Harnessing the Power of Sphingolipids: Prospects for Acute Myeloid Leukemia

Abstract

Acute myeloid leukemia (AML) is an aggressive, heterogenous malignancy characterized by clonal expansion of bone marrow-derived myeloid progenitor cells. While our current understanding of the molecular and genomic landscape of AML has evolved dramatically and opened avenues for molecularly targeted therapeutics to improve upon standard intensive induction chemotherapy, curative treatments are elusive, particularly in older patients. Responses to current AML treatments are transient and incomplete, necessitating the development of novel treatment strategies to improve outcomes. To this end, harnessing the power of bioactive sphingolipids to treat cancer shows great promise. Sphingolipids are involved in many hallmarks of cancer of paramount importance in AML. Leukemic blast survival is influenced by cellular levels of ceramide, a bona fide pro-death molecule, and its conversion to signaling molecules such as sphingosine-1-phosphate and glycosphingolipids. Preclinical studies demonstrate the efficacy of therapeutics that target dysregulated sphingolipid metabolism as well as their combinatorial synergy with clinically-relevant therapeutics. Thus, increased understanding of sphingolipid dysregulation may be exploited to improve AML patient care and outcomes. This review summarizes the current knowledge of dysregulated sphingolipid metabolism in AML, evaluates how pro-survival sphingolipids promote AML pathogenesis, and discusses the therapeutic potential of targeting these dysregulated sphingolipid pathways.

1. Introduction

Acute myeloid leukemia (AML) is the most common acute leukemia in adults and comprises a heterogeneous group of hematological malignancies characterized by abnormal proliferation of immature hematopoietic progenitor cells of the myeloid lineage [1]. An uncommon disease, AML accounts for 1.1% of all cancers and has a 5-year survival rate of 29.5% following diagnosis. In the U.S., an estimated 20,240 new AML cases will be diagnosed and approximately 11,400 people will die due to AML in 2021, resulting in 1.9% of all cancer deaths [1]. The disease incidence increases while survival rates decline with ageing [1–3]. Moreover, the incidence of AML appears to be increasing in recent years due to the overall aging population, toxic exposures, and prior use of chemotherapy and/or radiation for the treatment of other malignancies [1, 4].

Risk classification for AML in adults at the time of diagnosis is defined by cytogenetic and mutation events [4]. Further risk assessment can be determined using the detection of measurable residual disease [5–7]. Additional classifications (e.g., somatic events, epigenetic, and gene expression signatures) have also been proposed, but are not in routine clinical use [8–14]. However, this hematopoietic cell malignancy is heterogeneous in cell phenotypes and molecular events [2, 15]. Cell phenotypes (e.g., cytogenetics, somatic variants, cell surface markers) are utilized for diagnostic and prognostic purposes as well as measurable residual disease monitoring [6, 7]. In recent years, genomic studies of AML patient specimens have significantly expanded the understanding of disease heterogeneity [16–19]. These studies have defined patient groups harboring distinct molecular alterations, such as cytogenetic events and somatic mutations in genes involved in epigenetic regulation, cell signaling, tumor suppression, and transcriptional network modulation. Furthermore, disease heterogeneity has expanded to include epigenetic landscapes where mutation-associated and independent DNA methylation patterning associates with distinct disease biology and clinical outcomes [9, 10, 18, 20].

For many decades the main treatment option for AML patients at the time of diagnosis was intensive 7+3 induction chemotherapy consisting of daily infusions of cytarabine (Ara-C) supplemented with an anthracycline on the first three days [15]. Most patients respond to this therapeutic approach and the 5-year survival rate is 40–50% for younger patients with de novo AML [21–25]. Exceptions are patients with acute promyelocytic leukemia, which has a cure rate upwards of 90% when using differentiating agents such as all-trans-retinoic acid and arsenic trioxide [15]. Ultimately, if patients respond to treatment and achieve remission, further treatment with allogeneic hematopoietic stem cell transplantation can be considered for all eligible patients. If patients achieve remission after induction chemotherapy, intensive post-remission consolidation therapy with Ara-C can also be considered [26–29]. Unfortunately, most patients will relapse following induction therapy, and age-related toxicities preclude older patients from receiving intensive chemotherapy, highlighting the need for alternative therapeutic approaches [30].

Therapeutic options have evolved to target molecular and epigenetic characteristics of the disease due to the availability of novel agents and a better understanding of disease pathogenesis [15, 23, 24, 31]. Venetoclax, an orally available selective Bcl-2 inhibitor, is now routinely combined with the traditional single-agent treatments of hypomethylating agents (azacitidine or decitabine) or low-dose Ara-C (LDAC) for frontline AML treatment [32–34]. Glasdegib, an oral hedgehog pathway inhibitor, is also approved combined with LDAC in the same treatment setting [35, 36]. Patients who have specific mutations or molecular aberrations can be considered for other newly approved drugs. These targetable aberrations include AML expressing: CD33 (gemtuzumab ozogamicin [31]); FMS-related tyrosine kinase 3 gene (FLT3) internal tandem duplication mutation (ITD subtype) or point mutation in the tyrosine kinase domain (TKD subtype) (midostaurin [37] and gilteritinib [38]); and isocitrate dehydrogenase mutations (IDH1 (ivosidenib [39]) and IDH2 (enasidenib [40])). These agents can be considered in the frontline and/or relapsed settings. Although there are many recent drug approvals for AML, relapse rates remain high [41, 42].

Current treatments need to be adapted and combined with novel therapeutic regimens to improve clinical outcomes. Bioactive sphingolipids (SLs), once overlooked as just a structural component of membranes, control many cellular processes of paramount importance in cancer progression and represent promising therapeutic targets in cancer. Hence, there is an opportunity to develop SL therapeutics in AML, especially for non-responsive and relapsed patients. In this review, we summarize the current knowledge of SL dysregulation in AML, highlight SL alterations downstream of current AML therapeutics, and discuss the prospects of harnessing SL-based treatments to improve AML patient care.

2. Sphingolipids: Metabolism, clinical relevance, and therapeutic potential

SLs are a unique class of bioactive lipids that regulate cellular processes such as growth, proliferation, apoptosis, autophagy, chemotherapeutic response, stemness, and immune cell activation. They accomplish these functions both directly, as messenger molecules, and indirectly, through lipid raft formation and receptor component stabilization within membranes [43–48]. To effectively regulate the potent bioactive effects of SL signaling, cells utilize a complex network of enzymes that catalyze the generation, breakdown, and interconversion of SL species (depicted as subway systems in Figure 1). The complexity of SL metabolism and biology is further compounded by families of enzymes with distinct as well as redundant functional specificity (e.g., ceramide synthases, ceramidases), varying biological effects of SL species based upon fatty acyl chain length (e.g., C16 to C24) as well as trafficking and sequestration of SL within distinct cellular compartments. Considering the size of the SL network and the numerous signaling pathways it regulates, it is not surprising that dysregulation of this network can be catastrophic to cells, often causing or exacerbating diseases such as lipid-storage disorders and cancer [49]. Fortunately, an improved understanding of these pathways has led to clinical advancements for treating patients with lipid disorders, and efforts to advance SL-based cancer treatments into the clinic are ongoing. In this section, we briefly summarize key SL metabolic pathways and highlight noteworthy findings relevant to cancer biology and therapeutic intervention. For a complete summary of SL biochemistry, metabolism, and signaling, refer to these comprehensive reviews: [43, 45, 47, 50, 51].

Figure 1. Transit map of sphingolipid metabolism.

Ceramides reside at the canonical center of the sphingolipid metabolism transit map and serve as the central hub for the other sphingolipids. De novo synthesis begins with the condensation of serine and palmitoyl-CoA by the serine palmitoyltransferase (SPT) complex and further processing by 3-ketodihydrosphingosine reductase (KDSR), dihydroceramide synthases (CERS1–6), and dihydroceramide desaturases (DEGS1–2) to generate ceramide (D train). Other methods of generating ceramides include sphingomyelin hydrolysis (S train) by sphingomyelin phosphodiesterases (SMPD1–4) or CERS activity as part of the exit/salvage pathway (E train). Conversely, there are multiple routes to decrease cellular ceramide content. The addition of a phosphocholine head group by the sphingomyelin synthase enzymes (SGMS1–2) generates sphingomyelin (S train). Ceramidases (ASAH1–2 and ACER1–3) hydrolyze ceramide into sphingosine and free fatty acids, and sphingosine kinases (SPHK1–2) phosphorylate sphingosine into sphingosine-1-phosphate (E train). S1P breakdown by sphingosine-1-phosphate lyase (SGPL1) generates phosphoethanolamine and hexadecenal. Ceramide kinase (CERK) phosphorylates ceramide to generate ceramide-1-phosphate (P train). The synthesis of complex glycosphingolipids begins with the action of galactosylceramide synthase (UGT8) to generate galactosylceramides (G train north) or glucosylceramide synthase (UGCG) to generate glucosylceramides (G train south).

