*Article* **Stable Isotope Labeling Highlights Enhanced Fatty Acid and Lipid Metabolism in Human Acute Myeloid Leukemia**

**Lucille Stuani 1,2, Fabien Riols 3, Pierre Millard 4, Marie Sabatier 1,2, Aurélie Batut 3, Estelle Saland 1,2, Fanny Viars 3, Laure Tonini 2,5, Sonia Zaghdoudi 1,2, Laetitia K. Linares 6, Jean-Charles Portais 2,7, Jean-Emmanuel Sarry 1,2 and Justine Bertrand-Michel 2,3,\***


Received: 7 September 2018; Accepted: 22 October 2018; Published: 25 October 2018

**Abstract: Background**: In Acute Myeloid Leukemia (AML), a complete response to chemotherapy is usually obtained after conventional chemotherapy but overall patient survival is poor due to highly frequent relapses. As opposed to chronic myeloid leukemia, B lymphoma or multiple myeloma, AML is one of the rare malignant hemopathies the therapy of which has not significantly improved during the past 30 years despite intense research efforts. One promising approach is to determine metabolic dependencies in AML cells. Moreover, two key metabolic enzymes, isocitrate dehydrogenases (IDH1/2), are mutated in more than 15% of AML patient, reinforcing the interest in studying metabolic reprogramming, in particular in this subgroup of patients. **Methods**: Using a multi-omics approach combining proteomics, lipidomics, and isotopic profiling of [U-13C] glucose and [U-13C] glutamine cultures with more classical biochemical analyses, we studied the impact of the IDH1 R132H mutation in AML cells on lipid biosynthesis. **Results**: Global proteomic and lipidomic approaches showed a dysregulation of lipid metabolism, especially an increase of phosphatidylinositol, sphingolipids (especially few species of ceramide, sphingosine, and sphinganine), free cholesterol and monounsaturated fatty acids in IDH1 mutant cells. Isotopic profiling of fatty acids revealed that higher lipid anabolism in IDH1 mutant cells corroborated with an increase in lipogenesis fluxes. **Conclusions**: This integrative approach was efficient to gain insight into metabolism and dynamics of lipid species in leukemic cells. Therefore, we have determined that lipid anabolism is strongly reprogrammed in IDH1 mutant AML cells with a crucial dysregulation of fatty acid metabolism and fluxes, both being mediated by 2-HG (2-Hydroxyglutarate) production.

**Keywords:** lipidomics; isotopic profiling; metabolic reprogramming; IDH mutation; leukemia

#### **1. Introduction**

Cancer cells, including acute myeloid leukemia (AML) cells, grow and divide faster and more efficiently than normal cells, which increase their demand for energy, biosynthetic precursors, and macromolecular synthesis [1–4]. Most of them reprogram their metabolism from oxidative phosphorylation to aerobic glycolysis. The finding of this phenomenon, termed the "Warburg effect", stimulated much research on tumorigenesis [4–6]. Over the past twenty years, advanced developments in genetic, omics and high-throughput screening methods have revealed that many of oncogenic signaling pathways regulate cell metabolism in cancer. Therefore, changes in cell metabolism represent a key hallmark in cancer biology [3] and it has been largely demonstrated that metabolic reprogramming in cancer cells occurs far beyond the Warburg Effect [7,8]. Indeed, cancer cells activate various metabolic pathways—e.g., glutaminolysis, amino acid degradation, fatty acid β-oxidation (FAO)—to generate the numerous precursors that are required for macromolecule biosynthesis, such as ribose for nucleic acids and glycerol for lipid synthesis. Dysregulation of lipid-associated pathways is increasingly described in tumors [9–12], and different studies have demonstrated that lipogenesis is significantly up-regulated in human cancers, in particular to respond to higher demands for membrane biogenesis [13–15] or to serve as energy source when nutrients are limited [16,17].

Few years ago, mutations in two key metabolic enzymes, isocitrate dehydrogenases (IDH1 and IDH2) have been discovered in gliomas and myeloid malignancies [18–21]. In AML, 15–20% of patients carry mutations in IDH1 or IDH2 [19,22–25]. This finding has reinforced the interest in studying cell metabolism in this pathology. IDH mutations induce a neomorphic activity resulting from a rearrangement of the enzyme active site favoring the reduction of α-ketoglutarate (α-KG) to D-2-hydroxyglutarate (2-HG) oncometabolite [26,27]. 2-HG strongly structurally resembles α-KG and can function as a potent competitive inhibitor of α-ketoglutarate-dependent enzyme reactions, including dehydrogenases, transaminases, and dioxygenases [28–32]. On the other side, the wild type enzyme catalyzes the interconversion between isocitrate and α-KG and produces NADPH, an essential cofactor that is required for numerous anabolic pathways (nucleotide, fatty acid elongation, lipid synthesis, and cholesterol synthesis) to sustain cell growth and proliferation [2], especially in cancer cells exhibiting aerobic glycolysis (Warburg phenotype), in hypoxia or with defective mitochondria [33–35]. IDH mutations are heterozygous with the conservation of a wild-type allele, suggesting the importance of this wild-type protein to favor the mutant activity. However, while the role of wild type IDH is well documented in normal and cancer cells, the impact of IDH mutation on lipid metabolism, and especially on its respective metabolic fluxes in cancer, is still largely unknown.

