Summary
Acute myeloid leukaemia (AML) cells interact and modulate components of their surrounding microenvironment into their own benefit. Stromal cells have been shown to support AML survival and progression through various mechanisms. Nonetheless, it is unclear whether AML cells could establish beneficial metabolic interactions with stromal cells. Here, we identify a novel metabolic crosstalk between AML and stromal cells where AML cells prompt stromal cells to secrete acetate for their own consumption. By performing transcriptome analysis and tracer-based NMR studies, we show that stromal cells present a higher rate of glycolysis, and that the secreted acetate derives from pyruvate via a reactive oxygen species (ROS)-mediated process. Our data also reveals that AML cells transfer ROS to stromal cells using gap junctions. Overall, we present a unique metabolic communication between AML and stromal cells that could be exploited as adjuvant therapy.
Introduction
Acute myeloid leukaemia (AML) is a heterogeneous clonal disease characterised by a rapid proliferation of aberrant immature myeloid cells that accumulate in the bone marrow, and eventually in the blood and other organs, severely impairing normal haematopoiesis. AML cells show a highly flexible metabolism that allows them to efficiently use a variety of nutrients to obtain energy and generate biomass (reviewed in [1]). This high metabolic plasticity confers AML cells a strong advantage against normal haematopoietic cells and has been related to AML aggressiveness. Although the metabolism in AML cells has been broadly investigated [1], fewer studies have focused on identifying metabolic alterations related to the interaction between niche and AML cells. For instance, AML cells are known to interact and modulate niche components for their own support by secreting soluble factors [2–6], via exosomes [7–9] or by establishing direct interactions, mediated by gap junctions [10] or tunnelling nanotubes [11]. These interactions with components of the niche provide AML cells with survival cues, chemoresistance and increase AML relapse [2, 4, 7, 10, 12-14]. Furthermore, it has been reported that adipocytes in the niche secrete fatty acids, which are metabolised by AML cells through β-oxidation to obtain energy, protecting AML cells from apoptosis and ROS [15,16].
As a consequence of their highly proliferative demand, leukaemic stem cells (LSCs) [17–20] and, particularly, chemotherapy-resistant AML cells [21] present abnormally high levels of reactive oxygen species (ROS) [22, 23]. How AML cells cope with high ROS levels has been intriguing and recent reports are shedding some light on whether the microenvironment plays a role in the redox metabolism of AML cells. For instance, it was reported that Nestin+ bone marrow mesenchymal stem cells (BMSCs) support AML progression by increasing the bioenergetic capacity of AML cells and providing them with glutathione (GSH)-mediated antioxidant defence to balance the excess ROS [14]. Similarly, a recent study showed that co-culturing BMSCs with AML cells leads to a decrease in AML ROS levels due to an activation of the antioxidant enzyme GPx-3 in AML cells [24].
Our work provides new insight into the metabolic and redox crosstalk between AML and stromal cells, revealing a new metabolic interaction between AML and stromal cells. By combining transcriptomic and nuclear magnetic resonance (NMR) data, our results demonstrate that stromal cell metabolism is rewired in co-culture giving higher glycolysis and pyruvate decarboxylation, leading to acetate secretion. Our results also show that AML cells are able to transfer ROS to stromal cells by direct interaction through gap junctions and that this ROS can be used by stromal cells to generate and secrete acetate, which is utilised by AML cells as a biofuel. Targeting ROS transfer via modulation of gap junctions to suppress the antioxidant protection provided by stromal cells could serve as an adjuvant therapy to eradicate AML.
Results
Co-culturing AML and stromal cells in direct contact triggers acetate secretion by stromal cells
We first sought to determine whether interactions between AML and stromal cells in co-culture would result in differences in the consumption or production of extracellular metabolites. For this purpose, three AML cell lines (SKM-1, Kasumi-1 and HL-60) and the MS-5 mouse stromal cell line were cultured separately and in co-culture, and the metabolic composition of the extracellular medium was analysed by 1H-NMR and compared at different time points. The most striking difference found in co-culture compared to cells cultured separately was an increased secretion of acetate, which was common for the three cell lines used (Fig. 1A and B). Moreover, only stromal cells secreted acetate to a lower extent when cultured alone whereas AML cells did not secrete any acetate when cultured alone. Altogether, these findings suggest that acetate secretion is indeed a result of a direct interaction between AML and stromal cells. In addition, the observation that only stromal cells secrete acetate under these conditions suggests that stromal cells could be responsible for the increased acetate secretion found in co-culture.
Moreover, we examined the levels of other common extracellular metabolites, including glucose, lactate, glutamate and glutamine, as shown in Fig S1. A higher consumption of glucose, corresponding to a higher secretion of lactate, was observed in AML and stromal cells in co-culture compared to single cultures, suggesting a higher glycolytic flux in co-culture. However, the levels of glucose consumption and lactate production in co-culture were similar to the sum of the glucose consumption and lactate production levels of the AML and stromal cells in single cultures, suggesting that the overall increase in glycolysis in co-culture was just a result of culturing both cell types together. Additionally, we observed no variation in glutamate and glutamine levels suggesting that these metabolites are not involved in interactions that result from co-culture and are utilised depending on their availability.