2.1. De novo sphingolipid synthesis and generation of ceramide

The de novo SL synthesis pathway begins with the serine palmitoyl transferase-mediated condensation of serine and palmitoyl-CoA to form 3-ketodihydrosphingosine, which contains the sphingoid backbone common to all SLs. 3-ketodihydrosphingosine reductase (KDSR protein and gene abbreviation) then generates dihydrosphingosine. De novo synthesis culminates with the addition of a fatty acid moiety to the sphingoid backbone by one of six ceramide synthases (CERS1–6 protein and gene abbreviation) and subsequent desaturation of the sphingoid backbone by the desaturase enzymes (DEGS1–2 gene, DES1–2 protein abbreviation) to generate ceramide [43, 45]. While DES1-mediated ceramide generation from dihydroceramide is well characterized, the function of DES2 is less clear and has been implicated in phytoceramide synthesis from dihydroceramides [52]. However, the tissue specificity of DES2 expression is highly restricted and likely excludes hematopoietic tissue (proteinatlas.org), which limits its relevance in leukemia [53]. The six different CERS isoforms preferentially add distinct fatty acids to the sphingoid backbone (Fig. 1, D train) [54–56]. The resulting ceramides sit at the canonical center of the SL metabolic network and can be modified and/or broken down to form nearly every other SL species.

2.2. Ceramide catabolism

Ceramidase enzymes are lipid hydrolases that initiate the breakdown of ceramide via the “exit/salvage pathway.” Although roles for neutral (ASAH2 gene) and alkaline (ACER1–3 gene) ceramidases have been described in cancer, the most studied ceramidase enzyme in cancer is acid ceramidase (ASAH1 gene, AC protein abbreviation) [45]. Working at an acidic pH primarily within the lysosome, AC hydrolyzes ceramides into sphingosine and free fatty acids that can then be utilized as nutrients for the cell or recycled to form other complex lipids [45]. Sphingosine generated from this reaction can either be re-acylated via CERS to regenerate ceramides or phosphorylated by sphingosine kinases (SPHK1–2 protein and gene abbreviation) to generate sphingosine-1-phosphate (S1P) (Fig. 1, E train). S1P is often viewed as antithetical to ceramide and exerts pro-survival effects, in part, by directly binding S1P G protein-coupled receptors (S1PR1–5 protein and gene abbreviation) to regulate autophagy, T-cell migration, and cell survival [57–60]. Additional regulatory enzymes include S1P phosphatase (SGPP1–2 gene), which dephosphorylates S1P to re-generate sphingosine, and S1P lyase (SGPL1 gene), which catabolizes S1P into hexadecenal and ethanolamine-phosphate to exit the SL pathway [45]. Of clinical relevance, several immunomodulating S1PR targeting drugs are clinically approved to target dysregulated T-cell activation and inflammatory signaling (fingolimod and ozanimod) [45, 61, 62].

2.3. Ceramide modifications and conversion into higher-order sphingolipids

Apart from ceramide catabolism, three additional primary pathways deplete ceramides and/or use them as scaffolds to build higher-order SLs: phosphorylation, glycosylation, and sphingomyelin synthesis. Phosphorylation by ceramide kinase (CERK protein and gene abbreviation) generates ceramide-1-phosphate (C1P) (Fig. 1, P train). C1P is thought to induce both cell proliferation and migration, but receptors are poorly defined [45, 58]. One or more sugars may be added to the hydroxyl group of ceramides to generate glycosphingolipids. Galactosylceramide synthase (UGT8 gene) regulates the synthesis of galactosylceramides (Fig. 1, G train north). In contrast, the “gatekeeper” enzyme, glucosylceramide synthase (UGCG gene, GCS protein abbreviation), adds a glucose moiety to form glucosylceramides, which can be further processed to increasingly complex glycosphingolipids (e.g., gangliosides and globosides) (Fig. 1, G train south) [63]. The enzymes GBA1–2 and GALC remove sugar groups from glucosylceramides and galactosylceramides, respectively, to re-generate ceramides [63]. Many lysosomal storage disorders such as Tay-Sachs and Gaucher disease result from a build-up of glycosphingolipids, highlighting the clinical significance of these lipids and their regulating enzymes [64]. Recent studies have elucidated structural roles for various glycosphingolipids in other pathologies as well. Specifically, glucosylceramides stabilize membranes in response to bacterial infection and prevent viral uptake [65, 66]. Others have also linked galactosylceramides and lactosylceramides as secondary messengers to inflammatory pathways [67].

Cellular ceramide levels can also be reduced via the sphingomyelin synthesis pathway. Sphingomyelin synthases (SGMS1–2 gene, SMS1–2 protein abbreviation) transfer a phosphocholine headgroup to ceramide, generating sphingomyelin and diacylglycerol as a byproduct [43]. In some instances, phosphoethanolamine can substitute for phosphocholine [45]. Meanwhile, sphingomyelinase enzymes (SMPD1–4 gene, SMase protein abbreviation) catabolize sphingomyelin back into ceramide (Fig. 1, S train) [43, 45]. Sphingomyelins are of immense structural and signaling importance. Sphingomyelins, along with glycosphingolipids, stabilize cholesterol-dependent raft domains that dictate the thickness and curvature of the membrane, bring together receptors and other components to facilitate signal transduction, and alter interactions with nearby cells and extracellular pathogens [48]. Similar to the lysosomal storage diseases caused by glycosphingolipid dysregulation, mutations in the lysosomal acid SMase (SMPD1 gene) cause Niemann-Pick Disease through the accumulation of sphingomyelin [43]. Moreover, sphingomyelin metabolism is highly relevant to cancer biology, as radiotherapy and a multitude of chemotherapies induce sphingomyelin breakdown into pro-death ceramides to promote cancer cell death [43].

Although the central dogma of dysregulated SL metabolism in cancer states that decreased ceramides and increased S1P promote cell survival, SL signaling is complex and requires further investigation [68]. Indeed, many aspects of the regulation of the complex SL network described above and the specific function of each lipid species therein remain incompletely defined. However, the immense impact of these bioactive molecules on health and disease is exemplified by their linkage with nearly every major metabolic pathway and illness. For this reason, a better understanding of these metabolic pathways, enabled by sphingolipidomics, may redefine the 21^st century approach to medicine as the SL field matures from mere membrane components to bona fide prognostic disease biomarkers and/or promising therapeutic targets. The clinical and pathological significance of SLs in AML are best viewed through their roles in many critical aspects of cell biology, as described below.

3. Sphingolipid dysregulation in AML

SL dysregulation plays a major role in cancer through the aberrant expression and localization of SL enzymes, metabolites, and receptors, which results in increased pro-survival lipids and/or blockade of death pathways [43, 46, 47]. The characterization of SL dysregulation in AML has gained traction within the last five to ten years. In general, ceramide-catabolizing enzymes, S1P-synthesizing enzymes, and S1PRs are aberrantly expressed in AML and promote proliferation, survival, drug resistance, and leukemogenesis [69–78]. One notable example showed that increased S1P signaling through elevated S1PR3 expression induced AML in mice [78]. Overall, cumulative evidence strongly supports a higher pro-survival/pro-death lipid ratio in AML and thus, presents a relatively untapped source of both prognostic and diagnostic information [43, 78–80].

Given their key role in cancer progression, many approaches seek to target SLs and their dysregulation for clinical benefit. Strategies aim to either inhibit the generation of pro-survival S1P or promote the formation and accumulation of pro-death ceramides [81]. Noteworthy examples include studies that target aberrant SL players to significantly reduce leukemic burden and increase survival in leukemic mice [70, 73, 74, 82], as well as examples demonstrating improved drug efficiency by combining SL-based therapies with current AML treatments [69, 74]. Importantly, chemotherapy and radiotherapy increase endogenous ceramide levels in cancer cells, which is necessary for their therapeutic effects [83, 84]. SL dysregulation also plays a vital role in drug resistance [69, 85] and leukemic stem cell renewal [86]. Taken together, the SL pathway is a promising target for therapeutic intervention [43, 79–81, 87, 88]. A summary of dysregulated SL signaling in AML is depicted in Figure 2. We also refer readers to important work by Tan et al. [79] and Lewis et al. [80, 88], which have previously reviewed specific aspects of SL metabolism and targeting in AML. The following sections provide the first comprehensive review of the role of SLs in AML survival, drug resistance, and leukemic stem cells, as well as a discussion of SL enzymes and metabolites as predictive biomarkers.

Figure 2. Summary of dysregulated sphingolipid signaling in AML.

