viability-Annexin V/PI
assay NB-4, HL-60 and KG-1 cells were plated at 0.5 9 10 6 cells/mL/well, cultured for 24 hours with or without the respective treatment(s) and collected. The cells were then washed with 800 lL of phosphate-buffered saline (PBS) followed by the addition of 100 lL of binding buffer to each sample. An incubation with 5 lL of annexin V (BD Biosciences â ) and 10 lL of propidium iodide (PI) at 50 lg/mL (Invitrogen â ) was then performed for 15 minutes at room temperature in the dark. Two hundred microlitres of binding buffer was added once again to each sample. PI signal was measured using the FACS LSRII flow cytometer (BD Biosciences â ) with a 488-nm excitation laser. The annexin V signal was collected through a 488-nm blocking filter, a 550-nm long-pass dichroic with a 525-nm band pass. Signals from 10 000 cells/sample were captured, and FACS Diva was used as the acquisition software. Analysis of the results was performed using the FlowJo 7.6 (Tree Star â ) software. At least, 3 independent biological replicates were performed.  nuclei. An epifluorescence microscope (BX61 microscope with an Olympus DP70 camera) was used to slide visualization, and images were analysed with ImageJ â Software (National Institutes of Health).

| Immunofluorescence assay
At least, 3 independent biological replicates were performed.

| Statistical analysis
All data are reported as the mean AE standard error of the mean (SEM). Statistical analysis was performed using the 2-away ANOVA and Bonferroni's post hoc tests to denote significant differences between the tested groups for the annexin V/PI approach. Student's t test was applied to compare the extracellular glucose and lactate levels between untreated and MK-2206 treated HL-60 and NB-4 cells. The one-way ANOVA and Tukey's post hoc tests were used to compare the tested groups for all the other approaches. A P-value lower than 0.05 was assumed to denote a significant difference.

| Glycolytic versus oxidative metabolism of AML cells
Metabolic reprogramming, the switch from oxidative to glycolytic metabolism, ensures a rapid production of ATP and biosynthetic precursors that confer adaptive advantages and long-term maintenance to cancer cells. 10,35 Although enhanced glucose metabolism was recently described in AML, 12 | 4809 showed that the tested AML cells display a distinct energetic metabolism, with NB-4 and HL-60 cells being highly dependent on the glycolytic metabolism while KG-1 cells appear to be more dependent on OXPHOS metabolism.

| Complexity of the mTORC1 activation network and autophagy regulation in AML cells
The reprogramming of energetic metabolism in tumour cells is mainly driven by the deregulation of the nutrient-sensing pathways. 14  concerning AMPK activation agreed with the detected intracellular ATP levels ( Figure 1D), as AMPK activation occurs in the context of energy stress (high AMP/ATP ratio) 40 and KG-1 cells were those displaying the lowest intracellular ATP levels ( Figure 1D). Knowing that AMPK may directly inhibit mTORC1 activity, 41 the concomitant AMPK and mTORC1 activation appears to indicate a dissociative AMPK-mTORC1 axis in KG-1 cells.
The orchestrated metabolic network perpetuated by the nutrient-sensing pathways converges on the control of cellular catabolic processes required to maintain cellular homeostasis, such as autophagy. 42 Given the central, although controversial, role of autophagy in the AML pathogenesis, 33,43 it is critical to comprehend not only its regulation but also its crosstalk with the metabolic signals. Therefore, autophagy was evaluated in the AML cells by immunoblotting analysis of the Atg5-Atg12 complex, LC3 processing (I and II) and LC3 puncta. 44 KG-1 cells presented the highest Atg5-Atg12 complex protein levels ( Figure 3A) associated with the highest autophagy flux, as reflected by the LC3 processing ( Figure 3B).
Immunoblotting results of the LC3 processing were corroborated by the immunostaining of LC3 showing higher number of LC3 puncta in KG-1 cells than in NB-4 and HL-60 cells ( Figure 3C), which strengthens our hypothesis that autophagy is up-regulated in KG-1 cells.
Overall, the data herein presented showed a mTORC1 activation in all tested AML cells, whereas AKT activation was mainly observed in NB-4 and HL-60 cells. Importantly, our data strongly suggest that AMPK and mTORC1 are constitutively activated in KG-1 cells. This distinct nutrient-sensing pathway activation profile is associated with F I G U R E 2 NB-4 and HL-60 cells exhibit AKT activation while KG-1 cells display a constitutive AMPK and mTORC1 co-activation. NB-4, HL-60 and KG-1 cells were maintained for 24 h in normal growth medium. (A) Activation of AKT was determined by immunoblotting analysis of phosphorylated (Ser473) AKT levels. Activation of (B) mTORC1 and (C) S6K was also assessed by immunoblotting analysis of phosphorylated (Ser2448) mTORC1 and phosphorylated (Thr389) S6K levels, respectively. Activation of (D) AMPK and (E) ACC was evaluated by immunoblotting analysis of phosphorylated (Thr172) AMPK and phosphorylated (Ser79) ACC levels, respectively. Actin was used as loading control. Densitometric analysis was performed, and bands were quantified using the ImageLab4.1 TM software. The results presented as mean AE SEM of, at least, 3 independent biological replicates. One-way ANOVA and Tukey's post hoc test were used to compare the densitometric analysis of pAKT/AKT, pmTORC1/mTORC1, pS6K/S6K, pAMPK/AMPK and pACC/ACC ratios between NB-4, HL-60 and KG-1 cells. *P < .05; ***P < .001 an up-regulation of autophagy in KG-1 cells independently of mTORC1.  Figure 4C showed that CC promoted a reduction in the AMPK activation with no major impact on the mTORC1 activity (detected by the S6K phosphorylated levels), which was accompanied by a significant decline on the autophagy flux ( Figure 4C). These data favour, once again, a dissociation of the AMPK-mTORC1 axis and reinforce AMPK as the major regulator of

