BMSCs are a heterogeneous population within the stroma and make up ~20% of the cellular volume in the bone marrow (84). According to single-cell RNAseq of human bone marrow aspirates, BMSCs are predicted to be the main source of cellular signaling in the BMME (84). BMSCs span multiple cell subtypes based on expressions of markers such as Leptin receptors (Lepr), Adiponectin (Adipoq), NG2 (Cspg4), and CxCL12, and can differentiate into osteoblasts, adipocytes, and chondrocytes (85). BMSC regulates the hematopoietic compartment through several key niche factors such as CxCL12, Kitl (SCF), IL-7, angiopoietin-1 (Ang-1) (84, 85) ( Figure 1A ). Moreover, these cells can confer chemoresistance to AML cells as reported in vitro and in vivo through VLA-4/VCAM-1 signaling which activates the nuclear factor κB (NF-κB) (21) ( Figure 1B ). This mechanism is termed cell adhesion-mediated drug resistance (CAM-DR) and is also associated N-cadherin and E-selectin (70). Although the 2D culture of BMSCs with AML cells may involve these pathways, it lacks other adhesion-mediated interactions with niche cells and from the 3D ECM itself. These limitations contribute to the loss of ‘stem-ness’ of LSCs in conventional stroma cultures which can impact the translation of in vitro findings to in vivo and human studies. Additionally, BMSCs give rise to osteolineage cells that include osteoprogenitors, preosteoblasts, and mature osteoblasts (85–88). Osteolineage cells regulate HSPC homing, maturation, and function, and the loss of osteoblastic function is associated with the development of myelodysplasia and leukemia (2, 21).

Osteoblastic cells mineralize the collagen fiber network to synthesize new bone matrix and regulate electrolyte homeostasis between extracellular fluid and bones (30). Osteoblasts that line up the endosteum play a crucial role in bone remodeling and differentiation of osteoclast precursor cells into mature osteoclasts (30). When terminally differentiated, osteoblasts become osteocytes and account for up to 95% of the total cells in the bone (30). Osteoblasts in the endosteal niche promote HSC quiescence through the Notch pathway, Ang-1, and thrombopoietin (TPO) (70). CXCL12, osteopontin and SCF secreted by osteoblasts are also associated with the maintenance of HSPCs (79). HSPCs retention in the bone marrow by osteoblasts is associated with CXCL12 signaling (79). When G-CSF is reduced, the number of osteoblasts and CXCL12 expression decreases which results in the egress of HSPCs into the blood circulation (79). During AML, leukemic cells can disrupt BMSC differentiation into osteolineage cells ( Figure 1B ). Single-cell RNA sequencing (scRNA-seq) analysis showed that emerging AML cells impaired osteogenesis in BMSCs and reduced regulatory molecules that support hematopoiesis (85). Additionally, osteoblasts can also regulate the protection of leukemic cells against chemotherapy. Particularly, these cells can secrete osteopontin which binds to the αVβ3 receptor on AML cells and protects the cells from apoptosis (36, 57). Our group also showed that AML can functionally inhibit osteoblastic cell activity through CCL3 signaling as seen in the leukemic mice model, blast crisis chronic myelogenous leukemia (bcCML) (14) ( Figure 1B ).

BMSCs can also give rise to fibroblasts that support the hematopoietic microenvironment. A subset of fibroblasts in murine BM is shown to express Cxcl12 and Ang-1 which indicate a potential niche regulatory function similar to cancer associated-fibroblasts (CAFs) (85). In solid tumor microenvironment and B cell acute lymphoblastic leukemia (B-ALL), CAFs can support chemoresistance which is associated with poor prognosis (21). In addition, malignant CAFs may remodel the BMME and transform it into a more aggressive niche (89). In a retrospective study by Zhai et al. of 63 primary AML patient bone marrow biopsies, immunohistochemistry showed the abundance of reticular fibers in the bone marrow of AML patients which is associated with a higher frequency of relapse and mortality (90). In the bone marrow samples, CAFs widely expressed high levels of fibroblast-specific protein 1 (FSP-1), alpha smooth muscle actin (α-SMA), and fibroblast activation protein (FAP) (90). When the CAFs were cultured with leukemia cell lines, THP-1 or K562, leukemia cells were seen to be less sensitive to chemotherapy. Previously, In vitro experiments by Ryningen et al. demonstrated that coculturing human AML blast cells with fibroblasts resulted in increased blast cell proliferation, and reduced apoptosis mediated by dysregulated cytokine levels such as IL-8 (18). Although the mechanism underlying the role of CAFs in AML is still being explored, pieces of evidence in AML and other cancers showed that the presence of CAFs in the BMME is associated with bone marrow fibrosis, leukemia progression, and sensitivity to chemotherapies (89).

Bone marrow is highly vascularized allowing hematopoietic cells to leave or enter the circulation (2). Imaging studies show that HSPCs were found co-localizing with vascular structures within the bone marrow (71). In addition, perivascular endothelial cells were shown to be important for HSC activation during bone marrow stess (2, 13, 91). The types of endothelial cells present in the BMME include arteriolar endothelial cells (AECs) and sinusoid endothelial cells (SECs) that vary between marker expressions (2). Evidence has shown that E-selectin coupled with chemotherapy doubled the survival time of leukemic mice compared to receiving chemotherapy alone and thus acts as a potential BMME-targeted therapy to improve AML survival (21). Furthermore, evidence shows remodeling of the vasculature during AML indicated by increased permeability, increased nitric oxide (NO), reduced perfusion (2). In mouse and human AML, an increase in angiogenesis is seen however this result is contradicted by intravital imaging studies that showed a selective reduction of blood vessels near the endosteum (2). The reduction is seen due to the accumulation of pro-inflammatory and antiangiogenic cytokines secreted by AML cells in the BMME such as TNF-α and CXCL12 (21) ( Figure 1B ). When the vessel defects were rescued, the loss of healthy HSC number was prevented and blast cells were more sensitive to chemotherapy (2, 92). Similarly, when endothelial cell-derived NO is inhibited, vascular leakiness is rescued in AML xenograft experiments (93).

