Summary
T-cell acute lymphoblastic leukaemia (T-ALL) is a cancer of the immune system. Approximately 20% of paediatric and 50% of adult T-ALL patients relapse and die from the disease. To improve patient outcome new drugs are needed. With the aim to identify new therapeutic targets, we integrated transcriptomics and metabolomics data, including live-cell NMR-spectroscopy, of cell lines and patient samples. We found that T-ALL cells have limited energy availability, resulting in down-regulated mTOR-signalling and reduced autophagy. Despite this, mTOR kinase remains active and essential for the glutamine-uptake and rapid proliferation, as glutamine supplies three nitrogen atoms in purines and all atoms but one carbon in pyrimidines. We show that EAAT1, a glutamate-aspartate transporter normally only expressed in the CNS, is crucial for glutamine conversion to nucleotides and that T-ALL cell proliferation depends on EAAT1 function, identifying it as a target for T-ALL treatment. Finally, we performed an in silico screen and identified a novel EAAT1-specific allosteric inhibitor.
Introduction
For a cancer to proliferate, cells need to increase their size and replicate their DNA, which requires large quantities of proteins, lipids and nucleotides, as well as energy. Metabolic changes and adaptations during tumorigenesis enable tumour cell growth and survival by altering metabolic pathways, such as glycolysis and glutaminolysis [1–3] and signal transduction cascades that regulate metabolic processes, such as the PI3K-AKT-mTOR pathway [4–6]. The compilation of the metabolic changes results in a distinct metabolic phenotype. Annotation of the rate-limiting steps that support the oncogenic metabolic adaptations can lead to the identification of useful therapeutic targets for the development of new anti-cancer therapies.
In order to understand and identify oncogenic processes that support rapid proliferation, we developed a methodology that assessed and integrated malignant gene expression, signal transduction and metabolism. We used T-cell acute lymphoblastic leukaemia (T-ALL) as a model. T-ALL is a haematological malignancy of the T-cell lineage that occurs in adults and children. Gene expression profiling studies showed that paediatric T-ALL patients can be grouped into four major clusters: TAL/LMO, TLX/HOXA, Proliferative and Immature (ETP ALL) [7, 8]. Most T-ALL have additional mutations in genes instrumental for signal transduction pathways such as NOTCH (>60%), PTEN-PI3K-AKT, JAK-STAT and FBXW7 [9, 10]. Monitoring of minimal residual disease (MRD) is a response based treatment protocol that is used for T-ALL patient stratification into standard, intermediate and high risk groups [11]. Even though treatment outcome has improved, approximately 20% of paediatric and 50% of adult T-ALL patients have refractory disease or relapse and most of these die from the disease [12, 13]. Refractory disease is more frequent in T-ALL displaying an early T-cell phenotype, which most commonly occurs in ETP and TAL/LMO T-ALL [14, 15]. SIL/TAL chromosomal microdeletion (1p32q) is the most common genetic rearrangement in T-ALL [7]. SIL-TAL deletion positions the promoter of the constitutively active STIL gene in front of the coding sequences of the TAL1 transcription factor that is crucial for haematopoietic and T-cell development and differentiation [16].
In this study we integrated gene expression data, signal transduction activity and metabolic studies, using cell lines and purified primary patient samples. We found that T-ALL cell cycle progression depends on mTOR-dependent glutamine uptake. Employing metabolic labelling we found that glutamine provides all but one carbon in the pyrimidine ring and three of the four nitrogen atoms in purines. Furthermore, analyses of the metabolic processes supporting glutamine metabolic pathways and the gene expression of T-ALL cell lines and patient samples showed that T-ALL cells express the glutamate-aspartate antiporter EAAT1 (encoded by SLC1A3) [8]. In the healthy adult body, EAAT1 is only present in the neurons and glia of the central nervous system where it is required for the uptake of the cytotoxic glutamate from the glutamatergic synapses in exchange for aspartate [17, 18]. In T-ALL, EAAT1 is required for glutamate import into mitochondria in exchange for glutamine-derived aspartate that is used as a substrate for nucleotide production in the cytoplasm. We show that T-ALL survival depends on EAAT1 function and validates EAAT1 as a therapeutic target for treating T-ALL. Subsequently, we performed a drug screen and identified a novel potent EAAT1-specific allosteric inhibitor.
Results
SIL-TAL cells exhibit suppressed carbohydrate and upregulated nucleotide metabolism
Based on the prevalence and the chance of refraction/relapse, we have decided to focus our efforts on ETP and SIL-TAL T-ALL. We chose three cell lines that carry the SIL-TAL rearrangement (DU.528, HSB2 and CCRF-CEM) and one cell line without that immuno-phenotypically belongs to the ETP T-ALL group (ARR) [19–23]. RNA-seq identified 4508 differentially expressed genes between the cell lines. Principal component analysis and hierarchical clustering found that the three SIL-TAL cell lines clustered separately from ARR, with HSB2 and CCRF_CEM being most similar (Figure 1A,S1) Hierarchical clustering grouped differentially expressed genes into 11 groups (Figure 1A). Genes suppressed in the SIL-TAL cells grouped in cluster 1 (C1), whereas C9 and C10 comprised upregulated genes. Gene functional annotation analyses indicated that ARR and SIL-TAL cells had a number of differentially regulated metabolic pathways (Figure 1B,S2). C1 was associated with carbohydrate catabolism, protein processing in the endoplasmic reticulum and amino-acid biosynthesis, whereas C10 involved purine and pyrimidine synthesis, sphingolipid metabolism and C9 T cell receptor and Rap1 signalling (Figure 1B). Additionally, genes that were upregulated in at least two of the SIL-TAL cell lines (C8 and C11) were implicated in signal transduction pathways such as MAPK, Hippo, Rap-1 (C8) and TGF-β, Ras, Rap-1, JAK-STAT, AMPK, mTOR (C11) (Figure S2). Based on this we conclude that the SIL-TAL metabolic phenotype in comparison to ARR is characterised by suppressed utilisation of carbohydrates and amino acid/protein synthesis, and upregulated nucleotide metabolism.