The intersections between dysregulated sphingolipid metabolism and oncogenic signaling in AML are many. Multiple ceramide catabolizing enzymes (AC, SPHK, ACER3) are elevated in de novo AML. Multidrug resistance also upregulates the metabolite C1P and the enzymes AC, SPHK, GCS, and SMS. Overexpression of AC, SPHK, or GCS is sufficient to induce upregulation of MDR1, likely through glycosphingolipids or S1P-mediated NF-κB signaling. Oncogenic FLT3-signaling inhibits ceramide generation by antagonizing CERS1. Overexpression of AC, ACER3, SPHK, and the S1P transporter SPNS2, correlate with poorer overall survival. Bcl-2 and Bcl-xL inhibit the formation of ceramide pores. Ceramides and S1P signaling regulate levels of the pro-survival Bcl-2 family protein, Mcl-1. In addition, S1P-S1PR signaling upregulates levels of phospho-Akt and phospho-ERK. Signaling through S1PR3 promotes myeloid differentiation. KDSR, involved in de novo SL synthesis, prevents ER stress to promote leukemic maintenance. Black arrows = metabolic pathway; green arrows = activating; red inhibitory arrow = inactivating; red text = overexpressed and/or oncogenic proteins.

3.1. Sphingolipid metabolism and AML mutations/chromosomal abnormalities

AML is highly heterogeneous in disease presentation, mutational burden, and prognosis [8, 16, 17, 89]. Although efforts to link aberrant SLs and specific AML molecular alterations are just beginning, a few examples have emerged. In a lipidomic analysis of bone marrow aspirates from AML patients with translocation t(8;21), inversion (inv16), or normal karyotype, Stefanko et al. found that t(8;21) patients have decreased sphingomyelin content and increased membrane fluidity through the shift to polyunsaturated fatty acids. Compared to normal karyotype, both t(8;21) and inv16 AML samples had higher levels of glucosylceramides and decreased core binding factor expression, which regulates several genes in SL metabolism [90].

FLT3

FLT3 is a frequently mutated gene in AML encoding a receptor tyrosine kinase that canonically promotes proper hematopoietic stem cell (HSC) and progenitor cell development. Extracellular FLT3 ligand binds to and activates FLT3 receptor to promote proliferation and differentiation via PI3K, RAS, and STAT5 signaling. FLT3 mutations lead to constitutively activated signaling to promote leukemogenesis and cell survival [91]. Dany et al. published that aberrant FLT3-ITD signaling prevented the formation of endogenous pro-death C18-ceramide by antagonizing CERS1 [92].

IDH1

IDH1 is another frequently mutated gene in AML and the mutant enzyme mediates the conversion of α-ketoglutarate to D-2-hydroxyglutarate. This oncometabolite inhibits α-ketoglutarate-dependent enzymes that regulate epigenetic and signaling pathways [93]. Stuani et al. showed that IDH1^R132H mutant expression in the HL-60 AML cell line changed the global SL profile, including upregulation of sphingosines and ceramides [94]. In glioma cells, IDH1^R132H expression was associated with decreased S1P, increased ceramides, and increased N,N-dimethyl sphingosine, an endogenous SPHK inhibitor [95]. Interestingly, supplementing IDH1-mutant glioma cells, but not IDH1 wild-type cells, with N,N-dimethyl sphingosine and sphingosine C17 decreased cell growth. These findings highlight a SL-based therapeutic vulnerability that is selectively detected in IDH1-mutant glioma cells. This suggests that the presence of the IDH1^R132H mutation in glioma induces a growth disadvantage compared to IDH1 wild-type cells due to the decrease in mitogenic S1P and increase in ceramides. It remains to be seen whether this vulnerability exists in AML. Further studies are needed to define the mechanism and extent of SL changes associated with common AML somatic mutations and their contribution to disease pathology. Moreover, the IDH1 mutation-specific vulnerability to a specific SL inhibitor identified in glioma may represent a new paradigm for SL targeted therapeutics in molecularly defined AML subtypes.

3.2. Sphingolipid dysregulation controls mechanisms of cell death

Dysregulation of bioactive SLs influences cellular signaling and survival pathways, and targeting this dysregulation has revealed novel mechanisms underlying AML survival (summarized in Figure 2 and detailed in the following sections). Here, we discuss the current knowledge and unresolved questions of SLs regulating cell death.

3.2.1. Apoptotic cell death

Leukemic development and progression relies on evasion of apoptotic stimuli triggered by tumor suppressor activation or chemotherapy treatment, which typically culminate in the activation of executioner caspases and apoptotic cell death [96]. Intrinsic apoptosis is primarily regulated at the mitochondria where a delicate balance of SLs and pro-apoptotic (Bim, Bax, Bak, etc.) vs anti-apoptotic (Bcl-2, Mcl-1, Bcl-xL, etc.) Bcl-2 family proteins regulate mitochondrial outer-membrane permeabilization (MOMP) and the release of pro-apoptotic effectors [46, 51]. The ability to resist apoptosis is a classic hallmark of cancer that is regulated in part by SL metabolism. The roles of ceramide as an important signaling molecule and pro-death lipid were initially described by Kolesnick et al. and Obeid et al. in their seminal manuscripts published in 1992 and 1993, respectively [97, 98]. Since then, the role of SLs in intrinsic apoptosis has been evaluated further and key examples of this relationship include: (1) ceramides increase following chemotherapy and are necessary for cell death [83, 84, 99, 100]; (2) inhibition of anti-apoptotic Bcl-2 family proteins increases ceramides [44]; (3) ceramides putatively form “water-filled” channels that permeabilize the mitochondrial outer membrane [101]; (4) SLs are required for proper Bax/Bak function to permeabilize the mitochondrial outer membrane [44]; and (5) ceramide conversion into sphingomyelin, glycosphingolipids, and S1P typically promotes cell survival [43].

Apoptosis resistance is a major mechanism of relapse in AML, which is mediated in part by the overexpression of anti-apoptotic Bcl-2 family proteins [102]. Venetoclax is a selective Bcl-2 antagonist that indirectly activates pro-apoptotic effector proteins (Bim, Bax, Bak) by preventing their association with Bcl-2 [15, 103]. The recent approval of venetoclax in combination with azacitidine or LDAC for elderly AML patients and AML patients unable to undergo intensive chemotherapy highlights the importance of targeting Bcl-2 family proteins [32, 33]. Of additional importance, Mcl-1 is a potent survival factor and mediator of venetoclax resistance in AML due to its shared ability to sequester pro-apoptotic effectors [104, 105]. We have shown that ceramide metabolism modulates expression of the anti-apoptotic Bcl-2 family protein, Mcl-1. Tan et al. demonstrated that AC gene expression is elevated in TCGA AML samples, and validated this finding via gene expression micro-array [70]. AC knockdown in AML cell lines, including drug-resistant variants, reduced total Mcl-1 levels and decreased cell viability [70, 82]. This finding was corroborated with pharmacological inhibition of AC with the lysosomotropic B-13 inhibitor analog, LCL-204. Exogenous addition of ceramide was sufficient to reduce Mcl-1 levels. Conversely, overexpression of AC increased Mcl-1 levels [70]. This study details important links between Mcl-1 and SL biology although the molecular mechanism by which AC and ceramide levels modulate Mcl-1 is unknown. Of note, the alkaline ceramidase, ACER3, is also elevated in AML and its knockdown suppresses cell viability and colony-forming potential in multiple AML cell lines through mechanisms yet to be determined [77].

Ceramidases catabolize ceramides to sphingosine, which can then by phosphorylated by sphingosine kinases SPHK1 and SPHK2 to generate S1P. Pro-survival properties of SPHKs and their product, S1P, have also been identified in AML. SPHK1 likely promotes cell survival through ERK1/2 and Akt signaling since inhibition of SPHK1 with SK1-I significantly downregulated phospho-ERK1/2 and phospho-Akt [106]. A newly developed dual Akt and SPHK1 inhibitor, A-674563, also demonstrated anti-leukemic efficacy [107]. Powell et al. recently demonstrated that pharmacological and molecular inhibition of SPHK1 by MP-A08 and shRNA, respectively, depleted Mcl-1 levels. Signaling through S1P is mediated in part by binding the S1PR G-protein coupled receptors. Treating cells with the putative S1PR2 antagonist, JTE-013, phenocopied the MP-A08-mediated depletion of Mcl-1, suggesting that signaling through S1PR2 modulates Mcl-1 protein levels [74]. However, S1PR2 knockdown did not alter AML cell survival, and additional mechanistic studies of JTE-013 suggest off-target inhibition of DES1 and SPHK1/2, which may be contributory [108]. Taken together, the multiple studies demonstrating that SL signaling modulates levels of Mcl-1, a known mediator of venetoclax resistance [105], suggest that co-targeting AC or SPHK1 may improve venetoclax efficacy in AML.

