Recent advances in drug delivery and treatment strategies for acute myeloid leukemia

Abstract

Acute myeloid leukemia (AML) remains a highly heterogeneous and aggressive hematologic malignancy with a poor prognosis. Although significant advancements have been made in chemotherapeutic regimens, targeted therapies (e.g., FLT3 and BCL2 inhibitors), and immunotherapies such as antibody-drug conjugates and CAR-T cells, treatment outcomes remain unsatisfactory, due to chemoresistance, off-target toxicities, disease relapse, and limited bioavailability. To address these limitations, nanomedicine and drug delivery systems have emerged as a promising approach to enhance therapeutic efficacy and minimizing adverse effects. This review provides a comprehensive overview of current AML treatments, highlighting both achievements and persistent limitations, with a particular focus on gene therapies under investigation. We then delve into various nanocarrier platforms, mainly lipid-based and polymer-based nanoparticles (NPs), examining their potential to overcome existing clinical challenges in AML therapy by improving drug stability, bioavailability, and leukemic cell targeting. Recent innovations in targeted formulations, such as antibody-, peptide-, and nanobody-conjugated drug delivery systems, have been designed to improve AML specificity. Finally, we discuss the key challenges and future directions in AML treatment, emphasizing the need for continued research in biomaterial innovation, formulation optimization, and precision-targeted approaches to enhance patient outcomes.

1. Introduction

Acute myeloid leukemia (AML) is an aggressive hematological malignancy, characterized by the rapid and uncontrolled growth of non-functional myeloblasts in the bone marrow (BM) with blockaded of normal myeloid precursor differentiation (Wu et al., 2024a). It is the most prevalent type of acute leukemia in adults, with estimated 20,800 new cases and 11,200 deaths in the United States in 2024 (Siegel et al., 2024). Although AML can occur at any age, its incidence increases significantly in older populations, with a median age at diagnosis of 69 years. Compared to younger patients, patients aged ≥ 60 years are three to five times more likely to develop and die from AML. Older patients typically exhibit intolerance to intensive chemotherapy, thereby contributing to notably poor outcomes in this population. Overall, the five-year survival rate for all AML patients is 31.9 %, while it drops to 11.2 % in individuals aged 65 years or older (Institute, 2024). These statistics underscore the urgent need for more effective and better tolerated therapeutic strategies, especially given the large number of older patients with AML.

Beyond age-related challenges, AML treatment is further complicated by the substantial biological complexity, driven by heterogeneity, relapse, and the protective bone marrow microenvironment (BMM). The disease often exhibits significant inter- and intra-patient heterogeneity, arising from the malignant transformation of hematopoietic stem cells (HSCs) (Desai et al., 2022; Klco et al., 2014; Zeng et al., 2022). Leukemic subclones typically contain distinct sets of cytogenetic abnormalities and somatic mutations, resulting in considerable genetic complexity (Li et al., 2016). Large-scale genomic sequencing studies in patient samples have identified recurrent somatic mutations in more than 200 genes, many of which are implicated in transcriptional regulation (e.g. RUNX1, CEBPA), signaling pathways (e.g. FLT3, RAS, KIT), tumor suppression (e.g. TP53, WTA, PHF6), chromatin remodeling (e.g. ASXL1, EZH2), and epigenetic modifications (e.g. DNMT3A, TET2, IDH1/2) (Desai et al., 2022; Ley et al., 2013; Meyer and Levine, 2014). This genomic complexity results in considerable variability, making it challenging to establish universally effective treatment strategies.

Another great challenge in AML treatment is its relapse, particularly in patients with high-risk cytogenetic or molecular profiles, which remains a major cause of treatment failure. Among those who achieve complete remission (CR) following initial therapy, more than 50 % experience relapse due to the persistence of minimal residual disease (MRD), often originated from therapy-resistant leukemic stem cells (LSCs). Residual leukemic clones can survive at undetectable levels (1 in 10⁴ to 10⁶ cells), escaping standard detection methods such as optical microscopic quantification or morphological assessment of BM aspirate (Srinivasan Rajsri et al., 2023). Additionally, the BMM, or BM niche, where LSCs residue, provides a protective environment for leukemia cells, thereby contributing to disease relapse and progression (Shafat et al., 2017a).

Since the late 1970s, chemotherapy for AML has been developed, and continues to serve as the frontline therapy for younger and fit AML patients, often followed by hematopoietic stem cell transplantation (HSCT). For patients who are ineligible for intensive chemotherapy, lower-intensity regimens, such as low-dose cytarabine (LDAC) and hypomethylating agents (HMAs), are commonly applied. Although high-intensity therapies are associated with better clinical outcomes, their use is limited by substantial toxicity. On the other hand, while low-intensity regimens are better tolerated, they often demonstrate limited therapeutic efficacy (Forsberg and Konopleva, 2024; Heuser et al., 2023). As a result, many patients continue to face poor long-term survival, highlighting the need for more effective and safer treatment strategies. Nanomedicine and advanced drug delivery systems have emerged as a promising strategy to overcome existing challenges in AML management. Specifically, nanoparticle (NP)-based platforms can potentially improve drug pharmacokinetics, improve drug targeting and reduce off-target toxicities, facilitate the controlled release of therapeutic agents, and enhanced efficacy, compared to conventional chemotherapy. Additionally, nanotechnology-based platforms have the potential to overcome chemoresistance by facilitating the co-delivery of multiple drugs, increasing drug retention within the BM, and modulating the tumor microenvironment (Yao et al., 2020). This review aims to provide a comprehensive analysis of the evolving landscape of AML classification, discuss current treatment paradigms and the challenges they are facing, and most importantly, provide a detailed analysis of how nanotechnology-based formulations can reshape the future of AML management.

2. AML classification

Multiple classification systems for AML are now widely recognized and used in the clinical practice (Table 1). In 1976, the French-American-British (FAB) classification was proposed, offering a foundation of AML categorization based on the blast morphology and cytochemical staining of myeloblasts (Bennett et al., 1976). Subsequent updates have progressively incorporated new diagnostic criteria that integrate molecular, pathological, and clinical variables into the morphological classification. The myeloblast threshold in diagnostic criteria, once set at 30 %, has gradually decreased under the influence of the next-generation sequencing (NSG) technique, while the specific genetic abnormalities, some of them have an evident impact on outcome, have emerged as a critical criterion (Park, 2024). As AML can be broadly classified into de novo AML (dnAML) and secondary AML (sAML), the latter arising either from an antecedent hematologic disorder (AHD) such as myelodysplastic syndrome (MDS) or myeloproliferative neoplasm (MPN), or from prior exposure to chemotherapy or radiation therapy (therapy-related AML, t-AML). Historically, sAML has been associated with a worse prognosis, yet recent studies suggested that genetic abnormalities should play a more important role in risk stratification and clinical trial eligibility in AML, as dnAML cases carrying secondary-type mutations or cytogenetic aberrations have poor clinical outcomes (Hochman et al., 2023; Lehmann et al., 2012; Othman et al., 2023; Parmar et al., 2024; Weinberg et al., 2022).

Table 1.

AML classifications.

FAB	WHO2022	ICC2022	ELN
Morphology and cytochemical properties of AML cells	AML with defining genetic abnormalities	AML with recurrent genetic abnormalities	Risk stratification for AML
M0: Undifferentiated acute myeloblastic leukemia	APL with PML::RARA fusion	APL with t(15;17) (q24.1;q21.2)/PML::RARA APL with other RARA rearrangement	Favorable
M1: AML with minimal maturation	AML with RUNX1::RUNX1T1 fusion	AML with t(8;21) (q22;q22.1)/RUNX1::RUNX1T1	t(8;21) (q22;q22.1)/RUNX1::RUNX1T1
M2: AML with maturation	AML with CBFB::MYH11 fusion	AML with inv(16)(p13.1q22) or t(16;16) (p13.1;q22)/CBFB::MYH11	inv(16)(p13.1q22) or t(16;16) (p13.1;q22)/CBFB::MYH11
M3: Acute promyelocytic leukemia (APL)	AML with KMT2A rearrangement	AML with t(9;11) (p21.3;q23.3)/MLLT3::KMT2A AML with other KMT2A rearrangement	Mutated NPM1 without FLT3-ITD
M4: Acute myelomonocytic leukemia	AML with DEK::NUP214 fusion	AML with t(6;9) (p23;q34.1)/DEK::NUP14	bZIP in-frame mutated CEBPA
M4 eos: Acute myelomonocytic leukemia with eosinophilia	AML with MECOM rearrangement	AML with inv(3)(q21.3q26.2) or t(3;3) (q21.3;q26.2)/GATA2;MECOM(EVI1) AML with other MECOM rearrangement	Intermediate
M5: Acute monocytic leukemia	AML with BCR::ABL1 fusion	AML with t(9;22)(22) (q34.1;q11.2)/BCR::ABL1	Mutated NPM1 with FLT3-ITD
M6: Acute erythroid leukemia	AML with RBM15::MRTFA fusion	AML with mutated NPM1	Wild-type NPM1 with FLT3-ITD (without adverse-risk genetic lesions)
M7: Acute megakaryoblastic leukemia	AML with NPM1 mutation	AML with mutated bZIP CEBPA	t(9;11) (p21.3;q23.3)/MLLT3::KMT2A
	AML with CEBPA mutation	AML with TP53	Cytogenetic and/or molecular abnormalities not classified as favorable or adverse
	AML with NUP98 rearrangement	AML with other rare recurring translocations	t(6;9)(p23.3;q34.1)/DEK::NUP214
	AML myeloid leukemia, myelodysplasia-related	AML with t(1;3) (p36.3;q21.3)/PRDM16::RPN1	t(v;11q23.3)/KMT2A-rearranged
	AML with other defned genetic alteration	AML with t(3;5)(q25.3) (q25.3;q35.1)/NPM1::MLF1	Adverse
		AML with t(8;16) (p11.2;p13.3)/KAT6A::CREBBP	t(9;22)(q34.1;q11.2)/BCR::ABL1
		AML with t(1;22) (p13.3;q13.1)/RBM15::MRTF1	t(8;16)(p11.2;p13.3)/KAT6A::CREBBP
		AML with t(5;11) (q35.2;p15.4)/NUP98:NSD1	inv(3)(q21.3q26.2) or t(3;3) (q21.3;q26.2)/GATA2, MECOM(EVI1)
		AML with t(11;12) (p15.4;p13.3)/NUP98::KMD5A	t(3q26.2;v)/MECOM(EVI1)-rearranged
		AML with NUP98 and other partners	−5 or del(5q); −7; −17/abn(17p)
		AML with t(7;12) (q36.3;p13.2)/ETV6::MNX	Complex karyotype, monosomal karyotype
		AML with t(10;11)(p12.3);q14.2)/PICA LM::MLLT10	Mutated ASXL1, BCOR, EZH2, RUNX1, SF3B1, SRSF2, STAG2, U2AF1, and/or ZRSR2
		AML with t(16;21) (p11.2;q22.2)/FUS::ERG	Mutated TP53
		AML with t(16;21)(q24.3;q22.1)/RUNX 1::CBFA2T3
		AML with inv(16)(p13.3;q24.3)/CBFA 2 T3::GLIS2

The World Health Organization (WHO) classification of hematolymphoid tumors has long served as the global standard for AML diagnosis. The latest 5th edition of the WHO classification of hematolymphoid tumors (WHO-HAEM5) updated in 2022 revised previous criteria, expanding the category of AML with defining genetic abnormalities (DGA) and introducing modification to the AML with myelodysplasia-related changes (AML-MRC) subset (Haferlach et al., 2024; Huber et al., 2023) In parallel, the International Consensus Classification (ICC 2022) has been proposed, setting a lower 10 % blast threshold for AML-DGA, and assigning cases with 10–19 % blasts without DGA to a new category MDS/AML(Arber et al., 2023; Huber et al., 2023; Park, 2024; Weinberg et al., 2023; Xiao et al., 2024). Another key ICC update is the inclusion of AML with mutated TP53 as a distinct, high-risk subgroup, requiring to be managed as a unique MDS/AML and AML subcategory (DiNardo et al., 2023).

Beyond WHO and ICC frameworks, European LeukemiaNet (ELN) classifications have been used widely in clinical practice and in clinical trials, categorizing AML into favorable, intermediate, and adverse risk groups, based on certain genetic abnormalities (Döhner et al., 2024; Döhner, Wei, et al., 2022; Salman, 2024). Although widely used, the ELN system was based exclusively on data from younger patients who received intensive chemotherapy, which limits its applicability to the growing proportion of patients treated with less intensive approaches (Bataller et al., 2024; Döhner et al., 2024; Döhner, Pratz, et al., 2022). Thus, new prognostic tools, for example, a four gene molecular prognostic risk signature (mPRS), have been proposed to assess outcomes more effectively in the context of lower-intensity therapy (Bataller et al., 2024; Döhner et al., 2024; Döhner, Pratz, et al., 2022).

In addition to the WHO, ICC, and ELN systems, the National Comprehensive Cancer Network (NCCN), European Society for Medical Oncology (ESMO) and other classifications also provided clinical guidelines for AML treatment (Fenaux et al., 2021; Pollyea et al., 2023). These evolving classifications highlight the complexity and heterogeneity of AML, reflecting the importance of molecular diagnostics in guiding mutation-specific and risk-adapted treatment strategies.

3. Bone marrow microenvironment

The BMM, also known as the BM niche, was first described in 1978 as a specialized microenvironment where HSCs reside, self-renew, and differentiated (Schofield, 1978; Xiao et al., 2022). Within the BM, two anatomically distinct but functionally cooperative niches have been described: the endosteum and the vascular niche (Fig. 1). Both are interconnected and collectively regulate normal hematopoiesis. In AML, however, the normal homeostatic environment is profoundly reprogrammed. Leukemic cells interact with various cellular and non-cellular components, including mesenchymal stem/stromal cells (MSCs), osteoblasts (OBs), osteoclasts (OCs), fibroblasts, adipocytes, immune cells, endothelial cells (ECs), and molecular components like cytokines, chemokines, growth factors, and extracellular matrix (ECM) proteins (Bakhtiyari et al., 2023). Accumulating evidence have suggested that AML cells could reshape a supportive microenvironment to accelerate leukemia progression, suppress the normal hematopoiesis, and protect leukemic cells from chemotherapy (Yao et al., 2021).

Fig. 1.

