Overcoming therapeutic challenges in acute myeloid leukemia: active targeting strategies by nano-drug delivery systems

Abstract

Background

Acute myeloid leukemia (AML) is a highly aggressive hematological malignancy characterized by poor overall survival and high relapse rates. The standard chemotherapy remains the conventional “7+3” regimens, while the suboptimal pharmacokinetics and significant systemic toxicity in AML present ongoing challenges to long-term disease control. Nano-drug delivery systems (NDDSs) have emerged as a promising strategy to overcome these barriers by enabling enhanced drug stability, targeted delivery, and specific distribution. Although several NDDS-based therapies have been approved by FDA, the clinical translation of nanomedicine in AML remains limited. This is largely due to the unique pathophysiology of AML, which lacks the vascular structures found in solid tumors, resulting in a limited and atypical enhanced permeability and retention (EPR) effect. Active targeting strategies, including antibody, aptamer, and peptide-based ligand modifications, offer a compelling approach to improve cellular specificity and therapeutic efficacy.

Methods

In this review, we provide a comprehensive overview of NDDSs engineered for AML, focusing on recent advances in active targeting approaches, their mechanistic advantages, and translational challenges.

Results

Current active-targeting NDDSs in AML generally follow two major directions. One direction focuses on surface receptors that are aberrantly overexpressed on AML cells, thereby improving payload specificity. The other direction focuses on bone marrow (BM)-targeted nanocarriers that utilize cell homing mechanisms and disease-associated markers of the BM microenvironment.

Conclusion

NDDSs designed for different targets, carrier materials, and release mechanisms have demonstrated improved pharmacodynamic effects, but they remain at the preclinical stage. Based on a summary of the current challenges facing NDDSs, this review further discusses key directions for next-generation system design, such as the development of personalized carriers, reduction of off-target effects, and more effective delivery to leukemia stem cells.

Introduction

Acute myeloid leukemia (AML) is a highly aggressive hematologic malignancy characterized by clonal expansion of undifferentiated myeloid precursors, leading to the disruption of normal hematopoiesis. The disease progresses rapidly, with high relapse rates and poor overall survival, especially among elderly and unfit patients [1–3]. The standard first-line “7+3” regimen, consisting of 7 days of cytarabine (Ara-C) and 3 days of an anthracycline, has remained largely unchanged for decades [4, 5]. Despite the incorporation of molecularly targeted agents, including FLT3 inhibitors (midostaurin, gilteritinib, quizartinib) [6–8], IDH1/2 inhibitors (ivosidenib, enasidenib) [9, 10], and BCL-2 inhibitors (venetoclax) [11], long-term outcomes remain unfavorable. Hematopoietic stem cell transplantation (HSCT) offers curative potential but is limited by donor availability, toxicity, and cost [12, 13]. Immunotherapeutic approaches such as monoclonal antibodies (mAbs), antibody-drug conjugates (ADCs), and CAR-T cell therapy are emerging, yet they are associated with significant adverse events and variable efficacy in AML due to antigen heterogeneity and immune evasion [14, 15].

Key challenges in AML therapy are the systemic dissemination, as well as the pronounced genetic and phenotypic heterogeneity. Traditional chemotherapy lacks specificity and causes off-target toxicities, while small-molecule inhibitors are only effective in patients with select mutations and may face resistance. These limitations underscore the urgent need for more precise and personalized therapeutic strategies.

In this context, nano-drug delivery systems (NDDSs) have gained significant attention in oncology for their ability to improve pharmacokinetics, enhance tumor specificity, and minimize systemic toxicity [16–18]. NDDSs such as liposomes, polymeric micelles, dendrimers, and antibody-conjugated nanocarriers have demonstrated therapeutic potential in various tumors, including AML. For example, CPX-351, a liposomal co-formulation of Ara-C and daunorubicin (DNR), has been approved by the FDA for high-risk AML subtypes, demonstrating improved survival compared to conventional chemotherapy [19, 20].

In AML, passive targeting through the enhanced permeability and retention (EPR) effect can result in nanoparticle accumulation in the splenic and hepatic sinusoids, as well as inflamed bone marrow niches [21–23]. Nevertheless, this level of passive targeting is insufficient to ensure effective drug delivery, as leukemia cells (LCs) circulate freely in the bloodstream and infiltrate organs rather than forming solid tumor masses. Therefore, this fundamental difference necessitates the development of active targeting strategies. AML cells aberrantly overexpress specific surface receptors, allowing NDDSs to achieve precise drug delivery through receptor-ligand interactions. This approach overcomes the non-specific distribution of conventional chemotherapeutic agents. In parallel, by moving beyond a precision medicine paradigm that relies solely on intracellular genetic mutations, it also offers a new strategy to address the marked heterogeneity of AML. Consequently, this strategy has the potential to shift the current subtype-specific treatment model toward a more broadly applicable therapeutic platform with improved efficacy and reduced toxicity, ultimately increasing remission rates, lowering relapse risk, and improving long-term survival.

In this review, we summarize the current landscape of active targeting strategies employed in NDDSs for AML treatment (Fig. 1), including available target receptor, nanocarrier engineering, and delivery mechanisms. We further discuss preclinical progress, translational obstacles, and potential solutions for promoting nanomedicine into AML clinical practice. Our aim is to provide a comprehensive framework for understanding and guiding the future development of precision-targeted nanomedicines in the treatment of AML.

Fig. 1.

Schematic overview of targeted drug delivery strategies for the treatment of acute myeloid leukemia

NDDSs for AML cells-specific delivery

Active targeting strategies utilizing AML surface receptors have emerged as a promising approach to achieve precise therapy. This section focuses on specific receptors that are aberrantly overexpressed on AML blasts but are low or absent on normal cells. The corresponding receptors and their nanocarrier applications are summarized in Table 1.

Table 1.