Endoplasmic reticulum (ER) stress is frequently upregulated in leukemias and results from the accumulation of misfolded proteins as a consequence of rapid proliferation [109]. Cumulative evidence suggests that the unfolded protein response (UPR) is upregulated in AML, restores ER homeostasis, and maintains leukemic survival [109, 110]. Recent work by Liu et al. demonstrated that dysregulated SL metabolism promotes UPR in AML to maintain cell survival [111]. CRISPR knockout of KDSR, the enzyme responsible for the reduction of 3-ketodihydrosphingosine during de novo SL synthesis, results in apoptotic cell death and cell cycle arrest in multiple AML cell lines. Mechanistically, KDSR loss increased 3-ketodihydrosphingosine levels, which downregulated UPR checkpoint proteins, disrupted ER structure, and elicited cell death [111].

Protein phosphatase 2 (PP2A) is a cellular serine-threonine phosphatase that negatively regulates oncogenic signaling, apoptotic signaling, and cell cycle progression. The enzyme is recurrently inactivated in AML via endogenous PP2A inhibitory proteins such as CIP2A, I1PP2A, and I2PP2A (SET) [112, 113]. PP2A regulation is also dependent on SL metabolism [114–116] and ceramides were described as an activator of PP2A [117]. Interestingly, ceramides are also direct ligands of SET, which is a putative inhibitor of PP2A, and thus loss of ceramide correlated with PP2A inactivation [114]. Chen et al. demonstrated that treating Kasumi-1 AML cells with the S1PR-modulating drug and SPHK1 inhibitor, fingolimod (FTY720), induced caspase-dependent cell death coinciding with upregulation of ceramide and PP2A activity. Knockdown of PP2A activity partially inhibited cell death [118]. FTY720 was also recently described as a ligand of SET that can activate PP2A by directly blocking the SET-PP2A interaction [119], which would complement ceramide-mediated SET inactivation.

3.2.2. Other mechanisms of cell death

Autophagy is a process by which cellular materials are sequestered in double-layer membranes that contain microtubule-associated proteins 1A/1B light chain 3B (LC3B) and are termed autophagosomes [120]. Autophagosomes traffic to the lysosome and are degraded as a form of cellular recycling. Mitophagy is a specialized form of autophagy that involves the recycling of dysfunctional or damaged mitochondria to preserve energy metabolism [121].

The role of autophagy in AML appears to be multifaceted and variable based on the context. Numerous studies have shown that pharmacological, molecular, and genetic manipulation of normal autophagic pathways promote leukemia [122]. Intriguingly, modulating these same autophagic pathways appears to enhance certain AML therapeutics – especially FLT3-targeted therapies such as quizartinib and crenolanib. Thus, it can be appreciated that manipulation of the autophagic pathway holds therapeutic promise [122].

It is well established that various SLs and SL analogs can induce both autophagy and mitophagy via a myriad of mechanisms including increased expression of autophagic proteins (LC3B, Beclin 1, and BNIP3), downregulation of nutrient transporters, or direct binding to LC3B-II, which promotes the formation of autophagosomes [43, 46, 120]. Importantly, just as autophagy itself can be beneficial or detrimental in AML, these SL effects can be either protective or lethal depending on cell type, with growing evidence for the latter in AML.

Previous AML data showed that ceramide and ceramide analogs directly bind LC3B-II and target autophagosomes to the mitochondria [123]. Further studies highlighted that the FLT3 inhibitor crenolanib increased ceramide synthesis, mitophagy, and cell death in MV4–11 and MOLM-14 AML cell lines [92]. Also in this study, the positively-charged mitochondria-targeting pyridinium-ceramide analog, LCL-461, induced autophagy and mitophagy. Finally, C6-ceramide co-treatment with tamoxifen increased autophagic flux and synergistic cell death in multiple AML cell lines in a FLT3-ITD-independent manner [124]. Importantly, inhibition of autophagy via knockdown of LC3B and ATG5 reduced cell death from LCL-461 and ceramide plus tamoxifen, respectively. Therefore, ceramide-induced autophagy and mitophagy appear lethal in AML.

In summary, the links between SLs and cell death in AML are many. The recurring theme is that ceramide accumulation is blunted by overexpression of ceramide-catabolizing ceramidases and downregulation of ceramide synthases. This, accompanied with overexpression of SPHKs, promotes a pro-survival phenotype in AML. Indeed, pharmacological and genetic inhibition of these enzymes provide therapeutic benefit in preclinical models. The link between SL metabolism and Mcl-1 levels provides rationale for SL-targeting therapies in combination with venetoclax. However, much remains unknown regarding SLs and AML pathogenicity. For instance, the consequence of less-studied SL biology (e.g., ceramide-1-phosphate, dihydrosphingolipids, sphingoid backbone chain lengths, fatty acyl chain lengths) in AML is not known. While more studies are needed to address these and other uncertainties, the current links between SLs and AML cell death encourage future therapeutic development to target this vulnerability.

3.3. Drug resistance

Drug resistance arising in biological systems is ever-present and widespread. Owing to evolutionary processes and/or sudden selection pressures, nature has provided defenses, some beneficial, some not. These defenses protect from toxins of all kinds, protection that unfortunately extends to chemotherapy drugs. In AML, despite advances in cure rate, a considerable number of patients die from the disease due to the occurrence of multidrug resistance. Drug efflux driven by membrane transport proteins such as multidrug resistance protein 1 (MDR1, also known as P-glycoprotein) [125–129] and multidrug resistance-associated protein 1 (MRP1) [130–132] have been established as the most significant multidrug resistance mechanisms found in AML. Of clinical relevance, a negative correlation between MDR1 and/or MRP1 overexpression, remission rate, and disease-free survival has been demonstrated in patients with AML [130, 133–136]. Even with the known clinical relevance of these pervasive, hallmark proteins, targeting drug transporters has proven elusive [137, 138]. For example, the novel MDR1 antagonist, zosuquidar, in combination with conventional induction chemotherapy, failed to improve AML patient outcomes in a randomized, placebo-controlled trial for newly diagnosed AML [139].

Recent studies in chemotherapy selection pressure have revealed a curious link between MDR1 and SL metabolism. Specifically, wild-type, drug-sensitive AML variants challenged with heterocyclic anticancer agents (daunorubicin, vincristine) generated drug-resistant cells that demonstrated synchronous upregulation of MDR1 as well as several key enzymes of SL metabolism: GCS, AC, SPHK1, and SMS [85, 140–143]. Therefore, selection pressure upregulates MDR1 [128, 144, 145], but additionally enhances expression of enzymes that decrease ceramide, steps that cooperatively contribute to anti-apoptotic, mitogenic behavior that sabotages ceramide-driven cancer cell death. This is significant because mechanisms of cytotoxicity of heterocyclic anti-cancer drugs are, in part, driven by their capacity to generate cytotoxic levels of ceramides [146]. This unique association of SL metabolism with drug resistance has been supported in various works [147–149]. Interestingly, we and others have shown that forced overexpression of GCS, AC, and SPHK1 is sufficient to confer multidrug resistance through direct modulation of MDR1 expression and activity. Importantly, inhibition of these enzymes attenuated multidrug resistance and, in some cases, improved drug retention in AML [69, 71, 150]. In addition, we recently interrogated the consequence of multidrug resistance on mitochondrial bioenergetics [151]. While drug-resistant AML cell lines show increased basal and maximal respiration, they demonstrate a significantly lower fractional oxidative phosphorylation attributed to impaired respiratory complex I compared to wild-type cells. Highlighting the therapeutic implications of this finding are the superb synergistic killing effects observed when SL-modulating agents and the respiratory complex I inhibitor, phenformin, are combined [151].

In addition to its canonical function as a drug exporter at the plasma membrane, MDR1 can act as a “flippase”, transporting glucosylceramides into the Golgi lumen [152]. This raises questions regarding its possible role in ceramide metabolism. Studies in AML have shown that MDR1-expressing cells are resistant to ceramide-induced apoptosis. However, this resistance was averted by exposure to either elacridar or cyclosporin A, MDR1 antagonists [153]. Work of Morad et al. [154] in multidrug resistance AML strongly substantiates the cooperation of MDR1 in the metabolism of ceramide by showing that zosuquidar, a specific MDR1 antagonist, enhanced ceramide-based therapeutics via blockade of ceramide glycosylation. Intriguingly, these works propose that MDR1 contributes to multidrug resistance by acting as an intracellular “ceramide neutralizer”, and prescribe novel indications for MDR1 antagonists in the enhancement of ceramide-centric treatment of AML [43, 154–156]. For example, therapeutic MDR1 approaches were shown to decrease mitochondrial membrane potential, inhibit complex I respiration, and generate ROS in a model of drug-resistant AML [76]. Appropriately, as leukemic stem cells and chemotherapy resistant leukemic cells are thought to play a pivotal role in relapse and exhibit high multidrug resistance transporter expression [157] as well as high oxidative phosphorylation status [158], it is tempting to speculate that these self-renewing and drug-resistant cells could be prime targets for agents that enhance ceramide-driven apoptosis via MDR1 modulation and oxidative stress.