The metabolic differences found in co-culture, could be due to altered proliferation under these conditions. To determine whether this was the case, a CFSE-based proliferation assay was performed in which AML cells and not stromal cells were stained and their growth was compared when cultured separately vs in co-culture. Over 48 hours, none of the AML cell lines tested presented differences in their proliferation rates (Fig S2A), thus, confirming that the metabolic changes found in co-culture were caused by a mechanism independent from changes in proliferation.
Next, we aimed to determine whether cell-to-cell contact could play a role in the increased acetate secretion found in co-culture. We decided to co-culture AML and stromal cells separated by a permeable membrane, allowing cells to share the extracellular medium but impeding cell-to-cell contact. Co-culturing cells using a permeable membrane blocked the increased acetate secretion observed under direct contact conditions (Fig. 1C). In fact, cells in co-culture presented lower levels of acetate than MS-5 cells cultured alone revealing that direct cell-to-cell contact is required for acetate secretion in co-culture.
We further investigated whether increased acetate secretion could take place in primary co-cultures using AML cells derived from patients and whether acetate secretion in co-culture could be specific of AML cells. To address this question, we isolated the CD34+ population of four primary AML patient samples, cultured them alone vs in co-culture with MS-5 cells, and analyzed the composition of the extracellular medium at different time points. We found that three out of four primary AML samples presented higher levels of acetate when co-cultured with MS-5 cells compared to both MS-5 and AML cells cultured alone (Fig. 1D).
Additionally, to assess whether healthy haematopoietic cells might cause a similar increase in acetate secretion, we performed the same experiment using CD34+ cells from three independent healthy donors (Fig. 1E). None of these showed an increased acetate secretion in co-culture suggesting that acetate secretion is specific of AML cells in co-culture.
To further examine whether this increased acetate secretion was specific for the interaction of AML cells with stromal cells, we co-cultured AML cells with the cervical cancer cell line HeLa and compared the levels of acetate of each cell type cultured alone. We found that there was no acetate secretion in co-culture nor by HeLa and AML cells cultured alone (Fig. S2B) suggesting that increased acetate secretion could be specific of an AML-stromal interaction.
Considering that our previous data seemed to indicate that MS-5 cells were responsible for acetate secretion when co-cultured with AML cell lines and primary AML-derived cells, we decided to investigate how the levels of extracellular acetate would vary after separating cells from co-culture. The three AML cell lines were co-cultured with MS-5 cells prior to separation and culture in the same spent media. Extracellular acetate levels in previously co-cultured MS-5 cells followed a similar trend as before separation, suggesting that it is most likely that MS-5 cells are responsible for the increased acetate secretion found in co-culture (Fig. 1F). Moreover, AML cells did not follow this trend as they either maintained the levels of acetate seen prior to separating the cells (Kasumi-1 and HL-60) or presented only a moderate increase (SKM-1) after being separated from co-culture.
MS-5 cells secrete acetate in co-culture whilst AML cells consume and use it to flux the TCA cycle
Following the finding that stromal cells are responsible for acetate secretion in co-culture, we decided to determine whether AML cells might be able to metabolise the secreted acetate. We first sought to define the concentration of acetate generated and present in the extracellular medium in co-culture. For this, we compared a sample of extracellular medium from a co-culture of SKM-1 and MS-5 cells after 24 hours to a calibration curve (Fig. S2C) allowing us to determine that the concentration of acetate in co-culture was approximately of 3 mM. We then investigated whether SKM-1 cells can consume acetate both in physiological and co-culture concentrations. In both cases SKM-1 cells consumed acetate after 48 hours, and in physiological conditions also after 24 hours (Fig. 2A).
We used a tracer-based approach using [2-13C]acetate to assess whether SKM-1 or MS-5 cells can utilize acetate (Fig. 2B,C and Fig. S3A). Both cell types consume 13C labelled acetate as observed in NMR spectra (Fig. 2C). However, only SKM-1 cells metabolise acetate, which is found in several TCA cycle related metabolites, including aspartate, citrate, glutamate, 2-oxoglutarate, glutathione and proline, as well as in acetylcarnitine. Acetylcarnitine had been reported to be produced by cells when large amounts of acetyl-CoA are present in the mitochondria [25, 26], which could be in line with this experiment as a 4 mM concentration of [2-13C]acetate was used. Overall, this data suggests that acetate in co-culture could be preferentially imported and used by SKM-1 cells but not by MS-5 cells.
In order to determine whether AML cells can import and metabolise the secreted acetate in coculture, we co-cultured AML and MS-5 cells for 24 hours, we then added [2-13C]acetate to the spent extracellular medium and cultured cells for 8 hours before analysing the intracellular metabolites in each cell type (Fig. 2D). We found that only AML1 cells presented 13C labelling in intracellular metabolites (Fig. 2E, Fig. S3B). Moreover, the metabolisation pattern for all AML cells was similar to that observed for SKM-1 cells alone (Fig. 2C), where label was incorporated into TCA cycle-related metabolites. Overall, these results revealed that AML cells uptake and utilise the acetate secreted by stromal cells in co-culture as a substrate to feed into the TCA cycle.
