Classic and targeted anti‐leukaemic agents interfere with the cholesterol biogenesis metagene in acute myeloid leukaemia: Therapeutic implications

Abstract Despite significant advances in deciphering the molecular landscape of acute myeloid leukaemia (AML), therapeutic outcomes of this haematological malignancy have only modestly improved over the past decades. Drug resistance and disease recurrence almost invariably occur, highlighting the need for a deeper understanding of these processes. While low O2 compartments, such as bone marrow (BM) niches, are well‐recognized hosts of drug‐resistant leukaemic cells, standard in vitro studies are routinely performed under supra‐physiologic (21% O2, ambient air) conditions, which limits clinical translatability. We hereby identify molecular pathways enriched in AML cells that survive acute challenges with classic or targeted therapeutic agents. Experiments took into account variations in O2 tension encountered by leukaemic cells in clinical settings. Integrated RNA and protein profiles revealed that lipid biosynthesis, and particularly the cholesterol biogenesis branch, is a particularly therapy‐induced vulnerability in AML cells under low O2 states. We also demonstrate that the impact of the cytotoxic agent cytarabine is selectively enhanced by a high‐potency statin. The cholesterol biosynthesis programme is amenable to additional translational opportunities within the expanding AML therapeutic landscape. Our findings support the further investigation of higher‐potency statin (eg rosuvastatin)–based combination therapies to enhance targeting residual AML cells that reside in low O2 environments.


| INTRODUC TI ON
The dramatic progress in deciphering the molecular architecture of acute myeloid leukaemia (AML) has led to the development of a new generation of targeted therapeutic agents, such as the multitargeted kinase inhibitor midostaurin, the isocitrate dehydrogenase (IDH) inhibitors ivosidenib and enasidenib, and the B-cell lymphoma (Bcl)-2 and Hedgehog pathway inhibitors venetoclax and glasdegib, respectively, which recently received FDA approval. [1][2][3][4][5] Despite these undeniable advances, only a limited subset of patients is expected to benefit from such agents, due to the vast molecular heterogeneity and complex clonal architecture of AML. 6,7 Based on extensive pre-clinical and clinical literature, single anti-AML drugs, whether targeted or not, fail to eliminate minimal residual disease (MRD), with relapse generally occurring within three years from diagnosis. [8][9][10][11][12] Overall, the disconnect between the wealth of basic knowledge and dismal clinical outcome remains a defining feature of AML, with the vast majority of patients still relying on cytarabine, a nonspecific nucleoside analog introduced more than four decades ago, as the backbone of most chemotherapeutic regimens. 13 Our efforts concentrate on the development of combinations based on existing AML agents by identifying and exploiting vulnerabilities associated with neoplastic microenvironmental hallmarks. While the impact of O 2 abundance on cell physiology and drug responses has been extensively studied in relationship to solid tumours, 14,15 similar studies in the context of haematological malignancies are comparatively rare. Several lines of evidence suggest that O 2 -deprived (hypoxic) niches within the bone marrow (BM) play a central role in AML drug resistance and disease relapse. 16,17 In tissue culture experiments, O 2 concentrations between 1% and 5% are thought to better mimic oxygenation levels experienced by leukaemic cells in their natural environment. [18][19][20][21] Despite accumulating evidence for diverse artefact-producing effects of excessive oxygenation, 22,23 AML in vitro studies are still routinely performed under atmospheric oxygenation (21% O 2 , ambient air), with potentially significant translational consequences.
In the present study, we investigated the effects of two clinically relevant agents, cytarabine and quizartinib, the former representing the classic anti-AML cytotoxic drug, while the latter being representative for the new wave of highly specific kinase inhibitors, 24 against AML cells under O 2 -controlled conditions. Drug doses chosen for the proteo-transcriptomic studies (cytarabine 1 µmol/L; quizartinib 1 nmol/L) represent a compromise between sublethal and clinical relevance: sufficient for a detectable expected response (eg apoptosis), but still compatible with a majority of cells remaining viability. Our working model is that the surviving pool exhibits shifts in molecular programmes that are critical for surviving the first therapeutic hit. Of note, we used a conceptually similar strategy to extract information about potential synergistic therapeutic partnerships in the context of solid tumours. 25,26 For practical reasons, the initial screens were performed in an established AML cell line (Molm14, M14), with subsequent validation experiments extended to a diverse panel, including primary leukaemic cells.
The results of our integrated screens converged towards a key role of lipid biosynthetic programmes, and in particular cholesterol biogenesis, during AML response to therapy. The results presented herein have significant translational implications, as they provide support for the value of adding higher-potency statins (eg rosuvastatin) to classical and emerging AML therapies.

| Cell culture and reagents
Human AML cell lines, Molm14, and the relatively cytarabine-resistant cell lines OCI-AML3 and THP-1 27 ( Figure S2)  Fluorescein isothiocyanate-Annexin V antibody and propidium iodide were purchased from BD Pharmingen and Sigma-Aldrich.