Validation of the RNA-seq by qPCR confirmed differential gene expression for argininosuccinate synthase 1 (ASS1; cluster C1) and Acyl-CoA synthetase short-chain family member 1 (ACSS1; cluster C10), two genes whose products have metabolic functions (Figure 1C,D). ASS1 converts citrulline and aspartate into L-argininosuccinate, an intermediate in arginine biosynthesis, which is also important for alanine, aspartate and glutamine metabolism. ACSS1 encodes a protein that catalyses the conversion of acetate to acetyl-CoA [24–26]. ASS1 was only expressed in ARR cells, whereas ACSS1 had the opposite pattern, with higher expression levels found in the SIL-TAL cells (Figure 1C,D).
Protein synthesis is supressed in SIL-TAL cells
We decided to assess whether the findings of the gene expression analyses were reflected in the T-ALL metabolism. To test if the amino acid synthesis is suppressed in SIL-TAL cell lines, we measured de novo protein synthesis using Click-iT OP-puro methodology [27]. The quantification revealed that SIL-TAL cells exhibited 30-40% lower protein synthesis than ARR (Figure 2A,B).
The mTOR signalling pathway is the main regulator of protein synthesis based on the availability of amino acids and cellular energy status [28–30]. Leucine, arginine, and glutamine, have been identified as effective activators of mTORC1 and their depletion leads to the inhibition of the mTOR pathway and suppression of de novo protein synthesis [31, 32]. Based on (i) the downregulated protein synthesis in SIL-TAL, (ii) their inability to produce arginine due to the absence of the ASS1 enzyme and (iii) suppressed amino acid biosynthesis based on the gene ontology, we expected attenuated signalling through the mTOR pathway. To test the importance of the mTOR pathway on T-ALL cells we employed rapamycin, an inhibitor of the mTOR kinase activity and found that upon treatment SIL-TAL cell proliferation was reduced, so that after 4 days the number of treated cells was 40% compared to the vehicle-treated control (Figure 2C) [33, 34]. Rapamycin treatment of ARR did not affect cell proliferation, demonstrating that ARR was insensitive to mTOR inhibition, whereas SIL-TAL cell lines required active mTOR kinase activity for their proliferation.
In order to better understand mTOR signalling in the T-ALL cell lines, we assayed the endogenous levels and phosphorylation status of proteins constituting the PI3K-AKT-mTOR signalling cascade under normal growth conditions and in rapamycin treated cells. Binding of growth factors to G-protein coupled receptors results in PI3K activation and generation of PI3P that further activates PDK1 to phosphorylate itself (S241) and AKT1 (T308) [35–37]. Opposing this is the phosphatase PTEN, which dephosphorylates PI3P and inhibits AKT1 signalling [38–40]. Western blot analyses revealed that T-ALL cells had active PDK-1 (p-PDK1S241), only CCRF-CEM, which harbours a PTEN deletion, contained active AKT1 (p-AKT1T308) (Figure 2D) [41]. In CCRF-CEM, AKT1 was additionally phosphorylated at the mTORC1 target site S473 and this activating mark was absent in the rapamycin treated cells, showing that mTOR has intact kinase activity in this cell line (Figure 2D) [42, 43]. The presence of inactive AKT1 in ARR, DU.528 and HSB2 and constitutively active AKT1 in CCRF-CEM cells implies that mTOR signalling is uncoupled from growth factor-AKT stimulation in all four T-ALL cell lines.
Next, we examined the AKT substrates c-RAFS259 and AKT1S1/PRAS40T246 [44–46]. As expected, both sites were phosphorylated in CCRF-CEM cells, but also in ARR (Figure 2D). Since PRAS40 is an inhibitor of the mTOR signalling cascade and the phosphorylation of PRAS40 by AKT1, at T246, relieves mTORC1 from PRAS40 inhibition, our results imply that the mTORC1 signalling is not suppressed by PRAS40 in ARR cells.
RHEB is a small GTPase required for mTOR lysosomal activation, which is negatively regulated by the tuberous sclerosis complex (TSC) complex [47–50]. AKT inhibits the TSC complex and allows RHEB to stimulate mTOR signalling [49, 51–54]. The loss of any of the members of the TSC-complex renders mTOR insensitive to regulation through amino acid availability and growth factors [55–57]. Examination of the gene expression data for the genes encoding the proteins of the mTOR signalling pathway revealed defective TSC1 gene expression in ARR cells. Due to alternative splicing, TSC1 transcripts lack four exons (exons 3-7) including the translation start site (Figure S3A-C) and consequently ARR fails to generate TSC1 protein (Figure 2E). The lack of TSC1 implies that mTOR signalling in ARR cells cannot be inhibited by the TSC complex. Additionally, rapamycin failed to supress mTOR and 4E-BP phosphorylation (mTORS2448, mTORS2481, 4E-BP1T37/46, 4E-BP1T70 and 4E-BP1S65) but did reduce the phosphorylation of p70S6KT389 and S6 ribosomal protein (rpS6S235) (Figure 2D,E). Taken together, ARR cells are characterised by PRAS40 and TSC-independent mTOR signalling with rapamycin-insensitive phosphorylation of 4E-BP. Based on this we conclude that in ARR cells mTOR and protein synthesis regulated by 4E-BP is independent of growth factor and amino acid signalling.
Another regulator of the mTOR signalling is the binding partner Raptor, which mediates mTOR interaction with 4E-BP1 and p70S6K [58–61]. As a result of low cellular energy status, AMP kinase (AMPK) phosphorylates S792 on Raptor leading to mTORC1 inhibition and cell cycle arrest [62]. Surprisingly, p-RaptorS792 was observed in all cell lines implying active AMPK and reduced energy availability. In SIL-TAL cells p-RaptorS792 occurred together with reduced RHEB-driven mTOR auto-phosphorylation (S2481) and mTOR phosphorylation by AKT1 at S2448 (Figure 2E) [63, 64]. The commonalities found between the T-ALL cells suggest that T-ALL cells have low energy status that activates AMPK to phosphorylate Raptor and suppress mTOR signalling. Active AMPK and p-RaptorS792 had limited effect on mTOR activity in ARR cells showing that mTOR is constitutively active. The result of supressed mTOR signalling in SIL-TAL was reduced p-4E-BP1T37/46 in comparison to ARR. Additionally, SIL-TAL cells had reduced phosphorylation on S65 or T70, with DU.528 exhibiting low phosphorylation on both residues. Since the function of 4E-BP is to inhibit cap-dependent translation by binding to the translation initiation factor eIF4E, hyper-phosphorylation of 4E-BP1 by AKT and mTOR disrupts the interaction with eIF4E resulting in activation of cap-dependent mRNA translation [65–68]. 4E-BP phosphorylation at T37 and T46 is required for subsequent S65 and T70 phosphorylation and the release of eIF4E [67, 69, 70]. Therefore, we conclude that the SIL-TAL 4E-BP phosphorylation status is not optimal for supporting mRNA translation and it corroborates our observation that SIL-TALs have reduced de novo protein synthesis.