Transcriptomic data highlights a metabolic rewiring of stromal cells in co-culture characterised by upregulation of glycolysis and downregulation of pyruvate dehydrogenase
After establishing that stromal cells are responsible for acetate secretion (Fig. 2A) and that AML cells consumed the secreted acetate (Fig. 2F), we sought to elucidate the mechanism behind acetate secretion by MS-5 cells in co-culture. Thus, we set out to perform global gene expression profiling by RNA-seq comparing cells cultured alone and in co-culture (SUPPLEMENTAL FIG 4A). With this approach we identified 587 differentially expressed genes (q-value<0.1) (SUPPLEMENTAL FIG 4B); with 476 genes being upregulated and 111 genes being downregulated in co-culture.
Following clustering of differentially expressed genes, Gene set enrichment analysis (GSEA) was performed (Fig. 3A) revealing a positive correlation with the expression of genes that are part of the glycolysis pathway (MSigDB: M5937) as well as the reactive oxygen species pathway (MSigDB: M5938) (FIG.3B).
An examination of the genes involved in the glycolysis pathway revealed a major upregulation of MS-5 cells in co-culture (Fig. 3C). Genes involved in glucose transport (Slc2a1 and Slc2a4) and glucose breakdown to pyruvate (Pgm1, Hk2, Gpi1, Pfkl, Gapdh, Pgk1, Pgam1-2, Eno1-1b-2 and Pkm) were upregulated, including the gene encoding for the 6-phosphofructo-2-kinase/fructose-2,6-biphosphatase 3 (Pfkfb3), a well-known activator of glycolysis.
In contrast, individually examining genes related to pyruvate metabolism revealed that the genes related to pyruvate conversion to acetyl-CoA were downregulated (Fig. 3C). Pyruvate transport into the mitochondria (Mpc1-2) and several pyruvate dehydrogenase (PDH) complex-related genes (Pdha1, Pdhb and Pdhx), involved in the conversion of pyruvate to acetyl-CoA, remained largely unaltered or were slightly downregulated by co-culture. Additionally, the pyruvate dehydrogenase kinases (Pdk1, Pdk2 and Pdk4), which inhibit the activity of the PDH complex, were found to be upregulated by co-culture. Interestingly, the acetyl-CoA synthetases (ACSs), which can generate acetyl-CoA from acetate but have also been reported to perform the reverse reaction, encoded by Acss2 and Acss3, were found to be slightly upregulated in co-culture. Pyruvate can also be converted to 2-oxaloacetate via pyruvate carboxylase (Pcx), to lactate by lactate dehydrogenase (Ldha), and to alanine via alanine transaminase (Gpt). Only Ldha was found to be moderately upregulated in co-culture. However, we did not observe a substantial increase in lactate production experimentally in co-culture. Altogether, transcriptomic data suggests that MS-5 cells in co-culture present a major upregulation of glycolysis and downregulation of the PDH complex.
To examine whether glycolysis is upregulated at a metabolic level in MS-5 cells in co-culture at the metabolic level and whether acetate could derive from glucose, we decided to perform [U-13C]glucose tracing on AML and MS-5 cells cultured alone and in co-culture and analysed the label incorporation in glycolysis-related extracellular metabolites (Fig. 3D). For the three AML cell lines tested, extracellular acetate presented significantly higher label incorporations from [U-13C]glucose in co-culture compared cell types cultured alone, providing evidence that the secreted acetate in co-culture derives from glucose. Lactate and alanine, which can be synthesised from pyruvate, did not show significant increases in label incorporation from [U-13C]glucose in co-culture compared to each cell type cultured alone for all the AML cell lines. Overall, these results are in agreement with the transcriptomic data (Fig. 3C), highlighting that glucose metabolism is upregulated in co-culture but also confirming that acetate derives from glucose metabolism.
AML cells rewire stromal cell metabolism transferring ROS via gap junctions to obtain acetates
Tracer-based data on MS-5 cells in co-culture revealed that the acetate secreted in co-culture derives from glucose (Fig. 3D). However, transcriptomic data on MS-5 cells in co-culture suggested that acetate did not derive from pyruvate via acetyl-CoA (Fig. 3C), thus, suggesting that a different mechanism could be responsible. An alternative mechanism of acetate synthesis involving a non-enzymatic oxidative decarboxylation of pyruvate into acetate had previously been described [27-29]. This mechanism was reported to be mediated by ROS in mammalian cells and was linked to cells prone to overflow metabolism under the influence of high rates of glycolysis and excess pyruvate. Hence, we decided to investigate whether ROS might play a role in acetate secretion in our co-culture system.
We first modulated ROS levels in SKM-1 and MS-5 cells cultured alone and in co-culture and measured acetate production. Hydrogen peroxide was used to increase ROS levels and N-acetylcysteine (NAC) was used as a ROS scavenger. Extracellular acetate levels were measured by 1H-NMR after 24 hours. Increasing ROS levels with peroxide resulted in a significant increase in acetate secretion, in particular in cells in co-culture (Fig. 4A). Additionally, when the ROS scavenger NAC was used a significant decrease in the levels of acetate in both MS-5 cells cultured alone and in the co-cultured cells was observed. Interestingly, the production of acetate in cells in co-culture treated with NAC recovered the phenotype of MS-5 cells cultured alone, indicating that acetate synthesis in MS-5 cells in co-culture is mediated by ROS.
