Metabolic pattern is markedly different between the tissues of naive NSG mice. To understand how the xenotransplantation of human AML cells rewire tissue metabolism, we first characterized the metabolic profiles of ten different tissues (plasma, bone marrow (BM), spleen (SP), liver, subcutaneous white adipose tissue (SWAT), pancreas, lung, kidney, heart and muscle) from 4 naïve 8 weeks-old male and female NSG mice (Fig. 1a). Using two complementary LC-MS methods, 48 metabolites were quantified in absolute or relative amounts in each tissue. The measured metabolites were representative of the major metabolite classes (e.g. phosphorylated sugars, organic acids, amino acids (AA), nucleotides), and covered central and energy metabolic pathways (glycolysis, pentoses-phosphate pathway, TCA cycle and related processes, amino acid metabolism, nucleotide metabolism). Metabolic profiles of plasma from the naïve mice showed high levels of amino acids (range from 0.1 to 1mM), especially valine, glutamine, glycine and alanine. AA are key components of metabolism, as they constitute the main product of digested proteins and can be fuels for energy and biomass production. In contrast, other metabolites (glycolytic and TCA cycle intermediates, nucleotides and phosphorylated sugars) were observed in lower concentrations (range from 0.1 to 1.0 mM) (Fig. 1b). We were able to compare AA concentrations in NSG plasma with the data found in the MMMDB29 and observed that they were found in similar ranges. Other metabolites could not be compared.
Due to the vast range of concentrations in central metabolites in the different solid tissues, we normalized metabolite concentrations using a z-score (Fig. 1c). AA absolute amounts were also graphed in the different tissues (Additional Fig. 1a). Overall, most tissues had similar profiles, with the exception of two peculiar tissues (SWATs and SP). SWATs were characterized by the lowest metabolites concentrations and SP showed high levels in AA, early glycolytic intermediates and phosphorylated sugars. Low amounts of metabolites were expected in SWATs, as they are the major lipid reservoir. Surprisingly, AA and glycolytic intermediates in SP was over-abundant in this tissue. The SP is a hematopoietic organ with cellular composition similar to that of the BM and the overabundance of nutrients in this organ might explain the high propensity of leukemic cells to engraft this tissue. Because of the differences in AA content in SP, we further assessed AA distribution and concentration in all tissues (Fig. 1d). AA distribution was found quite similar in most tissues, with the exception of SWATs, showing undetectable amounts of glycine, and plasma, showing the lowest amounts of aspartate. This indicates that proteinogenic amino acids balance is conserved in the different organs.
Since gender is one of intrinsic factor regulating host metabolism, we also wanted to evaluate the impact of sex on metabolic profiles (Additional Fig. 1b). To our knowledge, no study was done on this aspect, and significant changes could bias further results. Interestingly, we found very few sex-specific changes in metabolite quantities in most tissues. The most impacted tissue was the BM with 6 amino acids (arginine, leucine, phenylalanine, serine, tyrosine and valine) significantly increased in female mice. We concluded that sex-driven changes were minimal in this cohort. Indeed, we studied 7–8 weeks-aged animals. These mice were at an early stage of puberty, indicating that sex-dependent metabolic impacts were minimized. Knowing that mice life span can extend to at least one year after cell engraftment for patient-derived xenografts, sex-driven metabolic changes would be uncovered at later stages.
AML xenotransplantation induces a profound reprogramming of local and systemic metabolism in vivo. We investigated the impact of leukemic engraftment on murine tissue metabolic profiles. NSG mice were engrafted with two metabolically different AML cell lines: high OxPHOS AraC-resistant MOLM14 cells and low OxPHOS AraC-sensitive U937 cells4. We have previously shown that the distribution of AML cells was variable in hematopoietic tissues of NSG mice post-transplantation24. MOLM14 and U937 cells mostly invade BM and SP, respectively. After two weeks of transplantation, we sacrificed mice and assessed leukemic engraftment in these two hematopoietic tissues by flow cytometry (Additional Fig. 2a-c). We confirmed significant differences in leukemic infiltration between MOLM14 and U937 cells. Other organs were not analyzed by cytometry, and possible engraftment could not be assessed. Then, metabolic profiles were done for the 9 solid tissues (Fig. 2a).
Metabolome of MOLM14-engrafted tissues showed little changes compared to their respective naïve metabolome. 24 metabolites were differentially changed mostly in BM (Fig. 2b). Asparagine, aspartate and phosphoenolpyruvate were depleted in MOLM14-engrafted mice, whereas glutamate, UMP, CMP, 2-hydroxyglutarate, galactosamine-1-phosphate (GalN-1P), N-acetylglucosamine-1-phosphate (GlcNAc-1P) and N-acetylglucosamine-6-phosphate (GlcNAc-6P) were accumulated. Changes were also observed in muscles with depletion of CMP, GMP, alanine, glutamate, glutamine, and pentoses-5-phosphate, in the pancreas with the depletion of GMP and pentoses-5-phosphate, in kidney with the depletion of glutamine and accumulation of ribose-1-phosphate, in the lung with an accumulation of UTP and glutamine, and in the heart with an accumulation of lactate and pentoses-5-phosphate (Fig. 2a).