The consequence of supressed mTOR signalling on p70S6K in SIL-TAL cells was only observed is HSB2 cells, where p-p70S6KT389 was the lowest. Despite this, phosphorylation of the p70S6K target rpS6 was observed and remained sensitive to rapamycin showing that AMPK-mediated mTOR suppression does not affect rpS6 phosphorylation in SIL-TALs.
Lower protein synthesis and reduced phosphorylation of mTOR and 4E-BP or p70S6K was found in all three SIL-TAL cell lines. A possible reason for this is limited amino acid availability. Amino acids are substrates for protein synthesis and drivers for the Rag GTPase-dependent recruitment of mTORC1 to lysosomes, where growth factor-controlled RHEB activates mTORC1 [47, 71, 72]. Lysosomal co-localisation of mTORC1 and its interacting partners on the lysosomal membrane is an additional level of mTOR regulation, which is required for correct mTORC1 signalling and is independent of mTOR kinase activity [47, 71, 73]. To examine mTOR localisation under normal growth conditions and upon mTOR inhibition by rapamycin, we imaged localisation of mTOR and the lysosomal marker LAMP1 in parallel with monitoring lysosomal acidification with a live-cell LysoTracker dye (Figure 2F) [74–76]. In all the cell lines mTOR partly co-localised with LAMP1 staining but also had nuclear and cytoplasmic localisation. Surprisingly, upon rapamycin treatment, mTOR was almost exclusively co-localised with punctual LAMP1 staining (Figure 2F). This increase in lysosomal mTOR localisation upon rapamycin treatment reveals that the regulation of mTOR localisation was functional and suggests that the T-ALL cells had limited amino acid availability, which was improved upon mTOR inhibition, most likely as a consequence of reduced protein synthesis and cell proliferation, which would lead to increase in intracellular amino acid availability.
While ARR had intense LysoTracker staining that overlapped with LAMP1 and was increased by rapamycin treatment, SIL-TAL cells had weak to absent staining that was further reduced upon rapamycin treatment (Figure 2F). This lack of lysosomal acidification could be a possible cause of limited amino acid availability as acidified lysosomes are required for protein degradation, supplying the cell with free amino acids.
Activated AMPK in all T-ALL cell lines
Lysosomal acidification is facilitated by vacuolar-type H+-ATPase (v-ATPase) that hydrolyses the ATP and transports H+ across the lysosome membrane, resulting in lysosomal acidification [77]. v-ATPase function is regulated by the cellular energy status, where an increase in the AMP/ATP ratio activates AMPK, which in turn phosphorylates and inhibits the v-ATPase, leading to preservation of the cellular energy stores [47, 71, 73, 78]. The observed inadequate lysosomal acidification and p-RaptorS792 suggests active AMPK in T-ALL cells (Figure 2D,F). Indeed, p-AMPKT172, a mark of AMPK activation was detected as well as the phosphorylation of the AMPK targets, ULK1S555 and Beclin-1S93 (Figure 3A) [79–83]. ULK1 and Beclin-1 are important for autophagosome formation and initiation of autophagy, a regulated self-digestion that is induced by starvation [84, 85]. AMPK stimulates autophagy directly by phosphorylating ULK-1(S555) and Beclin-1(S93) and indirectly by inhibiting mTOR through p-RaptorS792 ([62, 82, 83]). Our results demonstrate that all T-ALL cells have activated AMPK, which induces autophagy by relaying the signal to ULK-1 and Beclin-1.
Impaired autophagy and lysosomal function in SIL-TAL
Even though autophagy is induced in T-ALL, the observed impaired lysosomal acidification in SIL-TAL cells implies that these cells have a limited ability to hydrolyse cargo delivered by autophagosomes, which would cause an obstructed autophagic flux. ULK-1 and Beclin-1 activate the autophagy-related (ATG) protein machinery, resulting in the lipidation of the cytosolic protein LC3-I to LC3-II and integration into autophagosomes, double-layered cytoplasm-engulfing vesicles (reviewed in [86]). LC3-II interaction with the cargo receptors, such as p62, allows autophagosomes to engulf cytoplasmic components, including cytosolic proteins and organelles. Subsequent fusion with lysosomes facilitates cargo degradation by lysosomal hydrolases and the release of nutrients [86]. In order to determine whether T-ALL cells have effective autophagy, we examined the levels of LC3I, LC3II and the cargo receptor p62, as well as the localisation of LC3II in the control and rapamycin treated cells. All four cell lines exhibited higher LC3I than LC3II levels and punctual cytoplasmic LC3II, which only in ARR overlapped with lysosomal LAMP1 staining, revealing the presence of autophagosomes in all T-ALL cells, but fusion with lysosomes only in ARR cells (Figure 3B,C). Rapamycin-induced autophagy in ARR cells lead to elevated production of LC3I and LC3II, hence autophagosomes and their fusion with lysosomes, as seen by increased punctual LC3II staining that co-localised with LAMP1 (Figure 3B,C). ARR also show the lowest levels of p62, confirming effective cargo clearance (Figure 3C). Altogether, these results demonstrate functional autophagy only in ARR. Rapamycin treatment had no noticeable effect on LC3II and p62 abundance in DU.528 and HSB2 cells albeit some increase of LC3II was observed in CCRF-CEM cells (Figure 3B,C). SIL-TAL cells exhibited punctual LC3II staining, which did not co-localise with LAMP1 and even after rapamycin treatment co-localisation was sporadic. The limited LC3II-LAMP1 co-localisation and reduced lysosomal acidification indicate that SIL-TAL cells have obstructed autophagic clearance.