Next, we compared intracellular ROS levels in AML and stromal cells in co-culture and cultured alone by labelling cells with the H2DCFDA fluorescent dye. Analysis of fluorescence revealed that ROS levels in the three AML cell lines used were significantly decreased in co-culture, whilst in the stromal cells ROS levels were significantly increased in two of the three co-cultures (Fig. 4B). These results suggested that AML cells might transfer ROS to stromal cells. We also performed the same experiment using three primary AML samples to corroborate the previous result. Fluorescence analysis showed in decreased levels of ROS in AML samples in co-culture and increase in MS-5 cells ROS levels in co-culture for the three primary AML samples analysed (Fig. 4C), indicative of ROS transfer from AML cells to stromal cells in co-culture.
To further test whether AML cells might transfer ROS through a contact-depend mechanism, we compared the intracellular ROS levels in AML and stromal cells cultured alone and co-cultured without direct contact using a permeable membrane. Analysis of fluorescence in both cell types revealed that ROS levels remained unaltered in contact-free co-cultures (Fig. 4D), indicating that ROS transfer could only occur via a contact-dependent mechanism.
As it had been previously reported that haematopoietic stem cells can transfer ROS to stromal cells via gap junctiond to prevent senescence [30], we decided to examine gap junction genes in the transcriptome of MS-5 cultured alone and in co-culture. When individually examining the gap junction genes in MS-5 cells cultured alone vs co-culture, we found several gap junction genes upregulated in co-culture such as Gja5, Gja8, Gjb5 and Gjc2 (Fig. 4E). These results suggested that AML cells might establish gap junction interactions to transfer ROS to MS-5 cells when in co-culture. To test this hypothesis, we used the calcein-AM dye, which can only be transferred via gap junctions [10]. We labelled stromal cells with calcein-AM, cultured them with unlabelled AML cells and analysed the fluorescence of AML cells after three hours. We found that, for the three AML cell lines tested, the percentage of cells that had incorporated the calcein-AM dye from MS-5 cells was higher than 80% of the total number of cells (Fig. 4F, Supplemental Fig. 5), indicating that AML cells can establish gap junctions with stromal cells when co-cultured in direct contact.
Next, we decided to confirm that AML cells can transfer ROS via gap junctions by inhibiting the gap junctions using carbenoxolone (CBX) [10, 31, 32]. We first confirmed that efficiency of inhibition by analysing the calcein-AM dye transfer in a control and treated co-culture of SKM-1 and MS-5 cells. The CBX treatment reduced the calcein-AM transfer more than 60% in co-culture (Fig. 4G). We then compared intracellular ROS levels in both AML and stromal cells treated with CBX and control. CBX treatment abrogated the decrease in ROS levels in AML cells and the increase of ROS levels in stromal cells, indicating that ROS transfer was inhibited in CBX treated co-cultures. Overall, this result provides strong evidence that ROS transfer occurs via gap junctions established between AML and stromal cells.
Discussion
AML cells are known to interact and remodel niche cells through various mechanisms, including the secretion of soluble factors, cytokines or metabolites, resulting in a better support of AML cells at the expense of normal haematopoiesis. Yet, metabolic crosstalk between AML and stromal cells has not been reported before. Herein, we have identified a novel metabolic communication between AML and stromal cells mediated by acetate. Our data suggests that AML cells can modulate stromal cells into secreting acetate in co-culture, by rewiring stromal cell metabolism, and can utilise back the secreted acetate to feed their TCA cycle and obtain energy. Mechanistically, our data revealed that acetate secretion involves not only the higher glycolytic rate but also the non-enzymatic ROS-mediated conversion of pyruvate to acetate. Furthermore, our data indicates that AML cells can diminish their ROS levels by establishing gap junctions with stromal cells facilitating ROS transfer to stromal cells.
Studying the interactions between cancer cells and their microenvironment in terms of metabolism has become an exciting new field. Our data indicates that AML cells can influence the metabolism of stromal cells causing increased acetate secretion, which was not observed in healthy PB counterparts. We presented several lines of evidence suggesting that stromal cells and not AML cells are responsible for acetate secretion: (i) stromal cells cultured alone secreted acetate, whilst AML cell lines and primary AML cells did not; (ii) after separating stromal cells from co-culture with AML cell lines, stromal cells continued to secrete acetate at a similar rate to that observed in co-culture; and, (iii) glycolysis was upregulated in stromal cells in co-culture, and glucose was found to be the precursor of the secreted acetate in co-culture.
Interestingly, acetate has been reported as an alternative fuel for cancer cells [33, 34], especially under low oxygen conditions or lipid depletion [35-37], but it has not previously been reported to be part of a crosstalk communication between any type of cancer cells and their microenvironment. Nonetheless, other monocarboxylate metabolites, such as lactate and alanine, have been reported to participate in different types of metabolic interplay between stromal and cancer cells [38-40]. Our report reveals that aside from lactate and alanine, other monocarboxylate metabolites, such as acetate, can be utilized by leukaemic cells as a biofuel. Acetate usage to feed the TCA cycle had already been described in AML [41] and other types of cancer [36, 42], However, this is, to our knowledge, the first report of AML cells utilising acetate secreted by stromal cells in co-culture.