| Patient samples
Acute myeloid leukaemia blasts derived from whole blood or the bone marrow from 10 newly diagnosed (ND), previously untreated AML patients who consecutively presented to our institution were obtained under guidelines approved by the Institutional Review

| Apoptosis assay
A total of 5 × 10 5 leukaemia cells were seeded in 6-well plates and incubated with cytarabine and/or rosuvastatin under 21% or 1% O 2 conditions as outlined for each experiment. After 48 hours, cells were harvested from the plates, washed twice with cold PBS and then re-suspended in 100 µL binding buffer. 5 µL FITC-Annexin V and 10 µL propidium iodide stock solution (50 µg/mL) were added to each sample and incubated at room temperature in the dark for 15 minutes. Cells were analysed on a BD Accuri (BD Biosciences).

| RNA isolation and quantitative PCR
Molm14 and THP-1 cells were treated with cytarabine for 48 hours.

| Gene set enrichment analysis (GSEA)
Gene Set Enrichment Analysis was performed following GSEA User Guide (The Broad Institute). Briefly, each condition was considered as a group and gene list was ranked with GSEA default ranking metrics. Gene sets from Molecular Signature Database were used in the analysis to identify the pathways significantly enriched in each group. Gene sets were permutated 1000 times to obtain empirical FDR-corrected P-values.

| Reverse protein array (RPPA)
In brief, cell lysate samples were serially diluted twofold for 5 dilutions (undiluted, 1:2, 1:4, 1:8 and 1:16) and arrayed on nitrocellulosecoated slides in an 11 × 11 format to produce sample spots. Sample spots were then probed with antibodies by a tyramide-based signal amplification approach and visualized by DAB colorimetric reaction to produce stained slides. Stained slides were scanned on a Huron TissueScope scanner to produce 16-bit tiff images. Sample spots in tiff images were identified and their densities quantified by Array-Pro Analyzer. Relative protein levels for each sample were determined by interpolating each dilution curve produced from the densities of the 5-dilution sample spots using a 'standard curve' (SuperCurve) for each slide (antibody). All relative protein level data points were normalized for protein loading and transformed to linear values.

| High-performance liquid chromatography (HPLC)
Rosuvastatin and pravastatin were quantified per high-performance

| Statistical analysis
Statistical significance of differences was determined by Student's t test. Difference was considered statistically significant when Pvalue was <.05. Data are presented as mean ± standard error of the mean (SEM) unless stated otherwise. . Statistically significant differences in mRNA expression are indicated as follows: * or ‡ P < .05; ** or ‡ ‡ P < .01; *** or ‡ ‡ ‡ P < .001

| Exposure to cytarabine and quizartinib leads to coordinated down-regulation of the cholesterol biosynthesis pathway
We employed a transcriptome-proteome profiling strategy based on RNAseq and RPPA for a preliminary portrait of the global impact of cytarabine and quizartinib on Molm14 cells. The former drug was chosen as it remains the staple of AML therapy, 13 while the latter is a member of the new class of highly selective targeted agents transitioning from bench to bedside. 24 Our premise was that an in vitro pharmacological treatment is particularly informative if a drug is used at a concentration that leaves the majority of cells viable, but sufficient to elicit molecular

| Assessment of intracellular cholesterol levels in AML cell lines and primary cells after treatment with cytarabine in vitro
Appelbaum, Penn and colleagues noted that high mevalonate pathway activity and cholesterol abundance are hallmarks of AML. 33   This is again reminiscent of the differences between the residual versus original population in the recent in vivo study by Farge et al. 32

| Translational applications of cholesterol biosynthesis blockade using a potent HMG-CoA reductase inhibitor
Applebaum, Penn and colleagues established a proof of concept for mevalonate pathway inhibition using statins as a viable strategy to sensitize AML cells to cytotoxic agents, in addition to providing the first connection between cytarabine treatment and total cholesterol content. [33][34][35] However, in a clinical setting, while some evidence of benefit was cited, phase I-II studies of pravastatin in combination with chemotherapy failed to meet the predefined efficacy criteria for success, with the caveat of a relatively low number of subjects. 36,37 As a result, we utilized rosuvastatin, one of the most potent clinically available HMG-CoA reductase inhibitor. 38 We assessed the anti-leukaemic effects of rosuvastatin and pravas-     with high-dose pravastatin in newly diagnosed and relapsed/refractory AML. While acceptable toxicity and promising CR/CRp rates of up to 80% and 66% were reported in a phase 1 trial of newly diagnosed and relapsed/refractory AML, respectively, a recent phase 2 trial did not meet the criteria for a positive study. 37,41 In consonance with this finding, pravastatin has previously been reported to exert only minimal activity against a variety of cancer cells in vitro. 39,40 While the exact mechanisms underlying this process remain to be elucidated, it has been postulated that the organic anion transporter peptide 1B1 (OATP1B1), which is required for the transmembrane passage of several statins, and to which pravastatin has low affinity, might play a key role in this process. 47 Of note, a case for combining standard anti-AML agents with more potent statins was made previously by Penn and colleagues almost two decades ago. 48,49 However, the promising in vitro effects of cerivastatin were not clinically translatable, due to the withdrawal of this agent from the market. 50 Rosuvastatin, the statin used in our study, has distinct advantages over other statins in that it carries the highest affinity for HMG-CoA reductase and is only minimally metabolized by the liver with almost no interaction with CYP 3A4. 38

ACK N OWLED G EM ENTS
The authors thank Alex Farmer for excellent technical assistance

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.