Schematic illustration of the complex cellular architecture and signaling dynamics within the bone marrow microenvironment. The sinusoids are lined by ECs, which separate the bloodstream from the bone marrow stroma. Within the vascular niche, leukemic cells interact with ECs, MSCs, fibroblasts, OBs, OCs, adipocytes, and immune cells, which collectively secrete cytokines, chemokines, and adhesion molecules that promote leukemic cell survival, retention, and resistance to therapy. The endosteum, lined by OBs and OCs, regulates bone formation and resorption. In multiple myeloma, cancer cells adhere to BM stromal cells, triggering osteoclastogenesis through osteoprotegerin (OPG), macrophage inflammatory protein (MIP)-1α, IL-6, IL-3, and RANK ligand (RANKL). Immune cells, including macrophages, T cells, Tregs, and NK cells, are modulated by leukemic cells to suppress anti-tumor immunity and reinforce a leukemia-supportive environment. Together, these components create a protective and dynamic niche that facilitates AML progression and therapeutic resistance.

3.1. Stromal cell-derived supportive niche cell regulation

Regulation of the BMM is largely attributed to MSCs (Le et al., 2016). These pluripotent cells contribute to leukemia progression by altering cytokine and chemokine secretion, such as CXCL12 (also known as stromal cell-derived factor SDF-1), interleukin (IL)-6, IL-8, and monocyte chemoattractant protein-1 (MCP-1), to create a niche that favors leukemic cell survival and therapy resistance. (Ladikou et al., 2020a; Ladikou et al., 2020b; Øystein et al., 2007; Shafat et al., 2017a).

ECs line the BM sinusoids, and function as gate keepers regulating cell trafficking between the BM and circulation (Shafat et al., 2017a). The vascular network provides a protective niche for AML cells, which exploit cytokines like TNF-α and IL-1β, as well as direct adhesion receptor interactions to activate endothelial cells via E-selectin, P selectin, and vascular cell adhesion molecule 1 (VCAM-1) (Olson, 2019; Stucki et al., 2001). Moreover, elevated proangiogenic factor vascular endothelial growth factor (VEGF) was found in AML patients, to promote angiogenesis and inhibit apoptosis through the VEGF-activated Notch/DII4 pathway. Culturing endothelial cells with VEGF induced granulocyte–macrophage colony-stimulating factor (GM-CSF) secretin by endothelial cells, which is known to stimulate growth in AML cells (Zhang et al., 2012). These endothelial-driven processes are frequently accompanied by structural alterations within the marrow vasculature, sometimes referred as “vascular remodeling”, creating additional routes for nutrient supply and immune evasion (Mosteo et al., 2021).

Fibroblasts provide physical support to hematopoietic stem and progenitor cells (HSPCs) during their self-renewal and differentiation, through the ongoing synthesis, secretion, and remodeling of ECM components into a highly organized meshwork in normal BM (Miari and Williams, 2024). Similarly, fibroblast remodeling is acquired in AML, known as cancer-associated fibroblasts (CAFs), including subpopulations that favor leukemic expansion by secreting additional ECM proteins, upregulating pro-angiogenic cytokines like IL-8, and displaying enhanced pro-survival signals upon contact with AML blasts like growth differentiation factor 15 (GDF15) (Zhai et al., 2016).

Adipocytes are MSC derivatives and make up the majority of yellow marrow, which has been observed to expand with age (Justesen et al., 2001). Physiologically, obesity is associated with poor clinical outcomes in leukemic patients, suggesting a role for adipose tissue in AML progression (Tsilingiris et al., 2024). Adipocytes promote AML cell survival via a metabolic shift from pyruvate oxidation to fatty acid β-oxidation (FAO), which causes mitochondrial uncoupling that diminishes reactive oxygen species (ROS), decreases intracellular oxidative stress, and leads to upregulation of pro-survival genes (PPARγ, FABP4, CD36, and BCL-2) (Tabe et al., 2018; Tabe et al., 2017). Thus, inhibition of FAO can be a therapeutic strategy for AML, upon activation of AdipoR1 and the downstream AMPK pathway in response to adipocyte-secreted adiponectin (Tabe et al., 2017). For example, Avocatin B, a FAO inhibitor, induced cytotoxicity in AML by upregulating the stress-induced transcription factor ATF4, AMPK signaling and ROS (Lee et al., 2015; Tabe et al., 2018). Besides, free fatty acids (FFAs) released by adipocytes also enhances AML survival via activating transcriptional network (Tabe et al., 2017). AML cells, in turn, induce lipase phosphorylation in adipocytes, which consequently activates lipolysis (Tabe et al., 2020). Co-culturing AML cells with adipocytes upregulates the fatty acid-binding protein-4 (FABP4) mRNA, and the knockdown of FABP4 reversed the AML cell protection from adipocytes, highlighting its role in leukemia progression (Shafat et al., 2017b).

3.2. Immune cell regulation

Other than stromal cell-derived supportive niche cells, immune cells that reside in BMM also play important roles in AML development (Fig. 1). Leukemic cells actively suppress immune activation and modulate the BMM to evade immune surveillance. AML cells downregulate major histocompatibility complex (MHC) class I and II expression, produce indolamine-2,3-dioxygenase (IDO), and upregulate inhibitory ligands such as PD-L1, B7-H3, and Galectin 9 (Gal), all of which lead to immune evasion and T cell exhaustion (Sendker et al., 2021). Compared to healthy individuals, AML patients have experienced a notable decrease in CD4⁺ and CD8⁺ T cells, leading to decreased recognition and elimination of AML cells (Tang et al., 2020; Wu et al., 2024c). The protective effects of regulatory T cells (Tregs) on the immune system are also crucial in AML, which can limit activation and proliferation of cytotoxic lymphocytes through the secretion of anti-inflammatory cytokines, competition for cytokines and co-stimulatory signals, and direct contact-dependent suppression (Guo et al., 2021b). For example, AML can express a Treg co-stimulatory ligand ICOS-L, activating Tregs directly, and they can also induce some CD4⁺ cells to become inducible Tregs. These ICOS⁺ T cells can then produce IL-10 which encourage the proliferation of AML in turn (Han et al., 2018).

Natural killer (NK) cells, essential components of the innate immune system, also experience profound functional impairment in AML. Several mechanisms contribute to this dysfunction, including inhibitory signaling through MHC interactions with killer immunoglobulin-like receptors (KIR) and NKG2D, as well as downregulation of essential cytokines such as IFN-γ and TNF-α. Also, expression of key activating receptors such as NKp30, NKp44, and NKp46 is reduced, increasing the resistance of AML cells to NK cell-mediated cytotoxicity, as well as suppression by immunosuppressive cell types such as Tregs (Guo et al., 2021b; Gurney and O’Dwyer, 2021). Receptor activator of nuclear factor κB ligand (RANKL) is a key regulator of bone metabolism, is also found to be elevated in AML patients. As its receptor, RANK, is also expressed on NK cells, the RANK-RANKL interaction facilitates leukemic immune evasion by impairing NK-cell mediated immunosurveillance (Schmiedel et al., 2013). These mechanisms collectively weaken NK cell-mediated immune surveillance, allowing AML cells to proliferate unchecked.

Macrophages, primarily originated from monocytes, usually contribute to cancer cell elimination, osteoclastogenesis, and HSC homeostasis through CXCL12 secretion by MSCs, undergo a profound phenotypic shift in AML (Menter and Tzankov, 2022). AML blasts induce macrophage polarization toward an M2-like immunosuppressive state, promoting angiogenesis and tissue repair while inhibiting anti-leukemic immune responses (Miari and Williams, 2024; Yahya et al., 2016). Tumor-associated macrophages (TAMs) within the AML BMM stimulate leukemic cell growth and support immunosuppressive, protumorigenic niche. Macrophages also play a role in maintaining leukemic progenitor cells, which could be an important mechanism in the early stages of AML (Bakhtiyari et al., 2023).

3.3. BMM-related resistance and relapse

The aim of AML therapies is to eradicate the LSC population, but at the same time, it damages the other cells of the niche, triggering regeneration. Prolonged treatment induces the development of resistance mechanisms, some of which are mediated by stromal or endothelial cells and results in LSCs which persist after chemotherapy and mediate disease relapse (Ladikou et al., 2020b). Relapse remains a major challenge in AML treatment, largely due to the persistence of LSCs within the BMM. LSCs interact with BM stromal cells, osteoblasts, and ECM components, forming protective niches that shield them from chemotherapy-induced apoptosis. Adhesion molecules such as CXCR4 and integrins anchor LSCs within the BM, creating a microenvironment that sustains MRD and facilitates disease recurrence (Bolandi et al., 2021). Additionally, the metabolic reprogramming of the BMM, characterized by increased oxidative phosphorylation (OXPHOS) activity and excessive ATP production, promotes chemoresistance of AML cells through AMPK inhibition or IL-6/OXPHOS/STAT3 axis (Hou et al., 2020; Hou et al., 2023; You et al., 2021). Understanding these intricate interactions within the BM niche provides a foundation for developing novel therapeutic strategies aimed at disrupting leukemia-supportive signals, enhancing immune reactivation, and improving long-term disease control in AML.

4. Current treatments and challenges

Multiple novel drugs and combinations have been approved by FDA over the past decade for AML treatment (Table 2). Determining the molecular profile of AML is crucial, as it carries substantial prognostic relevance, and serves as a pivotal predictive parameter for selecting the most appropriate treatment (Jaramillo and Schlenk, 2023). Despite recent advances, however, significant gaps persist in achieving durable disease control.

Table 2.

FDA-approved AML drugs.

Drug (Brand Name)	Target	Year of approval	Approved indications
Chemotherapy Agents
Cytarabine	–	Early 1970 s	Standard chemotherapies for younger AML patients
Anthracycline		1979
Idarubicin		1990
Targeted Therapies and Newer Approvals
Midostaurin (Rydapt)	FLT3	2017	Newly diagnosed AML with FLT3 mutation, use with standard 7 + 3 induction and high-dose cyrarabine consolidation
Enasidenib (Idhifa)	IDH2	2017	Relapsed/refractory AML with IDH2 mutation
Gemtuzumab Ozogamicin (Mylotarg)	CD33	2017	Adults with newly diagnosed CD33-positive AML; adults or children aged ≥ 2 y with relapsed/refractory CD33-positive AML
Glasdegib (Daurismo)	Hedgehog pathway	2018	Newly diagnosed AML aged 75 years and older, or with comorbidities that preclude intensive induction chemotherapy, use in combination with low-dose cytarabine
Gilteritinib (Xospata)	FLT3	2018	Relapsed/refractory, FLT3-mutated AML
Ivosidenib (Tibsovo)	IDH1	2018	Relapsed/refractory, IDH1-mutated AML; or newly diagnosed, IDH1-mutated AML patients aged 75 years and older or patients ineligible to receive intensive chemotherapy
Venetoclax (Venclexta)	BCL-2	2018	Newly diagnosed AML; use with azacitidine/decitabine or low-dose cytarabine in patients aged 75 years and older or with comorbidities that preclude the use of itensive chemotherapy
Oral Azacitidine (CC-486)	Hypomethylating agent	2020	Continued treatment of AML in CR/CRi post intensive induction chemotherapy, not able to complete intensive curative therapy
Olutasidenib (Rezlidhia)	IDH1	2022	Relapsed/refractory, IDH1-mutated AML
Quizartinib (Vanflyta)	FLT3-ITD	2023	Newly diagnosed FLT3-ITD AML, use with standard cytarabine and anthracycline induction and cytarabine consolidation; and maintainance monotherapy after consolidation
Revumenib (Revuforj)	Menin inhibitor	2024	Patients aged 1 year and older with relapsed/refractory KMT2A-rearranged AML
Formulated Therapies
CPX-351 (Vyxeos)		2017	Newly diagnosed t-AML or MDS-related AML

4.1. Chemotherapy

Since early 1980 s, the backbone of induction chemotherapy in the United States has been the 7 + 3 regimen, consisting of continuous infusion of cytarabine (ara-C) for 7 days and intravenous injections of an anthracycline (like daunorubicin or idarubicin) for 3 days (Arellano and Carlisle, 2018; Rowe, 2022). Approximately 35–40 % of young adults can now achieve long-term survival after standard 7 + 3 treatment, although it carries considerable toxicity risks, including life-threatening infections and bleedings due to cytopenia, decreased cardiac function, and others (Roman Diaz et al., 2024; Rowe, 2022). Another high-intensity regimen using high-dose cytarabine (HDAC) and mitoxantrone has shown higher response rates, but tolerability remains a major issue, particularly in older patients (Choi et al., 2023; Larson et al., 2012).

In recent years, CPX-351, a liposomal formulation delivering cytarabine and daunorubicin in a fixed 5:1 M ratio, has emerged as a key intensive chemotherapy option for patients with newly diagnosed t-AML or AML-MRC (Choi et al., 2023). Compared to the conventional 7 + 3 regimen, CPX-351 has demonstrated superior clinical outcomes, including high response rate, prolonged overall survival (OS), and the ability to facilitate allogeneic hematopoietic stem cell transplantation (allo-HSCT), which contributes to improved long-term remission, particularly in patients achieving MRD negativity before transplantation (Lancet et al., 2018; Lancet et al., 2021; Rautenberg et al., 2021). While CPX-351 has been associated with improved efficacy, the safety concern still exists. Its safety profile was similar to that of the 7 + 3 regimen, with similar rates of febrile neutropenia, pneumonia (Lancet et al., 2018; Lemoli et al., 2023; Matthews et al., 2021; Matthews et al., 2022). Conversely, data from a French study reported a more favorable safety profile in terms of lower incidences of alopecia, gastrointestinal side effects, and potential cutaneous toxicity (Chiche et al., 2021). These findings demonstrated that CPX-351 could serve as a preferred treatment option over the 7 + 3 regimen for older patients with high-risk AML; however, further strategies are required to optimize patient selection, manage prolonged myelosuppression, and mitigate infection-related complications.

4.2. Specific therapy

Less intensive therapies have drawn increasing attention for the treatment of AML, particularly for patients older than 65 years, due to concerns about their intolerance to intensive chemotherapy regimens. Historically, patients aged 70 years and older, or those judged unfit for intensive chemotherapy, has limited treatment options beyond supportive or hospice care, with a median OS of fewer than 3 to 6 months (Kantarjian et al., 2025). However, due to a rapidly advancing understanding of the pathophysiologic molecular abnormalities in AML, targeted therapies have made their way into clinic. Since 2017, the FDA has approved multiple novel agents for various AML subtypes, including FLT3 inhibitors (midostaurin, gilteritinib, quizartinib), IDH1/2 inhibitors (enasidenib, ivosidenib, olutasidenib), BCL-2 inhibitor (VEN), hedgehog pathway inhibitor (glasdegib), and most recently, the menin inhibitor revumenib (Fig. 2).

Fig. 2.

AML drugs approved by the Food and Drug Administration (FDA) since 2017. The name and structure of each specific drugs are illustrated.