Actively targeted drug delivery systems based on leukemic cell surface receptors

Target	Ligand	Cargo	Carrier	Stimuli	Animal model			Year	Ref
					Type	Cell	Site
CXCR4	E5	E5	Micelle	-	CDX (SRM)	AE&CKITD816V-transduced splenocytes	Disseminated	2020	[24]
CXCR4	E5	E5+DOX	Micelle	-	CDX (SRM)	AE&CKITD816V-transduced splenocytes	Disseminated	2022	[25]
CXCR4	E5	Fe3O4 and Pt	Nanozyme	pH	CDX	HL-60	Disseminated	2021	[26]
CXCR4/CD44	E5/HA	DNR	PBNP	-	CDX	HL-60	Disseminated	2021	[27]
CXCR4	E5	DNR + Ara-C	HMPBs	Laser	-	-	-	2021	[28]
CXCR4	E5	E5+AraN	Liposome	-	CDX	HL-60-Luc	Disseminated	2024	[29]
CXCR4	T22	mRTA	NCs	-	CDX	THP-1-Luc	Disseminated	2018	[30]
CXCR4	T22	MMAE	NCs	-	CDX	THP-1-Luc	Disseminated	2020	[31]
CXCR4	T22	Ara-C	NCs	-	CDX	THP-1-Luc	Disseminated	2022	[32]
CXCR4	T22	DITOX	NCs	-	CDX	THP-1-Luc	Disseminated	2021	[33]
CXCR4	T22	PE24	NCs	-	CDX	MONO-MAC-6-Luc	Disseminated	2023	[34]
CXCR4	T22	MMAE	NCs	-	CDX	THP-1- Luc	Disseminated	2022	[35]
CXCR4/MUC1	T22-NLS/aptarmer	CRISPR-Case9 plasmid	Composite nanocarrier	-	-	-	-	2022	[36]
CXCR4	T22	VEN + SOR	Micelle	-	CDX	MV4-11-Luc	Disseminated	2024	[37]
CXCR4	AMD3100	AMD3100 + siRNA	PCX	-	GEM	Cbfb-MYH11+ cells	Disseminated	2020	[38]
CXCR4	CXCL12	Pt + DOX	Biomimetic NPs	-	CDX	HL-60	Disseminated	2022	[39]
CXCR4	CXCL12	DNR	Biomimetic NPs	-	-	-	-	2023	[40]
CXCR4	CXCL12/AMD3100	AMD3100+DOX+PDT	Biomimetic NPs	-	CDX	HL-60-Luc	Disseminated	2024	[41]
CD71	Tf	c-myb ASOs	Tf-polylysine	-	-	-	-	1992	[42]
CD71	Tf	GTI-2040	Tf-LPPs	pH	-	-	-	2010	[43]
CD71	Tf	miR-29b	Tf-LPPs	pH	CDX	MV4-11	Disseminated	2013	[44]
CD71	Tf	LOR-1248	Tf-NPs	-	CDX	MV4-11	Subcutaneous	2014	[45]
CD71	Tf	DOX	LPNs	-	-	-	-	2017	[46]
CD71	Tf	EDS	PEG-PLL-PLGA	-	-	-	-	2017	[47]
CD71/CD117	Tf/aptamer	DRN and Lut	Micelle	-	CDX	HL-60	Subcutaneous	2023	[48]
CD71/CD11b	Tf/C3b	DOX	FH	-	CDX	MOLM-13	Disseminated	2025	[49]
CD71	HumFt	Cytochrome C	HumFt	-	-	-	-	2019	[50]
CD71	Ferritin	miRNA-145-5p	HumFt-PAMAM	-	-	-	-	2021	[51]
CD71	Ferritin	ATO	Ferritin	-	CDX	Luc-HL-60	Disseminated	2021	[52]
CD71	HFn	Ara-C	HFn	-	CDX	Luc-HL-60	Disseminated	2023	[53]
CD71	HFn	siRNA	HFn	-	-	-	-	2024	[54]
CD71	OKT9	DOX	Liposome	-	-	-	-	1997	[55]
CD123	PO-6	PO-6	Micelle	-	CDX (SRM)	AE&CKITD816V-transduced splenocytes	Disseminated	2021	[56, 57]
CD123	mAb	DNR	Liposome	-	-	-	-	2017	[58]
CD123	mAb	DNR	Niosome	-	CDX	THP-1	Disseminated	2017	[59]
CD123/CD33	mAb	DNR	Liposome	-	-	-	-	2019	[60]
CD123	Fab’ fragment	siRNA	Micelle	-	-	-	-	2017	[61]
CD123	ZW25	DNR	TDT	-	CDX	MOLM-13	Subcutaneous	2017	[62]
CD123	SS30	SS30	Hydrogel	Cas9/sgRNA	CDX	MOLM-13	Subcutaneous	2021	[63]
CD44	HA	Cur	Liposome	-	CDX	KG-1	Disseminated	2017	[64]
CD44	HA	6-MP	Prodrug	GSH	CDX	OCI-AML-2	Subcutaneous	2017	[65]
CD44	HA	DOX	LACHA	GSH	CDX	AML-2	Subcutaneous	2017	[66]
CD44	HA	HDC	GDH	-	-	-	-	2017	[67]
CD44	HA	DOX+GA	LPHNs	-	CDX	HL-60	Subcutaneous	2019	[68]
CD44	mAbs		PLGA NPs	-	-	-	-	2019	[69]
CD33	scFv	siRNA	Liposome	-	-	-	-	2010	[70]
CD33	mAbs	PMI	LONp	-	-	-	-	2018	[71]
CD33	Fab’ fragment	Ara-C	Liposome	pH	CDX	HL-60	Disseminated	2009	[72, 73]
CD33	scFv	GTI-2040	LNPs	-	CDX	Kasumi-1	Subcutaneous	2015	[74]
CD33	mAbs	G3139	aCD33-NKSN	-	CDX	Kasumi-1	Subcutaneous	2022	[75]
CD33	Aptamer	ASO	Gold NPs	-	-	-	-	2016	[76]
CD33	mAbs	ASO	RBCEVs	-	CDX	MOLM13-Luc-GFP	Disseminated	2022	[77]
CD33	mAbs	-	MoS2	-	-	-	-	2021	[78]
CD38	Dara	TRAIL	Liposome	-	CDX	OCI-AML2	Disseminated	2023	[79]
CD38	Dara	Vincristine	Polymersomal	-	CDX	MOLM-13-Luc	Disseminated	2023	[80]
FR	FA	DOX	Liposomes	-	CDX	KG-1	Ascites	2002/2007/2010	[81–83]
FR	FA	SiRNA	Albumin	-	CDX	KG-1	Subcutaneous	2021	[84]
FR	FA	DNR + Eme	Liposomes	-	-	-	-	2014	[85]
FR	FA	PTX TanIIA	LB-MSN	-	CDX	NB4	Subcutaneous	2020	[86]
FLT3	FLT3L	miR-150	Dendrimers	-	GEM	MLL-AF9-transduced BM	Disseminated	2016	[87]
FLT3	scFv	scFv	ELP	-	CDX	MOLM-13	Disseminated	2020	[88]
FLT3/CD99	scFv	scFv	ELP	-	CDX	MV4-11	Disseminated	2024	[89]
CD117	mAb	DM1	ADC	-	-	-	-	2018	[90]
CD117	mAb	DM1	ADC	-	CDX	Kasumi-1	Subcutaneous	2022	[91]
CD117	Aptamer	MTX	Aptamer-drug conjugate	-	-	-	-	2015	[92]
CLL-1	CLL1-L1	DNR	Micelle	-	-	-	-	2012	[93]
CLL-1	CLL1-L1	DNR	Micelle	-	PDX	Primary cells	Disseminated	2019	[94]
CD34	mAb	DOX	Liposome	-	-	-	-	2004	[95]
CD99	scFv	scFv	ELP	-	CDX	MOLM-13	Disseminated	2020	[96]
CD96	mAbs	ICG	CPSNPs	-	CDX	32D-p210-GFP	Disseminated	2011	[97]

Abbreviations: CDX, cell-derived xenograft; SRM, secondary recipient model; GEM, genetically engineered mouse. HFn, heavy ferritin chain; ICG, indocyanine green. Others are defined in the Abbreviations section

Among the best-characterized targets, CXCR4, CD123, CD44, and CD71 have garnered significant attention due to their roles in leukemic cell survival, proliferation. CXCR4, CD123, and CD44 are involved in intracellular signaling pathways critical to AML pathogenesis. Notably, drug delivery systems directed at these receptors not only facilitate targeted payload delivery but may also confer direct antileukemic effects via receptor blockade. In contrast, CD71 (transferrin receptor 1) is primarily associated with cellular iron metabolism and proliferation. Its high expression in AML, particularly among aggressive and proliferative subtypes, renders it an attractive target for therapeutic intervention.

Various targeting ligands, including monoclonal mAbs, peptides, aptamers, and polysaccharide derivatives, have been designed to engage these receptors, with each type offering distinct advantages in terms of affinity, stability, and translational feasibility. For example, antibodies exhibit high affinity but are limited by high immunogenicity, large molecular size, and high cost. Nucleic acid aptamers are easy to synthesize but are prone to degradation in vivo. In comparison, peptides combine high affinity with ease of synthesis, while also offering low immunogenicity, controlled chemical synthesis, good batch-to-batch consistency, and facile modification. However, peptides can be vulnerable to proteolytic cleavage. The selection of an optimal ligand thus depends on a comprehensive consideration of the intended application, required pharmacokinetics, and translational practicality.

C-X-C chemokine receptor 4 (CXCR4)

CXCR4 is a G-protein coupled receptor with seven transmembrane domains that binds stromal derived factor 1α (SDF-1α or CXCL12) [98]. This receptor-ligand interaction plays a crucial role in directing LCs homing and retention in BM and spleen niches, where LCs receive survival and anti-apoptotic signals from stromal support [99, 100]. Overexpression of CXCR4 is frequently observed in AML and is associated with poor prognosis and chemotherapy resistance [101, 102]. Consequently, the CXCR4/CXCL12 axis has become an attractive therapeutic target for active-targeting nanomedicine strategies.

One of the most widely studied approaches involves the CXCR4-antagonistic peptide E5. Meng et al. synthesized E5 and fabricated micelles (M-E5) with DSPE-PEG [24]. M-E5 bound efficiently to CXCR4 at the ECL1/ECL2 and N-terminal domains, which overlap with CXCL12 binding sites. By downregulating proliferation and adhesion genes and upregulating apoptosis and differentiation genes, M-E5 blocked the CXCR4/CXCL12 axis and mobilized LCs into peripheral blood (PB). Building on this, a co-delivery micelle (M-E5-DOX) was created by loading doxorubicin (DOX) with E5 using DSPE-mPEG2000 (Fig. 2A) [25]. This co-delivery system integrates three key functions: CXCR4 targeting, CXCR4/CXCL12 axis inhibition and direct LCs elimination.

Fig. 2.

CXCR4-targeted nano-drug delivery systems (NDDSs) for AML therapy. (A) E5-modified micelles co-loaded with doxorubicin (M-E5-DOX), enabling CXCR4-targeted delivery, inhibition of the CXCR4/CXCL12 axis, and direct elimination of leukemic cells [25]. (B) CXCR4-targeted, photothermal-responsive HMPB co-loaded with DNR and Ara-C, surface-modified with zwitterionic sulfobetaine (ZS) to reduce nonspecific interactions, and conjugated with E5 peptide for leukemia cell targeting [28]. (C) E5-modified liposomes (LipoAran) co-loaded with a hydrophobic Ara-C derivative, enabling CXCR4-targeted delivery and enhanced liposomal stability. (Adapted with permission from [29]. Copyright 2024 American chemical society.) (D) T22-based protein nanoconjugates incorporating human-derived scaffolds for CXCR4-targeted delivery of cytotoxic agents with enhanced in vivo stability and reduced immunogenicity [35]. (E) T22-decorated, disulfide-crosslinked polymeric micelles for CXCR4-targeted co-delivery of venetoclax and sorafenib with enhanced synergistic efficacy. (Adapted with permission from [37]. Copyright 2024 American chemical society.) (F) Cholesterol-modified polymeric CXCR4 antagonist (PCX) enables CXCR4 blockade and siRNA delivery with nuclease protection [38]. (G) Biomimetic nanoplatform (BOC@PLGA@DG@ACM/A) for AML-targeted co-delivery of DOX, GOX, and PDT agents with enhanced therapeutic efficacy. (Adapted with permission from [41]. Copyright 2024 American chemical society.)