In order to find out whether reduced lysosomal acidification is the consequence of a low energy status/AMPK activity, or impaired v-ATPase function, we inhibited AMPK using dorsomorphin [87]. Treatment reduced the abundance of LC3I and LCII in all four T-ALL cell lines. Additionally, LC3 protein level was elevated in ARR and to a lower extent in HSB2 cells while p62 remained constant, except in DU.528 cells where dorsomorphin caused a mild reduction (Figure 3D). Immuno-fluorescent staining of the treated cells showed that the SIL-TAL cells had functional v-ATPase and the capacity to acidify their lysosomes, as seen by LysoTracker staining accumulated in LAMP1 positive lysosomes, a proportion of which was co-localised with autophagosomal LC3II (Figure 3E). The ARR cell line displayed an increase in the amount of punctual, co-localised LC3II and LAMP1 staining. In line with the western blot analyses these results show that inhibition of AMPK activity leads to lysosomal acidification and activated autophagy in SIL-TAL cells and to a further increase in autophagic activity in ARR cells. Therefore, we conclude that all T-ALL cell lines have functional v-ATPase and that reduced autophagy is caused by the unfavourable energy status and active AMPK.
mTOR Regulates Glutamine and Glucose Utilisation
So far, we observed the consequences of limited ATP and amino acid availability on the mTOR/AMPK signalling, autophagy and protein synthesis. Despite this, the T-ALL cell lines undergo rapid cellular proliferation. In order to better understand how the T-ALL cells source metabolites to fuel rapid proliferation, we measured their metabolism by establishing their metabolite uptake/release and their intracellular metabolite levels.
Metabolite uptake was assessed by measuring metabolite levels in the media, 24h after medium change, by NMR spectroscopy. We identified 22 metabolites, whose levels were significantly changing in at least one of the T-ALL cell lines. Hierarchical clustering of the observed changes showed that all four cell lines uptake lysine, glucose, phenylalanine and glutamine and release lactate, pyruvate, glutamate and pyroglutamate (Figure 4A). mTOR inhibition by rapamycin resulted in reduced glucose and glutamine uptake and lactate release in all T-ALL cells, indicating that glucose and glutamine uptake depended on an active mTOR pathway (Figure 4B).
T-ALL metabolic phenotype
NMR analysis of intracellular metabolite levels was used to determine differences in metabolite levels between the cell lines, 24h after changing the growth medium. Hierarchical clustering of the metabolites that showed significantly different levels between any two cell lines, identified four groups of metabolites (Figure 4C, S4A; metabolic clusters MC1-4). MC1 and MC3 are comprised of metabolites more abundant in SIL-TAL cells and ARR, respectively. Phosphatidylcholine, glycerophosphatidylcholine (GPC) and the amino acids alanine, glutamate, aspartate and cysteine-derived taurine are constituents of MC1, while AMP, succinate, NAD+, glutathione, as well as creatine and phosphocreatine are members of MC3. Therefore, direct measurements show that ARR cells have the highest AMP and the lowest ATP/ADP level, confirming that ARR have a low energy status, in line with the active AMPK seen in Figure 3A. Metabolites with low abundance in CCRF-CEM cells were featured in the MC2 cluster. This cluster included glutamine, which similar to MC3 metabolites, was found at the highest level in ARR (Figure 4C). We therefore observe that SIL-TAL have a specific metabolic phenotype, characterised by lower glutamine levels and higher levels of the glutamine-derived metabolites glutamate, aspartate and alanine. Taking into consideration that (i) T-ALL uptake glutamine, which is a precursor for glutamate and aspartate, (ii) rapamycin reduces glutamine utilisation and (iii) that glutamine, glycine and aspartate are substrates for de novo nucleotide synthesis, glutamine could be the main source for de novo nucleotide synthesis in T-ALL.
Intracellular metabolite levels are very dynamic and change as a response to metabolite uptake and cellular demand. In order to address the dynamics of the metabolite levels, we compared the abundance of intracellular metabolites at the time of growth-media change (0h) to that found 8h and 24h later (Figure 4D). A similar response to media change in all four cell lines was found for glutamate and myoinositol, whose availability decreased upon media supplementation. Elevated intracellular concentration of glutamate is consistent with our previous finding that T-ALL release glutamate into the media, as well as glutamate-derived pyroglutamate (Figure 4A). A polarised response between SIL-TAL and ARR was observed for AMP, which was the only metabolite with increasing levels in SIL-TAL cells while decreasing in ARR (Figure 4D). Based on the finding that in SIL-TAL cells media change led to increasing AMP levels, while ATP/ADP levels were overall maintained, we conclude that SIL-TAL cells have an increasing AMP to ATP/ADP ratio upon medium change. This supports our finding that SIL-TAL cells had active AMPK and reduced energy availability, as described above (Figure 3A).
Opposite dynamics were observed for ARR and SIL-TALs for NAD+, tyrosine, isoleucine, leucine, phenylalanine, valine and glutamine, whose concentrations were lower in SIL-TAL at 8h and/or 24h, and higher in ARR. This reduction of the intracellular abundance of essential amino acids is an experimental observation that directly corroborates our previous conclusion that SIL-TAL have reduced amino acid availability (Figure 4D,2A,D,F). In ARR, the levels of these metabolites were elevated at 8h and for a few at 24h, and additionally included aspartate, alanine, GPC and nucleotide sugars (UDP-glucose, UDP-Galactose and UDP-GalNAc). Lower availability and only a transient increase after medium supplementation of alanine, glutamate, aspartate and taurine in ARR compared to SIL-TAL cells show that ARR have a high requirement for amino acids to facilitate higher rate of de novo protein synthesis (Figure 4C,D,2A,B). The SIL-TAL cells rapidly utilise essential amino acids upon medium change, followed by their intracellular accumulation. Altogether, measurements of the intracellular metabolite levels and the changes occurring during the cell growth and proliferation confirm that T-ALL have an unfavourable energy status with high AMP/ATP ratios and that SIL-TAL cells have limited availability of essential amino acids in line with the previously observed reduced de novo protein synthesis.