We have also proposed a mechanism for the altered stromal cellular metabolism, involving increased glycolysis and the ROS-mediated synthesis of acetate from pyruvate. Glycolysis was found to be an upregulated pathway in the transcriptional data of stromal cells in co-culture, and tracer-based metabolism using [U-13C]glucose, demonstrated that more label was incorporated in pyruvate and acetate in stromal cells in co-culture compared to either AML or stromal cells cultured alone. Moreover, enrichment of hypoxia genes and elevated Pdk expression have been reported to be related to higher glycolytic activity [43–45]. Similarly, other cancer cells [46–49] have also been shown to modulate stromal metabolism by increasing higher glycolysis.
We have also provided evidence for a link between ROS and acetate synthesis in which acetate is synthesised from pyruvate in the presence of ROS. Increasing ROS levels in co-culture yielded higher acetate secretion and lowering ROS levels recovered the acetate secretion phenotype of MS-5 cells cultured alone.
AML cells are known to exhibit high levels of ROS [50, 51]. However, our data has shown that AML cells in co-culture present lower levels of ROS than AML cells cultured alone, which is in agreement with recent studies in primary AML and mesenchymal stromal cell co-cultures [14, 24]. The current mechanisms to describe this phenomenon involve mitochondrial transfer and activation of glutathione-related antioxidant pathways [14, 24], although previous data on haematopoietic stem cells (HSCs) revealed that HSCs can directly transfer ROS via gap junctions to stromal cells [30]. Interestingly, our data showed that the decrease in ROS levels was counteracted by inhibiting contact using a permeable membrane. We also found several gap junction genes upregulated in stromal cell types in co-culture. Moreover, we prevented ROS transfer from AML to stromal cells by using the gap junction inhibitor CBX [10, 31, 32], Thus, our results suggest that ROS transfer via gap junctions, at least partially, mediates the mechanism behind AML cells presenting lower levels of ROS in co-culture.
Overall, this work reveals a unique metabolic communication between AML and stromal cells that involves acetate as the main crosstalk metabolite. We showed that AML cells are capable of modulating the metabolism of stromal cells by transferring ROS via gap junctions resulting in an increased secretion of acetate, which accumulates in the extracellular medium. Furthermore, we found that AML cells consume the secreted acetate and use it as an energy source, by fluxing it into the TCA cycle. We believe our findings provide a better understanding of how AML cells communicate with stromal cells and could serve as a basis for the development of novel therapeutic strategies to target AML cells by modulating gap junction formation as an adjuvant therapy.
Author Contributions
N.V-L. conceived and performed the experiments, acquired and analyzed the data, interpret the data and wrote the manuscript. R.A., E.G., A.C., A.E., F.S., M.R., performed the experiments and edited the manuscript. G.P. analyzed the RNA-seq data. J.B helped with the experimental design. S.P. and M.R. provided with patient samples. J.J.S. provided patient samples, helped with the experimental design, interpretation, and critical discussion of the data and edited the manuscript. U.G and P.G. conceived and performed the experiments, acquired and interpreted the data, wrote the manuscript, and managed the project.
Methods
Cell lines
The human AML cell lines (SKM-1, Kasumi-1 and HL-60), the mouse stromal cell line (MS-5) and the human cervical cancer cell line (HeLa) were cultured in RPMI 1640 media supplemented with 15% (v/v) FBS, 2 mmol/L L-glutamine and 100 U/ml Penicillin/Streptomycin (all from Thermo Fisher Scientific). Co-cultures were plated in a 4:1 AML-stromal ratio, and 750.000 cells/ml density of AML cells over confluent stromal cells. Prior to cell extraction for NMR, RNA collection or protein extraction, a suspension of 10 million AML cells was collected, the stromal layer was washed with PBS and was subject to mild trypsinisation with 1:5 dilution of 0.25% Trypsin 1 mM EDTA (Thermofisher) to remove attached residual leukaemic cells before completely detaching stromal cells with 0.25% Trypsin-EDTA.
Primary patient samples
AML and PBMC primary specimens’ procedures were obtained in accordance with the Declaration of Helsinki at the University Medical Center Groningen, approved by the UMCG Medical Ethical Committee or at the University Hospital Birmingham NHS Foundation Trust, approved by the West Midlands-Solihull Research Ethics Committee (10/H1206//58).
Additional information about the primary AML samples used in this study can be found in Annex 1.
Peripheral blood and bone marrow samples from AML patients and healthy donors were obtained in heparin-coated vacutainers. Mononuclear cells were isolated using Ficoll-Paque (GE Healthcare) and stored at −80°C.
For AML1-2 and PBMC1-2, samples were thawed and resuspended in newborn calf serum (NCS) supplemented with DNase I (20 Units/mL), 4 mM MgSO4 and heparin (5 Units/mL) and incubated on 37°C for 15 minutes (min). For AML1 and PBMC1, cells were sorted after thawing by fluorescence-activated cell sorting (FACS) for the CD34+CD38-population using 10 μL of anti-CD34 APC (560940, BD Biosciences), 10 μL of anti-CD38 FITC (555459, BD Biosciences) and 10 μL of DAPI (D1306, Thermo Fisher Scientific) per 10 million cells. Cells were sorted using a Sony SH800S (Sony) sorter. For AML2 and PBMC2, cells were thawed and the CD34+ population was sorted by magnetic-activated cell sorting (MACS) using 10 μL of anti-CD34 microbeads (130-046-702, Miltenyi) per million of expected CD34 cells following manufacturer’s protocol.