Zhang’s group further advanced this strategy by combining E5 with nanozymes to form Fe3O4@Pt@E5 [26]. Fe₃O₄ and Platinum (Pt) catalyzed H₂O₂ into ·OH under acidic pH lysosomal conditions, inducing LCs death. They also designed a CXCR4/CD44 dual-target Prussian blue nanoparticles (PBNPs) to deliver DNR [27]. Further, a multifunctional system, HMPBs (DNR + AraC)@PEI-ZS-E5, was developed. It incorporated hollow mesoporous Prussian blue (HMPB) NPs with zwitterionic modification for stability and E5 for targeting (Fig. 2B) [28]. Despite its promising design, this system was only validated in vitro. Challenges for in vivo translation include blood dilution effects, optimization of laser power for photothermal therapy, and evaluation of whether localized bone marrow exposure can achieve sufficient therapeutic efficacy in the context of disseminated leukemia cells.

Other formulations have explored E5 as both targeting ligand and CXCR4 blocker. For example, LipoAraN-E5 was developed as a dual-function liposomal formulation that co-loads E5 and AraN, an amphiphilic Ara-C derivative generated by C14 conjugation (Fig. 2C). This design resulted in a structurally stable liposomal system [29].

Another notable ligand is T22, derived from polyphemusin II of the horseshoe crab. T22 functions both as a CXCR4 antagonist and a structural element in self-assembling nanoparticles. Compared to the existing protein drug delivery system, mostly ADCs, T22-based NPs offer superior in vivo stability, lower renal filtration, and reduced proteolysis. A Spanish group engineered T22-tagged nanoconjugates (NCs) such as T22-mRTA-H6 [30], T22-GFP-H6-MMAE [31] and T22-GFP-H6-Ara-C [32], achieving potent antitumor effects with low systemic toxicity in CXCR4-overexpressing AML cells. Further systems, including T22-DITOX-H6 [33] and T22-PE24-H6 [34], incorporated bacterial toxins but faced immunogenicity issues. To overcome this, human protein scaffolds such as Stefin A, chorionic gonadotropin, and Nidogen G2 were employed, generating NPs like T22-STM-H6-MMAE, which retained efficacy with reduced off-target effects (Fig. 2D) [35].

Beyond peptide-based designs, Ren et al. developed a dual-target nanomedicine (P@PPM) combining CXCR4 and MUC1 aptamers with CRISPR-associated protein 9 (Cas9) cargo to downregulate CXCR4 expression and induce LC death [36]. T22 was also recently incorporated into a disulfide cross-linked polymeric micelle designed to co-deliver VEN and sorafenib (SOR) (Fig. 2E) [37], achieving enhanced therapeutic effects at reduced drugs dosages by optimizing the drug ratio.

Small-molecule CXCR4 antagonists have also been repurposed. AMD3100 (Plerixafor), modified with cholesterol, yielded polymeric CXCR4 antagonist (PCX) (Fig. 2F) [38]. PCX both inhibited CXCR4 to sensitize LCs and served as a carrier for small interference RNA (siRNA), protecting it from nuclease degradation.

In addition to these synthetic systems, biomimetic strategies have also been developed for CXCR4-targeted delivery. Kong et al. constructed cell membrane-derived nanoparticles encapsulating PFOB@PLGA@Pt and DOX (PFOB@PLGA@Pt@DOX-CM) [39]. These particles converted H₂O₂ into reactive oxygen species (ROS), killed LCs, inhibited BM infiltration, and homed to BM niches. Another design, HMPB(DNR)@CM NPs, combined DNR-loaded Prussian blue cores with BMSC membranes, achieving CXCR4 targeting and ROS scavenging to protect the liver from DNR [40]. Recently, a multi-mechanism platform, BOC@PLGA@DG@ACM/A, integrated photodynamic therapy (PDT), chemotherapy, and biomimetic targeting (Fig. 2G). It employed AML cell membranes decorated with AMD3100, DOX, and glucose oxidase, alongside a PLGA core encapsulating CPPO, PFOB, and the photosensitizer Ce6 [41].

In summary, CXCR4-targeted NDDSs span a diverse range of delivery platforms, from polymeric micelles and protein-based self-assembling systems to gene-editing nanocarriers and biomimetic constructs. These formulations collectively offer promising strategies for disrupting LC-niche interactions and enhancing therapeutic precision in AML.

CD71

CD71, or transferrin receptor 1, is a 97-kDa type-II transmembrane glycoprotein that binds transferrin (Tf) and ferritin to mediate cellular iron uptake [103, 104]. Iron serves as an essential mental cofactor to regulate important cell process, such as cellular respiration, DNA synthesis and proliferation [105, 106]. Tumor cells, including AML cells, exhibit a heightened requirement for iron to support DNA synthesis during active proliferation [105–107]. Notably, CD71 expression in AML cells surpasses that observed in other types of leukemia [108–111], making it an attractive target for drug delivery. Moreover, anti-CD71 therapies have demonstrated anti-leukemia efficacy [112, 113], further supporting the feasibility of CD71 as a therapeutic target.

Tf is a key ligand for targeting CD71-high cells. In 1992, researchers used transferrin-polylysine to deliver c-myb antisense oligodeoxynucleotides (ASOs) [42]. Marcucci’s team further developed a series of Tf-based NPs for non-coding RNAs delivery. They first used Tf-modified pH-sensitive lipopolyplex nanoparticles (LPPs) to deliver GTI-2040, an ASO targeting the R2 subunit of ribonucleotide reductase (RNR), which later entered Phase I and II clinical trials (Fig. 3A) [43]. They later delivered miR-29b using transferrin-conjugated anionic LPPs [44], and utilized Tf-modified lipid nanoparticles (LNPs) to carry LOR-1248, a siRNA targeting RNR which showed strong antitumor activity [45].

Fig. 3.

Nanocarriers for CD71-targeted drug delivery. (A) Tf-modified pH-sensitive lipopolyplex nanoparticles (LPPs) for targeted delivery of GTI-2040, an ASO targeting R2 subunit of RNR. (Adapted with permission from [43]. Copyright 2010 American chemical society.) (B) Feraheme (FH)-based nanoparticles with a protein corona of transferrin and C3b for dual targeting of CD71 and CD11b, enhancing DOX delivery and reducing AML burden in vitro and in vivo. (Adapted with permission from [49]. Copyright 2025 American chemical society.) (C) AfFt, a humanized archaeoglobus ferritin, enables CD71 targeting and efficient encapsulation of positively charged cargos, including cytochrome C [50]. (D) As@Fn nanoparticles, thermotolerant apoferritin loaded with ATO and iron, retain ATO in its trivalent form while exhibiting high CD71 affinity, highlighting their potential for large-scale production and clinical application [52]. (E) ferritin nanocages loaded with ara-C, demonstrating stable CD71 expression and effective delivery for leukemia treatment [53]

Other Tf-modified systems also demonstrated efficacy. Zhu et al. synthesized a novel target ligand, transferrin-polyethylene glycol-oleic acid (Tf-PEG-OA), and prepared Tf-DOX P85/LPNs which effectively killed LCs while overcoming multidrug resistance [46]. Sun et al. developed Tf-modified PEG-PLL-PLGA micelles for edelfosine (EDS) delivery, achieving significant anti-leukemia effect [47]. Zhu et al. designed a binary nanodrug, combining Tf-Lut (Tf-modified, luteolin-loaded NPs) with AP-Drn (aptamer-decorated, DNR-loaded NPs), targeting CD71 and CD117 respectively [48]. Wu et al. exploited protein corona formation on feraheme (FH) to achieve dual targeting, as the corona contained both Tf and C3b. Using FH as a DOX carrier (DOX@FH), they achieved dual targeting of CD71 and CD11b and reduced AML burden in vitro and in vivo (Fig. 3B) [49].

Ferritins are emerging as promising nanocarriers for drug delivery due to their hollow structure and targeting ability to CD71. Bonamore’s team engineered humanized Archaeglobus fulgidus (AfFt) by grafting a loop from human H-ferritin onto AfFt (humanized Archaeoglobus ferritin, HumFt), enabling efficient CD71-targeting capability. HumFt retained the ability to efficiently encapsulate positively charged cargos, such as full-length cytochrome C (Fig. 3C) [50]. To enable the loading of negatively charged nucleic acids, polyamine dendrimers were incorporated into the HumFt cavity, enabling the delivery of miRNA-145-5p [51]. Considering the hollow cage and the natural absorptivity to Fe ions of apoferritin (ferritin without an iron core), Wang et al. utilized thermotolerant apoferritin to load arsenic trioxide (ATO) and iron, forming Fe-O-As cores inside the nanocages (Fig. 3D) [52]. The resultant As@Fn nanoparticles retained ATO in its medicinal trivalent form while exhibiting high CD71 affinity. This streamlined and safe preparation procedure highlighted the potential for large-scale production and clinical application. This team also leveraged ferritin nanocages to load Ara-C, demonstrating stable CD71 expression levels during treatment (Fig. 3E) [53].