Inhibition of the mTOR Pathway Restores Metabolite Levels in T-ALL Cell Lines
As we showed that mTOR inhibition reduced glucose/glutamine uptake and blocked SIL-TAL proliferation (Figure 4B,2C), we next measured the effect of rapamycin on intracellular metabolite levels. NMR measurements identified 28 metabolites that had a significantly different abundance after 24h rapamycin treatment in at least one of the cell lines (Figure 4E). Hierarchical clustering of the observed log2 fold changes divided these metabolites into six rapamycin-responsive metabolic clusters (RMC1-6). The most pronounced effect of rapamycin was increased intracellular levels of the majority of metabolites. In all four cell lines the largest increase was observed for glutamine and alanine levels, while AMP and succinate levels had the opposite response (Figure 4E,S4B). A differential effect of rapamycin on SIL-TAL in comparison to ARR was particularly visible in RMC2 and RMC4. The levels of the amino acids phenylalanine, leucine, tyrosine, isoleucine and valine (RMC2) increased to a greater degree in SIL-TAL cells, whereas the increase in aspartate, glycine, glutamate and alanine (RMC4) was more pronounced in ARR (Figure 4E). The observed increase in the amino acids availability upon mTOR inhibition explains the mTOR recruitment to lysosomes upon rapamycin treatment and proves that T-ALL have limited amino acid availability.
Glutamine is Essential for SIL-TAL Proliferation and Survival and Fuels Aspartate and Nucleotide Biosynthesis
The polarised effect of mTOR inhibition on ARR vs SIL-TAL, seen primarily as an increase in aspartate, glycine, glutamate and alanine levels could be dueto the continuous proliferation in ARR and suppressed proliferation in SIL-TAL cells. As all four amino acids can be derived from glutamine and glutamine, glycine and aspartate are used as substrates for nucleotide synthesis, we tested the effect of glutamine withdrawal on cell proliferation. Omitting glutamine from the growth medium, while maintaining all other nutrients, including 10% foetal bovine serum, SIL-TAL cell numbers were reduced by 95% after 8 days of treatment (Figure 5A). Glutamine deprivation caused a decrease in ARR numbers over the first four days of culture, after which ARR recovered and resumed proliferation at the original rate. Apoptosis assays of the cells grown in glutamine-free medium revealed that the SIL-TAL cell lines had an increased rate of apoptosis and cell death (Figure S5). For ARR, this was initially also the case, but normalised after four days (Figure S5). Therefore, we conclude that glutamine is important for T-ALL cell proliferation.
Glutamine has several significant roles in metabolic processes such as fuelling the TCA cycle through glutaminolysis and acting as a nitrogen donor in transamination and transamidation reactions that are used for the production of non-essential amino acids. Also, together with aspartate and glycine, glutamine feeds de novo nucleotide synthesis. In order to assess glutamine contribution to T-ALL metabolic processes we performed tracer experiments using [2,5-15N] Glutamine and [3-13C] Glutamine. When cells were grown in the presence of [2,5-15N] Glutamine, in addition to the expected 15N labelled glutamate (data not shown), we observed 15N incorporation into the amino group of aspartate and alanine within 8h of treatment in all four cell lines (Figure 5B and S6A). Resonances arising from Hβs of aspartate (2.80-2.81ppm) and alanine (1.48-1.49ppm) are shown. At 0h, a doublet was seen due to coupling to the Hα. However, at later time points (Figure 5B, 8 and 24h) there was increased spectral complexity because Hβs were also weakly coupled to 15N. The resulting spectrum was a weighted average of a doublet of doublets (from 15N-labelled aspartate) and a simple doublet (unlabelled aspartate). Both aspartate and alanine were found at the lowest level prior to labelling in ARR and CCRF-CEM cells. While CCRF-CEM signal intensity was maintained at a similar level at 8h and 24h, ARR showed a transient increase at 8h and a reduction to the original level by 24h (Figure 5B). A similar effect was observed for alanine, although the signal intensity in ARR remained the same at 8 and 24h (Figure S6A). This finding demonstrates that glutamine-derived 15N was incorporated into aspartate and alanine by transamination of oxaloacetate and pyruvate, respectively.
Glutamine, glycine and aspartate are amino acids used as substrates for nucleotide synthesis, so in order to assess the glutamine contribution to nucleotide biosynthesis we first acquired 2D-1H,15N heteronuclear single quantum coherence (HSQC) NMR spectra of 15N uniformly labelled ATP, GTP and UTP standards, where we observed five 1H-15N interactions for [U-15N] ATP (a-d), three for [U-15N] GTP (a-c) and three for [U-15N] UTP (a-c) (Figure 5C). Spectra acquired from [2,5-15N] Glutamine-labelled T-ALL cells showed 15N incorporation into nucleotides at the 1, 3 and 9 positions, the 3 and 9 positions and the 1 and 3 positions in adenine, guanine and uracil respectively (Figure 5C). As expected, based on the known contribution of glutamine to nucleotide synthesis, transamidation using 15N-glutamine resulted in label incorporation at N-3 and N-9 in purines and the N-3 in pyrimidines (Supplementary Illustration SI1,2). However, label incorporation at N-1 in purines and N-3 in pyrimidines is sourced from glutamine-derived 15N-aspartate and shows that glutamine ultimately supplies two out of three nitrogens in purines and both nitrogens in pyrimidines.
So far, we demonstrated that glutamine transamination of oxaloacetate gives rise to aspartate. As it is possible that glutamine supplies oxaloacetate as well through glutaminolysis, which would subsequently mean that aspartate is completely derived from glutamine, we used [3-13C] Glutamine in labelling experiments to test this. In the course of the TCA cycle, [3-13C] Glutamine is converted to [2-13C] Fumarate, a symmetrical molecule which is hydrated equally to [2-13C] Malate and [3-13C] Malate, and further to [2-13C] or [3-13C] oxaloacetate and aspartate (Supplementary Illustration SI3). Acquired 2D-1H,13C HSQC NMR spectra revealed 13C incorporation in fumarate, malate, oxaloacetate and aspartate (Figure 5D,S6B and data not shown). Resonances for 1H-13C moieties were derived from both [2-13C] and [3-13C] labelling, as shown for aspartate and malate (a and b respectively, Fig 5D,S6B). Quantification of the signals from the labelled samples, relative to the naturally occurring 13C in the control samples, showed that within 24h, 50% of all aspartate was [2-13C] or [3-13C] Aspartate (Figure S6E). These findings confirmed our hypothesis that two molecules of glutamine give rise to a single aspartate molecule; first glutamine supplies the backbone via the TCA cycle, while the second is used for transamination.