For patients ineligible for intensive chemotherapy, the combination of venetoclax (VEN) and HMA, including azacitidine (AZA) and decitabine (DAC), has become a new standard of care, significantly improving therapeutic outcomes (Borlenghi et al., 2024; Cherry et al., 2021; Vachhani et al., 2022). While the HMA incorporates into DNA or RNA in AML cells, VEN induces AML cell apoptosis as a BCL-2 inhibitor (Fig. 3). The addition of VEN to AZA increased median OS from 9.6 months to 14.7 months, with the complete remission/complete remission with incomplete hematologic recovery (CR/CRi) rate rising from 28.3 % to 66.4 % (DiNardo et al., 2020). However, longer follow-up data revealed that the 3-year OS rate remained low at approximately 25 %, indicating that achieving durable long-term remission remains a challenge (Kantarjian et al., 2025; Pratz et al., 2022). In subgroups of older/unfit patients with sAML or those classified under ELN adverse-risk category, particularly those harboring TP53 mutation, the combination did not offer a survival advantage over HMA monotherapy (Borlenghi et al., 2024). This highlights the requirements for novel therapies for these high-risk populations. Additionally, while the combination demonstrated superior efficacy, it was also associated with notable adverse effects, including cytopenia and cytopenia-related complications such as neutropenic fevers, especially in the first one to two treatment cycles (DiNardo and Wei, 2020). The more mature results highlighted that although this new standard of care in older/unfit AML represented a great advancement, further improvements are required to enhance long-term outcomes and mitigate toxicity. Some studies suggested that shortening the durations of venetoclax (7–14 days per cycle) may achieve similar efficacy while reducing toxicity, compared to the original 21–28 days of venetoclax schedule (Aiba et al., 2023; Bazinet et al., 2024; Karrar et al., 2024).

Fig. 3.

Mechanisms of action of currently approved targeted therapies for AML. BCL-2 inhibitors (e.g., venetoclax) promote mitochondrial apoptosis by inhibiting anti-apoptotic BCL-2 and facilitating activation of pro-apoptotic proteins BAX and BAK. Isocitrate dehydrogenase (IDH) inhibitors (enasidenib, ivosidenib, and olutasidenib) block mutant IDH1/2 enzymes, reducing the oncometabolite 2-hydroxyglutarate (2-HG) and restoring normal TCA cycle and epigenetic function. Hypomethylating agents (HMAs) such as azacitidine and decitabine are incorporated into DNA/RNA as nucleoside analogs, leading to inhibition of DNA methyltransferases (DNMTs) and reactivation of silenced genes, including pro-apoptotic mediators like BIM. FLT3 inhibitors (e.g., midostaurin, gilteritinib, quizartinib) target aberrant FLT3 signaling driven by internal tandem duplications (ITDs) or tyrosine kinase domain (TKD) mutations, blocking downstream activation of JAK/STAT, PI3K/Akt/mTOR, and RAS/MAPK pathways, reducing leukemic cell proliferation and survival. Menin inhibitors (e.g., revumenib) disrupt the interaction between menin and KMT2A fusion proteins, leading to transcriptional repression of leukemogenic target genes. Together, these therapeutic strategies target metabolic, epigenetic, and signaling dependencies in AML cells, illustrating the rationale for pathway-specific treatment approaches.

In parallel, oral DAC/cedazuridine, a bioequivalent oral formulation of decitabine, has been introduced as an alternative to parenteral HMAs (Garcia-Manero et al., 2024). This total oral therapy could reduce hospitalizations and clinic visits, improve patients’ quality of life, and reduce the cost of care. While oral DAC/cedazuridine has been approved by the European Medicines Agency (EMA) for AML, it has not received FDA approval in the United States to the date (Kantarjian et al., 2025). Nonetheless, it represents an important step toward improving patient convenience and access to treatment.

Isocitrate dehydrogenase 1 and 2 (IDH1 and IDH2) and fms-like tyrosine kinase 3 (FLT3) mutations are two most common mutations in AML. IDH1/2 are critical enzymes that catalyze the oxidative decarboxylation of isocitrate to α-ketoglutarate (α-KG). Mutations in IDH1/2 result in aberrant conversion of α-KG to the oncometabolite 2-hydroxyglutrrate (2-HG), leading to its accumulation. Elevated 2-HG levels induce DNA and histone hypermethylation, which contribute to blocked cellular differentiation and tumorigenesis (DiNardo et al., 2013). IDH inhibitors target mutated IDH1 or IDH2, suppressing their catalytic activity, lowering 2-HG levels, and promoting cell differentiation or slowing tumor growth (Fig. 3) (Zhuang et al., 2022). FLT3 is a receptor tyrosine kinase crucial in cellular signaling. Upon binding of the FLT3 ligand to its extracellular domain, FLT3 undergoes dimerization with another FLT3 receptor, activating their intracellular kinase domains. This activation triggers phosphorylation of downstream proteins and the initiation of signaling cascades. FLT3 inhibitors work by binding to the FLT3 receptors at kinase-active or inactive structures, to prevent activation and disrupt oncogenic signaling pathways (Fig. 3) (Negotei et al., 2023).

Revumenib, a menin inhibitor, is the latest FDA-approved drug for AML treatment, receiving approval in November 2024. Lysine methyltransferase 2A (KMT2A) is a large, multi-protein chromatin modifier. KMT2A rearrangements (KMT2A-r) lead to oncogenic fusion proteins that upregulate leukemogenic homeobox (HOX) genes and their DNA-binding cofactor Meis homeobox 1 (MEIS1) (Issa et al., 2025). Menin is a scaffold protein that interacts with both the wild-type and rearranged KMT2A, regardless of its fusion partner. Preclinical studies identified menin as a critical regulator of KMT2A activity and HOX, especially HOXA gene expression, but found it is not essential for hematopoiesis in healthy cells. In contrast, menin proved necessary for ongoing leukemogenesis in KMT2A-r models, making it a promising therapeutic target (Martínez-Gamboa and Kaner, 2025; Salman and Stein, 2024).

Taken together, the expansion of targeted agents has significantly improved the therapeutic landscape for older/unfit AML patients, but challenges such as prolonged cytopenia, treatment-related toxicities, and limited long-term survival still exist. To address these issues, the development of novel drug delivery systems offers a promising strategy to enhance drug efficacy, reduce toxicity, and optimize treatment outcomes.

4.3. Gene therapy

As our understanding of AML’s genetic landscape continues to evolve, gene therapies have emerged as promising approaches for correcting or modulating genetic abnormalities that drive leukemogenesis. These strategies aim to restore normal gene function, silence oncogenic drivers, or enhance the immune system’s ability to target leukemia cells. Current gene therapies and genome-editing technologies both have demonstrated potential in preclinical and early clinical studies.

4.3.1. RNA-based therapy

The discovery of noncoding RNAs (ncRNAs), which are functional small RNA molecules unable to be translated into proteins, suggesting new prospects for AML diagnosis, prognosis, and treatments. ncRNAs are broadly categorized into housekeeping and regulatory ncRNAs, and the latter includes microRNAs (miRNAs), circular RNAs (circRNAs), and long noncoding RNAs (lncRNAs) (Liu et al., 2019). All together, they formed a regulatory network called competing endogenous RNA (ceRNA) firstly proposed in 2011, whose imbalance promotes tumorigenesis and progression (Fig. 4) (Cheng et al., 2020). As lncRNAs, circRNAs, and mRNAs all have miRNA response element (MRE) regions that can bind miRNAs, they compete for limited miRNA and form a ceRNA regulatory network. When mRNA competes to bind miRNAs, its stability decreases, the translation process is blocked, and gene expression is affected, in which way a variety of ncRNAs participate in regulation of mRNA coding function.

Fig. 4.

Biogenesis and functions of ncRNAs in the nucleus and cytoplasm. This schematic summarizes the processing and molecular roles of key ncRNAs, including miRNAs, lncRNAs, and circRNAs. pri-miRNAs are transcribed by RNA Pol II and processed by the Drosha-DGCR8 complex into pre-miRNAs, followed by cytoplasmic cleavage by DICER/TRBP into mature miRNAs. These guide the RISC complex (via AGO2) to regulate mRNA stability, translation, and splicing. lncRNAs and circRNAs, transcribed by RNA Pol I/II/III, act through diverse mechanisms such as miRNA sponging, protein scaffolding, and direct RNA/protein interactions. circRNAs arise from back-splicing events and include ecircRNAs, ciRNAs, and EIciRNAs. Both lncRNAs and circRNAs can serve in ceRNAs regulatory network by competitively binding miRNAs, thus modulating the availability of miRNAs for target mRNAs and fine-tuning gene expression.

4.3.1.1. MicroRNA.

miRNAs are highly explored group of ncRNAs containing around 22 nucleotides that bind to the 3′-untranslated region (3′-UTR) of the target mRNA and regulate the protein expression at the post-transcriptional level. The mature miRNA in cytoplasm complexes with Argonaut (AGO) family proteins to form the miRNA-induced silencing complex (miRISC) and eventually stop the protein synthesis process (Fig. 4). In AML, miRNAs play important roles in modulating differentiation, proliferation, apoptosis, and chemotherapy resistance (Table 3). Oncogenic miRNAs (oncomiRs) are often overexpressed in AML, promoting leukemogenesis by suppressing tumor suppressor genes. One of the most extensively studied oncomiRs in AML is miR-155, which is significantly elevated in both adult and pediatric AML, particularly in FLT3-ITD-positive cytogenetic normal AML (Koolivand et al., 2018; Narayan et al., 2017; Nguyen et al., 2021; Salemi et al., 2015). miR-155 enhances leukemic cell proliferation and survival by direct repressing of genes associated with myeloid differentiation and tumor suppression, such as SPI1, CEBPB, PML, and TLE2 (Narayan et al., 2018; Salemi et al., 2015). Inhibition of miR-155 using sulforaphane or anti-miR-155 vector both exhibited anti-cancer effect in AML cell lines (Koolivand et al., 2018; Salemi et al., 2015). The miR-125 family, comprising miR-125a, miR-125b1, and miR-125b2, is also essential in AML (Fletcher et al., 2022). Recent studies have linked miR-125 overexpression to leukemogenesis and HSC self-renewal in mixed lineage leukemia (MLL)-translocated AML. Specifically, miR-215a overexpression upregulates vascular endothelial growth factor A (VEGFA), correlating with a significantly reduced median survival in MLL/miR-125 transduced mice (Liu et al., 2017). An investigation of AML with t(2;11) in a murine model showed up to a 90-fold increase in miR-125b expression, with miR-125b-transplanted mice developing myeloproliferative disorders that progressed to AML. It was noted that different levels of miR-125b correlated with different leukemic phenotypes, suggesting that the level of overexpression not only contributes to the disease state but also to the disease phenotype. Beyond miR-155 and miR-125, other commonly dysregulated oncomiRs in AML include mir-196b, miR-21, miR-10a/b, miR-126, miR-223, miR-29b, and miR-181, all of which play significant roles in leukemogenesis and therapeutic resistance.

Table 3.

Recently reported miRNAs, circRNAs, and lncRNAs used for AML treatment.

miRNAs	Pathways involved/Target genes	Application	References
miR-155	CEBPb, TLE2, SPI1	FLT3 mutated AML	(Koolivand et al., 2018; Narayan et al., 2017; Nguyen et al., 2021; Salemi et al., 2015)
miR-125	VEGFA, PNPT18	–	(Chen et al., 2022; Liu et al., 2017; Liu et al., 2020a; Peng et al., 2021)
miR-193a	PTEN/PI3K	t(8;21) AML	(Li et al., 2013)
miR-23a	TOP2B	cytarabine-resistant AML	(Hatzl et al., 2020)
miR-29c	–	–	(Tang et al., 2019)
miR-493–5p	METTL3/MYC	–	(Wang et al., 2022)
miR-409–3p	CCNG2, CREB1, RAB10	–	(Li et al., 2020)
miR-214–3p	ATGL/PPARα	–	(Li et al., 2022)
miR-1303	TPTEP1/JNK/c-JUN	–	(Li and Zhao, 2021)
miR-143	ATG7, ATG2B	–	(Zhang et al., 2020a)
miR-193b-3p	MAPK/ERK	–	(Issa et al., 2023)
let-7b	AML1-ETO	t(8;21) AML	(Johnson et al., 2021)
miR-124	NRAS/ERK	AML with high EVI-1 expression	(Lang et al., 2022)
miR-486	SOCS2/JAK/STAT	–	(Sha et al., 2021)
miR-100	ATM	pediatric AML	(Sun et al., 2020)
miR-143–3p	KAT6A	pediatric AML	(Zhang et al., 2022)
miR-148a	DNMT1	–	(Wang et al., 2019d)
miR-342	Naa10p	–	(Wang et al., 2019a)
miR-146a	CNTFR/JAK2/STAT3, NF-κB	–	(Wang et al., 2019c)
miR-582–3p	CCNB2	–	(Li et al., 2019a)
miR-222–3p	IRF2	–	(Yuan et al., 2023)
miR-199a-5p	DRAM1	adriamycin-resistant AML	(Li et al., 2019b)
miR-144	APP/p-ERK/c-Myc/MMP-2	AML1-ETO + AML	(Jiang et al., 2019)
miR-9	EIF5A2/MCL-1	daunorubicin-resistant AML	(Liu et al., 2019)
miR-107	RAD51	–	(Huang et al., 2021)
miR-497–5p	MAP2K1/IL-6/JAK/STAT	–	(Pan et al., 2024)
circRNAs	Pathways involved/Target genes	Application	References
circ_0000370	miR-1299, miR-370– 3p, miR-502–5p, miR-1281 and miR-640	FLT3-ITD AML	(Zhang et al., 2020b)
circMYBL2	MYBL2	FLT3-ITD AML	(Sun et al., 2019b)
circPVT1	c-MYC, miR-455–3p/MCL1	–	(Sheng et al., 2023)
circ_0005774	miR-192–5p/ULK1	pediatric AML	(Li et al., 2021a)
circ_0006896	HDAC1	–	(Can et al., 2025)
circPAN3	AMPK/mTOR	doxorubicin-resistant AML	(Shang et al., 2019)
circ_001264	miR-502–5p/RAF1	–	(Du et al., 2023)
circ_0009910	miR-20a-5p	–	(Ping et al., 2019)
circPTK2	miR-330–5p/FOXM1	–	(Yi et al., 2021)
circ_0003602	miR-502–5p/IGF1R	–	(Ye et al., 2022)
circZBTB46	SCD	–	(Long et al., 2023)
circ_0004136	miR570–3p/TSPAN3	–	(Bi et al., 2021)
circ_MYC	miR-516a-5p/AKT3	–	(Zou and Jiang, 2021)
circ_0104700	miR-665/MCM2	–	(Chen et al., 2023)
circ_0079480	miR-654–3p/HDGF	–	(Hu et al., 2020)
circ_0001187	miR-499a-5p/RNF113A/METTL3	–	(Yang et al., 2023a)
circ_0010984	miR375/YAP1	–	(Yang et al., 2023b)
circ_0081188	miR-616–3p/ADCY2	–	(Wei et al., 2023)
circRNF220	SRSF1	–	(He et al., 2022)
circ_0035381	miR-582–3p/YWHAZ	–	(Xue et al., 2023)
lncRNAs	Pathways involved/Target genes	Application	References
LINC00641	miR-378a/ZBTB20	–	(Wang et al., 2019b)
LINC01018	miR-499a-5p/PDCD4	–	(Zhou et al., 2021b)
DARS-AS1	miR-425/TGF-β1	–	(Dou et al., 2020)
LINC1268	miR-217/SOS1	–	(Chen et al., 2020)
ZEB2–AS1	miR–122–5p/PLK1	–	(Guan et al., 2020)
TUG1	miR-193a-5p/Rab10	–	(Li and Wang, 2020)
DUBR	miR-142–3p/HOXA/A10 & FUS	–	(Yin et al., 2021)
SNHG5	miR-26b/CTGF/VEGFA	–	(Li et al., 2021b)
MAFG–AS1	miR–147b/HOXA9	–	(Yao et al., 2023)
LINC00460	miR–320b/PBX3	–	(Zhuang et al., 2021)
CEBPA–AS1	miR–24–3p/CTBP2	–	(Wang et al., 2023)
RPPH1	miR-330–5p	–	(Lei et al., 2019)
SNHG3	miR-758–3p/SRGN	–	(Xu et al., 2020)
LINC00675	miR-6809/CDK6	–	(Long et al., 2024)
LINC00618	SLC7A11	–	(Wang et al., 2021)
LINC00987	miR-4458/HMGA2	–	(Liu et al., 2024b)
LINC00963	miR-608/MMP-15	–	(Zuo et al., 2020)
SUCLG2-AS1	miR-17–5p/JAK1	–	(Liu et al., 2024a)
LIN28B	let-7/IGF2BP1	–	(Zhou et al., 2017)
GAS6-AS1	YBX1/MYC	–	(Zhou et al., 2021a)