Anti-transferrin receptor (TfR) mAbs offer another approach for CD71 targeting. Suzuki’s group developed OKT9-modified chemoimmunoliposomes (OKT9-CIL) encapsulating DOX [55], which specifically bound to CD71-positive LCs and maintained high DOX levels in K562/ADM cells, which are resistant to conventional DOX treatment.

Overall, diverse CD71-targeted delivery strategies have demonstrated effective drug delivery and antitumor activity in leukemia models. Among them, ferritins stand out for their dual role as carriers and targeting ligands, positioning CD71-targeted systems as a promising platform for AML therapy.

CD123

CD123, also known as the interleukin-3 receptor alpha subunit (IL-3Rα), is a type-I cytokine receptor coded in the pseudo-autosomal region of Xp22.3 and Yp22.3 [114–116]. Upon binding IL-3, CD123 heterodimerizes with the common β-subunit of the GM-CSF/IL-5/IL-3 receptor complex, activating JAK/STAT and PI3K/mTOR pathways and upregulating anti-apoptotic proteins, thereby promoting cell differentiation and proliferation [116, 117]. CD123 is expressed in 40%–93% of leukemia cell samples, with variations likely due to detection methods or patient cohorts [118–121]. It is aberrantly overexpressed by CD34⁺CD38^- AML cells but undetectable or low in their normal BM counterparts [118, 119], enabling proliferation even under low IL-3 conditions [122]. Notably, only the CD34⁺/CD38^-/CD123⁺ subpopulation is capable of initiating leukemia in immunodeficient mice, establishing CD123 as a marker of leukemia stem cells (LSCs) [123]. Moreover, CD123 expression is closely associated with poor prognosis in AML [124–128]. Given these features, CD123 is a compelling therapeutic target in AML.

Xu et al. successfully identified a CD123-selective binding peptide (PO-6) [56, 57], which competes with IL-3 for CD123 binding. When encapsulated in amphiphilic polymeric micelles (^mPO-6, Fig. 4A), PO-6 exhibited improved solubility and cellular uptake. Beyond targeting, ^mPO-6 mimicked a CD123 antibody, effectively blocking the CD123/IL-3 axis, resulting a significant anti-leukemia effect and prolonged survival time in mice. Besides blocking the CD123/IL-3 axis, PO-6’s targeting capability may also serve as a ligand for future drug delivery systems.

Fig. 4.

CD123-targeted nanomedicines for enhanced drug delivery and therapy. (A) The amphiphilic polymeric micelle (^mPO-6) encapsulating the CD123-selective peptide PO-6 enhances cellular uptake and CD123/IL-3 axis blockade. (Adapted with permission from [57]. Copyright 2022 American chemical society.) (B) CD123-LIP, a CD123 antibody-conjugated DNR-loaded immunoliposome [58]. (C) CD.DSPE-PEG-Fab, a CD123 Fab-conjugated cyclodextrin-based NP for siRNA delivery and BRD4 silencing. (Adapted with permission from [61]. Copyright 2017 American chemical society.) (D) The CD123 aptamer-conjugated TDT enables selective targeting and high-efficiency DOX loading [62]. (E) SS30 polyaptamer hydrogel (SSFH) system releases SS30 via CRISPR/Cas9, enhancing therapeutic efficacy in AML cells [63]

As an alternative strategy, Wang et al. developed antibody-based NDDSs. They conjugated a CD123 antibody onto a DNR-loaded liposome via a PEGylated linker, generating CD123-LIP (Fig. 4B) [58], and similarly designed an anti-CD123 niosome (CD123-NS) for DNR delivery [59]. To overcome limitations of single-targeting, such as relapse from target-negative cells, they further constructed a CD123/CD33 dual-antibody-modified liposome to broaden the targeting spectrum [60]. In parallel, Gou’s team constructed another CD123 antibody-decorated NDDS by conjugating the Fab fragment onto cyclodextrins through a PEGylated linker (CD.DSPE-PEG-Fab, Fig. 4C) [61]. This system enabled efficient siRNA delivery, overcoming challenges such as rapid clearance and poor intracellular trafficking, thereby achieving effective BRD4 silencing and notable antitumor activity in vitro.

Given the sensitivity of antibodies to temperature, pH, and freeze-thaw cycles, Wu et al. applied systematic evolution of ligand exponential enrichment (SELEX) to generate high-affinity CD123 aptamers ZW25 and CY30 [62]. ZW25 was then incorporated into an aptamer-mediated targeted drug train (TDT) containing two C/G-rich probes and a pair of aptamer-linked trigger probes (Fig. 4D). This system exhibited high loading efficiency, as DOX efficiently intercalates into the C/G-rich regions at high concentration. The aptamers demonstrated selective binding to CD123, and the TDT based on them showed promising therapeutic efficiency. Furthermore, this team developed the first CD123 thioaptamer, SS30, exhibited high affinity for CD123 and compete with IL-3 at the CD123 binding cite, blocking multiple signaling pathways and inhibiting AML cells growth [129, 130]. To overcome SS30 instability in vivo, they designed a DNA hydrogel delivery platform, termed SS30 polyaptamer hydrogel (SSFH) (Fig. 4E). This hydrogel system enabled CRISPR-Cas9-mediated controlled release and enhanced therapeutic efficacy [63].

CD44

CD44 is a non-kinase transmembrane glycoprotein encoded on chromosome 11 [131, 132], and it plays a critical role in normal myelopoiesis [133]. On progenitor cells, CD44 mediates adhesion to hyaluronic acid (HA), a glycosaminoglycan component of the extracellular matrix [134–136]. Blocking CD44 with mAbs has been shown to completely inhibit long-term myelopoiesis in bone marrow (BM) culture in vitro [135, 137]. Notably, CD44 is significantly overexpressed in various malignancies, including AML [138, 139]. Upon HA binding, CD44 becomes activated and recruits cytoplasmic adaptor proteins, initiating signaling pathways that support tumor cell survival [138]. Inhibition of this signaling consistently exerts anti-AML effects [140–142]. Together, these findings highlight CD44 as both a promising therapeutic target and a potential mediator for active drug delivery.

HA, the natural ligand of CD44, has been widely applied in AML drug delivery. Sun et al. developed HA-modified liposomes for curcumin delivery (HA-Cur-LPs, Fig. 5A) [64], which exhibited high affinity and significant therapeutic potential. Qiu et al. designed HA-mercaptopurine prodrug (HA-GS-MP) to specifically target CD44-positive LCs [65]. HA-GS-MP was composed of HA conjugated to 6-Mercaptopurine (6-MP) via a glutathione (GSH)-responsive carbonyl vinyl sulfide linker, which enhanced the stability and water solubility of 6-MP while enabling rapid release and effective tumor inhibition (Fig. 5B). Zhong et al. developed another CD44-targeted nanocarrier by encapsulating DOX in lipoic acid-crosslinked HA (LACHA-DOX) (Fig. 5C), which demonstrated potent anti-AML effect both in vitro and in vivo [66]. Cherukula et al. developed a graphene quantum dots (GDH)-based system for histamine dihydrochloride (HDC) delivery to suppress ROS in LCs [67], where HA was conjugated to 3,4-dihydroxy-L-phenylalanine (DA) for stability, and then anchored onto HDC loaded GDH. Shao’s group designed lipid-polymer hybrid NPs for co-delivery of DOX and gallic acid (GA), using DSPE-PEG-HA to enable CD44 targeting [68].

Fig. 5.

CD44- and CD33-targeted nanomedicines for AML therapy. (A) HA-Cur-LPs enable CD44-targeted curcumin delivery with therapeutic potential in AML. (Adapted with permission from [64]. Copyright 2017 American chemical society.) (B) HA-GS-MP enables GSH-responsive 6-MP release with improved solubility and AML inhibition. (Adapted with permission from [65]. Copyright 2017 American chemical society.) (C) LACHA-DOX delivers DOX via CD44 targeting, showing strong anti-AML effects in vitro and in vivo [66]. (D) Anti-CD33 fab-conjugated pH-sensitive liposomes enhance Ara-C delivery and retention in AML cells [72]. (E) CD33 antibody-conjugated RBCEVs enable targeted delivery of ASOs against FLT3-ITD and miR-125b while preserving vesicle integrity [77]. (F) Anti-CD33 MoS₂ nanoflakes for potential AML diagnosis and therapy [78]

In addition to HA-based approaches, CD44 mAbs have been employed to enhance target drug delivery. Noureldien’s team conjugated anti-CD44 mAbs to PLGA NPs encapsulating parthenolide (PTL) [69]. Their study showed that elevated NF-κB activity correlates with poor AML prognosis and that PTL, a natural NF-κB inhibitor, can be effectively delivered using this targeted approach.