A further implication of these findings was that carbons derived from glutamine would, via aspartate, get incorporated into pyrimidines, giving rise to [5-13C] or [6-13C] Uridine (Supplementary illustration SI2). Resonances for 1H-13C moieties arising from [5-13C] Uridine or [6-13C] Uridine were indeed observed in 2D-1H,13C HSQC NMR spectra (peaks a and b, Figure 5D). This strongly implies that the carbon at position 4 also originates from glutamine via aspartate. Together, our results suggest that glutamine serves as a source for all but one of the atoms in the pyrimidine ring.
Examination of the 13C incorporation in other detectable metabolites found a significant label accumulation into proline, as would be expected since glutamate is used as a substrate for proline synthesis (Figure S6C). Surprisingly, [3-13C] Proline was detected in all four cell lines, but [2-13C] Proline was found only in the SIL-TAL cells, indicating that SIL-TAL derived some glutamate from the TCA-cycle while ARR used glutamate almost exclusively derived directly from glutamine. Additional differences were observed between ARR and the SIL-TAL, in the area of the spectra that is characteristic for the aspartate-derived metabolites arginine, ornithine, citrulline and argininosuccinate (13C- ppm 30.2-30.8, Figure S6D). Due to overlaps between their spectra we were not able to identify the individual contributions of these metabolites. However, it is likely that the observed differences can be attributed to low argininosuccinate and arginine levels in SIL-TAL cells due to the finding that SIL-TAL cells lack ASS1 expression, an enzyme essential for their biosynthesis.
Patient-derived T-ALL and T-ALL cell lines have similar metabolic uptake
To determine if glutamine is not only used by patient derived cell lines, but also by primary T-ALL patient isolates, we measured live-cell real-time metabolite uptake by NMR. T-ALL CD34+/CD7+ cells were isolated from two T-ALL patients at presentation. The flow cytometry profile revealed that both samples had cytoplasmic CD3 expression, but lacked CD3 surface expression. T-ALL_1 consisted of 60% CD34+ blasts, half of which were CD7+. T-ALL_2 was predominantly CD7+, as well as CD5+, CD2+, CD38+ and CD4+. CD34+/CD7+ cells were isolated and 106 cells were resuspended in RPMI medium supplemented with GlutaMAX (L-glutamine/L-alanine dipeptide) and used for measuring metabolite uptake. GlutaMAX is a temperature-stable source of glutamine [88]. The L-glutamine/L-alanine dipeptide is hydrolysed, by peptidases located on the plasma membrane, into L-Glutamine and L-Alanine, which can then be taken up by the cells [89]. Therefore, the extracellular glutamine and alanine concentration depends on GlutaMAX hydrolysis and their utilisation by the cells. Using live-cell NMR we observed decreasing GlutaMAX availability in both patient samples with a greater rate of hydrolysis observed in T-ALL_2 (Figure 6A). Additionally, T-ALL_1 was taking up glucose and pyruvate and releasing lactate. Glutamine and alanine had bi-modal kinetics with an initial decrease in concentration followed by a transient increase and a final period of more rapid uptake. Changing metabolite availability in T-ALL_2 culture was very similar for all the observed metabolites with concentrations steadily increasing by up to 25% during the first 10 hours of the experiment followed by 20% decrease in the final 3h of the measurements (Figure 6A). Live-cell NMR was also performed with ARR and DU528 cell lines and similar dynamics of GlutaMAX hydrolysis were observed (Figure S7A). These results confirm that primary T-ALL cells utilise glutamine similarly to T-ALL cell lines.
T-ALL cells express SLC1A3 and its protein product EAAT1 is localised in the mitochondria
We demonstrated that the contribution of glutamine, directly and through aspartate, is crucial for T-ALL nucleotide synthesis and proliferation. As glutamine uptake and its direct contribution to nucleotide metabolism and TCA cycle is indispensable for whole organism survival, we focused on identifying a useful therapeutic target for T-ALL treatment on metabolic processes supporting glutamine-derived aspartate synthesis (Supplementary Illustration SI3). In particular, the glutamate-aspartate anti-port across the mitochondrial membrane would be a particularly interesting target as it would not target enzymatic conversions directly. Aspartate export and glutamate import into mitochondria is known to be facilitated by 3 different antiporters SLC1A3, SLC25A12 and SLC25A13 [90, 91]. While SLC25A12 and SLC25A13 are ubiquitously expressed in all tissues, the expression of SLC1A3 gene is mainly restricted to CNS [92]. Analyses of our RNAseq data, as well as recently published gene expression data of 265 T-ALL patient samples, revealed that SLC1A3 is expressed in our model cell lines and in 95% of patient samples (Figure 6B,C)[8]. Additionally, EAAT1, the protein encoded by SLC1A3, was detected in the mitochondria of the T-ALL cell lines. SLC1A3/EAAT1 was not detected in the acute myeloid leukaemia (AML) cell line Kasumi-1, which does not express SLC1A3 (Figure 6D,E) [93]. Together, these results show that, in addition to the common glutamate-aspartate antiporters, T-ALL express the high affinity antiporter EAAT1.