Conversely, tumor-suppressive miRNAs are typically downregulated in AML, leading to unchecked cell growth and survival. For instance, miR-29b targets DNA methyltransferases, thereby altering the epigenetic landscape of leukemic cells; miR-34a induces apoptosis and inhibits cells proliferation by targeting multiple oncogenes; miR-223 plays a critical role in myeloid differentiation, with its reduced expression contributing to AML pathogenesis (Wallace and O’Connell, 2017). Among the tumor-suppressive miRNAs, miR-497–5p have gained attention for its tole in inhibiting AML progression. It has been shown to target multiple oncogenic pathways, including those involved in cell cycle regulation and apoptosis (Chen et al., 2021b; Nie et al., 2020; Pan et al., 2024). Downregulation of miR-497–5p is associated with increased leukemic cell proliferation and reduced sensitivity to chemotherapy. Studies suggest that restoring miR-497–5p expression may enhance apoptosis and sensitize AML cells to treatment, highlighting its potential as a therapeutic target.

As mentioned, studies have shown that miRNAs are involved in AML chemotherapy resistance in many ways, such as apoptosis, cell cycle, and ATP-binding cassette (ABC) transporter-mediated multidrug resistance. The differential expression of miRNAs contributes to either enhance sensitivity or resistance to chemotherapy. OncomiRs, such as miR-125b and miR-181, have been shown to promote drug resistance by inhibiting apoptosis and enhancing cell survival. miR-125b represses the pro-apoptotic genes p53 and BAK1, allowing leukemic cells to evade cell death in response to chemotherapy. Secondary AML caused by doxycycline-induced miR-125 expression was reversed by the inhibition of miR-125b, resulting in improved OS and event free survival. Likewise, miR-181 enhances resistance to cytarabine and anthracyclines by upregulating anti-apoptotic proteins. On the other hand, tumor-suppressive miRNAs, like miR-29b and miR-34a, have been linked to increased chemosensitivity by downregulating oncogenic pathways. miR-29b inhibits DNA methyltransferases (DNMTs), reversing epigenetic silencing and improving response to hypomethylating agents, while miR-34a restores apoptosis by targeting BCL-2, thereby enhancing the efficacy of venetoclax and cytarabine (Liu et al., 2019). Notably, miR-497–5p has also been implicated in AML chemoresistance. Its downregulation correlates with reduced apoptosis and increased survival of leukemic cells under chemotherapy, as we have reported that the introduction of miR-497–5p can reverse venetoclax-resistance in AML in vitro and in vivo models (Pan et al., 2024).

Although miRNAs are promising therapeutic agent in AML, several significant obstacles hinder their clinical application. One major challenge is off-target effects, as miRNAs can simultaneously regulate multiple mRNA targets. This pleiotropic nature, while beneficial for targeting multiple oncogenic pathways, raises concerns about unintended gene silencing, which can lead to unpredictable toxicity and adverse effects. Another critical limitation is the intrinsic instability of miRNAs in vivo, as they are highly susceptible to degradation by nucleases in the bloodstream, leading to a short half-life and poor bioavailability.

4.3.1.2. Circular RNA.

CircRNAs are single-stranded ncRNAs, characterized by a closed-loop structure formed through a phosphodiester bond between the 3′ and 5′ ends. This unique structure, along with the absence of a 3′ polyadenyl tail, grants circRNAs increased stability and a longer half-life compared to linear RNAs, as they are resistant to exonuclease degradation (Singh et al., 2021). This structure was first described in pathogens like viroids (virus-like infectious particles), circular satellite viruses, and hepatitis delta virus (Wang et al., 1986). Later, they were also identified in eukaryotic cells (Wilusz and Sharp, 2013). CircRNAs produced from back splicing of the primary transcripts can be classified into exonic circRNA (ecircRNA), intronic circRNA (ciRNA), and exon–intron circRNA (ElciRNA). EcircRNAs are predominantly located in the cytoplasm, whereas ciRNAs and ElciRNAs are usually located in the nucleus (Fig. 4) (He et al., 2021).

Emerging evidence suggests the crucial role of circRNAs in AML, where they regulate gene expression through diverse mechanisms, among which the most extensive one is miRNA sponging to reduce miRNA-mediated mRNA degradation. Many circRNAs have been implicated in AML pathogenesis, either promoting or suppressing disease progression. For instance, circANXA2 has been found to enhance leukemic cell proliferation and inhibit apoptosis by regulating the miR-532–3p/SOX4 axis. Similarly, circ-FOXO3 plays an oncogenic role by sponging miR-143–3p, which in turn upregulates cyclin-dependent kinase 1 (CDK1), leading to increased AML cell proliferation. Among oncogenic circRNAs, circPVT1 has drawn particular interest (Ghetti et al., 2023; Jia and Gu, 2021; Palcau et al., 2022; Sheng et al., 2023; Wu et al., 2024a). Derived from the PVT1 gene, circPVT1 has been found to act as a miRNA sponge, promoting the expression of anti-apoptotic proteins like myeloid cell leukemia-1 (MCL-1). Studies have shown that circPVT1 is highly expressed in AML blasts and can be detected in plasma, making it a potential biomarker for disease progression and prognosis. Additionally, its involvement in mediating drug resistance and LSC survival suggests that targeting circPVT1 may have therapeutic implications. Conversely, certain circRNAs exhibit tumor-suppressive functions; for example, circFBXW7 encodes a functional peptide that inhibits c-Myc, which is a proto-oncogene that encodes a transcription factor crucial for cell growth, differentiation, and apoptosis. The dual nature of circRNAs highlights their complex role in AML biology and underscores their potential as both therapeutic targets and diagnostic biomarkers (Jamal et al., 2019).

CricRNA can also directly interact with proteins and modulate their functions, including RNA binding proteins (RBPs), which are a class of proteins associated with the metabolic processing (Huang et al., 2020). One example is circMYBL2, derived from the cell-cycle checkpoint gene MYBL-2, which is highly expressed in AML patients with FLT3-ITD mutations. Knocking down circMYBL2 inhibits FLT3-ITD AML cell proliferation both in vitro and in vivo, by binding to polypyrimidine tract-binding protein 1 (PTBP1) to increase its affinity to FLT3 mRNA, thereby enhancing translational efficiency of FLT3 kinase (Sun et al., 2019b).

Additionally, circRNA also facilitates co-localization of enzymes by serving as a protein scaffold (Zeng et al., 2017). While circRNAs are considered to be ncRNA, recent studies have confirmed that some circRNAs can be translated into functional peptides under certain conditions (Fig. 4) (Ma et al., 2023; Singh et al., 2021). However, to the best to our knowledge, no circRNAs have been identified that inhibit AML progression via protein scaffolding or direct translation.

4.3.1.3. Long noncoding RNA.

Long noncoding RNAs (lncRNAs) are noncoding RNAs that are more than 200 nucleotides in length and lack a meaningful open reading frame. They regulate gene expression at multiple levels by interacting with RNAs. Though lncRNA exhibit poor sequence conservation across different species, it demonstrates higher tissue and development-specific expression than mRNA. Notably, lncRNA can bind to the DNA, RNA and proteins, there by regulating diverse cellular processes (Fig. 4) (Mishra et al., 2022). Recent studies have demonstrated specific lncRNAs that contribute to the AML, presenting potential targets for therapeutic intervention. Extensive profiling studies have been a popular strategy to identify distinct lncRNAs associated with AML pathogenesis, prognosis, and resistance.

In AML, several lncRNAs have been characterized and described as oncogenes or tumor suppressors. Mechanically, lncRNAs can serve to regulate gene transcription by chromatin or transcription factor modification in the nucleus, or interact with miRNA, mRNA or protein in the cytoplasm. HOX gene clusters encode transcription factors (TFs) essential for myeloid and lymphoid differentiation, as well as lncRNAs facilitating leukemogenesis (Wilson et al., 2024). HOXA transcript antisense RNA myeloid-specific 1 (HOTAIRM1) is the first lncRNA found to be associated with myeloid differentiation, with its expression restricted to myeloid cells. HOTAIRM1 plays a dual role in AML, regulating proximal HOX gene expression by sequestering either silencing protein complexes such as Polycomb Repressive Complex 2 (PRC2), or activating protein complexes such as UTX/MLL (Wilson et al., 2024). Knocking down HOTAIRM1 in NB4 AML cells retarded induced granulocytic differentiation, resulting in a significantly larger population of immature and proliferating cells that maintained cell-cycle progression from G1 to S phase (Zhang et al., 2009). Moreover, in nucleophosmin (NPM1)-mutated AML, HOTAIRM1 is upregulated, nucleus, it facilitates the ubiquitin-mediated degradation of early growth response-1 (EGR1) TF by acting as a scaffold to recruit E3 ubiquitin-protein ligase MDM2 to EGR1 protein, while in cytoplasm it acts as a sponge for the miR-152–3p, which increases Unc-51-like kinase 3 (ULK3) expression. Both nuclear and cytoplasmic activities enhance leukemogenesis, indicating HOTAIRM1 may be a promising therapeutic target for this distinct AML subtype (Jing et al., 2021).

In addition, several recent studies revealed the potential role of lncRNA signatures in AML risk stratification for patients across a diverse age range. In AML patients with the intermediate cytogenetic risk group, high HOTAIRM1 expression is associated with poor prognosis (Díaz-Beyá et al., 2015). Other than this, poor prognostic lncRNA biomarkers include PANDAR, SNHG5, H19, TUG1, CCAT1 and PVT1, while good prognostic lncRNA biomarkers include IRAIN, cancer susceptibility 15 (CASC15), and maternally expressed 3 (MEG3) (Gourvest et al., 2019; Priya et al., 2024).

The involvement of lncRNAs in AML chemoresistance underscores the need for further research to elucidate their specific mechanisms and interactions. Understanding these processes will not only enhance our knowledge of leukemogenesis but also pave the way for innovative therapeutic approaches that could improve treatment outcomes for AML patients. One lncRNA, differentiation antagonizing noncoding RNA (DANCR), has been identified as a positive regulator of cytarabine resistance in AML. By enhancing autophagy through modulation of the miR-874–3p/ATG16L1 axis, DANCR promotes leukemia cell survival during chemotherapy, suggesting that targeting this lncRNA could help overcome cytarabine-resistance (Zhang et al., 2021a). Another lncRNA, AC026150.8, is upregulated in AML and is associated with poor prognosis. Overexpression of AC026150.8 increases drug resistance in AML cells, suggesting its potential as a therapeutic target (Zhang et al., 2021b). Similarly, LINC00152 is highly expressed in CD34⁺ LSCs and regulates their self-renewal. Knockdown of LINC00152 significantly increases the drug sensitivity of leukemia cells, indicating its role in chemoresistance (Cui et al., 2021).

4.3.2. Gene-editing therapy

Genome-editing technologies, particularly clustered regularly interspaced short palindromic repeats (CRISPR) and its associated protein 9 (Cas9), have provided a precise method for modifying leukemic driver mutations. Originally derived from a bacterial adaptive immune system, this RNA-guided genome-editing platform utilizes a programmable endonuclease directed by a guide RNA (gRNA), to a specific genomic locus through complementary base-pairing with the target DNA sequence. Cas9 mediated cleavage requires the presence of a protospacer adjacent motif (PAM) immediately downstream of the target site, typically 5′-NGG-3′. Upon successful binding of the Cas9-gRNA complex to the target DNA, Cas9 introduces a double-strand break (DSB) at the destinated location. The cell then repairs the DSB through one of its inherent DNA repair mechanisms: non-homologous end joining (NHEJ) or homology-directed repair (HDR). NHEJ is an error-prone process that frequently introduces insertions or deletions (indels), potentially disrupting gene function, making it useful for gene knockout strategies. In contrast, HDR enables precise genetic modifications when a donor DNA template is supplied, allowing for gene knock-in or precise gene editing (Fig. 5A) (Alafeefi and Raza, 2025). This technology has been successfully applied to model or correct key mutations in AML, including FLT3, NPM1, and RUNX1::RUNX1T1 fusion (Brunetti et al., 2018; Rivera-Torres et al., 2020; Soerensen et al., 2024). CRISPR-based screening has emerged as a powerful tool in AML, offering an unbiased and high-throughput strategy to systematically interrogate gene function, discover therapeutic targets, and unravel mechanisms of drug resistance. Depending on the experimental setting, CRISPR screen can be performed in vitro, in vivo, or at the single-cell level, each offering unique insights into AML biology (Fig. 5B) (Tian et al., 2023).