CD33

CD33, or Siglecs-3, is a 67-kDa transmembrane glycoprotein selectively expressed on myeloid linage, and its activation inhibits tyrosine phosphorylation and calcium mobilization [143–145]. In AML, CD33 is overexpressed in 85–90% of cases [146–149], with a molecule density in BM cells approximately 3.5-fold higher than in normal counterparts [150]. High CD33 expression is associated with poor prognosis [151, 152]. Therapeutic strategies targeting CD33, including mAbs and CD33-specific immunotoxin, have shown clinical potential in AML [153–155]. Notably, CD33-targeting therapies restore normal clonal hematopoiesis without impairing differentiation of CD33^–CD34⁺ myeloid progenitors [154], supporting its value as a selective drug delivery target.

The most common strategy for targeting CD33-positive leukemic cells is antibody-based drug delivery. Rothdiener et al. developed siRNA-loaded liposomes targeting the AML1/MTG8 fusion protein, with an anti-CD33 single-chain variable fragment (scFv) conjugated to the liposomal surface to enhance specificity [70]. Anti-CD33 mAbs have also been applied to construct multifunctional carriers [71]. For instance, a humanized anti-CD33 mAbs was conjugated to fluorescent lanthanide oxyfluoride nanoparticles (LONps) co-loaded with a dual MDM2/MDMX peptide inhibitor (PMI), allowing disease tracking and treatment monitoring.

Given that the efficacy of Ara-C depends on intracellular bioavailability and sustained exposure, rapid clearance of free Ara-C increases the risk of drug resistance. To address this, anti-CD33 Fab fragments were conjugated to pH-sensitive liposomes for targeted Ara-C delivery (Fig. 5D) [72, 73]. Another approach modified LNP carrying GTI-2040, an ASO targeting mRNA of R2 subunit of RNR, with anti-CD33 scFv to overcome Ara-C resistance [74].

Additional CD33-targeted gene delivery strategies have also been explored. A vector (aCD33-NKSN) was designed to deliver the BCL2 ASOs (G3139), incorporating a nuclear localization signal, fusion peptide, and stearic acid to enhance nuclear transport [75]. Considering the therapeutic potential of ASOs in AML, Zaimy et al. developed gold nanoparticles functionalized with five ASOs targeting key AML oncogenes, integrated with an anti-CD33⁺/CD34^- aptamer for precise targeting [76]. Furthermore, red blood cell-derived extracellular vesicles (RBCEVs) were engineered to deliver ASOs against FLT3-ITD and miR-125b [77]. In this system, CD33 antibodies were attached to RBCEVs through a biotin-streptavidin interaction while preserving vesicle membrane integrity (Fig. 5E).

Beyond drug and gene delivery, CD33-targeting strategies have also been applied in diagnostics. For example, an anti-CD33 antibody was conjugated to MoS₂ nanoflake using biotin-avidin interactions, demonstrating potential for AML diagnosis and therapy (Fig. 5F) [78].

These advancements highlight the versatility of CD33-targeted approaches in both therapeutic and diagnostic applications for AML.

CD38

CD38, also known as cyclic ADP ribose hydrolase, is a 45-kD type II transmembrane glycoprotein [156]. As an exoenzyme, it converts nicotinamide adenine dinucleotide (NAD) into adenosine diphosphate ribose (ADPR) and cyclic ADP ribose (cADPR), the latter acting as a key second messenger in Ca²⁺ signaling [157–159]. It is highly expressed in hematologic malignancies, including myeloma, chronic myeloid leukemia (CML), and AML, but shows low expression in normal tissues [160]. Experimental studies indicate that CD38 activation promotes proliferative signaling, whereas anti-CD38 antibodies inhibit its enzymatic activity and suppress cell growth [156]. Daratumumab, an anti-CD38 antibodies approved by FDA for treatment of multiple myeloma (MM) [161, 162], has recently been applied in NDDS both as a therapeutic agent and as a targeting ligand for AML nanomedicines.

A research team in Ireland conjugated Daratumumab to PEG-lysed liposomes, which were functionalized with recombinant human TNF-related apoptosis-inducing ligand (TRAIL) (Fig. 6A) [79]. This approach was inspired by the finding that soluble TRAIL showed limited cytotoxicity in clinical trials, while membrane-bound TRAIL on NK cells exhibited potent antitumor activity. The NK·NPs maintained strong tumoricidal effects and avoided inactivation by tumor-derived immunosuppressive factors. The inclusion of daratumumab also enabled active CD38 targeting.

Fig. 6.

Other targeted nanomedicines for leukemia cells. (A) NK·NPs co-displaying TRAIL and daratumumab mimic NK cells for CD38-targeted AML therapy with enhanced cytotoxicity [79]. (B) ATRA combined with daratumumab-decorated polymersomal vincristine (DPV) induces CD38 upregulation and improves therapeutic efficacy in CD38-low AML [80]. (C) ATRA-loaded albumin NPs enhance FR-β expression and improve targeting efficiency of FA-siRNA nanocarriers in AML [84]. (D) Anti-FLT3 scFv fused with A192 polypeptide enables stable and cost-effective FLT3-targeted delivery [88]. (E) Dual-targeting NPs co-assembled from CD99–A192 and FLT3–A192 fusion proteins enhance AML therapy via CD99 and FLT3 recognition [89]. (F) CD117-specific ssDNA aptamer-drug conjugate (apt-MTX) enables targeted methotrexate (MTX) delivery with enhanced intracellular accumulation and reduced systemic toxicity [92]. (G) CLL1-L1 peptide-functionalized telodendrimer micelles enable selective DNR delivery to CLL-1⁺ LSCs [93]. (H) disulfide-crosslinked CLL1-targeted micelles [94]

Given the heterogenous expressive of CD38 level in AML patients, CD38-targeted nanomedicine may face translational challenges. To overcome this limitation, researchers introduced all-trans retinoic acid (ATRA) to induce CD38 upregulation, thereby ensuring sustained molecular target. ATRA treatment significantly increased CD38 expression in AML cell lines and primary LCs, regardless of baseline levels [163, 164]. Subsequently, a daratumumab-decorated polymersomal vincristine sulfate (DPV) was developed and combined with ATRA for AML treatment in a CD38-low animal model (Fig. 6B) [80]. While DPV monotherapy produced modest benefits, the combination therapy tripled the median survival time, demonstrating the synergistic therapeutic potential of this strategy.

Folate receptor

Human folate receptors (FR) mediate folic acid (FA) internalization [165, 166]. FR-β is expressed at low level in the placenta and hematopoietic cells [167–169]. Although it serves as a neutrophilic lineage marker, FR-β on granulocytes is incapable of binding FA due to post-translational modifications, whereas FR-β overexpressed on AML cells retains this ability [81, 170]. In addition to being a precursor for nucleic acid synthesis, FA acts as a cofactor in multiple metabolic reactions [167, 171, 172]. The high proliferative state of AML cells leads to FR-β overexpression to meet increased FA demand [173–175], making it an attractive target for drug delivery.

ATRA can upregulate FR-β expression in AML cells. Based on this mechanism, FA-modified, DOX-loaded liposomes (f-L-DOX) were developed for AML therapy [81]. f-L-DOX enhanced FR-specific DOX uptake, which was unaffected by serum folate levels, and combination with ATRA further increased cytotoxicity. These results demonstrate the scalability and clinical potential of f-L-DOX [82, 83]. This ATRA-mediated FR-β upregulation was also utilized by Wang’s team [84]. They designed an ATRA-loaded albumin nanoparticle system co-administered with FA-modified, siRNA-encapsulating albumin nanocarriers, wherein ATRA enhanced the targeting efficiency of siRNA carriers (Fig. 6C). Similarly, FA-modified liposomes were used for co-delivery of DNR and emetine (Eme) under methotrexate (MTX) promoted FR-β expression, thereby priming selective liposomal uptake [85]. Building on this strategy, lipid-bilayer-coated mesoporous silica nanoparticles (LB-MSNs) were designed to encapsulate paclitaxel (PTX) and tanshinone IIA [86]. The lipid coating effectively prevented premature drug release and reduced hemolysis by shielding the silica surface, which would interact with tetra-alkylammonium groups on erythrocyte membranes.

FLT3

FMS-like tyrosine kinase 3(FLT3) is a critical oncogene in AML, encoding a membrane-bound class III receptor tyrosine kinase [176, 177]. The internal tandem duplication (ITD) mutation within its juxtamembrane domain leads to ligand-independent activation, which associates with poor prognosis [178–180]. Although FLT3 mutations occur in approximately 30% of newly diagnosed AML cases [181, 182], aberrant FLT3 expression and activation are not restricted to FLT3-mutant AML. FLT3 overexpression has also been reported in FLT3 wild-type AML, where it undergoes phosphorylation and remains sensitive to FLT3 kinase inhibition [183, 184]. These findings suggest that FLT3 represents a relevant biological and therapeutic target, supporting the broader applicability of FLT3-targeted drug delivery strategies.