AML1-2, PBMC1-2 and MS-5 cells were cultured alone and in co-culture in a 4:1 AML/PBMC-stromal ratio and 500,000 cells/mL density in a-MEM (Gibco) with 12.5% (v/v) FCS (Sigma-Aldrich), 1% (v/v) Pen/Strep (Thermo Fisher Scientific), 12.5% (v/v) Horse serum (Invitrogen), 0.4% (v/v) β-mercaptoethanol (Merck Sharp & Dohme BV) and 0.1% hydrocortisone (H0888, Sigma-Aldrich). For AML1 and AML2 the medium was supplemented with 0.02 μg/mL of IL-3 (Sandoz), 0.02 μg/mL NPlate (Amgen) and 0.02 μg/mL of G-CSF (Amgen). For PBMC1 and PBMC2 the medium was supplemented with 100 ng/ml of human SCF (255-SC, Novus Biologicals), 100 ng/ml of NPlate, 100 ng/ml of FLT3 ligand (Amgen) and 20 ng/ml of IL-3. Samples of medium were collected at 0 and 48 hours.
AML3 - 7 and PBMC3 were thawed and kept in culture for 16-24 hours in Stem Span H3000 media (STEMCELL Technologies) supplemented with 50 μg/ml ascorbic acid (Sigma-Aldrich), 50 ng/ml human SCF (255-SC-010, R&D Systems), 10 ng/ml human IL-3 (203-IL-010, R&D Systems), 2 units/ml human-erythropoietin (100-64, PeproTech), 40 ng/ml insulin-like growth factor 1 (IGF-1) (100-12, PeproTech), 1 μM dexamethasone (D2915, Sigma-Aldrich) and 100 μg/ml primocin (ant-pm-2, lnvivogenCD34+ cKit+ cells were purified using magnetic microbeads (130-046-702 (CD34) and 130-091-332 (CD117), Miltenyi Biotec). Cells were cultured in the supplemented Stem Span media for 24 hours prior to co-culture setting. Co-cultures were plated with a 4:1 leukaemic to stromal cells ratio and a 300,000 cells/ml density in supplemented Stem Span medium. Samples of leukaemic/healthy cells and MS-5 cells alone were also cultured in supplemented Stem Span media as controls. Samples of medium were collected at 0 and 48 hours.
Proliferation analysis using CFSE
The CellTrace™ carboxyfluorescein succinimidyl ester (CFSE) Cell Proliferation Kit (C34570, Invitrogen) was used to assess proliferation of AML cells following the manufacturer’s protocol. AML cells were stained and their fluorescence was assessed before dividing the bulk of cells into culturing them alone or with MS-5 cells for 48 hours. Small aliquots of cells after 24 and 48 hours were analysed by flow cytometry. Flow cytometry analysis was carried out in a CyAn ADP flow cytometer (Beckman Coulter). Data analysis was performed using the FlowJo software package (BD).
Carbenoxolone treatment
Carbenoxolone (CBX) disodium salt (Sigma-Aldrich, C4790) was prepared fresh at 200 μM in cell culture medium. Cells were resuspended in the CBX medium and cultured alone or in co-culture for 24 hours prior to ROS or Calcein-AM dye transfer experiments.
Cellular ROS measurements using DCFH-DA
Cellular ROS was measured by incubating cells with 100 μM 2⍰,7⍰-Dichlorofluorescin diacetate (DCFH-DA) (D6883, Sigma-Aldrich) in Hank’s Balanced Salt Solution (HBSS) (Thermo Fisher Scientific) at 37°C for 30 min protected from light. Cells were then harvested and stained with 5 μg/μL antihuman CD33 eFluor 450 (eBioscience, P67.6) for 30 min at 4°C before flow cytometry analysis as previously described.
ROS-related treatments with H2O2 and NAC
SKM-1 and MS-5 cells cultured alone and in co-culture were incubated for 24 hours in 50 μM H2O2 (Merck) complete cell culture medium, 5 mM N-acetylcysteine (NAC) (106425, Merck) complete cell culture medium or control medium. Samples of medium were collected at 0 and 24 hours and kept at −80°C.
Calcein-AM dye transfer assay
Functional gap junction presence was evaluated using the fluorescent dye Calcein-AM green (Invitrogen, C1430) adapting a previously established protocol [10]. MS-5 cells were stained with 500 nM Calcein-AM dye in complete cell culture medium for 1 hour at 37°C. Stained cells were washed with serum-free medium for 30 min at 37°C before being co-cultured with AML cells for 3h. AML cells were then harvested and stained with 5 μg/pL anti-human CD33 eFluor 450 (eBioscience, P67.6) for 30 min at 4°C before flow cytometry analysis as previously described. Calcein-AM dye transfer was quantified as the frequency of CD33+ and Calcein-AM+ cells.