3.4. Leukemic stem cells

Leukemic stem cells (LSCs) are a hallmark of AML and a source of relapse in AML patients [159–161]. LSCs arise from hematopoietic stem cells (HSCs) after one or more mutation events and are generally found within the CD34+CD38− subpopulation, though not exclusively [162–164]. LSCs, like HSCs, have self-renewal capabilities and inherent drug-resistant properties [163]. Targeting LSCs has been challenging due to their poor response to treatment and heterogeneity, especially in relapsed/refractory patients after exposure to chemotherapy [165, 166]. Newer therapeutic strategies seek to exploit LSCs’ unique immunophenotypes, which include CD123 (IL3RA), CD96, C-type lectin-like molecule-1 (CLL-1), and TIM-3 [162, 167], as well as unique metabolic profiles, epigenetic properties [168] and transcriptional regulation [169]. In addition, combinatorial strategies involving LSC targeting with conventional chemotherapy appears beneficial [165]. Hence, exploring and exploiting LSCs’ unique SL profile could provide new approaches to LSC-targeted therapy.

To this point, different subpopulations of progenitor, mature, and HSCs were shown to have distinct SL profiles and these unique SL signatures, which center around the dihydroceramide to ceramide pathway, regulate stem cell renewal and lineage commitment [170]. Importantly, dihydroceramide desaturase 1 (DEGS1 gene, DES1 protein abbreviation) was significantly increased in long-term-HSCs, short-term HSCs, and granulocyte-monocyte progenitors after ex vivo cytokine stimulation. Inhibiting DES1 with fenretinide (4-HPR), a synthetic retinoid, and genetic knockdown induced cellular-stress signaling such as the autophagy pathway and the unfolded protein response to maintain HSC self-renewal [170].

In addition, S1P-S1PR signaling is dysregulated in AML and S1PR3 has been identified as the key player in regulating TNFα-NF-κB signaling in both HSCs and LSCs [86]. Treatment of leukemic mice with the S1PR antagonist, FTY720, decreased leukemic burden and reduced LSC frequency [86, 171]. More importantly, an eight-gene SL signature was identified as significantly associated with overall survival of AML patients, thereby providing an avenue to identify and target LSCs in AML patients. Three of the genes in the SL signature were involved in S1P signaling (SPHK1, S1PR3, and S1PR5). In addition, SPHK1 inhibition in CD34+/CD38−/CD123+ leukemic stem and progenitor cells synergized with Ara-C treatment to potently decrease cell survival over single agents [74]. Collectively, these works showed that targeting S1P signaling in LSCs has therapeutic benefits. Additional studies on LSC-SL relationships are warranted, such as utilizing the eight-gene SL signature to inform studies combining SL therapeutics and LSC-targeting therapies.

3.5. Utility of dysregulated sphingolipid enzymes and metabolites as predictive biomarkers

Investigations into the landscape of SL alterations in AML have revealed a complexity that has yet to reach a consensus. In general, malignancy is associated with increased S1P at the expense of ceramides, which leads to a pro-survival phenotype. Indeed, we and others have demonstrated dysregulated expression of pro-survival SL enzymes in AML and that these changes predict poorer patient outcomes. AC expression and activity are elevated in AML and patients with high AC activity exhibit significantly lower overall survival [70]. IFN regulatory factor 8 (IRF8), a potent tumor suppressor and important factor of proper myeloid cell differentiation, also functions as a negative regulator of AC expression. IRF8 expression is often impaired in AML, which may contribute to elevated AC [172]. High mRNA expression of the alkaline ceramidase, ACER3, also correlated with poor overall survival [77]. This high ceramidase activity produces more substrate for SPHKs to generate S1P. Sobue et al. observed significantly elevated SPHK1 mRNA, but not SPHK2 mRNA, in acute leukemia patient samples [173]. Specific upregulation of SPHK1 in AML was confirmed by Powell et al., who described elevated SPHK1 mRNA and protein levels, but not SPHK2, in AML patient samples of various cytogenetic subgroups [74]. In contrast, Ghazaly et al. reported no alterations in SPHK1 or SPHK2 mRNA within AML specimens [174]. The contradictory results may be due to differences between sample types as Ghazaly et al. primarily utilized unenriched peripheral blood cells with variable blast percentages [174]. The studies describing high SPHK1 levels and activity in acute leukemias utilized bone marrow aspirates [173] or lineage-depleted HSCs [74]. In addition, Sobue et al. demonstrated that increased SPHK1 mRNA and activity from 16 leukemia cell lines strongly correlated with decreasing sensitivity to the AML chemotherapeutic, daunorubicin [175]. Thus, the majority of literature supports the hypothesis that elevated SPHK1 promotes survival in AML and represents a promising therapeutic target [80, 176, 177]. High expression of the S1P transporter, spinster homolog 2 (SPNS2 gene), also correlated with poorer event-free survival and overall survival [178]. This could suggest increased export of cellular S1P in AML. One study indicated lower plasma S1P levels in AML than normal controls [179], which highlights variations in SL content depending on the tissue type being measured (e.g., AML blasts vs plasma).

Primary AML cells exhibit reduced cellular ceramides [174]. This is supported by the study of FLT3-ITD positive AML specimens, as mentioned above [92]. This group reported reduced CERS1 mRNA with FLT3-ITD mutation [92], though this would not fully account for the global reduction in ceramides that was observed. Reduced mRNA expression for the ceramide-generating neutral SMase 2 (SMPD3 gene) may also be contributory [173]. Our group recently showed that AML with myelodysplastic syndrome-related changes (AML-MRC) preferentially metabolizes ceramides into pro-death SL metabolites compared to de novo AML [180]. Thus, single agent ceramide-based therapies seem better suited for AML-MRC while ceramide-based therapies may need to be used in combination with other AML treatments for de novo AML. In contrast to cellular ceramides, circulating ceramides were reported to be elevated in adult AML [179] and female pediatric AML cases [181], though one group reported a decrease in the ceramide ratios as defined as C18:1/C24:1 in adult AML [182].

Others have shown that sphingomyelins were predictors of AML in pediatric males [181]. In adult AML, several circulating sphingomyelin species were reportedly decreased [182], and another report showed lower serum levels of odd-chain sphingomyelins, but not others [183]. Moreover, acid sphingomyelinase-like phosphodiesterase 3b mRNA was upregulated in human AML samples, correlated to cytogenetic risk and karyotypes, and associated with poor overall survival [184].

Metabolism of ceramides to and from glycosphingolipids has also been investigated in AML. One group reported that CD34+ AML myeloblasts had significantly increased levels of the ganglioside, monosialodihexosylganglioside (GM3) [185]. This increase in GM3 was supported by a separate group showing increased expression of the glycosphingolipids: GM3, lactotriaosylceramide, neolactotetraosylceramide, and reduced globosides in AML specimens [186]. Lastly, circulating lactosylceramide may be diminished in pediatric AML [181].

Together, these studies have begun to define SL dysregulation in AML, although some inconsistencies remain. In general, SL profiling suggests decreased ceramides and increased glycosphingolipids and sphingomyelin in AML. The majority of studies found that ceramide-depleting (AC, ACER3) and S1P-generating (SPHK1) enzymes are elevated and correlated with poorer survival in de novo AML while few studies have described decreased SPHK1 levels. These discrepancies likely arise from the population sizes, detection methodologies, sample types, model systems, and the heterogeneity of AML. SL metabolism may be differentially regulated by specific chromosomal rearrangements [90] and specific driver mutations [94]. Increased understanding requires standardized and comprehensive SL measuring techniques to characterize different sample types (e.g., AML blasts vs plasma) and AML subtypes (e.g., mutational profile, cytogenic profile, age). The specific distribution of SL subspecies, their role in leukemic development, and their utility as diagnostic or prognostic biomarkers represent areas of future study and opportunity.

4. Role of sphingolipids in standard AML treatments

Standard AML 7+3 intensive induction chemotherapy has evolved to also include more specific approaches such as epigenetic-modulating regimens and Bcl-2, CD33, FLT3, and IDH-targeted therapies, as described above. SLs control many facets of cell signaling and are important signaling molecules that facilitate cell death following cancer therapy. Here, we summarize what is known regarding the relationship between SLs and AML treatments.