Fig. 5.

Schematic overview of CRISPR-Cas9 gene editing and functional screening approaches in AML. (A) Mechanism of CRISPR/Cas9 gene editing. (B) Functional CRISPR screening in AML can be performed using in vitro, indirect in vivo, direct in vivo, or single-cell platforms.

In vitro CRISPR screens conducted in human AML cell lines have identified a broad spectrum of genetic vulnerabilities. A genome-wide CRISPR-Cas9 dropout screen in AML cell line revealed hundreds of AML-specific genetic dependencies, key among which include DOT1L, BCL2, MEN1, BRD4, and the histone acetyltransferase KAT2A (Liu et al., 2020b; Tzelepis et al., 2016). Targeting KAT2A was shown to impair leukemic cell growth, promote myeloid differentiation, and spare normal hematopoietic progenitors, suggesting a high therapeutic index (Tzelepis et al., 2016). Other in vitro screening efforts have highlighted regulators such as CDK12, ZFP36L2, IKAROS, PIK3C3, CRKL, and metabolic enzymes like DHODH (Aubrey et al., 2022; Guo et al., 2024; Takacs et al., 2019). In addition to dependency mapping, resistance-focused screens uncovered gene deletions (Kurata et al., 2016). For example, the loss of SPRY3 and GSK3 was found to promote resistance to FLT3 inhibitors, while deletion of CDKN2A conferred cytarabine resistance (Hou et al., 2017; Ling et al., 2023). A recent resistance-focused CRISPR screen identified BEND3, a transcriptional repressor and regulator of chromatin organization, as a top gene whose knockout confers AML cells resistant to TAK-243, a first-in-class inhibitor of the ubiquitin-activating enzyme UBA1. Functional validation showed that BEND3 knockout led to increased expression of the efflux transporter BCRP, resulting in reduced intracellular TAK-243 accumulation and dampened drug-induced proteotoxic and DNA damage stress. This effect was reversible by pharmacological inhibition of BCRP, highlighting a clinically actionable mechanism of resistance to TAK-243 (Barghout et al., 2021).

While in vitro studies provide mechanistic insights, they lack the complexity of the BMM (Tian et al., 2023). To address this, in vivo CRISPR screening platforms have been developed. For example, SLC5A3, a sodium/myo-inositol cotransporter, was identified as essential in AML cells with impaired inositol biosynthesis, and its loss led to cell-cycle arrest and apoptosis. Another in vivo-validated target, MARCH5, a mitochondrial E3 ligase, was shown to stabilize MCL1 and protect AML cells from apoptosis. Depleting MARCH5 enhanced sensitivity to VEN, even in resistant models, while sparing normal hematopoietic progenitors, highlighting its therapeutic relevance within the BMM (Lin et al., 2022). Additional in vivo studies have identified FERMT3, a regulator of integrin signaling, as essential for leukemic engraftment and maintenance, and PBRM1 as a key suppressor of interferon signaling and MHC-II expression, enabling immune evasion in AML (Li et al., 2023a; Mercier et al., 2022). Collectively, these examples underscore the value of in vivo CRISPR screening for uncovering microenvironment-dependent vulnerabilities that are not apparent in vitro.

To further dissect gene function at single-cell resolution, CRISPR screening has been integrated with single-cell RNA sequencing. Using Cas13d-based Perturb-seq and CaRPool-seq, combinatorial perturbations of chromatin regulators revealed synergistic and antagonistic interactions that influence AML cell fate decisions and differentiation trajectories. These technologies enable the dissection of functional heterogeneity and gene regulatory networks at unprecedented resolution (Wessels et al., 2023).

Despite promising preclinical results, several challenges remain for translation of CRISPR-based strategies, including safety concerns over off-target effects, immunogenicity, and efficient in vivo delivery. Thus, recent efforts focus on integrating CRISPR with delivery platforms such as lipid nanoparticles and viral vectors to enhance in vivo applicability (Tzelepis et al., 2016).

4.4. Immunotherapy

The concept of adoptive immunotherapy was first described in 1960 s, in the context of HSCT for AML treatment (Mathé et al., 1965; Tettamanti et al., 2022). More recently, with advancing understand of the mechanisms underlying the ability of AML immune escape, different immunotherapies, including T-cell based therapies, which harness adaptive immunity, and T cell-independent therapies, which leverage innate immune mechanisms, have been investigated.

4.4.1. Hematopoietic stem cell transplantation

Hematopoietic stem cell transplantation (HSCT) has long been the standard post-remission strategy for AML, remaining the only potentially curative modality for non-favorable risk AML (Chen and Garcia, 2023). HSCT is broadly categorized into autologous HSCT (auto-HSCT) and allogeneic HSCT (allo-HSCT) transplantation, with allo-HSCT being more commonly employed due to its capacity for graft-versus-leukemia (GVL) effects. However, the procedure is associated with significant challenges, including transplant-related mortality, prolonged immunodeficiency, graft-versus-host disease (GVHD), and secondary malignancies (Blazar et al., 2020; Chabannon et al., 2018; Chen and Garcia, 2023). Additionally, accessibility is another major issue, as hematopoietic host-donor chimerism is required before BM transplantation. Recent advancements in conditioning regimens, donor selection strategies, including haploidentical transplantation, and post-transplant immune modulation, such as donor lymphocyte infusions (DLI), have expanded HSCT eligibility to older and frailer patients (Chabannon et al., 2018). Despite these improvements, many patients are still ineligible due to advanced age, frailty, or comorbidities, underscoring the need for alternative therapeutic strategies to improve long-term disease control and reduce toxicity (Chen and Garcia, 2023).

4.4.2. Target selection in AML

Identifying an antigen that is integral to AML biology and exclusively expressed on malignant cells remains a significant challenge (Vago and Gojo, 2020). While multiple antigenic targets have been explored for AML therapy, none is entirely leukemia-specific, as they are also present on normal hematopoietic cells and tissues (Table 4). This lack of exclusivity increases the risk of “on-target, off-tumor” toxicity, significantly narrowing the therapeutic window and limiting the clinical utility of many immunotherapeutic approaches.

Table 4.

Identified AML targets and their expression profiles.

Target antigen	Function	Expression on Normal Cells	Expression on HSCs	Expression on LSCs
CD33	SIGLEC family protein, Transmembrane receptor	Progenitor, myeloid, Kupffer cells	+	+
CD123	Type I cytokine receptor of IL-3, IL3 receptor subunit	Myeloid progenitors, DC, and basophils	+	+
CLL-1	Glycoprotein, Transmembrane receptor	Myeloid, lung, epithelial cells	−	+
TIM3	Type I trans-membrane protein	Bladder, lung, appendix, lymph nodes, tonsil, ovary	−	+
CD7	Ig superfamily/Glycoprotein, B and T cell lymphoid development, Transmembrane protein	T, NK cells, and myeloid progenitor	−	+
FLT3	Type III cytokine receptor, Tyrosine kinase receptor	Neurons, testis	+	+
CD38	Glycoprotein, Cyclic ADP ribose hydroxylase	B, T, NK cells	−	+
CD44v6	Glycoprotein, Transmembrane receptor	Keratinocytes	−	+
NKG2D	C-type lectin-like receptor protein, Activator receptor	NK, NKT, Tαδ, Th, and CTL	−	+
CD70	Glycoprotein from the TNF family, Transmembrane receptor	T and B cells	−	+
CD96	Member of immunoglobulin superfamily, adhesion of activated T and NK cells	T cells and NK cells	−	+

CD33 and CD123 are among the most extensively studied AML targets. While they are highly expressed on AML blasts, their presence on normal hematopoietic cells and other healthy tissues may result in unacceptable side effects like myelosuppression (Liu et al., 2022; Pizzitola et al., 2014; Uy et al., 2021). Newly proposed targets, such as CLL1, TIM3 and CD244, display more selective expression on leukemic cells including LSCs, yet they still exhibit low-level presence in healthy tissues and thus do not fully eliminate toxicity concerns (Haubner et al., 2019). To improve specificity and minimize toxicity, strategies such as combinatorial targeting and the incorporation of “suicide switch” mechanisms in engineered immune therapies are being explored.

The immune system activity is closely regulated by interactions between co-inhibitory molecules and their ligands, many of which are upregulated in hematological cancers to facilitate immune evasion (Damiani and Tiribelli, 2023). CTLA-4 (CD152), PD1 (CD279), CD47, TIM3, LAG4 (CD223), and CD200, and are all transmembrane proteins that deliver inhibitory signals to terminate immune responses and contribute to AML progression. Targeting these pathways has become a promising strategy to restore immune surveillance and enhance anti-leukemic immunity.

4.4.3. T cell-based immunotherapies

In the presence of AML, multiple clinical studies have demonstrated various disruptions in T cell immunity, including augmented T regulatory cells and reduced T helper cells, T cell exhaustion, and dysregulated activity of transcription factors. Although T cell immunity is compromised in AML, cytotoxicity against AML cells is still exhibited, as patients with higher BM T cell percentages (≥78.5 % of total lymphocytes) were reported to have increased OS (Ismail and Abdulateef, 2017; Lamble and Lind, 2018). Given the effective antileukemic properties of T cells, current immunotherapies aim to restore weakened T cell activity in AML (Hao et al., 2021).

Antibody-based therapies exert anti-tumor effect primarily through antibody-dependent cellular cytotoxicity (ADCC) and complement-mediated cytotoxicity (CDC) (Lichtenegger et al., 2017; Yu et al., 2019). While monoclonal antibodies (mAbs) can target tumor surface antigens, their efficacy is often limited due to the suboptimal lethality of ADCC alone (Fig. 6A). To enhance therapeutic potency, antigen-drug conjugates (ADCs), which link antibodies to cytotoxic payloads, have been developed (Fig. 6B). This approach combines highly specific targeting ability and highly potent killing effect to achieve accurate and efficient elimination of cancer cells (Fu et al., 2022). Gemtuzumab ozogamicin (GO) is an ADC consisting of anti-CD33 mAb and calicheamicin. Initially approved in 2001 for AML patients aged ≥ 60 years, it was withdrawn in 2010 due to the safety concern. In 2017, it was re-approved to be used in combination with conventional chemotherapy in CD33-postitive AML (Selby et al., 2019). To further optimize its clinical benefits, several clinical trials are carried out to evaluate GO in combination with other agents, such as tagraxofusp (NCT05716009), midostaurin (NCT04385290), CPX-351 (NCT05558124), and others (per clinicaltrials.gov).

Fig. 6.

Immunotherapeutic strategies targeting AML. (A) Cytotoxic monoclonal antibodies (mAbs) bind to surface antigens on AML blasts and mediate immune cell–dependent killing. (B) Antibody drug conjugates (ADCs) are internalized upon antigen binding and release cytotoxic payloads that induce DNA damage and apoptosis. (C) CAR T cells are engineered to recognize AML-associated antigens and directly lyse leukemic cells. (D) Immune Checkpoint Inhibitors (ICIs) block suppressive receptor–ligand interactions between AML blasts and T cells, restoring T cell activation. (E) Dendritic cells (DC)-based vaccines present AML antigens via major histocompatibility complex (MHC) to stimulate antigen-specific T cell responses. (F) Innate immune strategies involve CD47-SIRPα blockade to disrupt AML immune evasion and promote phagocytosis by macrophages and recognition by natural killer (NK) cells.

Chimeric antigen receptor (CAR) T cell therapy has revolutionized the treatment of certain hematologic malignancies, by using genetically engineered T cells that express tumor antigens. They have the potential to persist after infusion and induce a long-term antileukemic memory. The binding between CAR and its antigen on a tumor cell trigger a signal transduction cascade through signaling domains that then activate T cells to kill the target wither directly or by harnessing other components of the immune system (Fig. 6C). CARs bind to their tumor antigens in an MHC-independent manner, which is their main advantage over regular TCRs (Daver et al., 2021). Translating CAR-T cell therapy to AML is complicated by the non-restricted expression of AML-associated antigens. To date, multiple antigens, including CD33, CD123, CLL1, TIM3, NKG2D, CD38, and FLT3, are investigated as modifications to CAR T cells; however, none of them have been approved, as toxicity remains a concern (Arcangeli et al., 2017; Baumeister et al., 2019; Carol et al., 2015; Gill et al., 2014; Glisovic-Aplenc et al., 2023; He et al., 2020; Jetani et al., 2018; Kenderian et al., 2015; Kikushige et al., 2015; Lee et al., 2021; Pei et al., 2023; Sánchez Martínez et al., 2022; Tettamanti et al., 2013; Wang et al., 2018). To improve safety, novel strategies have been investigated, including the incorporation of suicide switches such as the inducible caspase-9 system (iCasp9), combinatorial target-antigen recognition, synthetic notch receptors, and logic gates (NOT, AND, OR) (Yu et al., 2019). Despite these advancements, CAR-T therapy for AML remains in early clinical development, and further refinements are necessary to improve its therapeutic index.

Immune checkpoint inhibitors (ICIs) targeting immune evasion mechanisms have also gained interest in AML (Fig. 6D). Unlike solid tumors, where ICIs has achieved significant success, AML presents a more complex immunosuppressive microenvironment. The PD-1/PD-L1 and CTLA-4 pathways are among the most well-characterized immune checkpoints in AML, with several inhibitors under clinical investigation. Ipilimumab, a CTLA-4 inhibitor, has shown to efficacy in post-transplant relapse, in a phase I clinical trial, resulting in a 23 % CR (Davids et al., 2016). However, PD-1 inhibitors, such as nivolumab and pidilizumab, have demonstrated limited efficacy as monotherapies but have shown improved responses when combined with HMAs, yielding an overall response rate (ORR) of 33 % and a higher ORR of 58 % in HMA naïve patients (Daver et al., 2019). Despite the therapeutic efficacy of CPIs, potential immune-related adverse events, such as cytokine storms and GVHD, should not be ignored.

Dendritic cell (DC) vaccines represent another innovative immunotherapy strategy aimed at priming or enhancing AML-specific immune responses (Fig. 6E). As professional antigen-presenting cells, DCs play a crucial role in coordinating innate and adaptive immunity. By loading DCs with AML-associated antigens and reintroducing them into patients, this approach seeks to stimulate robust anti-leukemic T-cell responses. Although early-phase clinical trials have demonstrated promising immunogenicity, DC vaccines have yet to establish clear clinical efficacy, and further refinements in antigen selection and DC maturation protocols are needed to improve their therapeutic potential.