Tracer-based NMR experiments
[U-13C]Glucose (CC860P1, CortecNet) was added to RPMI 1640 medium without glucose (11879020, Merck) to a final concentration of 2 g/L (as in the complete cell culture medium) and was supplemented as usual with 15% (v/v) FBS, 2 mmol/L L-glutamine and 100 U/mL Pen/Strep. The medium was prepared fresh and was filtered with a 0.2 μm syringe filter (Sartorius) before each experiment.
4 mM sodium [2-13C]acetate (279315-1G, Sigma-Aldrich) was added to complete cell culture medium and the medium was filtered with a 0.2 μm syringe filter before each experiment. Cells were incubated for the time indicated in each experiment before separation of cells and/or metabolite extraction. Unlabelled samples were prepared as a control for 2D-NMR experiments and to measure acetate consumption by adding unlabelled sodium acetate trihydrate (1.37012, Merck) to complete cell culture medium before metabolite extraction or collection of medium samples.
Metabolite extraction
Suspension cells and detached adherent cells were washed with PBS before being rapidly resuspended in 400 μl of HPLC grade methanol pre-chilled on dry ice. Samples were transferred to glass vials and were subject to 10:8:10 methanol-water-chloroform extraction as described in (Saborano 2019). Polar phase samples were evaporated in a SpeedVac concentrator and were subsequently kept at −80°C.
Sample preparation for NMR
Medium samples were mixed with a 10% of 1 M phosphate D2O buffer with 3 mM NaN3 and 5 mM TMSP ((3-trimethylsilyl)propionic-(2,2,3,3-d4)-acid sodium salt) (all from Sigma) and transferred to 3 mm NMR tubes (Cortecnet).
Dried polar extracts for tracer-based NMR experiments were reconstituted in 50 μL of 0.1 M phosphate buffer in 100% D2O with 3 mM NaN3 and 0.5 mM TMSP. Samples were sonicated for 10 min and transferred to 1.7 mm NMR tubes (Cortectnet) with the Micro Pipet System 1.7. Samples were prepared fresh before the acquisition.
NMR acquisition and analysis
All NMR data was acquired on Bruker 600MHz spectrometers equipped with Avance-III consoles using cooled Bruker SampleJet autosamplers.
For media samples, a 5mm triple resonance cryoprobe (TCI) z-axis pulsed field gradient (PFG) cryogenic probe was used, and for cell extracts, a TCI 1.7mm z-PFG cryogenic probe was used. Probes were equipped with a cooled SampleJet autosampler (Bruker) and automated tuning and matching.
For medium samples, spectra were acquired at 300K using 1D 1H-NOESY (Nuclear Overhauser effect spectroscopy) pulse sequence with pre-saturation water suppression (noesygppr1d, standard pulse sequence from Bruker). The spectral width was 12 ppm, the number of data points was TD 32,768, the interscan delay was d1 4 sec and the NOE mixing time was d8 10 msec. The 1H carrier was set on the water frequency and the 1H 90º pulse was calibrated at a power of 0.256 W and had a typical length of ca 7-8 μs. 64 scans and 8 dummy scans were acquired and the total experimental time was 7.5 min. Spectra were processed using the MetaboLab [52] software within the MATLAB environment (MathWorks).). A 0.3Hz exponential apodization function was applied to FIDs followed by zero-filling to 131,072 data points was applied prior to Fourier transformation. Chemical shift was calibrated by referencing the TMSP signal to 0 ppm and spectra were manually phase corrected. The baseline was corrected by applying a spline baseline correction, the water and edge regions of the spectra were excluded before scaling the spectra using a probabilistic quotient normalization (PQN). Chenomx 7.0 software (Chenomx Inc.) and the human metabolome database (HMDB) were used to assign the metabolites present in the acquired spectra. Metabolite signal intensities were obtained directly from the spectra and were normalised to a control medium sample obtained at time 0 hours.
For 13C-filtered 1H-NMR experiments, [U-13C]glucose labelled medium samples were analysed with 13C-filtered 1H-NMR spectroscopy as described in [41], Spectra were acquired at 300 K using a double gradient BIRD filter pulse sequence developed in-house [41], A pulse program combining the 1H[12C] and the all-1H experiments in scan-interleaved mode was used. The difference between the two FIDs gave the 1H[13C] signal. The spectral width was 12 ppm, the number of data points was 16,384 in each dimension and the relaxation delay was 5.3 sec. 256 scans with 64 dummy scans were acquired and the experimental time was 15 min. 13C-filtered 1H-NMR spectra were processed in Topspin 4.0.5 (Bruker). Spectra were zero filled to 32,768 data points before Fourier transformation. Phase correction was applied to the 1H[-12C] and the all-1H spectra and the difference 1H[13C] spectrum was obtained by aligning on chemical shift. Metabolites were selected in the 1H[13C] spectrum and integrated in all the spectra (H[C], H[C] and all-H) using the tool available in the program. Label incorporations (13C percentages) were calculated by dividing the signal areas obtained in the 1H[13C] spectrum by the ones obtained in the all-1H spectrum.