FLT3 has been widely explored for targeted drug delivery. One strategy use FLT3 ligand (FLT3L) as a targeting ligand conjugated to poly(amidoamine) (PAMAM) dendrimers [87]. These nanocarriers were loaded with miR-150, a tumor-suppressive microRNA downregulated by MLL-fusion proteins and MYC/LIN28 signaling that induces FLT3 overexpression. Restoring miR-150 expression produced strong anti-leukemic effects in vivo, and co-administration with a bromodomain inhibitor showed synergistic inhibition of FLT3-overexpressing AML.

Another targeting approach utilizes the scFv of anti-FLT3 antibodies. Park et al. fused anti-FLT3 scFv with an elastin-like polypeptide (A192), generating a stable and cost-effective nanoplatform efficiently produced in Escherichia coli (Fig. 6D) [88]. Their study also revealed that FLT3-ITD AML exhibited elevated CD99 expression. Based on this, dual-targeting nanoparticles, were developed by co-assembling CD99–A192 and FLT3–A192 fusion proteins, significantly enhancing therapeutic efficacy (Fig. 6E) [89].

CD117

CD117, also known as c-KIT, is a receptor tyrosine kinase encoded by the proto-oncogene KIT on chromosome 4q12 [185, 186]. It is expressed in a subset of hematopoietic stem cells (HSCs), where it regulates normal hematopoiesis with expression decreasing during differentiation [187–189]. Upon binding to its ligand, stem cell factor (SCF), CD117 activates downstream signaling via tyrosine phosphorylation, promoting cell proliferation, differentiation, migration, and survival [189–192]. CD117 represents a promising therapeutic target in AML, as it is upregulated in approximately 80% of AML patients [193, 194], correlating with poor prognosis [193, 195].

To exploit this target, an ADC was developed by linking the anti-CD117 mAbs (LMJ729) to the cytotoxic maytansinoid DM1 [90]. Researchers demonstrated that non-ligand-blocking ADCs showed significantly c-KIT degradation efficiency, particularly in the presence of SCF. This effect was consistent in both wild-type and mutant c-KIT cell lines, indicating potential therapeutic value for CD117-positive AML. However, LMJ729 was reported to induce hypersensitivity reactions via Fc receptor interactions. To address this, Kim’s team developed NN2101, a highly fucosylated and galactose-deficient mAb, which was subsequently conjugated to DM1 [91].

Considering the time, cost and complexity associated with humanized antibody production and subsequent drug conjugation, an alternative approach used a CD117-specific single-stranded DNA (ssDNA) aptamer as a drug carrier [92]. The resulting Apt-MTX conjugate achieved significantly higher intracellular drug concentrations than free MTX, enabling sub-toxic dosing while maintaining therapeutic efficacy and minimizing systemic toxicity (Fig. 6F).

CLL-1

C-type lectin-like molecule-1 (CLL-1), or C-type lectin domain family 12 member A (CLEC12A), is involved in the establishment of innate and adaptive immunity [196–198]. As a type II transmembrane glycoprotein, CLL-1 is expressed on LSCs of AML and normal myeloid cells, but it is absent in normal HSCs [199, 200]. This selective expression enables the distinction between LSCs and HSCs, making CLL-1 an attractive therapeutic target for AML.

Through phage display screening, peptides containing the LR (S/T) motif were identified for specific binding to CLL-1. Among them, CLL1-L1 was selected for further modification and conjugated to the surface of telodendrimer-based micelles (Fig. 6G) [93]. These micelles encapsulated DNR for targeted delivery to LSCs. Compared to anti-CLL-1 antibodies, CLL1-L1 peptide-functionalized micelles reduced off-target toxicity and minimized damage to normal hematopoietic cells. To enhance structural stability and prevent premature drug release, disulfide crosslinking was introduced into the micellar framework (Fig. 6H), resulting in improved therapeutic efficacy against AML and LSCs in a patient-derived xenograft (PDX) model [94].

CD34

CD34, encoded on chromosome 1q, is a 45-kDa type I glycoprotein selectively expressed on hematopoietic progenitor cells [201–203]. CD34⁺CD38^- cells have been identified as LSCs or leukemia initiation cells [204, 205]. Moreover, CD34 expression correlates with poor chemotherapy response and shorter remission duration [206, 207]. To target CD34, Carrion et al. conjugated the My10 mAb onto DOX-loaded liposomes, generating CD34-specific immunoliposomes [95]. These nanocarriers selectively bound to CD34 but were not internalized by LCs, leading to localized extracellular DOX release near LCs. Notably, non-CD34-expressing cells showed a higher IC₅₀ for CD34-targeted liposomes than for free DOX, indicating enhanced selectivity and reduced systemic toxicity.

CD99

CD99 is a 32-kDa type II transmembrane glycoprotein essential for cell migration and adhesion [208, 209]. Its overexpression has been reported in myelodysplastic syndromes (MDS) and AML [210–212], and it serves as a reliable marker for LSCs [213]. Targeting CD99 with mAbs has been reported to reduce leukemic burden in AML xenograft models [212, 214]. However, the complex production process and high cost of mAbs remain major obstacles. To overcome these limitations, scFvs have been explored as an alternative due to their comparable specificity and rapid production in Escherichia coli. To further enhance therapeutic efficacy, an anti-CD99 scFv was fused with elastin-like polypeptide (ELP) to form a nanoworm, which selectively targeted CD99 and exhibited a prolonged pharmacokinetic half-life [96].

CD96

CD96, also known as Tactile, is an immunoglobulin superfamily member expressed on T cells and NK cells, where it mediates cell adhesion during the late immune response [215–217]. Subsequent studies identified that CD96 is selectively expressed on CD34⁺CD38⁻ AML cells, suggesting its potential as a therapeutic target for eliminating LSCs [213]. A study reported that approximately 14.5% of AML patients harbored CD34⁺ CD38⁻ CD96⁺ LSCs [97]. Based on this, researchers integrated targeted nanomedicine with PDT to overcome its inherent limitations, such as photosensitizer toxicity, poor targeting specificity, and limited photon penetration [97]. They synthesized PEGylated calcium phosphosilicate nanoparticles (CPSNPs) functionalized with sulfo-NHS, which subsequently conjugated with anti-CD96 antibodies for selective targeting.

Strategies for BM-targeted drug delivery

BM is the primary hematopoietic niche where blood cells originate and mature. However, it also functions as a protective reservoir for LCs, particularly after chemotherapy, allowing minimal residual disease (MRD) to persist. The LCs within BM microenvironment often develop drug resistance through interactions with stromal components, extracellular matrix proteins, and cytokine signaling networks, leading to disease relapse [218–220]. Therefore, eliminating residual LCs in the BM is critical to improving patient outcomes.

To achieve active BM-targeted drug delivery, two primary approaches have been investigated. The first exploits the natural homing ability of HSCs to the BM. This involves functionalizing carriers with ligands that mimic the interactions between HSCs and the BM niche, or by coating NPs with HSC membranes to enhance BM accumulation. The second strategy targets the BM microenvironment itself, focusing on receptors or molecular markers that are aberrantly expressed in the BM of AML patients. By integrating these targeting strategies, BM-specific drug delivery platforms can improve therapeutic efficacy while minimizing off-target toxicity. Continued investigation of BM niche interactions and the development of innovative delivery platforms will be essential to overcoming resistance and preventing relapse. Nanocarriers designed for BM-targeted drug delivery are summarized in Table 2.

Table 2.

Nanocarriers designed for BM targeting in AML treatment

Target	Targeting tool	Cargo	Carrier	Stimuli		Animal model			Year	Ref
						Type	Cell	Site
HRs	HSCM	aPD-1	Biomimetic NPs		-	CDX	C1498-Luc	Disseminated	2018	[221]
HRs	AMLCM	DOX	Biomimetic NPs		-	CDX	C1498-Luc	Disseminated	2020	[222]
HRs	HSPCM	Ara-C	Biomimetic NPs		-	CDX(SRM)	Ka539/MLL-AF9-GFP transduced BM	Disseminated	2024	[223]
HRs	HSCM	aTIM-3	CR-TNG		Hypoxia	CDX	Luc-HL-60	Disseminated	2024	[224]
E-selectin	ESTA	NPs	PSP		-	-	-	-	2011	[225]
E-selectin	ESTA	PTL	MSV		-	PDX	Primary cells	Disseminated	2016	[226]
VLA-4	LDV	siRNA	LNPs		-	PDX	Primary cells	Disseminated	2023	[227]
VLA-4/CXCR4	VCAM-1/CXCL12	Cas9 RNP	MSCM-NFs		-	CDX	THP-1	Disseminated	2021	[228]
HARE	HA	SOR	Micelle			PDX	Primary cells	Disseminated	2022	[229]
HAP	ALN	Ara-C	Liposome		GSH	CDX	C1498	Disseminated	2022	[230]
TRAP	TBP	MVC	TBP-NP		-	CDX (SRM)	MLL-AF9-transduced LSK cells	Disseminated	2021	[231]

Abbreviations: HSCM, hematopoietic stem cell membrane; AMLCM, acute myeloid leukemia cell membrane; HSPCM, hematopoietic stem/progenitor cell membrane; MSV, bone marrow-targeted multistage vector; MVC, Maraviroc. Others are defined in the Abbreviations section

Homing receptors

Hematopoietic stem/progenitor cells (HSPCs) and AML cells possess an intrinsic ability to home to the BM through a multistep process involving adhesion, rolling, and nesting. This process is mediated by homing receptors (HRs) such as CXCR4, VLA-4, VLA-5, CD44, and LFA-1, as well as selectins expressed on HSPCs or AML cells [232–234]. CXCL12 secreted by stromal cells recruits CXCR4-expressing cells to the BM sinusoids, where E-selectin and P-selectin on endothelial interact with glycoprotein ligands on cells to mediate rolling [235, 236]. Subsequent firm adhesion to the vessel wall is mediated by interactions such as VLA-4/LFA-1 – ICAM-1, CD44 – HA, and VLA-5 – fibronectin [237–239]. Finally, under the CXCL12 gradient, activated adhesion receptors such as VLA-4 and CD44 facilitate transendothelial migration into the BM [240, 241].