In this study we investigated lipid metabolism in AML cells harboring IDH1 R132H mutation, the most common IDH1 mutation. Lipids represent a very large class of molecules that show strong structural diversity (e.g., various combinations of fatty acyls and functional headgroups in phospholipids or various positions for hydroxyl groups on sterol). This chemical heterogeneity, together with the occurrence of many isomeric and isobaric lipid species and the large concentration range over which lipids are found, preclude the measurement of complete lipidomic profiles with a single analytical method. Specific methods are then used for each class of lipids: neutral lipids and fatty acids were analyzed by GC-FID, phospholipids and sphingolipids by LC-MS. Furthermore, the lack of analytical standards for complex lipids hampers the absolute quantification of many molecular species [36]. In this project, lipids were analyzed through different complementary approaches to get a broad coverage of the lipidome [37–39]. The data indicated significant changes in the lipidomic profile of IDH1 R132H cells as compared to WT cells, especially with the increase of phosphatidylinositol, ceramide, and monounsaturated fatty acid. These results encouraged us to investigate the dynamics of lipid synthesis in these cells. This was achieved by using 13C-labeling strategies in which the incorporation of 13C-label from [U-13C]-labeled glucose or [U-13C]-labeled glutamine into fatty acids was measured by GC-MS. The results showed increased de novo synthesis of fatty acids in IDH mutants through the production of 2-HG. Altogether, our investigations show that IDH1 mutation results in

significant reprogramming of lipid metabolism in AML cells and could represent an interesting therapeutic target for this subgroup of patients.

#### **2. Results and Discussion**

#### *2.1. Lipid Metabolism is Dysregulated in IDH1 Mutant Cells*

We first compared the proteome of IDH1 mutant HL60 cells to the one of IDH1 WT cells. A list of proteins that are significantly more abundant (fold change higher than 1.5 and FDR lower than 0.06) have been established (Supplementary Table S1B). Data mining of this specific protein set with Genomatix software revealed major changes in proteins that are associated to pathways of lipid biosynthesis and degradation (Figure 1), while proliferation rates remained unchanged and no significant differences in size, morphology, or doubling time for these cells have been observed. IDH1 R132H cells showed higher content in proteins that are involved in lipid synthesis, including cholesterol and sterol biosynthesis (IDI1, LSS, EBP; Supplementary Table S1). Interestingly, proteins involved in fatty acids (FA) oxidation were also significantly increased in IDH1 R132H cells (ACOX1/ACOX2, HSD17B4, Figure 1), suggesting higher FA catabolism to produce acetyl-CoA and feed the TCA cycle. FAO and lipogenesis are traditionally not described as being operating synchronously because they have opposite functions and are both regulated by ACC activity in opposite ways. However, some studies have demonstrated that FAO was essential to cell survival and metastasis in highly lipogenic solid cancers [40,41]. Of note, German and colleagues [42] have investigated this feature in AML cells. In fact, the authors demonstrated that under nutrient abundance, prolyl-hydroxylase 3 (PHD3) activates specifically ACC2 by hydroxylation, hence favoring malonyl-CoA formation and consequently inhibiting FAO. PHD3 does not act on ACC1, which could therefore maintain lipid synthesis while FAO is upregulated. Furthermore, 2-HG inhibition of α-KG-dependent dioxygenases, including PHDs, has been mainly described [28,43]. As a result, inhibition of PHD3 by 2-HG could prevent ACC2 hydroxylation and malonyl-CoA production to favor FAO in IDH1 mutant AML cells.

**Figure 1.** Enrichment in proteins associated with lipid metabolism in HL60 IDH1 R132H cells compared to WT cells (*n* = 4), based on GO biological processes. (Total) means all the genes encoding the proteins corresponding to the pathways described in the literature while (observed) refers to the genes encoding the proteins found more abundant in IDH1 R132H cells.

Moreover, we and others have shown that FAO and FA translocase/receptor CD36 played a crucial role in cell survival and drug resistance in AML in vitro and in vivo [44,45], thus reinforcing the key role of lipid metabolism in AML.