4.4.4. T cell-independent immunotherapy

Tumor microenvironment is immunosuppressive. For AML, the BM of a subgroup of patients is immune-depleted, which has been associated with reduced responses to T cell-based immunotherapies (Fig. 6F). Thus, T cell-independent immunotherapies are being explored, including NK cell-based therapies and macrophage-directed strategies. NK cell-based approaches leverage the innate cytotoxic potential of NK cells to target AML without the need for antigen specificity, making them particularly attractive for immunotherapy-resistant cases. Macrophage-directed therapies, such as CD47 blockade, aim to enhance innate immune clearance of AML cells by disrupting immune evasion mechanisms. Magrolimab is the first-in-class anti-CD47 blocking mAb that is being developed for myeloid neoplasms. A preclinical study used triplet of magrolimab, azacitidine and venetoclax promoted phagocytosis of TP53-mutant or VEN-resistant cell (Jia et al., 2021). These alternative immunotherapy strategies hold promise for improving outcomes in AML patients who do not respond to traditional T-cell-based therapies.

4.5. Challenges

Despite significant advances have been made for AML treatment, challenges still exist, limiting the efficacy and long-term success of existing therapies. One of the most pressing issues is the high relapse rate, driven largely by the persistence of LSCs within the BMM. These LSCs are inherently resistant to conventional chemotherapy and targeted therapies, leading to minimal residual disease (MRD) and disease recurrence. Additionally, the genetic and molecular heterogeneity of AML complicates treatment selection, as patients with different mutational profiles respond variably to existing therapies. For example, TP53-mutated AML remains particularly refractory to treatment, with poor responses to both intensive chemotherapy and venetoclax-based regimens (Shahzad et al., 2024). Similarly, adverse-risk cytogenetic subgroups continue to show inferior outcomes despite therapeutic advancements.

Beyond efficacy concerns, toxicity remains a major limitation. The standard 7 + 3 chemotherapy regimen, while effective in younger and fit patients, is associated with significant myelosuppression, infection risk, and multiorgan toxicity (Kantarjian et al., 2023). Even lower-intensity regimens, such as HMAs combined with VEN, frequently result in severe hematologic toxicities, including neutropenia and thrombocytopenia, which can compromise patient outcomes (Freeman et al., 2023). Immunotherapeutic approaches such as ICIs, while capable of restoring antitumor response, can also trigger a unique set of immune-related adverse events (IRAEs) (Gumusay et al., 2022). CAR-T therapies for AML remain experimental due to severe on-target, off-tumor toxicities affecting normal hematopoietic cells (Atilla and Benabdellah, 2023).

To address these challenges, advanced drug delivery vehicles are increasingly recognized as an important strategy to improve AML treatment efficacy and reduce toxicity. The following section will explore recent advances in drug formulations and how they are being developed to optimize AML treatment outcomes.

5. Formulations

Conventional drug delivery methods have been associated with excessive administrative dosages, off target effects on non-specific cells/organs and rapid degradation of systemic drugs by the reticuloendothelial system (RES) which could worsen the drug-associated side effects and reduce the efficacy (Abbaspour Sani et al., 2020). Nanotechnology-based drug delivery platforms, including liposomes, polymeric nanoparticles, and others, provide several advantages over conventional formulations. These systems can encapsulate therapeutic agents, modulate drug release kinetics, and facilitate selective targeting of AML cells, thereby reducing toxicity to normal hematopoietic cells. Moreover, nanocarriers can be engineered to penetrate the BMM more effectively, addressing a critical limitation in AML therapy. In addition to improving the pharmacokinetics and biodistribution of existing chemotherapeutics, emerging formulations also offer the potential for combination therapies, enhancing the synergistic effects of multiple agents while mitigating drug resistance. Notably, particle size plays a critical role in AML treatment, since the transcellular route takes place through the fenestrae between the endothelial cells in the BM. It has been reported that the sizes of the fenestrae in the endothelial wall are less than 150 nm which means the particles larger than 150 nm would be less likely to pass through (Sou et al., 2011). Further, NPs smaller than 60 nm can penetrate and distribute into the BM interstitial space, since reticuloendothelial sinusoidal blood capillaries consist of pores as large as 60 nm in diameter. Liposomes less than 100 nm in diameter circulate longer in the blood and have less interaction with plasma proteins. However, there is also a limitation of NPs with a small size, since NPs less than 50 nm limit the drug encapsulation efficiency (Bozzuto and Molinari, 2015). In this section, we will explore the advancements in formulation strategies designed to enhance drug efficacy and reduce toxicity in AML treatment, with a focus on lipid-based nanoparticle and polymeric nanoparticles (Table 5).

Table 5.

Recently reported nanoparticles used for AML treatment.

Lipid-based nanoparticles.
Materials	Loading	Modification	Target	Effect	Ref.
Liposomes/Lipoplexes
DSPC:Chol or DSPC:DSPE-PEG2000 with copper	flavopiridol	–	–	Flavopiridol can be encapsulated into liposome based on its copper-binding capabilities. The obtained liposomal flavopiridol demonstrated enhance pharmacokinetics and significant therapeutic activity in subcutaneous AML models.	(Chen et al., 2021a)
soybean PC: Chol	cytarabine	Alendronate and hyaluronic acid	bone mineral hydroxyapatite and CD44	The bone and CD44 dual targeting liposomes could target to AML including LSCs, with encouraging antitumor effect in AML mouse model.	(Wu et al., 2022)
POPC:DOPE:Chol:DSPE-PEG2000	–	DNA aptamer or Fab antibody	CD33	Fab-conjugated liposomal system offers enhanced precision over aptamer in targeting AML cells	(Jin et al., 2025)
S100PC:DOPC:Chol:DSPE-PEG2000	daunorubicin and homoharringtonine	Folic acid	Folate receptor	Folic acid-modified liposomes exhibited higher anti-leukemia activity in vitro and in AML subcutanous mouse model	(Liu et al., 2023b)
S100PC:Chol:DSPE-PEG2000	daunorubicin	CD33 and CD123 antibodies	CD33 and CD123	Dual targeted liposome enhanced the targeting ability against AML cells and potentially reduced the antigen-negative escape, with higher cytotoxicity found in vitro.	(Sun et al., 2019a)
DOTAP:DPPC:DSPE-PEG2000:Chol	miR-101	–	–	The lipoplex is stable and significantly prnetrate into AML cells	(Lotfabadi et al., 2018)
Lipofectamine	siR-SHARP1 and bortezomib	cyclic RGD	αvβ3 integrin	Co-delivery of siRNA and bortezomib led to targeted SHARP1 knockdown, demonstrating a potential therapeutic option for MLL-AF6 AML.	(Mohammed and Ju, 2022)
MDC2:HSPC:Chol:DSPE-PEG2000	miR-497–5p and VEN	anti-CLL1 peptide	CLL1	Anti-CLL1-decorated, miR-497–5p and VEN-loaded liposomes could inhibit AML growth in vitro and in vivo with minimal toxicity, and has potential to overcome chemoresistance induced by VEN.	(Pan et al., 2024)
DOTAP:DOPE:Chol:DMG-PEG2000	RUNX1 sgRNA and Cas9 mRNA	–	–	Depletion of RUNX1 gene inhibit AML cell growth in vitro, without incorporation into cord blood CD34 + cells.	(Iida et al., 2022)
Red blood cell-derived extracellular vesicles (RBSEVs)	ASO against FLT3-ITD or miR-125b	CD33 antibody	CD33	CD33-targeting RBCEVs improved AML suppression with ASO loadings.	(Chen et al., 2022)
Lipid Nanoparticles
Dlin-MC3-DMA: DSPC: cholesterol: DMG-PEG: DOPE-Rho	si-LINC01257	–	–	LNP-si-LINC01257 was able to inhibit t(8;21) AML growth with minimal uptake by health PBMCs.	(Connerty et al., 2021)
SM-102:DSPC:cholesterol:DMG-PEG2000	RUNX1::RUNX1T1 gRNAs and Cas9 mRNA	–	–	SM102 has higher efficiency than D-Lin-MC3-DMA in this case; SM102-LNP can delivery dual-gRNA and Cas9-mRNA to Kasumi-1 cells and disrupt the RUNX1::RUNX1T1 fusion gene.	(Soerensen et al., 2024)
O-14B:DOPE:Chol:DSPE-PEG2000	IL1RAP sgRNA and Cas9 mRNA	MSC membrane-coated nanofibril scaffolds (MSCM-NF) containing CXCL12α	CXCR4 on LSC surfaces	Sustained local delivery of Cas9/IL1RAP sgRNA via CXCL12-loaded LNP/MSCM-NF scaffolds provides an effective strategy for attenuating LSC growth to improve AML therapy.	(Ho et al., 2021)
Dlin-MC3-DMA:DSPC:Chol:DMG-PEG2000	miR-193b-3p	–	–	LNP/miR-193b-3p has strong anti-leukemic effects that can be leveraged to treat AML in a clinically relevant system based on PDX model.	(Issa et al., 2023)
SM-102:DOPE:Chol:DMG-PEG2000	UNC-GRK4-V mRNA	–	–	The SM-102 based UNC-GRK4-V mRNA LNP vaccine elicits robust, antigen-specific T-cell responses and could be a potent therapeutic for decreasing relapse rates post-alloSCT.	(Snow et al., 2022)
polysarcosine:β-sitosterol:sphingomyelin:Dlin-MC3-DMA	si-RUNX1::ETO	LDV peptide	VLA-4	Using polysarcosine as a replacement to PEG,:β-sitosterol as a replacement for cholesterol, and sphingomyelin as a helper lipid demonstrated superiority over a marketed Onpattro formulation, offering the potential for more effective RNA-based therapies with enhanced targeting and reduced side effects.	(Mata Casimiro et al., 2024)
Polymeric nanoparticles
Materials	Loading	Modification	Target	Effect	Ref.
Polymersomes
cystine	paclitaxel	–	–	GSH-responsive; specifically releasing paclitaxel and simultaneously inducing ferroptosis in AML cells with restricted myeloablation and tissue damage side effects	(Yu et al., 2023)
PAsp	vincristine sulfate	A6 peptide	CD44	Reduce AML burdens; easy fabrication	(Gu et al., 2021)
PEG and PAsp	volasertib	Transferrin	Transferrin-receptor	Reduce AML cells in mice and impede bone loss	(Xia et al., 2023)
PEG-P(TMC-DTC)-SP	tumor cell lysate	muramyl dipeptide and CpG	NOD2 and TLR9	Induce potent and broad immunity against AML	(Zhang et al., 2024)
PEG-P(TMC-DTC)-Ac-KD10	gilteritinib and palbociclib	daratumumab	CD38	Demonstrate significant leukemia inhibition with higher safety in CD38-upregulated AML model	(Zhai et al., 2025)
Polymeric micelles
PEO and PPO	curcumin	FLT3-specific peptide EVQ	FTL3	Enhance curcumin activity on FLT3-ITD overexpressing AML cells	(Tima et al., 2019)
PLGA and poloxamer	curcumin	CD123 antibody	CD123	Targeted NPs enhanced AML cell apoptosis and uptake, with some specificity to LSCs	(Nirachonkul et al., 2021)
Poloxamer 407	curcumin and doxorubicin	FLT3-specific peptide EVQ or CKR	FLT3	Using two targeting peptides improves the probability of micelles binding to the FLT3 receptor and induces cytotoxicity in LSCs	(Chueahongthong et al., 2024)
PVCL-PVA-PEG graft-copolymer	–	CD123 antagonistic peptide mPO-6 +	CD123	The novel CD123 antagonistic peptide micelle formulation mPO-6 can target AML cells as well as induce refractory AML cell apoptosis via CD123/IL-3 axis.	(Xu et al., 2022)
mPEG-b-PTMC	VEN	–	–	Polymeric micelle-loaded VEN has more vital anti-leukemic ability and less toxicity.	(Liang et al., 2022)
Dendrimers
PAMAM G4	pre-miRNA	Human H ferritin	CD71	The dendrimer presented pave the way for the design of a new family of protein-based transfecting agents.	(Palombarini et al., 2021)
PAMAM G3	cytarabine and dexamethasone	–.	–	The conjugation of cytarabine and fludarabine with PAMAM G3 dendrimer resulted in anticancer activity in AML cells.	(Wróbel et al., 2023)
PAMAM G7	miR-150	FLT3 ligand	FLT3	FLT3 ligand-guided miR-150-based Nanoparticles could treat FLT3-overexpressing AML with high efficacy and minimal side effects.	(Jiang et al., 2016)
Polymeric Scaffolds
Material	Loading/Cells	Modification	Target	Effect	Ref.
Scaffold for HSC expansion
PDMS	CD34+ mPB HSCs	–	–	The in vitro HSCs culture conditions were optimized, resulting in improved amplification, multipotency maintenance and vitality of HSCs.	(Marx-Blümel et al., 2021)
PLLA	CD133+ UCB MSCs	–	–	UCB-HSCs and MSCs coculturing on PLLA scaffold could provide a proper microenvironment that efficiently promotes UCB-HSCs expansion and UCB-MSCs can also be considered as a promising candidate for UCB-HSCTs.	(Darvish et al., 2019)
Gelatin and PMMA	Lin-c-Kit+ cells	–	–	Designed a 3D culture system for HSCs by mimicking the composition as well as the physical properties of the ECM.	(Li et al., 2023b)
Alginate hydrogel	Lin−Sca+cKit+ mouse BMMNCs	Angiopoietin-1	Tyrosine kinase receptor (Tie2)	Angiopoietin-1-coupled alginate gels were useful to provide a niche for HSC quiescence without a co-culture system.	(Lee et al., 2022)
PCL coated with collagen	CD34+ UCB HSCs	–	–	The scaffold and collagen have a synergistic effect on dimensionality protein, and this 3D scaffold coated with collagen has potential for ex vivo expansion of HSCs.	(Seyed Hadi et al., 2019)
Silk fibrin	CD133+ UCB HSCs	–	–	The silk scaffold could be used as a suitable substrate for UCB CD133+ stem cell expansion, as CD133+ stem cells have advantages over other MSCs.	(Mahdavi and Enderami, 2022)
Gelatin and sodium alginate	CD34+ HSCs	–	–	The gelatin powders created submillimeter-scale pores in alginate gel, enhancing the expansion efficiency of the HSCs.	(Liu et al., 2023a)
star-shaped PEG and heparin	MSCs	–	–	The ECM-functionalized cryogel could be used in the future for the ex vivo expansion of HSCs for transplantation and may also aid in the development of more realistic hematological disease models.	(Martínez-Vidal et al., 2023)
Scaffold for drug delivery
MSC membrane-coated PCL nanofibril	LNP loaded with CRISPR-Cas9	–	–	CXCL12α release induced migration of LSCs to the scaffolds, and LNP-Cas9 induced efficient gene editing.	(Ho et al., 2021)

5.1. Lipid-based nanoparticles

Lipid-based nanoparticles, including liposomes, lipoplexes, lipid nanoparticles (LNPs), and solid lipid nanoparticles (SLNs), have emerged as highly efficient carriers for intracellular drug and gene delivery. These systems have been successful in delivering small molecule drugs, RNA therapeutics, and gene-editing tools. Notably, LNPs have allowed siRNA therapy to achieve clinical success, with the FDA-approved siRNA-loaded LNP therapeutic, Patisiran, approved in 2018 for hereditary amyloidosis treatment (Connerty et al., 2021). The versatility of these nanoparticles lies in their composition, which may include cationic or ionizable lipids, helper lipids, PEGylated lipids, and cholesterol. Among them, cholesterol enhances structural stability, while PEGylation provides stealth property to prevent nonspecific serum protein adsorption and nanoparticle aggregation, ultimately improving pharmacokinetics (Fernandes et al., 2024).