For 1H-13C-HSQC experiments, spectra were acquired using a modification of the hsqcgphprsp from Bruker, with additional gradient pulses during the INEPT echo periods and using soft 180º pulses for 13C. For the 1H dimension, the spectral width was 13.018 ppm with 1024 complex points. For the 13C dimension, the spectral width was 160 ppm with 2048 complex points. 2 scans were acquired per spectrum and with an interscan delay of 1.5 sec. Non-uniform sampling (NUS) with a 25% sampling schedule (generated using the Wagner’s schedule generator (Gerhard Wagner Lab, Harvard Medical School) with a tolerance of 0.01 and default values for the other parameters) was used with 4096 increments yielding 8192 complex points after processing was. The total experimental time was 4 hours. Spectra were processed and phased with NMRPipe (National Institute of Standards and Technology of the U.S.) (Appendix 8.1). MetaboLab was used to reference using the signal for the methyl group of L-lactic acid, at 1.31/22.9 ppm in the 1H and 13C dimensions, respectively. Identification of metabolites in 2D spectra was carried out using the MetaboLab software which includes a chemical shift library for ca. 200 metabolites. Intensities were obtained from signals in the spectra and were corrected for differences in cell numbers contributing to the sample as follows: A 1H-NMR spectrum was acquired for each sample and the total metabolite area of this spectrum after removal of the water and TMSP reference signal was calculated in MetaboLab. The intensities in the 2D spectra were then divided by the total metabolite area of the corresponding 1H-NMR spectrum. To obtain the % of 13C in a metabolite, the normalized intensity of a certain carbon in the labelled sample was divided by the normalized intensity of the same carbon in the unlabelled sample and was multiplied by the natural abundance of 13C (1.1%).
RNA extraction and sequencing
MS-5 cells were cultured alone or in co-culture with SKM-1 cells for 24 hours. Cells were separated and washed with PBS prior to RNA extraction with TRIzol (Gibco) according to the manufacturer’s protocol. RNA was purified with RNeasy Plus Micro kit. Samples were sent to Theragenetex to be sequenced with Novaseq 150bp PE with 40M reads.
Real-time PCR
Samples of RNA from SKM-1 and MS-5 co-cultures were collected and extracted using Trizol (Invitrogen) following manufacturer’s protocol. cDNA was synthesized using the M-MLV reverse transcriptase (Promega) according to manufacturer’s instructions. For gene expression analysis, qRT-PCR of Hk2 (NM_013820), Pdhx (NM_175094), Pdk1 (NM_172665) and Pdk2 (NM_133667) (all KiCqStart™ primers KSPQ12012, Sigma Aldrich) were carried out using the SYBRGreen Master mix (Thermo Fisher Scientific) and qRT-PCR of B2m (NM_009735.3, TaqMan® assays, Thermo Fisher Scientific) was performed using TaqMan PCR Master Mix (Thermo Fisher Scientific). Reactions were performed in a Stratagene Mx3000P and run in triplicate. Relative gene expression was calculated following the 2-ΔΔct method relative to the expression of B2m.
RNA sequencing
RNA samples of MS-5 cells cultured alone and in co-culture with SKM-1 cells extracted using Trizol were purified using the RNeasy Plus Micro kit (Qiagen) according to manufacturer’s protocol. Transcriptome analysis was performed by Theragen Etex Co., LTD. (www.theragenetex.net). cDNA libraries were prepared with the TruSeq Stranded mRNA Sample Prep Kit (Illumina) and RNA sequencing was performed in a HiSeq2500 platform (Illumina). Quality control metrics were obtained with FastQC 0.11.7 software (Babraham Bioinformatics). To quantify transcript abundances using the Kallisto 0.43.0 software (Patcher Lab), read counts were aligned to the GRCm38 mouse reference genome cDNA index (Ensembl rel.67). Gene-level differential expression analysis was carried out with the R statistical package Sleuth 0.30.0 comparing the expression of cells cultured alone vs co-culture. Differentially expressed genes were calculated using the Wald statistical test, correcting for multiple testing with the Benjamin-Hochberg method. The false discovery rate (FDR) threshold was set at 1% (q-values < 0.01). Ensembl gene transcripts were annotated to Entrez IDs and official gene symbols with the R statistical package BioMart 2.40.3. To normalise for sequencing depth and gene length, transcripts per million (TPM) expression values were calculated. Gene Set Enrichment Analysis (GSEA) was performed with the R statistical package fgsea 1.10.0. The collection of hallmark gene sets from the Molecular Signature Database was used for the GSEA, setting the FDR threshold at 5%.
Data and Software Availability
The accession number for the RNA-seq data reported in this paper is GEO: GSE163478
Acknowledgements
N. Vilaplana-Lopera, G. Papatzikas, A. Cunningham, and A. Erdem were supported by the EU grant HaemMetabolome H2020-MSCA-ITN-2015-675790. U. Günther, P. Garcia, J. J. Schuringa, J.-B. Cazier, and F. Schnütgen acknowledge support from the European Commission (HaemMetabolome [EC-675790]). This work was further supported by the Deutsche Forschungsgemeinschaft (DFG, German Research foundation) - SFB815, TP A10 (F.S.). We also acknowledge the Wellcome Trust for supporting access to NMR instruments at the Henry Wellcome Building for Biomolecular NMR in Birmingham (grant number 208400/Z/17/Z).
Footnotes
↵* These two authors are co-senior authors.