4.1. Induction chemotherapy

Induction chemotherapy, consisting of Ara-C and topoisomerase II inhibition with an anthracycline like daunorubicin, has been the mainstay treatment for newly diagnosed AML for decades. Ceramide generation is important for apoptosis induced by both Ara-C and daunorubicin. Ceramide is upregulated due to increased SMase activity following Ara-C treatment of U-937 cells. This increase in SMase activity was dependent on the formation of reactive oxygen species and the Src-family kinase p53/p56 Lyn. Mechanistically, Ara-C and daunorubicin induce oxidative stress, which activates p53/p56 Lyn and leads to the recruitment and activation of neutral SMase to sphingomyelin-enriched membrane rafts [83, 84].

As discussed previously, AML cells resistant to Ara-C or daunorubicin demonstrated significantly decreased levels of total pro-death ceramides as a consequence of elevated activities of the ceramide catabolizing enzymes AC, GCS, SMS, and SPHK1 [85, 143]. Increases in these enzymes dampen the cytotoxic ceramide induction following SMase activation and contribute to drug resistance. As previously mentioned, increased SPHK1 mRNA and activity correlated strongly with decreased daunorubicin sensitivity in 16 leukemia cell lines. Pharmacological and genetic inhibition of SPHK1 increased sensitivity of K-562 cells to daunorubicin [175]. Interestingly, resistance to Ara-C and daunorubicin did not significantly alter the toxicity of multiple SL enzyme targeting agents (AC, GCS, and SPHK inhibitors), which suggests that these agents may be used to combat drug resistance following induction chemotherapy [85]. Taken altogether, these studies highlight the dichotomy between ceramide-generating processes and mitogenic S1P signaling in the context of AML induction chemotherapy.

4.2. Bcl-2 family-targeted therapeutics

Bcl-2 family proteins are key survival mediators and important therapeutic targets in AML. Several studies have linked SLs and Bcl-2 family proteins. As mentioned previously, AC and SPHK1 inhibition as well as ceramide induction negatively regulate Mcl-1 levels [70, 74]. Moreover, Bcl-2 and Bcl-xL directly inhibit the formation of “water-filled” ceramide pores within the mitochondria [187]. Treating leukemia cell lines U-937, K-562, and MV4–11 with navitoclax, a pan inhibitor targeting Bcl-2, Bcl-xL, and Bcl-w, resulted in cell death and the accumulation of pro-apoptotic C16-ceramide. Intriguingly, knockout of the pore-forming protein BAK prevented ceramide accumulation following navitoclax treatment [188]. SL metabolism also interfaces with the pore-forming proteins, BAK and BAX, to regulate apoptosis. BAK, but not BAX, is required for CERS-dependent long-chain ceramide generation during chemotherapy-induced apoptosis [189]. Hexadecenal is generated from the breakdown of S1P by S1P lyase, and BAK- and BAX-induced mitochondrial outer membrane permeabilization requires a specific lipid composition involving S1P and hexadecenal, respectively [190]. These studies collectively demonstrate key interactions between SL and apoptosis regulation by the Bcl-2 family.

4.3. FLT3-targeted therapy

Twenty-five percent of AML cases harbor internal tandem duplication in the juxtamembrane domain of the FLT3 receptor tyrosine kinase. However, its targeting has yielded limited clinical success due to acquired drug resistance [91]. Dany et al. provided evidence that the FLT3-ITD mutation suppresses ceramide levels and significantly downregulates CERS1 mRNA in CD34+ bone marrow from FLT3-mutated AML [92]. CERS1 protein levels were also reduced in TF-1 cells overexpressing FLT3-ITD. Knockdown of FLT3 in two human AML cell lines (MV4–11 and MOLM-14) harboring the FLT3-ITD mutation significantly rescued CERS1 levels. Pharmacological inhibition of FLT3 with the first-generation FLT3 inhibitor, sorafenib, and the second-generation FLT3 inhibitors, AC220 and crenolanib, significantly increased CERS1 mRNA, CERS1 protein, and C18-ceramide levels. Knockdown of CERS1 protected cells from crenolanib while reconstitution of wild-type CERS1, but not a catalytically inactive CERS1 mutant, sensitized cells to crenolanib [92]. Further studies are warranted to characterize the role of CERS and other SL enzymes in AML with additional molecular drivers.

4.4. Chemotherapy-free AML treatments

Acute promyelocytic leukemia (APL) accounts for 5–10% of AML cases [15] and is characterized by the t(15;17) translocation, which encodes the PML-RARα fusion oncogene. Chemotherapy-free treatment with all-trans-retinoic acid (ATRA) and arsenic trioxide (ATO) result in APL cure rates upwards of 90% [15]. ATRA and ATO increase endogenous ceramides primarily through enhanced activation of acid SMase activity and CERS activity, respectively [191, 192]. ATRA and ATO treatment also significantly decreased GCS activity.

To summarize, preclinical studies suggest that the efficacy of intensive chemotherapy, molecularly-targeted AML therapies, and differentiating agents are all influenced by SL metabolism. Specifically, ceramide accumulation is important for the cytotoxic effects of these treatments, and mechanisms that deplete ceramide promote drug resistance. An intimate understanding of these relationships is necessary to improve therapy responses and prevent drug resistance.

5. Therapeutic agents targeting sphingolipid metabolism

The importance of ceramide and ceramide-metabolizing enzymes in cell survival and drug resistance, as described above, justifies substantial effort to develop small-molecule inhibitors targeting SL enzymes. Below, we summarize the current state of SL enzyme-targeting drugs in AML, and Table 1 provides a representative list of those in preclinical development. Recent studies have identified FDA-approved drugs (tamoxifen [124, 193–195], desipramine [196], dacarbazine [197], and fluphenazine [198]) that have off-target effects on SL metabolism. A representative list of clinically-relevant SL-targeting agents is summarized in Table 2 and mapped onto the SL metabolic pathway in Figure 3. The following subsections focus on drugs that directly target SL metabolism. Please refer to these excellent articles for a comprehensive review of SL-targeting agents: [87, 199].

Table 1.

Summary of sphingolipid-targeting agents in preclinical development for AML. Compounds listed were used as anti-neoplastic agents or tool compounds in AML.

Target(s)	Sphingolipid Drug	References
Ceramide analogs and ceramide inducing compounds
Acid ceramidase	LCL-204	[70]
Acid ceramidase	SACLAC	[82, 151]
Glucosylceramide synthase	PDMP	[150, 151, 208, 222]
Glucosylceramide synthase	PPMP	[141, 208]
Mitochondria	Pyridinium-ceramide (LCL-461)	[92]
Antagonists of S1P synthesis and signaling
Sphingosine kinase 1 Akt	A-674563	[107]
Sphingosine kinase 1	SK1-I	[106, 151]
Sphingosine kinase 1 & 2	MP-A08	[74]
Sphingosine kinase 1 & 2	SKI-II	[200, 222]
Sphingosine kinase 1 & 2	SKI-178	[72, 73]
Sphingosine kinase 1 & 2	SKI-349	[201]
S1PR 2 & 4 DES1 Sphingosine kinase 1 & 2	JTE-013	[74, 108]

Table 2.

Summary of clinically relevant sphingolipid-modulating drugs.

Target(s)	Sphingolipid Drug	Developmental Stage	References
Ceramide analogs and ceramide inducing compounds
Acid ceramidase	Dacarbazine	Approved to treat melanoma and Hodgkin’s lymphoma	[197]
Acid ceramidase	Desipramine	Approved to treat depression	[196]
Acid sphingomyelinase	Fluphenazine	Approved to treat schizophrenia	[198]
Dihydroceramide desaturase 1	Fenretinide (4-HPR)	Phase I clinical trials	[211–215, 225, 226]
Estrogen receptor Acid ceramidase Glucosylceramide synthase	Tamoxifen	Approved for hormone receptor+ breast cancer	[193–195]
Exogenous ceramide	C6 ceramide nanoliposome	Phase I clinical trials	[151, 180, 209]
Glucosylceramide synthase	Eliglustat	Approved for Gaucher disease	[49, 85, 227]
Antagonists of S1P Synthesis and Signaling
Protein kinase C Sphingosine kinase 1 & 2	Safingol	Phase I clinical trials	[87, 203]
Sphingosine kinase 2 Dihydroceramide desaturase 1	ABC294640	Phase I/II clinical trials	[88, 202, 224]
S1PR modulator Sphingosine kinase 1 S1P lyase	Fingolimod (FTY720)	Approved to treat multiple sclerosis Phase I clinical trials	[118, 119, 218, 219]

Figure 3. Dysregulated sphingolipid metabolism in AML and potential therapeutic intervention.