SLC1A3/EAAT1 is essential for oncogenic de novo nucleotide synthesis
The mitochondrial localisation of EAAT1 in T-ALL cells supports the possibility that this protein is involved in an oncogenic glutamate-aspartate antiport in T-ALL and crucial for T-ALL proliferation and survival. In order to assess the importance of EAAT1 for T-ALL survival we performed SLC1A3 knock-down using shRNA. Five shSLC1A3 were designed and tested for their capacity to supresses the expression of SLC1A3 cDNA. Mouse fibroblasts were co-transduced with retrovirus expressing SLC1A3-IRES-GFP and shSLC1A3. The efficiency of the shRNA was measured relative to the negative control, shFF3, which targets firefly luciferase and the positive control, shGFP, that supresses GFP expression originating from the second cistron of the SLC1A3-IRES-GFP mRNA. shSLC1A3_1 and particularly _2 had the capacity to supress EAAT1 protein levels similar to shGFP, while the negative control shFF3 did not have any effect on EAAT1 protein levels (Figure 7A). shSLC1A3_1 and _2 were cloned into a piggybac backbone that supports doxycycline-inducible shRNA expression and integrated into the T-ALL cell genomes. Induction of shSLC1A3 lead to rapid cell death in ARR, DU.528 and CCRF_CEM cells within 5 days of the doxycycline treatment (Figure 7B). We were unable to produce the same results with HSB2 cells since these cells were affected by doxycycline treatment alone. These results corroborate that SLC1A3/EAAT1 is required for T-ALL cell proliferation and survival.
In healthy adults EAAT1 is normally expressed in the CNS [18, 92]. With the purpose to inhibit its function in the CNS, two specific allosteric inhibitors, UCPH-101 and UCPH-102 were developed [94, 95]. Both drugs exhibited inhibitory effects on T-ALL proliferation and, while UCPH-101 had a cytotoxic effect that reduced the survival of the control AML cell line Kasumi-1, UCPH-102 had a specific anti-proliferative effect only on T-ALL cells (Figure 7C). Cytotoxicity connected with UCPH-101 can be explained by the presence of known toxicophores. Taken together, SLC1A3 knock-down and pharmacological inhibition caused the supression of T-ALL proliferation. This effect is similar to the effect of glutamine removal from the growth medium (Figure 5A). Finally, in order to confirm the mechanism of UCPH-102 mediated SLC1A3 inhibition, we measured the metabolite uptake/release. UCPH-102 abolished glutamine uptake, while statistically significant changes in all four T-ALL cell lines were only observed for glutamate and aspartate (Figure 7D and S7B). EAAT1 inhibition increased glutamate release in the ARR cell culture and decrease in the others, while aspartate was reduced in the growth medium of all four cell lines (Figure 7D). Reduced apartate release into the medium implies limited intracellular availability of aspartate that would impact on nucleotide synthesis, hence T-ALL proliferation. This corroborates our finding that EAAT1 is required for T-ALL survival and proliferation.
Development of EAAT1 inhibitors to block T-ALL cell proliferation
UCPH-101 and UCPH-102 were developed with the aim to reach the brain and therefore to cross the blood-brain barrier. Despite extensive development from the original hit compounds, these two compounds still have many unfavourable characteristics, including toxicophores and a very short half life in vivo [96]. We therefore set out to identify new small molecule inhibitors of EAAT1 with potential to be be further developed for future T-ALL treatment. As the UCPH compounds show high specificity due to their allosteric inhibitory function, an in silico approach was chosen on basis of the available crystal structure of EAAT1 in complex with UCPH-101 [95]. The in silico screen was performed by Domainex Ltd, using their database of approximately 2M commercially available compounds with lead-like characteristics. Compounds were selected that have the potential to interact at the same position as UCPH-101 and were screened in an in vitro cell assay, using CCRF_CEM and the Kasumi-1 cell line as a negative control. This cellular proliferation screen resulted in compounds that showed an enhanced effect on the growth of CCRF_CEM when compared to the results from Kasumi-1. EAAT1inh_1 showed a significant effect at concentrations below 1 uM (Figure 7F).
Discussion
Describing metabolic pathways is pivotal for understanding the processes that are essential for embryonic development, homeostasis and disease progression. Metabolic processes are interconnected and many steps are supported by several different enzymes that can be located in different cellular compartments and even different tissues and organs. Understanding the capacity of the cells to utilise a certain metabolic pathway based on the gene expression and combining this with the measurements of cellular demand and output is essential for building a bigger picture of the metabolic processes that support cellular proliferation and survival. One of the main characteristics of cancer cells is increased proliferation. During the oncogenic transformation, changes in gene expression and cellular signalling allow uncontrolled proliferation. In order to successfully achieve this, cancer cells may use processes that are more commonly observed during embryonic development, when large quantities of nutrients are used for growth and organogenesis. Understanding the metabolic pathways that support oncogenic proliferation can help us identify cancer-specific processes and rate limiting steps that can be used for developing new therapies.
Our results demonstrate that integration of gene expression, signal transduction and metabolic studies provides a powerful approach for the dissection of key metabolic processes. We show that, in T-ALL cells, mTOR regulates glutamine uptake. Glutamine is converted to glutamate that enters the mitochondria where Glutamate Dehydrogenase (GLUD1,2) uses it to generate α-ketoglutarate that will enter the TCA cycle (Figure 7E). Additionally, mitochondrial Glutamate Oxaloacetate Transaminase (GOT2) uses glutamate as a donor for the amino group, which is transferred to oxaloacetate, another TCA cycle intermediate, resulting in generation of aspartate and α-ketoglutarate. Aspartate is then transported out of the mitochondria in exchange for a new glutamate molecule, continuing to fuel the TCA cycle and generating aspartate. Together, glutamine and aspartate are used as substrates for nucleotide synthesis (Figure 7E). Similar processes are likely to occur in other cancers. Indeed, recently EAAT1 has been implicated in supporting proliferation in several cancer cell lines representing solid cancers [97, 98]. Both studies showed a main role for EAAT1 in the uptake of aspartate from the medium, especially under conditions of glutamine deprivation or asparaginase treatment. In both studies removing EAAT1 under normal cell culture conditions showed little or no effect in contrast to our work on T-ALL cells. This difference can be explained as the solid tumour cell lines were mainly dependent on EAAT1 for aspartate/glutamate uptake from the environment and the lack thereof resulted in a combinatorial effect on the TCA cycle, the electron transport chain, and de novo glutamine/glutamate and nucleotide synthesis. T-ALL cells on the other hand are dependent on the availability of glutamine. They rely heavily on the function of EAAT1 on the inner membrane of the mitochondria, where glutamate/aspartate antiport function is required for de novo nucleotide production. This process is heavily dependent on the availability of glutamate, which is derived from glutamine. Our NMR tracer experiments show clearly how both the carbon and nitrogen atoms of glutamine are used for aspartate, purine and pyrimidine biosynthesis. Sun et al. also showed that asparaginase treatment resulted in a cellular depletion of glutamine/glutamate, possibly explaining why asparaginase treatment is often effective in treatment of T-ALL [99].