On the other hand, U937 engraftment led to major changes in metabolic profiles compared to naïve tissue metabolome (Fig. 2c). Lung was the most impacted tissue with depletion of aspartate and glycine, and accumulation of AAs, fructose-1.6-biphosphate (FBP), fructose-6-phosphate (Fru6P), 2-hydroxyglutarate (2-HG), and derivatives of uridine. BM was characterized by depletion of aspartate, asparagine, glycine, GalN-1P and GlcNAc-1P, and accumulation of pyruvate, CMP and 2-HG from U937-xenografted mice. SP was characterized with depletion of aspartate, asparagine, serine, methionine, orotate, glucose-6-phosphate and Fru6P. Liver showed depletion of succinate and GalN-1P, and accumulation of glutamate, phenylalanine and threonine, while pancreas was characterized by depletion of glutamine and sedoheptulose-7-phosphate (Sed7P), and accumulation of glutamate, branched chained AA (BCAA), tyrosine, phenylalanine and citrate & isocitrate. Pentose-5-phosphates, CMP and glutamate were depleted, and BCAA accumulated in muscles. Heart was significantly exhibited a depletion of serine, and an accumulation of BCAA, phenylalanine, 2-hydroxyglutarate, Sed7P and GlcNAc-6P. Finally, kidney showed depletion of glycine and alanine, and accumulation of UDP (Fig. 2c).
Importantly, comparing global metabolic profiles after the xenotransplantation of both AML cell lines in vivo showed the most changes (Fig. 2d). In brief, AA were accumulated in SP and kidneys of MOLM14-engrafted mice, whereas they were accumulated in the lungs, SWAT and pancreas of U937-engrafted mice. Glycolytic intermediates were accumulated in SP of MOLM14 xenografts and accumulated in lungs of U937 xenografts. This suggests that the metabolic impact of AML engraftment is mostly driven by the intrinsic characteristics of injected AML cells and their organ distribution (Fig. 2d). Of note, we also showed commonly affected metabolites in hematological tissues of both xenografts (BM of MOLM14 xenografted mice (Fig. 2e), and in both the BM and SP of U937 xenografted mice (Fig. 2f, 2g)). Aspartate, asparagine and orotate were depleted, whereas UMP and CMP were found accumulated in U937-xenografted tissues. Interestingly, these changes were not found in SP of MOLM14 CLDX (Fig. 2h). Since engraftment of MOLM14 cells in SP is too small (Additional Fig. 2b), this could indicate that this common modulation is in a blast-dependent manner. These biggest changes in metabolic pathways suggest a rewiring of aspartate metabolism in the nucleotide biosynthesis (Fig. 2i).
Cytarabine treatment modifies systemic metabolism in AML xenografted mice. To understand how chemotherapeutic agents might affect tissue metabolism of AML-engrafted mice, we treated mice with AraC (30 mg/kg/d), or its vehicle PBS for a week (Additional Fig. 2a). As previously described, chemotherapy did not induce any significant reduction of the total cell tumor burden in MOLM14-xenografted model, while it significantly reduced engraftment of U937 cells (fold-change of 13 and 276 in BM and SP, respectively; Additional Fig. 2b-c). Metabolic profiles of the 9 tissues of xenografted mice treated with AraC were performed as described above, with a normalization to the PBS-treated conditions (Fig. 3a).
MOLM14-engrafted mice treated with AraC showed modest modulations compared to PBS treatment (Fig. 3b). Those changes were mostly observed in SP with a net accumulation of AA after treatment (Fig. 3a). Lactate was increased after treatment in both SP and liver, and some nucleotides-related metabolites (orotate, cytidine monophosphate, guanosine diphosphate, and uridine triphosphate) were modulated in BM, SP, pancreas and heart. This absence of global changes in tissue metabolomes after AraC treatment was also observed when comparing naïve mice and AraC-treated MOLM14-engrafted mice (Fig. 3c). In addition, metabolic changes were similar to the one observed by comparing vehicle-treated mice to naïve NSG (Fig. 3c versus Fig. 2b). However, additional changes were observed with an accumulation of galactosamine-1-phosphate (SP and lung), of orotate (kidney), and UDP (muscle), of BCAA (isoleucine, valine and phenylalanine), and pyrimidines (CMP and UMP) in heart.