Refer to Figure 2 for a detailed summary of dysregulated sphingolipid metabolism in AML and Table 2 for a representative list of sphingolipid-modulating drugs that are FDA-approved or in clinical trials. In summary, multiple mechanisms contribute to ceramide depletion and S1P accumulation in AML, which are targeted by the indicated therapeutics. Dysregulated sphingolipid metabolites and enzymes are labeled in red. A larger node indicates upregulation of a particular metabolite. A large arrow indicates increased flux while a dashed arrow indicates decreased flux through a pathway. Clinically relevant compounds and their targets are labeled.

5.1. Sphingosine kinase (SPHK) inhibitors

Ceramide breakdown generates the substrate for S1P synthesis by the sphingosine kinases SPHK1 and SPHK2. SPHKs are attractive targets and several inhibitors targeting one or both isoforms have been evaluated in AML [75, 177]. The SPHK1 inhibitor, SK1-I, induced caspase-mediated apoptosis in U-937 cells, which is accompanied by decreased S1P and increased ceramides. Furthermore, SK1-I decreased phospho-ERK1/2 and phospho-Akt signaling and suppressed tumor growth in a U-937 xenograft model [106]. Corroborative findings were reported for the dual SPHK1 and SPHK2 inhibitors SKI-II [200], SKI-178 [72], and SKI-349 [201] in additional AML cell lines. Interestingly, SKI-178 and SKI-349 also target microtubule polymerization, which contributes to their cytotoxic effects [72, 201]. SKI-II was more potent at reducing cell viability and inducing apoptosis than SK1-I in vitro, which may indicate beneficial effects of targeting both SPHK isoforms [200]. An additional pan SPHK inhibitor, MP-A08, promoted caspase-dependent apoptosis in AML cell lines and patient samples. As discussed previously, MP-A08 treatment also induced pro-apoptotic Bcl-2 family proteins while depleting Mcl-1 [74]. These SPHK inhibitors have yet to progress to cancer-related clinical trials due to poor potency and a lack of enzyme specificity [176]. Two S1P-modulating therapies, ABC294640 and safingol, have progressed to clinical trials in advanced solid tumors [202–204]. It remains to be seen if these agents exhibit efficacy in AML.

5.2. Acid ceramidase (AC) inhibitors

We demonstrated that AC is elevated in AML and promotes a pro-survival state by depleting endogenous ceramides, promoting S1P synthesis, and upregulating Mcl-1 [79]. Several small-molecule inhibitors of AC are being explored in AML. We first characterized the effects of LCL-204, which elicited potent caspase-dependent cell death in multiple AML cell lines, patient samples, and a syngeneic mouse model. AC and Mcl-1 levels both decreased following LCL-204 treatment. The ceramide analog SACLAC is an irreversible inhibitor that forms a covalent adduct with Cys143 in the catalytic site of AC [70, 205]. SACLAC treatment also suppressed S1P, upregulated ceramides, led to caspase-dependent apoptosis in AML cells, and reduced leukemic burden in a human AML cell line xenograft model. SACLAC treatment induced alternative splicing of the anti-apoptotic Mcl-1 transcript via ceramide-mediated modulation of the levels of RNA splicing factor SF3B1. Alternative splicing of Mcl-1 resulted in the formation of the Mcl-1-short (Mcl-1S) isoform [82]. Intriguingly, while the full-length Mcl-1 is an anti-apoptotic protein capable of binding and inhibiting pore-forming Bcl-2 family proteins, Mcl-1S lacks the BH1 and BH2 domains, which are necessary to heterodimerize and inhibit pro-apoptotic proteins. Consequently, Mcl-1S is a pro-apoptotic BH3-only protein capable of binding full-length Mcl-1 exclusively [206]. Translation of these AC inhibitors to the clinic is precluded by their micromolar potency as well as poor solubility, pharmacokinetics, and bioavailability, which are likely attributed to their lipophilic nature.

5.3. Glucosylceramide synthase (GCS) inhibitors

The formation of glycosphingolipids promotes chemotherapeutic resistance in cancer [43, 147, 207]. Thus, GCS inhibitors may provide therapeutic benefit since drug resistance is a major factor in treatment failure and relapse in AML. To that end, Plo et al. treated leukemia cell lines with acquired daunorubicin resistance and high levels of the multidrug resistance protein, MDR1, with the GCS inhibitor 1-phenyl-2-decanoylamino-3-morpholino-1-propanolol (PDMP). Treatment improved daunorubicin retention and resensitized cells to daunorubicin-induced cell death while depleting glucosylceramides and other glycosylated SL species [150]. While several studies have characterized GCS in the context of drug resistance in cancer, few have identified the effects of GCS inhibition in drug-sensitive cells. Surprisingly, GCS inhibition with PDMP or 1-phenyl-2-palmitoylamino-3-morpholino-1-propanol (PPMP), in the drug-sensitive leukemic cell lines U-937 and HL-60 protected cells from daunorubicin-induced apoptosis. These protective effects were attributed to the induction of galactosylceramides instead of pro-apoptotic ceramides [208]. These divergent findings highlight the complexity of targeting SL metabolism in cancer. Since GCS inhibition sensitizes drug-resistant cells to chemotherapy while protecting drug-sensitive cells, glucosylceramide inhibition may be best utilized in relapsed/refractory AML. Of note, eliglustat is a potent GCS inhibitor in clinic for Gaucher disease as an effective lipid-reducing therapy. The therapeutic potential of eliglustat has not been extensively evaluated in AML, with one report citing its effectiveness in reducing cell viability in chemotherapy-sensitive and resistant HL-60 cells [85].

5.4. Ceramide nanoliposomes

Ceramides are canonically considered the center of SL signaling and are the most characterized SLs for their anti-cancer properties (Fig. 1). Therapies based on direct administration of exogenous ceramides are being evaluated in a variety of cancers including AML [209], but this approach is limited by poor delivery and solubility. The development of short-chain C6-ceramide nanoliposomes (CNL) was a key enabling technology to overcome delivery challenges of lipid-based drugs [209]. Our group compared the effects CNL as a single agent in de novo AML (DN-AML) versus AML with myelodysplastic-related changes (AML-MRC) [180]. As mentioned above, CNL preferentially exerts pro-apoptotic effects against AML-MRC over DN-AML patient samples. The colony-forming potential of AML-MRC patient samples was particularly sensitive to CNL compared to DN-AML. These findings were corroborated using transgenic mouse models representing AML-MRC versus DN-AML subtypes. The differential sensitivity to CNL between subtypes was attributed to their differences in SL metabolism following CNL treatment. AML patient samples with MDS-related mutations showed elevated pro-death SLs and decreased S1P compared to DN-AML samples, which showed increased pro-survival metabolites such as S1P and C6-hexosylceramides upon CNL treatment [180]. These results exemplify the need to characterize the uniformity of therapeutic responses to SL targeting agents across the heterogenous subtypes of AML. CNL demonstrated modest single agent effects on leukemic burden in immunodeficient mice engrafted with primary human AML cells [210]. Remarkably, CNL exhibited strong synergistic potency and “cures” in approximately 66% of treated leukemic mice when combined with the microtubule targeting agent vinblastine [180]. Our data suggest that this synergy is due, in part, to lysosomal disruption induced by vinblastine and resultant ceramide-induced autophagic cell death.

CNL has recently completed a phase I clinical trial for patients with advanced solid tumors (NCT02834611), exceeding all pharmacokinetic endpoints and showing evidence of efficacy (as defined by stable disease) in nearly 40% of the patients treated (manuscript in preparation). CNL is scheduled to enter a clinical trial for relapsed/refractory AML in combination with LDAC/venetoclax (NCT04716452).

5.5. “Ceramide generator” fenretinide

Fenretinide (4-HPR) is a synthetic retinoid being evaluated in several cancer types including AML. Treating HL-60 AML cells with 4-HPR induced apoptosis and ceramide accumulation. CERS inhibition suppressed ceramide synthesis and attenuated 4-HPR-induced apoptosis [211]. Jiang et al. confirmed these findings in two additional cell lines (NB-4 and U-937) and showed that vitamin C abrogated the apoptotic effects of 4-HPR, suggesting that oxidative stress is important for its anti-proliferative effects [212]. Our group confirmed that de novo ceramide synthesis and oxidative stress are necessary and demonstrated that sphingomyelin-derived ceramides contribute to 4-HPR cytotoxic effects [213]. Importantly, 4-HPR preferentially killed leukemic stem cells over normal stem cells, which was attributed to the induction of oxidative stress and downregulation of NF-κB and Wnt signaling [214]. Zhao et al. showed that AML blasts harboring FLT3 mutations were significantly more sensitive to 4-HPR than FLT3 wild-type AML. These findings were also confirmed using FLT3-mutant AML cell lines [215]. Interestingly, 4-HPR was also reported to inhibit DES1, which inserts a 4,5-trans double bond to the sphingoid backbone of dihydroceramides to generate ceramides [216]. These findings are consistent with 4-HPR treatment-induced ceramide accumulation via CERS and SMase activation, independent of de novo ceramide synthesis. Nanoformulations of 4-HPR are currently in clinical trials for relapsed/refractory T-cell non-Hodgkin lymphoma (NCT04234048) and represent intriguing options for development in AML.