Systemic targeting of the metabolic enzymes that conduct the glutamine to aspartate conversion or nucleotide synthesis would be detrimental. Our finding that the transport of aspartate to the cytoplasm is an essential step in nucleotide production and that it is facilitated by aberrantly expressed EAAT1, identifies this protein as a novel therapeutic target for developing treatments for T-ALL. EAAT1 is normally present in the CNS on the plasma membrane of neurons and glia where it uptakes glutamate from the glutamatergic synapses [17]. Based on the RNA expression and protein localisation assessed with three different specific antibodies, the Human Protein atlas Database reports that EAAT1 is not found outside of the CNS (https://www.proteinatlas.org/ENSG00000079215-SLC1A3/tissue) [100]. However, EAAT1 expression has been reported in neonatal cardiomyocytes where, similar to T-ALL, it is also localised in mitochondria together with two other glutamate-aspartate antiporters, i.e. ARALAR1 and CITRIN [90]. Altogether, this means that the aberrant expression of SLC1A3 in T-ALL renders this gene as an oncogene and EAAT1 a therapeutic target. We therefore performed a screen for novel small molecule inhibitors and we present EAAT1inh_1 compound as a potent EAAT1 inhibitor.
The rate limiting reaction in aspartate synthesis is a glutamate-aspartate antiport, which is normally carried out by ARALAR1 and CITRIN (SLC25A12 and SLC25A13) [91]. ARALAR1 and CITRIN are calcium-binding mitochondrial carrier proteins that import a glutamate molecule together with a H+ into the mitochondria in exchange for the export of an aspartate anion [101–103]. Their activity depends on the mitochondrial membrane potential that is maintained by cytoplasmic ATP [104]. Both transporters are present in the majority of cells and tissues (reviewed in [105]). We found that SLC25A12 is expressed in all four T-ALL cell lines, while SLC25A13 was found only in DU.528. With this in mind, an obvious question is why mitochondrial EAAT1 is necessary for T-ALL survival. One possibility is that EAAT1 has a higher Kd for glutamate than ARALAR1and CITRIN and therefore facilitates a higher rate of antiport across the mitochondrial membrane. Even though the kinetics of all three transporters are reported, the experiments were not performed in comparable experimental conditions and EAAT1 capacity was always addressed in respect to its membranous localisation [105–107]. The second possibility is that the intracellular conditions, such as the pH, ATP availability and mitochondrial action potential, are restrictive for ARALAR1 and CITRIN function, whereas EAAT1 is independent of these factors.
Authorship Contributions
Original Concept V.S.S., Project Planning, Experimental Design, Project supervision, V.S.S., M.H.; Genome-wide data acquisition and Bioinformatical analyses V.S.S., E.G., H.P., M.H.; Statistical analyses V.S.S.; NMR data acquisition and analyses M.A.C.R, J.R., V.S.S, U.G., C.L.; Drug screen V.S.S.; All other Experimentation, Data Acquisition, Processing and Analyses V.S.S., M.H.; Provision of patient samples S.P., G.P.; Reagents provision S.S.; Figures V.S.S.; Writing of the Manuscript, V.S.S., M.A.C.R. and M.H.
Acknowledgments
We would like to thank Dr. A.W. Langerak, (Erasmus Medical Centre, Rotterdam, NL) for the provision of ARR and DU.528 cell lines and Prof P.N. Cockerill (University of Birmingham, UK) for HSB2 and CCRF-CEM. We would like to thank Dr. M. McGrew (University of Edinburgh, UK) for the PB_tet-on_Apple_shGFP plasmid and Prof. L. Bunch for pcDNA3-EAAT1 plasmid. We would like to acknowledge BlueBEAR High Performance Computing (HPC) service for supporting the analyses of the genome-wide data and Biomolecular NMR Facility at the Henry Wellcome Building for Nuclear Magnetic Resonance (HWB-NMR), University of Birmingham. This work was supported by Bloodwise, through a Bennett Fellowship to M.H. [11002], the Medical Research Council and the University of Birmingham.
References
- 1.↵
- 2.
- 3.↵
- 4.↵
- 5.
- 6.↵
- 7.↵
- 8.↵
- 9.↵
- 10.↵
- 11.↵
- 12.↵
- 13.↵
- 14.↵
- 15.↵
- 16.↵
- 17.↵
- 18.↵
- 19.↵
- 20.
- 21.
- 22.
- 23.↵
- 24.↵
- 25.
- 26.↵
- 27.↵
- 28.↵
- 29.
- 30.↵
- 31.↵
- 32.↵
- 33.↵
- 34.↵
- 35.↵
- 36.
- 37.↵
- 38.↵
- 39.
- 40.↵
- 41.↵
- 42.↵
- 43.↵
- 44.↵
- 45.
- 46.↵
- 47.↵
- 48.
- 49.↵
- 50.↵
- 51.↵
- 52.
- 53.
- 54.↵
- 55.↵
- 56.
- 57.↵
- 58.↵
- 59.
- 60.
- 61.↵
- 62.↵
- 63.↵
- 64.↵
- 65.↵
- 66.
- 67.↵
- 68.↵
- 69.↵
- 70.↵
- 71.↵
- 72.↵
- 73.↵
- 74.↵
- 75.
- 76.↵
- 77.↵
- 78.↵
- 79.↵
- 80.
- 81.
- 82.↵
- 83.↵
- 84.↵
- 85.↵
- 86.↵
- 87.↵
- 88.↵
- 89.↵
- 90.↵
- 91.↵
- 92.↵
- 93.↵
- 94.↵
- 95.↵
- 96.↵
- 97.↵
- 98.↵
- 99.↵
- 100.↵
- 101.↵
- 102.
- 103.↵
- 104.↵
- 105.↵
- 106.
- 107.↵