By utilizing cell membranes as nanocarriers, NPs can mimic natural homing mechanisms while competitively binding adhesion molecules, thereby preventing LSCs from engaging these receptors. Hu et al. developed a biomimetic drug delivery system to transport PD-1 antibodies (aPD-1) to the BM [221]. In this system, HSC plasma membranes were covalently conjugated with aPD-1-decorated platelets through a click reaction, forming HSC-Platelet-aPD-1 (S-P-aPD-1) (Fig. 7A). S-P-aPD-1 was demonstrated enhanced BM accumulation, prolonged circulation time and reduced toxicity. Notably, it significantly improved T-cell-mediated anti-leukemia efficacy compared to the free aPD-1. However, unintended platelet activation during circulation remains a concern, as it may induce pro-inflammatory responses or systemic T-cell activation.

Fig. 7.

Schematic illustrations of diverse bone marrow (BM)-targeted nanocarriers for AML. (A) HSC membrane-platelet hybrid nanoparticles functionalized with aPD-1 for BM-targeted immune checkpoint blockade [221]. (B) Liquid nitrogen-treated (LNT) AML cell-derived carriers for BM-targeted doxorubicin delivery and AML vaccination [222]. (C) HSPC membrane-coated Ara-C-loaded liposomes enabling BM homing via adhesion molecules [223]. (D) Hypoxia-responsive albumin nanogels co-delivering aCD47 and R848, guided by TIM-3 targeting and BM accumulation [224]. (E) LDV peptide-modified lipid nanoparticles targeting VLA-4 receptor for BM-selective siRNA delivery [227]. (F) MSC membrane-coated nanofibrils loaded with CXCL12α and LNP-Cas9 for LSC-targeted CRISPR-Cas9 gene editing in the BM [228]. (G) HA-EGCG/SOR self-assembled micellar nanocomplex (sora-MNC) for BM-targeted delivery and selective uptake by CD44⁺ leukemic cells [229]. (H) ALN-HA-SS-AraC-Lip liposomes for hydroxyapatite-targeted BM delivery, CD44-mediated leukemic cell uptake, and redox-responsive Ara-C release [230]. (I) TRAP-binding peptide (TBP)-modified PSMA-b-PS micelles enable BM-targeted delivery of CCR1/5 inhibitors via TRAP-mediated endosteal accumulation [231]

A research team utilized liquid nitrogen-treated (LNT) AML cells as carriers for DOX delivery (Fig. 7B) [222]. LNT treatment eliminated leukemogenic potential while retaining cellular structure and BM-homing properties. In addition to serving as an efficient drug carrier that prolonged survival of AML mice, LNT cells also acted as a cancer vaccine, inducing protective immunity against AML when administered with an adjuvant.

Li et al. extracted membranes from HSPCs and fused them with Ara-C-loaded liposomes to generate Ara-C@HSPC-Lipo (Fig. 7C) [223]. Their study demonstrated that adhesion molecules such as CD44 and ITGB2 on HSPC-derived membranes conferred BM-targeting capability.

In another study, a hypoxia-responsive albumin nanogel was designed to enhance the efficacy of immune checkpoint blockade agents, aCD47 and resiquimod (R848) [224]. CD47/R848@HSA-TIM-3 (CR-TNG) was constructed by self-assembling human serum albumin (HSA), aCD47, R848, and the hypoxia-sensitive cross-linker (NHS-AZO-NHS), followed by covalent coating with aTIM-3. After intravenous administration, aTIM-3 specifically recognized TIM-3 on circulating AML cells, facilitating CR-TNG attachment. Exploiting the BM-homing property of AML cells, CR-TNGs preferentially accumulated in the BM and underwent cleavage in response to the hypoxic microenvironment (Fig. 7D).

E-selectin

As noted above, E-selectin (or CD62E) is specifically expressed on BM endothelial cells at tumor cell homing sites, making its ligand a promising target for BM-directed drug deliver [237, 240]. Building on this, Mann and colleagues identified an E-selectin-specific thioaptamer ligand (ESTA), and developed ESTA-functionalized porous silicon microparticles (ESTA-PSP) to modulate the biodistribution of payloads [225]. Porous silicon enables the loading of diverse cargoes, including proteins, drugs, and even nanoparticles. ESTA-PSP exhibited high nanoparticle-loading efficiency and significantly enhanced BM accumulation while reducing hepatic and splenic sequestration in vivo. Zong et al. further designed a multistage delivery system consisting of a PTL-loaded mPEG-PLA micellar core, encapsulated within an ESTA-modified porous silicon (pSi) shell [226].

Very late antigen-4 (VLA-4)

The VLA-4 is an adhesion receptor involved in HSCs homing and cells-matrix interactions. Although nanoparticles can pass through BM sinusoids during circulation, non-targeted LNPs exhibit insufficient retention to achieve therapeutic efficacy. To address this, VLA-4 targeted nanocarriers were designed to enable selective uptake by LSCs in the BM. The tripeptide sequence Leu-Asp-Val (LDV) in fibronectin, identified as the minimal motif required for VLA-4 binding, has been utilized to direct siRNA-loaded LNPs into the BM (Fig. 7E) [227].

Additionally, BM stromal cells secrete CXCL12 to support LSC survival and inhibit their mobilization into the bloodstream [241, 242]. Ho et al. utilized this property to develop a scaffold-mediated CRISPR-Cas9 delivery system which combined nanoparticle technology with gene-editing technology [228]. In this system, Cas9/single-guide RNA ribonucleoprotein targeting IL1RAP was encapsulated within LNPs (LNP-Cas9). Nanofibrils (NFs) were coated with mesenchymal stem cell membranes (MSCMs), onto which CXCL12α was subsequently loaded to enhance LSC recruitment. LNP-Cas9 was then electrostatically assembled onto the MSCM-coated NFs, forming the final delivery construct. After injection into the BM, CXCL12α and VCAM-1 on MSCM-NFs interacted with CXCR4 and VLA-4 on LSCs, promoting internalization of the gene-editing complex (Fig. 7F).

Hyaluronan receptor for endocytosis (HARE)

The HARE, distinct from CD44 and ICAM-1, is essential in systemic HA clearance [243, 244]. Intravenous administration of HA results in preferential accumulation in the BM [245, 246], suggesting its potential for BM-targeted drug delivery. Based on this, Bae et al. conjugated HA with the anti-leukemia agent epigallocatechin-3-O-gallate (EGCG) to form HA-EGCG. The resulting conjugate self-assembled with sorafenib (SOR) into a bone marrow-targeting micellar nanocomplex (Sora-MNC) (Fig. 7G) [229]. After reaching the BM, Sora-MNC selectively targeted leukemic blasts with high CD44 expression.

Hydroxyapatite (HAP)

Beyond HR-based strategies, the bone mineral HAP represents another effective target for BM-directed drug delivery. Alendronate (ALN), with high affinity for HAP, been widely applied as BM-targeting ligand in models of osteoporosis [247, 248], osteosarcoma [249, 250], CML [251], MDS [252], MM [253] and bone metastasis originated from breast cancer [254, 255] or prostate cancer [256, 257].

In AML, Wu et al. developed a multifunctional liposomal system (ALN-HA-SS-AraC-Lip) that demonstrated significant anti-leukemic efficacy [230]. ALN-HA was synthesized and conjugated to cholesterol via a bio-reducible disulfide (-SS-) linker. This design enabled BM targeting, sequential recognition of CD44⁺ leukemic cells, and redox-responsive drug release within the bone microenvironment (Fig. 7H).

Tartrate-resistant acid phosphatase (TRAP)

TRAP is a protein secreted by osteoclasts and deposited at endosteal bone resorption sites, making it a promising BM target. In leukemia, C-C chemokine ligand 3 (CCL3) plays a pivotal role in disease progression, but inhibitors of its receptors (CCR1 and CCR5) have shown limited efficacy in vivo due to poor BM accumulation [231]. To address this, poly (styrene-alt-maleic anhydride)-b-poly(styrene) (PSMA-b-PS) was used as the micellar skeleton due to its passive BM accumulation. To further enhance BM targeting, a TRAP-binding peptide (TBP) was conjugated onto the nanocarrier (Fig. 7I). TBP, a 13-amino acid oligopeptide, was identified through an M13 phage display library screening using TRAP as the target bait [258]. TBP has also been utilized to facilitate nanocarriers accumulation at bone fracture sites [259, 260], further demonstrating the confirm BM targeting ability.