5.1.1. Liposomes and lipoplexes

Liposomes, first described in 1965, are phospholipid bilayer vesicles with high biocompatibility. Their structure consists of an aqueous core enclosed by a lipid bilayer, allowing them to encapsulate both hydrophobic and hydrophilic drugs, thereby modifying their pharmacokinetics (PK) and pharmacodynamics (PD) profiles. More than 15 liposomal formulations are currently in clinical use, including CPX-351, a liposomal formulation of daunorubicin and cytarabine for AML treatment (Wang et al., 2024a). However, the application of liposomes for nucleic acid delivery is still in the preclinical or early clinical stages.

The development of liposomes for gene therapy has led to the formation of lipoplexes, which arise from the electrostatic interaction between cationic lipids and negatively charged nucleic acids. These complexes can either encapsulate nucleic acids within the liposomal structure or bind them externally. Commonly used cationic lipids in lipoplexes include 2,3-dioleoyloxy-propyl-trimethylammonium chloride (DOTAP), di-O-octadecenyl-3-trimethylammonium propane (DOTMA), and dimethyl-dioctadecyl ammonium bromide (DDA). DOTAP is a biodegradable analogue of DOTMA, and it is more extensively used in lipoplex formation due to its greater stability and transfection efficiency (Hou et al., 2021). For example, DOTAP, combined with 1,2-dipalmitoyl-sn-glycero-3-phosphocholine (DPPC), monomethoxy polyethylene glycol 2000-distearoyl phosphatidylethanolamine (mPEG₂₀₀₀-DSPE), and cholesterol, was used to formulate a cationic lipoplex via the thin-film hydration method for transfecting miR-101. This formulation demonstrated enhanced cytotoxicity and cellular uptake in AML cells, highlighting the potential of DOTAP-based lipoplexes in RNA-based therapies (Lotfabadi et al., 2018). Furthermore, we recently reported a novel di-cationic lipid, 1,3-bis(4,6-dimethyl-2-tetradecylpyridinium-1-yl)propane dihexafluorophosphate (MDC2). Its dual-positive charge offers a higher charge density, allowing effective miRNA binding and encapsulation with a lower amount of cationic lipid, potentially reducing toxicity (Kumar et al., 2023). Lipoplex containing MDC2 exhibited high therapeutic potential against AML in vitro and in vivo, with acceptable safety profile (Pan et al., 2024). However, though lipoplexes improve nucleic acid stability and facilitate cellular uptake, their application is still limited by toxicity concerns, including high clearance rates by the RES, complement activation, oxidative stress, and induction of apoptosis (Tsakiri et al., 2022). These limitations highlight the need for more sophisticated delivery platforms for RNA-based therapies.

For small-molecule drug delivery, liposomes remain one of the most robust and effective nano-scaled carriers, and various modifications have been explored to enhance their performance. One novel method of encapsulating drugs that can coordinate with metals, such as copper, held within liposomes (Chen, 2020; Leung et al., 2018). Flavones such as flavopiridol exhibit poor encapsulation in traditional liposomes, but when coordinated with copper, the encapsulation of flavopiridol into copper sulfate-containing liposomes is rapid with more than 98 % encapsulation efficiency. Compared to copper free formulation, the copper-containing liposomes exhibited 30-fold increase in area under curve (AUC) of flavopiridol, as well as significant therapeutic activity in subcutaneous AML models (Chen et al., 2021a). Moreover, stimuli-responsive drug release systems have also gained attention, including enzymatic, pH-sensitive, reactive oxygen species (ROS)-responsive, or redox-sensitive mechanisms. Among these, the redox-activated therapy has been studied due to the elevated levels of glutathione (GSH) found within tumor cells. Redox-sensitive liposomes incorporating disulfide bonds enable the rapid release of encapsulated drugs within the tumor microenvironment, enhancing therapeutic efficacy while minimizing off-target toxicity. With cytarabine loading, redox-sensitive liposomes exhibited significantly enhanced antitumor effects in mice compared to non-sensitive liposomes, reinforcing the potential of this approach for improving AML treatment outcomes (Wu et al., 2022).

5.1.2. Lipid nanoparticles

Recent advancements have led to the development of LNPs with amorphous or lipid-rich cores, which have demonstrated superior efficacy and safety in nucleic acid delivery. A key advantage of LNP lies in their use of ionizable cationic lipids. Unlike permanently charged cationic lipids, ionizable lipids remain electrically neutral at physiological pH but acquire a positive charge in acidic environments, such as the endosome, facilitating efficient endosomal escape and drug release to the cytoplasm (Hagedorn et al., 2024). This feature makes LNPs a safer and more efficient alternative to traditional lipoplexes. Their safety and scalability of LNPs have been well-established, as evidence by Pfizer/BioNTech and Moderna mRNA vaccines (Tsakiri et al., 2022).

Several ionizable lipids have been developed and commercialized, such as Dlin-MC3-DMA, TT3, C14–494, SS-OP, SM-102, and ALC-0315 (Bai et al., 2024; Maeki et al., 2022; Palanki et al., 2024; Snow et al., 2022). The choice of ionizable lipid significantly impacts key characteristics of LNPs, such as particle size, drug loading capacity, biodistribution, and therapeutic efficacy (Snow et al., 2022). The effectiveness of ionizable lipids is often linked to their pKa and 3D structure properties. An optimal pKa of 6.2–6.6 has been reported for intravenous mRNA delivery, whereas a slightly higher range of 6.6–6.9 is preferred for intramuscular vaccination. Another critical factor dictating lipid efficacy in LNPs is the molecular shape of ionizable lipids. Including branched tails within the ionizable lipid structure may lead to a con-shaped geometry, thereby promoting endosomal escape (Binici et al., 2025; Xu et al., 2023). Also, the structure plays a crucial role in organ-specific accumulation. For instance, OF-C4-Deg-Lin, an ionizable lipid with degradable linkers, has demonstrated more than 85 % accumulation in B lymphocytes in the spleen, while OF-Deg-Lin is mainly concentrated in the liver, highlighting the importance of designing ionizable lipid for specific nucleic acid delivery (Fenton et al., 2017; Fenton et al., 2018). Despite the growing diversity of ionizable lipids, no ionizable lipid has been specifically optimized for AML-targeted delivery based on, presenting a promising avenue for further research and development. Other than therapeutic RNA delivery, LNPs have recently been explored for the delivery of CRISPR-Cas9 gene editing tools. LNPs synthesized from SM-102 or Dlin-MC3-DMA, along with DSPC, cholesterol and DMG-PEG₂₀₀₀, were loaded with dual gRNAs targeting introns within RUNX1::RUNX1T1 and Cas9-mRNA. Treatment of Kasumi-1 cells with dual LNP/gRNA and LNP/Cas9-mRNA resulted in consistent disruption of RUNX1::RUNX1T1, although with lower absolute efficiency as compared to electroporation of the ribonucleoprotein complex. Both LNPs exhibited high transfection efficiency in patient blood and BM cells, while the transfection was not observed in B cells nor granulocytes. Notable, LNPs incorporating SM-102 demonstrated higher efficiency than those using Dlin-MC3-DMA, highlighting the importance of lipid selectin in optimizing gene editing efficacy (Soerensen et al., 2024). These findings emphasize the potential of LNPs as effective vectors for RNA-based therapies in hematologic disorders, provided that lipid composition and formulation parameters are carefully optimized to enhance targeted delivery, stability, and therapeutic outcomes. With the rise of machine learning and artificial intelligence, the rational design of ionizable lipids is becoming increasingly data-driven, enabling faster identification of lipid structures with optimized delivery performance and safety profiles (Li et al., 2024a; Wang et al., 2024b).

LNP production has been revolutionized by microfluidic-based systems, which allow for precise control over nanoparticle size, homogeneity, and encapsulation efficiency. (Connerty et al., 2021) Compared to conventional bulk mixing techniques, microfluidic methods enable high-throughput production, improved batch-to-batch reproducibility, and scalable manufacturing. The ethanol-dilution method, commonly employed in microfluidic systems, involves the rapid mixing of an ethanol phase (containing lipids) with an aqueous phase (containing nucleic acids), leading to spontaneous self-assembly of LNPs (Maeki et al., 2022). Despite these advances, challenges remain, as high-dose LNP administration can lead to undesirable side effects, including increased immunogenicity and cytotoxicity (Cheung and Shoichet, 2025).

5.1.3. Solid lipid nanoparticles

SLNs offer distinct advantages over liposomes and LNPs due to their solid lipid core, which enhances stability, controlled drug release, and bioavailability (Kadhim et al., 2025). Unlike liposomes, which have aqueous cores, SLNs consist of a solid lipid matrix stabilized by surfactants, providing better protection of encapsulated drugs from enzymatic degradation and oxidation. Additionally, SLNs can be produced without the use of organic solvents, making them a more environmentally friendly alternative (Akanda et al., 2023).

In AML therapy, SLNs have shown promise in overcoming chemoresistance. Co-encapsulation of mitoxantrone and β-elemene in SLNs composed of soybean lecithin phospholipids, trilaurin, cholesterol, S100, and SDS effectively reversed multidrug resistance (MDR) in AML by inhibiting P-glycoprotein (P-gp) efflux and increasing intracellular drug retention (Amerigos Daddy et al., 2020). Similarly, SLNs loaded with idarubicin and doxorubicin demonstrated enhanced cytotoxicity against P-gp-overexpressing leukemia cells compared to free drugs, indicating their potential to bypass MDR mechanisms (Ma et al., 2009).

5.2. Polymers

Polymers have gained significant attention as versatile carriers in drug delivery due to their tunable physicochemical properties, including monomer composition, hydrophobicity, biocompatibility, molecular weight, and mechanical strength. Based on their origin, polymers are classified into synthetic and natural categories. Common synthetic polymers include poly(lactic acid) (PLA), poly(l-lactide-co-glycolic acid) (PLGA), poly(vinyl alcohol) (PVA), and poly(butyl cyanoacrylate) (PBCA), while natural polymers mainly consist of chitosan, alginate, hyaluronic acid, and cellulose (Barmin et al., 2024). Beyond these established materials, numerous novel polymeric systems and formulation strategies are being developed to enhance the specificity, efficiency, and therapeutic efficacy of drug delivery to AML sites.

5.2.1. Polymeric nanoparticles

Polymeric nanoparticles (NPs) provide a multifunctional drug delivery platform with controlled release kinetics, increased cellular uptake, and potential for targeted drug delivery. These NPs can be broadly classified into polymersomes, micelles, and dendrimers (Waheed et al., 2022).

5.2.1.1. Polymersomes.

Polymersomes, structurally like liposomes, are self-assembled vesicles composed of amphiphilic block copolymers. Unlike liposomes, which are formed from phospholipids, polymersomes offer superior stability due to the tunable hydrophilic-to-hydrophobic ratio of the copolymers. This adaptability allows for precise control over vesicle size, morphology, rigidity, and responsiveness to environmental stimuli (Guo et al., 2021). Compared to other drug delivery platforms, polymersomes have been relatively underexplored, particularly in the context of AML therapies. However, over the past decade, they have gained increasing attention due to their advantages, including enhanced drug-loading capacity, tunable membrane properties, and improved in vivo behavior (Guo et al., 2021; Xu et al., 2024).

Ferroptosis, a non-apoptotic, iron-dependent form of programmed cell death, has emerged as a potential therapeutic strategy in AML. High GSH levels in leukemic cells support glutathione peroxidase 4 (GPX4) in neutralizing lipid peroxides, thereby preventing ferroptosis and contributing to chemoresistance. To counteract this, redox-sensitive polymersomes have been designed to deplete intracellular GSH, sensitizing AML cells to chemotherapy. A cystine-based polymersome, synthesized from a leucine polymer 8L2 and loaded with paclitaxel (PTX), was reported to function as both a ferroptosis inducer and a chemotherapeutic carrier, effectively enhancing AML treatment outcomes (Yu et al., 2023). Beyond ferroptosis induction, polymersomes have also been explored for the delivery of highly toxic chemotherapies, like vincristine sulfate (VCR), which is a highly toxic water-soluble drug approved for the treatment of acute lymphoblastic leukemia (ALL), lymphoma, and neuroblastoma, but not AML due to the systemic toxicity. A polymersome formulation constructed using PEG-p(TMC-DTC)-pAsp demonstrated ultra-small sizes (<40 nm), enhancing the AML tumor site penetration and cell uptake. Functionalization with CD44-targeting peptides further improved selectivity for AML cells, minimizing adverse effects and improving drug tolerability (Gu et al., 2021). Volasertib, a PLK1 inhibitor, is suffering from rapid clearance and dose-limiting toxicity. Thus, a transferrin-guided polymersome platform was proposed to encapsulate Vol, using dithiolane-functionalized polycarbonate as membrane, poly(ethylene glycol) as outer shell, and poly(aspartic acid) as inner shell. TPVol demonstrated superior cellular uptake in AML cells, effectively downregulated p-PLK1 and p-AKT, and improved survival in an orthotopic AML mouse model. Notably, TPVol accumulated preferentially in the BM, highlighting its potential for treating AML (Xia et al., 2023).

Additionally, polymersomes have been employed in AML immunotherapy. A vaccine delivery platform based on poly(ethylene glycol)-b-poly(trimethylene carbonate-co-dithiolane trimethylene carbonate)-b-spermine (PEG-p(TMC-DTC)-SP) has been designed to elicit potent immune responses. These polymersomes co-encapsulate nucleotide-binding oligomerization domain-containing protein 2 (NOD2) and Toll-like receptor 9 (TLR9) agonists, muramyl dipeptide (MDP) and CpG, respectively, along with leukemia cell lysates as tumor antigens. The resulting multifunctional vaccine (MCA-NV) effectively stimulates dendritic cells, enhances antigen presentation, and promotes a robust anti-AML immune response (Zhang et al., 2024).