Importantly, AraC-treated U937-xenografted mice were characterized by major metabolic changes (Fig. 3d). AA were globally accumulated in SP and liver, and depleted in pancreas and muscle. Some common difference in AA were seen in most organs, with an accumulation of aspartate, glutamate, glycine and serine. These AA can all fuel glutathione synthesis for reactive oxygen species (ROS) detoxification, suggesting that AraC treatment on this model globally impact ROS production and anti-oxidant defense. Glycolytic intermediates and phosphorylated sugars were accumulated in both muscle and lung. TCA cycle intermediates amounts were globally unchanged, with the exception of liver. In this tissue, succinate, fumarate and malate were accumulated, suggesting that AA such as glutamine, glutamate and aspartate were more fueled into energy production. Lastly, nucleotides were globally depleted in lung. All these changes led towards a ‘naïve-like’ condition, as tumor burden was greatly decreased upon AraC treatment (Additional Fig. 2b-c). When comparing AraC-treated U937-engrafted mice with naïve ones (Fig. 3e), most changes were reversed, with the exception of phosphorylated sugars, being depleted in SP and accumulated in BM and lung.
Of note, orotate pool was recovered in BM of both xenografted models and in SP of U937-engrafted mice post-AraC (Fig. 3f). AraC-treatment widely impact the energy metabolism and OxPHOS4, and dihydroorotate dehydrogenase (DHODH) one of the enzymes involved in the biosynthesis of nucleotides from aspartate is also dependent on the electron flux by reduction of ubiquinone41. Accordingly, orotate accumulation could imply an increase of the electron transfer chain activity in the whole tissue. Interestingly, orotate was detected in two other organs, lung and kidney, and no significant changes were seen in these organs (Additional Fig. 2d). Therefore, this phenotype is only found in engrafted hematological tissues. Altogether, we proposed that intrinsically chemoresistant MOLM14 cells do not induce global change of tissue metabolomes, whereas U937 significantly undergo a massive decrease in tumor burden with a profound reprogramming in metabolomes back to a ‘naïve-like’ state.
Plasma LC-MS profiling experiment provides metabolic biomarkers of the oxidative phenotype of tumor in vivo. To assess whether plasma metabolome might reflect metabolic signature of specific host tissues or/and AML, metabolic profiling of plasma was performed as described above for tissues (Fig. 4a). Except AA, level of all studied metabolites was changed post-xenotransplantation. TCA cycle intermediates were accumulated in MOLM14 xenograft and depleted in U937 xenograft. Glycolytic intermediates, nucleotides-related, and phosphorylated sugars were accumulated in U937 xenograft and depleted in MOLM14 xenograft. Since U937 and MOLM14 cells displays low and high OxPHOS dependency4, we hypothesized that plasma metabolic signatures might predict the oxidative state of AML tumors. We thus performed an unsupervised principal component analysis (PCA) of metabolomic data in the five conditions (naïve, MOLM14, U937, PBS or AraC; Fig. 4b), and found that 54% of the overall variation in plasma metabolic profiles allowed significant discrimination of the two types of AML xenograft with several metabolites as representative of each xenograft (Fig. 4a). These results suggest a specific signature of central metabolites as biomarkers of the metabolic state of the injected/xenografted AML cells.
To further explore this assumption, we analyzed a second independent cohort of mice (cohort 2), including 3 CLDX models (MOLM14, U937 or KG1a, another low OxPHOS AML cell line), and 8 different patient-derived xenograft (PDX) models (IM06, IM26, IM31, IM76, IM110, Ps325, Ps 6030 and Ps6312; Additional Table 2). Based on the transcriptome of primary patient AML blasts, the oxidative status of 4 PDX models (IM06, IM31, IM76 and Ps325) was characterized by single sample gene set enrichment analysis (ssGSEA) using curated OxPHOS gene signatures (Fig. 4c). Accordingly, Ps325 and IM31 were found to have a low OxPHOS status, whereas IM06 and IM76 display a high OXPHOS status. The plasma metabolic profiles of the cohort 2 were performed, and the results were normalized to the naive state as described in our first cohort (Fig. 4d). We then performed PCA on CLDX and PDX samples with known OxPHOS status, and found that 20% of the global variation allowed discrimination of plasma metabolomes according to the OxPHOS status of injected cells (Fig. 5e). When comparing the contributing metabolites for low and high OxPHOS status in both cohorts (Fig. 5f), we found 13 common metabolites accumulated in low OxPHOS CLDX and PDX models, including 2.3-diphosphoglycerate, 6-phosphogluconate, adenine mono/di/triphosphate, fructose-6-phosphate, guanosine mono-/di-phosphate, glucose-6-phosphate, mannose-6-phosphate, pentoses-5-phosphate and uridine mono/diphosphate. We also found 6 common metabolites accumulated in high OxPHOS CLDX and PDX models, including arginine, citrate/isocitrate, fumarate, lactate, malate and pyruvate. This indicates that this set of murine plasma metabolites might be predictive of the OxPHOS status of tumor cells.