Conclusion and future directions

From the early 2000s, nanotechnology has gradually evolved from basic research to applied science, promoting the application of NDDS in medical field. Since traditional chemotherapy remains the primary treatment for AML, NDDSs have emerged as great potential in improving the antitumor efficacy of free drugs. The predominant advantages applied in AML include [261, 262]: (1) Targeted drug delivery. Active targeting can be achieved by surface modification with specific ligands or using the ligands as carrier. Which endowers classical chemotherapeutic drugs target ability without chemical modification, avoiding non-specific distribution. (2) Prolonged circulation time. Drugs encapsulated in nanocarriers are shielded from premature enzymatic degradation in the bloodstream, thereby improving their stability and extending their half-life. (3) Controlled drug release. The nanocarrier materials can be engineered to respond to specific stimuli such as pH [26], hypoxia [224], laser [28], and GSH [65], which differ between healthy and tumor tissues. (4) Crossing biological barriers. The small size of nanoparticles allows them to penetrate blood-bone marrow barrier (BMB) [226]. (5) Multi-drug delivery. NDDSs can be engineered to co-deliver multiple drugs, which is particularly advantageous in AML, where combination therapy is always required.

Despite significant progress, all the studies are still in the preclinical stage. Challenges hinder the clinical translation of NDDSs include: (1) Toxicity and biocompatibility concerns. Some nanocarrier materials, particularly metals (such as copper, iron), silicon-based nanoparticles (such as SiO₂), and high-molecular-weight polymers (such as polystyrene, polylactic acid), are difficult to degrade and may accumulate in organs, causing systemic toxicity. (2) Impaired targeting efficiency. Some tumor-targeting receptors are also expressed on normal tissues, leading to off-target effects. (3) Manufacturing challenges. The complex fabrication process of active-targeted NDDSs presents scalability and quality control issues. (4) Complex approval process and supervision standards. NDDS are still a relatively new field. The complex structure and mechanisms of action need more toxicological and clinical trials to assess safety and efficacy. And the uniform international evaluation standards of preparation and characterization of NPs is deficient.

While these limitations present considerable obstacles, they do not overshadow the promising outcomes achieved in preclinical AML models. In response to the clinical translational obstacles mentioned above, several strategies can be proposed:

Future NDDS design should prioritize biodegradable and biomimetic materials. Inspired by the concept of CAR-T therapy, which involves extracting a patient’s own T cells, genetically modifying them, and reinfusing them back into the patient [263, 264], a similar strategy may be envisioned using AML cells. AML cells could be isolated from patients, and their membranes harnessed to coat anti-tumor drugs or drug-loaded nanoparticles. Since AML is a highly heterogeneous disease, with genetic, phenotypic, and microenvironmental differences among patients, which contribute to variable responses to standard chemotherapy regimens. These membrane-camouflaged nanoparticles could then be reinfused into the same patient, enabling a highly personalized drug delivery approach that leverages tumor-derived homing capabilities while minimizing immune response.

In addition, off-target effect remains a major obstacle in targeted therapy, particularly in the late treatment stage when tumor burden decreases and MRD shielding in the BM becoming hard to target. Therefore, targeting BM is crucial. Few studies have focused on this aspect, most of them are biomimetic NDDSs, demonstrating their great potential once again. Future research should focus on maintaining membrane function during extraction and preparation to ensure stable BM-targeting ability and batch-to-batch consistency. Meanwhile, it is also essential to maintain consistent and high receptor expression across all disease stages. When leukemic cells express significantly higher receptor levels than normal cells, NDDSs are more likely to accumulate selectively in LCs, reducing off-target effect. As presented in this review, combining NDDS with chemotherapeutic agents that upregulate receptor expression will enhance the specificity and efficacy of targeted NDDSs. Such as MTX upregulate FR expression [85], and ATRA increases FR [81, 84] and CD38 [163, 164]. This strategy shows clinical promise in reducing off-target toxicity and overcoming limitations caused by variable receptor expression, while expanding the therapeutic potential of these drugs.

Further, LSCs represent a rare subpopulation of leukemic cells with self-renewal capacity and resistance to conventional therapies [265, 266]. They mainly reside in protective BM niches, where the microenvironment provides anti-apoptotic and quiescence-maintaining signals. This shelter enables LSCs to evade immune surveillance and survive chemotherapy, leading to relapse [218, 267]. In addition, most chemotherapeutic agents act primarily on rapidly dividing cells, quiescent LSCs can escape cytotoxic damage. Consequently, NDDSs offer a promising strategy for eliminate LSCs by selectively targeting their specific surface markers. Targeted delivery strategies against CD34, CD123, CLL1, CD96, and CD99 have shown encouraging results, with CD123-targeted NDDSs being the most extensively studied. Considering that the BM niche supports LSCs survival and can be remodeled by LSCs, dual-targeting NDDSs that combine BM-homing mechanisms with LSC-specific ligands represent an appealing strategy. Incorporating stimuli-responsive components allows these systems to release drugs precisely in response to features of the LSC niche, such as hypoxia [268] and overexpression of matrix metalloproteinases (MMPs) [269]. This approach holds promise for improving the precision and efficacy of AML treatment in the future.

Finally, for clinical translation and large-scale production, NDDS design should be simplified. Complicated synthesis and preparation processes will impair batch consistency and increase cost. And ligands should be proven efficacy, and easy to synthesis or naturally available. Future research should focus on improving targeting precision, delivery efficiency, and in vivo validation of these systems in clinically relevant AML models. In summary, NDDSs represents not only a critical future trend but also a potential breakthrough in the treatment of AML.

Abbreviations

6-MP

6-Mercaptopurine

ADC

Antibody-drug conjugate

ALN

Alendronate

AML

Acute myeloid leukemia

aPD-1

PD-1 antibodies

Ara-C

Cytarabine

ASOs

Antisense oligodeoxynucleotides

ATO

Arsenic trioxide

ATRA

All-trans retinoic acid

BM

Bone marrow

CLL-1

C-type lectin-like molecule-1

Cur

Curcumin

CXCR4

C-X-C chemokine receptor 4

DNR

Daunorubicin

DOX

Doxorubicin

EDS

Edelfosine

Eme

Emetine

EPR

Enhanced permeation and retention

FA

Folic acid

FH

Feraheme

FLT3

FMS-like tyrosine kinase 3

FR

Folate receptors

GA

Gallic acid

GDH

Graphene quantum dots

GEM

Genetically engineered mouse

GSH

Glutathione

HA

Hyaluronic acid

HAP

Hydroxyapatite

HARE

Hyaluronan receptor for endocytosis

HDC

Histamine dihydrochloride

HMPB

Hollow mesoporous Prussian blue

HRs

Homing receptors

HSCs

Hematopoietic stem cells

HSPCs

Hematopoietic stem/progenitor cells

HSCT

Hematopoietic stem cell transplantation

HumF

Humanized Archaeoglobus ferritin

ICG

Indocyanine green

ITD

Internal tandem duplication

LC

Leukemia cells

LDV

Leu-Asp-Val

LNPs

Lipid nanoparticles

LONp

Lanthanide oxyfluoride nanoparticle

LPPs

Lipopolyplex nanoparticles

LSCs

Leukemia stem cells

Lut

Luteolin

mAbs

Monoclonal antibodies

MDS

Myelodysplastic syndromes

MM

Multiple myeloma

MRD

Minimal residual disease

MTX

Methotrexate

NCs

Nanoconjugates

NDDSs

Nano-drug delivery systems

NP

Nanoparticles

PB

Peripheral blood

PBNPs

Prussian blue nanoparticles

PDT

Photodynamic therapy

PDX

Patient-derived xenograft

PTX

Paclitaxel

PPNPs

Porous silicon microparticles

PSP

Porous silicon microparticles

PTL

Parthenolide

PTX

Paclitaxel

RNR

Ribonucleotide reductase

ROS

Reactive oxygen species

scFv

Single-chain variable fragment

siRNA

Small interference RNA

SOR

Sorafenib

TBP

TRAP-binding peptide

TDT

Targeted drug train

Tf

Transferrin

TRAIL

TNF-related apoptosis-inducing ligand

TRAP

Tartrate-resistant acid phosphatase

VEN

Venetoclax

VLA-4

Very late antigen-4

Author contributions

GYP and TYQ conceived the review idea. YW, DYW, ZY and YYX performed literature search and analysis. TYQ and LJX wrote the main manuscript. TYQ prepared the visualization. All authors have reviewed and approved the final version of the manuscript.

Funding

The authors acknowledge the institutional support provided by the Science and Technology Department of Sichuan Province (2023YFS0307).

Data availability

Not applicable.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.