Like liposomes, polymesomes can be constructed with positively charged or ionizable polymer blocks for nucleic acid delivery. Poly(ethyleneimine) (PEI) is a widely used cationic polymer, as its strong electrostatic interaction with siRNA could achieve nearly 100 % encapsulation efficiency and significantly improve RNA serum stability (Guo et al., 2021; Xu et al., 2019). However, its high positive charge also leads to cytotoxicity, as well as rapid clearance by macrophages and RES (Wang et al., 2016). To address these limitations, ionizable polymers have been developed. These polymers form complexes with nucleic acids under acidic conditions and subsequently self-assemble into polymersomes when the pH is adjusted to physiological levels (pH 7.4). Examples include neutralized PEI derivatives and imidazole-containing polymers, which improve encapsulation efficiency during self-assembly, with minimal toxicity during in vivo treatments (Gallon et al., 2015; Wang et al., 2016; Zheng et al., 2022). To the best of our knowledge, polymersomes have not yet been applied in AML therapy, highlighting their potential as promising future delivery systems.

Though polymersomes provide a more robust platform for drug delivery, accommodating both hydrophilic and hydrophobic therapeutic agents while offering enhanced resistance to premature degradation, their fabrication is more complex and requires precise optimization (Leong et al., 2018). Further research into optimizing polymersome compositions and surface modifications may improve their biocompatibility and therapeutic efficacy, making them a valuable tool for future AML treatment strategies.

5.2.1.2. Polymeric micelles.

Polymeric micelles are nanosized self-assembled structures composed of amphiphilic block copolymers, featuring a hydrophobic core for drug encapsulation and a hydrophilic shell for enhanced solubility and stability (Ghezzi et al., 2021). These micelles improve the solubility of poorly water-soluble drugs, prolong circulation times, and facilitate passive tumor targeting via the enhanced permeability and retention (EPR) effect (Zheng et al., 2024). Also, cationic polymeric micelles have demonstrated the ability to effectively protect nucleic acid molecules from degradation and thus enhance therapeutic efficacy (Cheng et al., 2026; Sinani et al., 2023). Recent investigations of polymeric micelles are listed in Table 5.

Curcumin, the major active compound in the turmeric rhizome, possesses numerous biological properties, including anti-leukemia activity. Studies have demonstrated that curcumin exhibits significant cytotoxicity against multiple AML cell lines. The underlying mechanisms may include the downregulation of WT1 gene expression and inhibition of FLT3 protein (Anuchapreeda et al., 2008; Tima et al., 2014). However, curcumin’s clinical application is limited by its low aqueous solubility and poor bioavailability. To overcome these challenges, various polymeric micelles have been investigated for the efficient delivery of curcumin, such as Poloxamer 407 (P407) and PLGA, aiming to enhance its anti-leukemic activity and suitability for further preclinical studies (Chueahongthong et al., 2024; Nirachonkul et al., 2021; Tima et al., 2019). Specifically, a functional mixed micellar system based on two co-assembled triblock copolymers, poly(2-(dimethylamino)ethyl methacrylate)-b-poly(ε-caprolactone)-b-poly(2-(dimethylamino)ethyl methacrylate) bearing triphenyl-phosphonium ligands (PDMAEMA₂₀^TPP+ -b-PCL₇₀-b- PDMAEMA₂₀^TPP+) and poly(ethylene oxide)-b-poly(ε-caprolactone)-b-poly(ethylene oxide) (PEO₁₁₃-b-PCL₇₀-b-PEO₁₁₃), has been assessed for mitochondria-targeted delivery of curcumin. This system exhibited superior pro-apoptotic activity in AML cells, highlighting its potential for AML therapy (Momekova et al., 2018).

Compared to polymersomes, micelles generally exhibit less drug loading and reduced stability (Alibolandi et al., 2015; Hu et al., 2017). Yet, though micelles face challenges such as instability and inability to carry hydrophilic drugs, their relatively straightforward preparation makes them widely used in drug delivery systems. Multiple strategies like crosslinking and macrocyclic host–guest complexation, are under investigation to improve the micelle stability (Lu et al., 2018).

5.2.1.3. Dendrimers.

Dendrimers are well defined, homogeneous nanostructures with tree-like branches (Sherje et al., 2018). Dendrimers have precise structural control and high drug-loading capacity due to their unique architecture, but toxicity, inadequate tumor accumulation due to rapid clearance, and synthetic complexity limit their clinical application for AML (Alven and Aderibigbe, 2020; Chis et al., 2020). Poly(amidoamine) (PAMAM) dendrimer, the most commonly studied dendrimer type, exhibit superior transfection efficiency, making them ideal for nucleic acid and chemotherapeutic agent delivery (Hassan and Zou, 2022; Jiang et al., 2016; Wróbel et al., 2023). However, the high positive charge density of PAMAM dendrimers can cause cytotoxicity through membrane disruption and oxidative stress (Chis et al., 2020). To mitigate these limitations, recent advances have explored hybrid systems, such as ferritin-dendrimer NPs, which encapsulate PAMAM dendrimers within ferritin protein cages to enhance biocompatibility and reduce cytotoxic effects. Ferritin, particularly human H ferritin, are attractive because of its ability to recognize the CD71 receptor, typically overexpressed in cancer cells, including AML (Acharya and Kala, 2019). This allows for targeted drug delivery with enhanced cellular uptake. A recent study demonstrated that a self-assembling molecular complex in which ferritin traps a positively charged PAMAM dendrimer facilitates the delivery of nucleic acids, including miRNA, into AML cells, effectively triggering differentiation-related phenotypic changes (Palombarini et al., 2021).

In summary, while polymersomes, polymeric micelles, and dendrimers each have their own advantages and disadvantages, the choice of which to use should be based on the specific context, including the nature of the loaded molecules, drug dose, administration methods, and the most critical delivery requirements.

5.2.2. Polymeric 3D scaffold

The AML BMM creates a protective niche responsible for frequent relapse and drug resistance phenomena (Pinho and Frenette, 2019; Sola et al., 2019). Thus, given the complex interactions between AML cells and BMM, alternative approaches are emerging that rely on new biocompatible materials and modern technologies to fabricate three-dimensional (3D) structures (Sharipol and Frisch, 2024). These scaffolds are a category of polymeric materials characterized by water-insoluble porous or network structures that hold promise for multidimensional functionalization, which could be used to mimic the BM niche. This makes them particularly promising for the in vitro and in vivo cell culture, expansion, and editing, enabling preclinical drug evaluation, CAR-T cell generation, leukemic cell targeting, and HSC expansion, for AML treatment (Li et al., 2024b).

5.2.2.1. 3D scaffold for ex vivo biomimicry of human bone marrow niche.

In regenerative medicine, porous biodegradable scaffolds are widely used to repair or regenerate damaged tissues. Similarly, in hematopoietic transplantation, polymeric 3D scaffolds provide an optimized environment for the ex vivo expansion of HSCs, which are the basis of BM transplantation. While BM transplantation remains a curative therapy for hematological diseases, the availability of healthy donor-derived HSCs, as well as potential GVHD, remain obstacles for prognosis of AML patients (Pinho and Frenette, 2019). For autologous transplantation, healthy HSCs are very rare in AML patients (Hong et al., 2022). Therefore, in vitro expansion of HSCs to obtain sufficient HSCs for transplantation is important. Compared to traditional two-dimensional (2D) cell culture, 3D scaffolds represent a powerful method to the achievement and development of tissue engineering applications, as they play an essential role in directing cell behaviors such as migration, proliferation and differentiation (Dozzo et al., 2023).

To reconstitute key features of the BMM, various biomaterials have been used in scaffold fabrication, including synthetic polymers like PEG, PLGA, poly(L-lactic acid) (PLLA) and polycaprolactone (PCL), as well as natural polymers like alginate, collagen, hyaluronic acid (HA), and chitosan (Darvish et al., 2019; Islami et al., 2018; Li et al., 2023b; Ottensmeyer et al., 2018; Seyed Hadi et al., 2019). Natural polymers support cell adhesion and signaling, while synthetic polymers provide tunable mechanical strength, degradation rates, and functionalization potential (Yadav et al., 2024). An ideal scaffold provides a microenvironment with appropriate porosity, mechanical stability, and biomolecular cues to maintain HSC multipotency and promote long-term self-renewal.

Beyond HSC expansion, 3D scaffolds offer a valuable tool for studying AML progression and drug resistance. The AML bone marrow niche provides protective signals that contribute to cell adhesion-mediated drug-resistance (CAMDR), a major barrier to effective chemotherapy (Nair et al., 2015). By integrating ECM proteins such as fibronectin and vitronectin, which mediate AML-stroma interactions, 3D scaffolds enable researchers to investigate the mechanisms underlying CAMDR and evaluate novel therapeutic strategies aimed at disrupting these interactions. The ability to culture AML cells within a physiologically relevant 3D microenvironment allows for a more accurate assessment of drug efficacy compared to traditional in vitro models.

5.2.2.2. Scaffold-based drug delivery.

The use of polymeric 3D scaffolds extends beyond cell culture and disease modeling to drug delivery applications. As traditional AML therapies fail to eliminate all leukemic cells, particularly LSCs that reside within the protective BM niche, scaffold-based drug delivery systems provide a potential solution by enabling localized and sustained drug release within the BM niche, thereby increasing drug bioavailability while minimizing systemic toxicity. Additionally, these scaffolds can be designed to incorporate NPs or targeted therapies, enhancing drug retention and selectivity toward AML cells (Blanco et al., 2010). MSC membrane-coated nanofibril scaffolds designed to mimic the BM niche, release CXCL12α to recruit LSCs to the scaffold for sustained, localized drug delivery that attenuates LSC proliferation and strengthens AML therapy (Ho et al., 2021).

Overall, polymeric 3D scaffolds offer a multifaceted approach to AML research and treatment. They provide an advanced platform for ex vivo HSC expansion, facilitate more clinically relevant disease modeling, and serve as innovative drug delivery systems. As biomaterial engineering and 3D fabrication technologies continue to advance, scaffold-based approaches are expected to play an increasingly prominent role in optimizing AML therapies and overcoming current treatment challenges.

5.3. Targeted formulations

Targeted drug delivery formulations offer a significant advantage over conventional nanocarriers by improving specificity and reducing systemic toxicity (Din et al., 2017). Various lipid-based formulations, polymeric NPs, and other delivery systems can be modified with targeting ligands to enhance selectivity for AML cells. Since numerous surface antigens on AML cells have been identified, targeting strategies typically employ antibodies, peptides, or nanobody-based approaches to improve therapeutic efficacy.

Antibodies remain the most widely used targeting moieties due to their high specificity and strong antigen-binding affinity (Bäumer et al., 2022; Nirachonkul et al., 2021). They can be conjugated to NPs to facilitate selective drug accumulation in AML cells. However, their large molecular size may limit tissue penetration and affect formulation stability. In contrast, peptide-based targeting offers an alternative approach with smaller molecular weight and improved tumor penetration (Gu et al., 2021). However, the specificity of peptide ligands still requires further validation before widespread clinical adoption.

Nanobodies, derived from camelid heavy-chain-only antibodies, represent a novel targeting strategy with several advantages over traditional antibodies. These single-domain antibody fragments exhibit high solubility, improved stability, and superior refolding capacity after denaturation (Lu et al., 2025; Romão et al., 2020; Zeng et al., 2024). Their smaller size enhances tumor penetration and facilitates multi-specific targeting approaches. For instance, a CD33/CD123-targeting nanobody T-cell engager (TCE) has demonstrated superior cytotoxicity against AML cells expressing either or both markers, outperforming single-targeted TCEs in preclinical models (Zeng et al., 2024). Given their modularity and potential for specific targeting, nanobody-based formulations hold great promise for next-generation AML therapies.

5.4. Oher formulations

In addition to lipid-based and polymer-based systems, a variety of alternative nanocarriers, such as niosomes, inorganic NPs, and protein-based nanocarriers, have also been explored for drug delivery in AML. Several reviews have provided comprehensive overviews of these platforms and their applications in AML therapy (Moghassemi and Hadjizadeh, 2014; Wu et al., 2024a; Zhao et al., 2025).

6. Conclusions and future directions

AML remains a complex and challenging hematologic malignancy with high relapse rates and limited treatment options, particularly for older and high-risk patients. While significant advancements have been made in the understanding of AML pathophysiology, including molecular classifications and targeted therapies, treatment resistance and disease recurrence remain major barriers to achieving long-term remission. The tumor microenvironment, genetic heterogeneity, and LSC persistence continue to complicate therapeutic efforts, necessitating the development of novel strategies to enhance treatment efficacy while minimizing toxicity.

Nanotechnology-based drug delivery systems have shown great promise in improving drug bioavailability, reducing off-target effects, and overcoming chemoresistance in AML. The incorporation of stimuli-responsive and targeted delivery systems has further enhanced the therapeutic potential of these nanocarriers. Additionally, advances in gene therapy, particularly RNA-based therapeutics such as microRNAs, long non-coding RNAs, and CRISPR-based gene editing, offer innovative avenues for AML treatment by modulating oncogenic pathways and overcoming drug resistance.

Despite these promising developments, several challenges remain. The heterogeneity of AML necessitates patient-specific therapeutic strategies, requiring further refinement of biomarker-driven treatment approaches. The clinical translation of nanomedicine and gene therapy faces hurdles related to large-scale production, regulatory approval, and long-term safety evaluation. Moreover, immunotherapies, including CAR-T cells and immune checkpoint inhibitors, are still in early-phase trials for AML, with efficacy limited by tumor immune evasion and the immunosuppressive BMM.

Future research should focus on optimizing nanocarrier formulations for AML, particularly those capable of delivering multiple therapeutic agents simultaneously to target leukemia at different levels. Personalized medicine approaches, incorporating patient-derived AML models and biomarker-driven drug selection, will be crucial in improving treatment precision. Additionally, further exploration of the BM niche and its role in AML resistance will provide insights into novel therapeutic targets. The integration of artificial intelligence and high-throughput screening in drug development may accelerate the discovery of next-generation therapeutics.

Ultimately, the convergence of nanomedicine, gene therapy, and immunotherapy holds great potential to revolutionize AML treatment. Future studies should aim to bridge the gap between preclinical discoveries and clinical application, ensuring that these innovative strategies translate into durable and effective therapies for AML patients.

Data availability

No data was used for the research described in the article.