The comparison of the proteomes of IDH1 R132H and WT AML cells suggests a major reprogramming in the pathways of lipid degradation and biosynthesis, with potentially different impacts considering the diverse classes of lipids affected: mainly with sterol and fatty acids metabolism. A better understanding of this reprogramming process could give a precious insight into the biology of AML and the consequences of IDH1 mutation.

#### *2.2. Lipidomic Experiments Confirmed Changes in Fatty Acid Reorganization in IDH1 R132H AML Cells*

The lipidome of IDH1 R132H cells was measured using a combination of LC-MS and GC methods and compared to the one of WT cells. Lipids were extracted using a universal liquid-liquid extraction method [46] in the presence of internal standards (one per family studied) to follow the sample preparation and to perform relative quantification of the molecular species. Relative quantification of most lipids has been achieved by expressing the intensities of the peaks of interest relative to the area of the internal standard (Arbitrary Unit AU/million of cells). Absolute quantification (μg/million of cells) has been performed for LPC and sphingoïd bases due to the availability of pure standards. Phospholipids were measured by LC-MS/MS [37]. The different classes (PE, PC, PI, PS) were separated by polar head on a HILIC column (except for LysoPC, which were analyzed on an apolar column) [47]. Molecular species (with their number of carbons and number of double bonds) were discriminated by MS/MS based on specific MRM transitions.

The relative amounts of PS, PE, and PC were similar in IDH1 WT and R132H cells (Figure 2A). This result is different from observations reported for gliomas harboring IDH mutation, suggesting a potential metabolic specificity of IDH mutation in AML cells. Indeed, Izquierdo-Garcia et al. [48] measured reduced PC levels while Reitman et al. [49] observed a decrease in PE levels in IDH1/2 mutant gliomas cells compared to IDH1 WT cells. More recently, Viswanath et al. [50] demonstrated that reduced PC and PE amounts in IDH mutant gliomas were due to a decrease in choline kinase and ethanolamine kinase, the enzymes that catalyze the production of PC and PE, respectively. However, we observed a significant increase for LysoPC (+15%) and mainly PI (+82%) families in AML mutant cells. The majority of the PI species were significantly increased, except 32:1; 34:2; 38:3; 38:4; 38:5; 40:5; and 40:6 (details of molecular species profiled for PI are listed in Supplementary Figure S1A). Modifications in intermediates of glycerophospholipid metabolism, such as LysoPC and PI, suggest that membrane trafficking and lipid signaling are stimulated in these cells [51].

Sphingolipids d18 (Cer and SM) were analyzed with the same method as phospholipids, while sphingoïds bases (sphinganine and sphingosine) were analyzed on a C8 column, and could be quantified due to appropriate standards [52]. Interestingly, the total amount of sphingolipids was increased by more than 40% in mutant cells and all of the four sphingolipids classes were enhanced (Figure 2B). For ceramides, if a global tendency to be increased in IDH1 mutant cells has been observed, significant changes have specifically been measured in *N*-(hexadecanoyl)-sphing-4-enine (Cer(d18:1/16:0)) and *N*-(docosanoyl)-sphing-4-enine (Cer(d18:1/22:0)) amounts (details of molecular species profiled for Cer are listed in Supplementary Figure S1B). Sphingomyelins are specific components of the cell membranes as they can form lipid rafts [53] that are essential for membrane protein dynamics and trafficking [51,54]. It is also well known that sphingolipids are key metabolites in oncogenic transformations [55].

Free and esterified cholesterol, as well as triacylglycerides (TG), were analyzed by GC-FID. The total amount of neutral lipids was unchanged between IDH1 WT and R132H cells, but the distribution of the molecular species was different. Indeed, we observed a decrease in esterified cholesterol in mutant cell counterbalanced by higher proportion of free cholesterol, while TG remained stable (Figure 2C).

Finally, the total FAs were profiled. Esterified FAs of the total extract (i.e., glycerolipids) were hydrolyzed in basic conditions and they were derivatized to be analyzed by GC-FID. A slight but significant increase in total FAs was observed in IDH1 mutant cells (+8%), which was mainly due to higher amounts of monounsaturated FAs (+17%). Polyunsaturated and saturated FAs remained stable (Figure 2D).

As expected regarding proteomics experiments, quantitative lipidomics data confirmed that IDH1 mutation leads to a re-organization of lipid metabolism, especially sphingolipids, lysoPC, the balance

between cholesterol and cholesterol esters, and total FAs. In order to better understand how IDH1 mutation could be involved in the accumulation of lipids, we decided to apply stable isotope labeling experiments to identify pathways leading to total FA accumulation.

**Figure 2.** Lipidomic analysis in HL60 AML IDH1 WT (blue dots) and IDH1 R132H cells (red dots) (*n* = 4). (**A**) Phospholipids; (**B**) Sphingolipids; (**C**) Neutral lipids and percentages of each of its constituents; and, (**D**) Total Fatty Acids.