5.6. Fingolimod

Fingolimod (FTY720) is an immunosuppressive drug that is FDA-approved for refractory multiple sclerosis. FTY720 antagonizes the sphingosine-1-phosphate receptor, S1PR1, to sequester T cells within lymph nodes [217]. At higher concentrations, FTY720 also has promising anti-cancer effects attributed to its ability to target and downregulate S1PR1 and SPHK1 [218]. FTY720 increases endogenous ceramide levels, decreases S1P levels, and elicits caspase-dependent apoptosis in Kasumi-1 AML cells, which was partially attributed to PP2A activation [118]. In addition, FTY720 activates PP2A directly via SET antagonism. A detailed mechanistic study of FTY720 in AML found that cytotoxicity was not mediated primarily by caspase activity, necroptosis, ferroptosis, ROS, or autophagy, but rather a form of non-canonical phosphatidylserine exposure that is PP2A dependent and requires cellular vacuolization [219]. It is likely that the therapeutic effects of FTY720 vary depending on the cellular SL metabolic state and are mediated by the combination of ceramide induction, SPHK1 antagonism, and direct SET interaction.

5.7. Novel sphingolipid drug combinations

SLs regulate many facets of cell survival. By defining these functions, synergistic combinatorial therapies can be rationally designed to target specific aspects of SL metabolism in conjunction with other anti-cancer agents. Our group demonstrated that exogenous C6-ceramide in combination with tamoxifen elicited synergistic killing in several AML cell lines that was independent of tamoxifen’s anti-estrogenic function [76]. The efficacy was associated with elevated reactive oxygen species, inhibited complex I respiration, and reduced mitochondrial membrane potential [76]. These effects may also be attributed to tamoxifen’s off-target inhibition of GCS, AC, and SPHK1 [124, 193, 195]. The combination of ceramide and tamoxifen also downregulated the inhibitors of apoptosis proteins survivin and XIAP [76]. Exogenous C6 ceramide and tamoxifen-driven cell death also coincided with upregulated autophagy in AML cell lines [124]. In addition, the combination of nanoliposomal safingol and CNL synergistically killed AML cell lines and patient samples [220]. As described above, we also showed respiratory complex I was impaired in drug-resistant AML and that further ablation of complex I function with phenformin, in combination with ceramide-promoting regimens (CNL treatment or GCS/AC/SPHK1 inhibition), resulted in impressive synergistic killing in multiple drug-sensitive and drug-resistant AML cell lines [151]. Moreover, the combination of CNL and vinblastine demonstrated potent synergy in in vitro and in vivo models of AML. Specifically, the combination synergistically killed primary AML patient samples and dramatically reduced leukemic burden and improved overall survival in multiple mouse models of AML [180]. Efforts are focused on improving CNL activity in combination with Ara-C, low dose liposomal vinblastine, and other agents including an aldehyde dehydrogenase inhibitor developed by our group [221].

Bcl-2 family proteins are important therapeutic targets due to their direct regulation of cellular apoptosis. Combinatorial studies with SL-modulating agents and Bcl-2 family inhibitors show promise in leukemia. Casson et al. demonstrated that inhibiting ceramide detoxification with the GCS inhibitor, PDMP, or the dual SPHK1 and SPHK2 inhibitor, SKI-II in combination with navitoclax resulted in increased ceramides and synergistic killing of U-937 and K-562 leukemia cell lines. Unexpectedly, the combination of PDMP and navitoclax also increased S1P level, which highlights the complexity of SL responses following treatment with anti-cancer therapeutics [222]. Synergistic effects were also observed between SPHK inhibition with MP-A08 and navitoclax treatment in MV4–11 AML cells [74]. Unfortunately, on-target thrombocytopenia induced by navitoclax precludes its use in AML [223]. Thus, studies exploring the therapeutic efficacy of SL-targeting agents in combination with the Bcl-2 inhibitor venetoclax are necessary, but none have been published to date in AML. However, this novel combination has been explored in other leukemias. For instance, the SPHK2 inhibitor, ABC294640, synergizes with venetoclax to reduce cell viability and increase apoptosis in myeloma cell line models. The combined treatment downregulated Mcl-1 and Bcl-xL, which likely contributed to the observed synergy [224]. Ongoing studies within our group are exploring the synergistic potential of CNL or AC inhibitors in combination with venetoclax-containing regimens.

6. Conclusions and future considerations

AML is a highly heterogeneous malignant disorder that continues to require urgent advances in therapeutic strategy and efficacy. In younger patients with newly diagnosed, de novo AML, the 5-year survival rate is only 50%, a discouraging statistic that is nonetheless far superior to the dismal 5-year survival rates seen in older patients or following relapse. Novel approaches based on immunogenic properties, mutated proteins, apoptotic signaling, and epigenetic dysregulation are raising the bar for “bench-to-bedside” AML therapy. However, the rise of relapse cases, the large population of patients unable to tolerate intensive induction chemotherapy, and the poor survival outcomes for the vulnerable older AML patient population collectively dictate that other avenues should be explored and exploited. SLs have emerged as key regulators of numerous aspects of AML biology. Many questions on the precise roles of SLs will need to be addressed to fully exploit the potential of SL modifying drugs to treat AML. Nonetheless, targeting SL dysregulation induces cell death mechanisms and reduces leukemic burden in human AML cell lines and murine models. SL inhibitors in conjunction with AML therapeutic agents have also shown considerable combinatorial synergy in preclinical studies.

Considering the impressive response to the combination of CNL with vinblastine in preclinical in vivo studies, current clinical studies offering CNL alone and in various combinations for relapsed and refractory AML patients are awaited with great interest. It will be of importance to determine whether AML patients exhibit similar promising therapeutic responses to combined cytotoxic and CNL therapy seen in animal models. Of additional importance will be the assessment of relative sensitivity of AML phenotypic and molecular subsets to SL directed therapeutics, including in combination with emerging AML therapeutic strategies including Bcl-2 inhibition, targeting mutant epigenetic/signaling effectors, and modalities which target cell surface molecules in AML. In addition, CNL and other SL directed agents should be studied in relapsed and refractory AML to define dose and toxicity of these agents. Subsequently, their use in combinations in newly diagnosed, poor prognosis patient subsets will also be of great interest. Several FDA-approved drugs have off-target SL modulating effects (tamoxifen, dacarbazine, FTY720) that may allow for repurposing to treat AML. Unfortunately, better SL therapeutic agents are needed to enable widespread clinical utilization. Some of the issues facing the current SL drugs include poor bioavailability, potential toxicity, high micromolar dosing, and inefficient delivery due to lipophilic properties. Further drug development will help resolve these issues to allow combinatorial studies to move forward. Supplementing current AML clinical treatments with SL-centric therapeutics presents a promising opportunity to meet the current challenges of treating AML.

7. Practice points

• Preclinical studies demonstrate synergistic efficacies when combining AML treatments with sphingolipid therapeutics.

• Sphingolipids and their regulatory enzymes are dysregulated in AML, and the use of sphingolipids as biomarkers is currently under investigation and shows promise.

• In vitro studies suggest that ceramide induction is necessary for chemotherapeutic effectiveness.

• Enzymes that decrease ceramides and/or increase S1P, glycosphingolipids, and sphingomyelins are frequently upregulated in multidrug-resistant AML.

• Sphingolipids modulate Mcl-1 levels in vitro, which should be exploited to improve current venetoclax standard-of-care regimens.

8. Research agenda

• Define additional relationships between common AML mutations and sphingolipid metabolism.

• Identify AML subtypes with differential sphingolipid metabolism, profiles, flux, and/or enzyme dependencies.

• Determine whether AML subtypes, as defined by sphingolipid profiles, exhibit differential sensitivity to AML therapeutics and/or sphingolipid targeting agents.

• Demonstrate the utility of sphingolipid enzymes and metabolites as predictive or therapeutic biomarkers in AML.

• Define the subcellular localization and flux of sphingolipid metabolites with mass spec-based whole-cell and organelle lipidomics.

• Uncover contributions to AML pathogenicity of less-studied sphingolipid biology such as ceramide-1-phosphate, dihydrosphingolipids, sphingoid backbone chain lengths, and fatty acyl chain lengths.

• Advance sphingolipid drug development and utilize nanotechnology to improve drug specificity, potency, and solubility.

• Evaluate and predict the efficacy of combinatorial drug studies in molecularly defined AML patient samples using bioinformatics and systems biology approaches.