NB-4 and HL-60 but has a minor impact on KG-1 cells
The impact of manipulating AKT, mTORC1 and AMPK on the survival of AML cells is still controversial. 18,20,21,43,45 Knowing that inhibition of these nutrient-sensing pathways has a major impact on autophagy and energetic metabolism of AML cells, the viability of these cells was evaluated. Data showed a significant decrease on the viability of NB-4 ( Figure 4M) and HL-60 ( Figure 4N Together with the distinct effects that these compounds had on autophagy flux ( Figure 4C) and with the independency of glycolysis (Figure 1), AMPK and mTORC1 do not seem to be an attractive target for KG-1 cells. Most probably this phenomenon reflects the conflicting metabolic signals resulting in the constitutive co-activation of AMPK and mTORC1.

| DISCUSSION
The genetic and epigenetic heterogeneity, compromising differentiation, proliferation and self-renewal of hematopoietic stem cells and myeloid progenitors, is a fundamental property of AML. This multitude of AML scenarios not only has hampering the understanding of AML's pathogenesis and classification but also the development of efficient therapeutic approaches. Different studies have been trying to establish a metabolic signature of AML cells 12,13,46,47   Overall the data presented on the inhibition of nutrient-sensing pathways and its impact on the AML cells' survival demonstrate that targeting nutrient-sensing pathways sensitizes NB-4 and HL-60 cells to chemotherapy but has a minor impact on KG-1 cells survival, which emphasizes the idea that nutrient-sensing pathways may not constitute a promising and effective therapeutic target.
In the present study, our results show that different AML cells have different energetic, metabolic and autophagy patterns that are tightly interconnected in the regulation of AML cells' survival.
Our data also point to AKT as the major regulator of energetic metabolism and autophagy in NB-4 and HL-60 cells. In KG-1 cells, the energetic metabolism and autophagy seem to be regulated by AMPK and mTORC1. These results highlight that the genetically, metabolically and clinically heterogeneity of AML should be considered and might justify the general modest growth-inhibitory effects in preclinical AML models and clinical trials of mTOR inhibition. 25,55 Furthermore, the results highlight the relevance that comparative studies implying AML cell lines have on the determination of the anti-leukaemia efficacy, particularly, of the effectiveness of combinatory therapy with conventional and new targeted agents.
The therapeutic approach to AML diseases must pass through personalized therapy adapted to the heterogeneity of this group of neoplasms.

ACKNOWLEDG EMENTS
We thank Sara Fernandes who assisted in the immunoblotting analysis.

CONFLI CT OF INTEREST
The authors declare that they have no conflict of interests concerning the contents of this article.

AUTHOR CONTRI BUTION
PL and IC designed the research study; BSM and OP analysed the data; PL, BSM and OP wrote the manuscript; OP, BSM and AT performed the research; IC and HG performed a critical revision of the manuscript. All authors have read and approved the manuscript.
F I G U R E 4 Autophagy and energetic metabolism are mainly regulated by AKT-mTORC1 axis in NB-4 and HL-60 cells and by AMPK in KG-1 cells. NB-4 and HL-60 cells were maintained for 24 h with or without MK-2206 20 lmol/L while KG-1 cells were cultured for 24 h with or without compound C (CC) 2.5 lmol/L or rapamycin (Rap) 2 lmol/L. (A-C) Activation of AMPK, AKT and S6K as well as autophagy flux were assessed by immunoblotting analysis. Activation of AMPK and AKT was evaluated by immunoblotting analysis of phosphorylated (Thr172) AMPK and phosphorylated (Ser473) AKT levels, respectively. Activation of S6K was also evaluated by immunoblotting analysis of phosphorylated (Thr389) S6K levels. Autophagy flux was assessed by immunoblotting analysis of LC3 processing (I and II; all samples were incubated for 2 h with bafilomycin A1 [10 nmol/L] before the end of the experiment to block autophagy flux and to allow LC3-II accumulation). GAPDH was used as loading control. The results are representative of, at least, 3 independent biological replicates. (D-F) Extracellular glucose and (G-I) lactate levels were determined using glucose and lactate enzymatic detection kits. (J-L) Ratio between the extracellular lactate and glucose levels ([Lactate]/[Glucose]). The results presented as mean AE SEM of, at least, 3 independent biological replicates. Student's t test was applied to compare the extracellular glucose and lactate levels as well as the [Lactate]/[Glucose] ratio between untreated and MK-2206-treated NB-4 and HL-60 cells. One-way ANOVA and Tukey's post hoc test were used to compare the extracellular glucose and lactate levels as well as the [Lactate]/[Glucose] ratio between untreated and CC-or Rap-treated KG-1 cells. *P < .05; **P < .01; ***P < .001. (M-O) Cell viability quantification was determined by flow cytometry analysis of annexin V and PI-stained NB-4, HL-60 or KG-1 cells. The results presented as mean AE SEM of, at least, 3 independent biological replicates. Annexin V/PI data were analysed using the 2-way ANOVA and Bonferroni's post hoc test. **P < .01; ***P < .001 PEREIRA ET AL.