The ECM anchors HSC to the endosteal and vascular niches and allows for the appropriate hematopoietic signaling to occur in the BMME. Hijacking of this ECM feature leads AML cells to develop resistance to chemotherapy and sequestering of minimal residual disease in the BMME (94). For example, CD44 expression on LSC and blast cells which bind to hyaluronic acid, osteopontin, fibronectin, selectins, laminins, collagens, and matrix metalloproteinases (MMPs), is associated with increased chemotherapy resistance and cancer aggressiveness (95). CD44 is also implicated in the migration, proliferation, differentiation, and survival of HSPCs. In addition, it also regulates the homing of HSPCs in the bone marrow and the homing of LSCs to intra and extra-medullary niches (96). On the other hand, MMP9 is typically downregulated in AML patients, and its high expression is potentially linked to a better prognosis (94). Furthermore, the biomechanical properties of the ECM may also be altered during AML. An in vitro study comparing ECM deposited by BMSCs from different MDS subtypes and healthy controls found that MDS-derived ECM was thicker and more compliant compared to, measured by atomic force microscopy (97). The ECM content differed from the control samples, exhibiting higher levels of glycosaminoglycans (GAGs), sulfated GAGs, and hyaluronic acid (97).

Tissue engineering strategies often involve the use of hydrogels and complex 3D culture devices to provide a matrix structure and dynamic microenvironment, similar to native tissue. Considerations for designing the BMME in vitro must include selecting the appropriate biomaterials, emulating the biophysical and biochemical cues, and incorporating niche cells ( Figure 2 ). The selection of biomaterials is contingent upon achieving biocompatibility to sustain cell growth, differentiation, and function without inducing immune responses or toxicity reactions (85). Particularly in modeling the BMME, where multiple cell types are involved, considerations must be made for cell-matrix interactions that can influence cellular functions, such as proangiogenic or osteogenic responses in BMSCs, immune cell responses, and HSPC differentiation (57, 98). Mechanical properties of biomaterials like stiffness, elasticity, and viscoelasticity must be considered as they can influence cellular behavior and response (75). Additionally, the porosity and permeability of hydrogels must be tuned specifically to the tissue of interest to ensure appropriate nutrient and oxygen diffusion, waste removal, and cell-cell interactions. A recent study by Li et al. showed that larger pore size (80 um) and higher Young’s modulus of 70 kPA were favorable for the proliferation of HSPCs and maintenance of stemness phenotypes (99). The biomaterials chosen should maintain structural integrity and mechanical properties over extended culture periods to support long-term studies. Various synthetic and naturally derived biomaterials have been used in modeling the bone marrow stroma ( Table 1 ). Decisions made on the use of the desired materials revolve around cost-effectiveness, ease of synthesis, and biocompatibility. The biomaterials include synthetically derived polymers such as polyacrylamide (pAAm), polyurethane (PU), poly-L-lactic acid (PLLA), polyglycolic acid, polystyrene (PS), poly(ethylene glycol) (PEG), and hydroxyapatite. Incorporating cell-adhesive peptides, growth factors, or signaling molecules can also promote cell attachment, differentiation, and tissue organization, thus increasing biocompatibility. On the other hand, various naturally occurring materials have been used to mitigate the biocompatibility of synthetic matrices by using alginate, decellularized Wharton’s jelly matrix, Matrigel, and fibrinogen.

Earlier efforts in culturing AML cells include using synthetically derived-biocompatible hydrogels (24, 37, 100). The benefits of using these hydrogels include the vast tunability of the mechanical properties that can be tailored based on the tissue or disease of interest. For example, bioactive peptides can be incorporated into the hydrogel backbone such as peptides containing the Arg-Gly-Asp (RGD) integrin-binding motif. Dainiak et al., 2008 fabricated a 3D hydrogel composed of pAAm crosslinked with either agmatine-based RGD mimetic or bovine-derived collagen type I (Col1) via cryogelation (100). The resulting hydrogels are macroporous, having a pore diameter of 10–100 μm and very elastic. pAAm-based hydrogel is used as it is drainage-protected, highly tunable, and provides mechanical strength that can resemble tissue-like elasticity. KG-1a cell line (ATCC CCL-246.1), a promyeloblast macrophage cell line isolated from an AML patient, was cultured either in hydrogel containing RGD, hydrogel containing Col 1, or in a 2D culture plate. The authors found that KG-1a cells form aggregates in Col1 and RGD-incorporated pAAM hydrogel while the cells stay as single cells in control unmodified hydrogel. Chemoresistance was tested by treating the culture systems with Ara-C. The authors found that the KG-1a cell aggregates in 3D conditions were more resistant to Ara-C treatment compared to single cells in the unmodified hydrogel. This higher resistance, however, may be due to the protective effect of aggregates. Critically, the high porosity of the hydrogels caused the release of non-adherent KG-1a cells from the unmodified hydrogel into the culture with every media change or addition of treatments. Therefore, the comparison of Ara-C effects on KG-1a cells between the control unmodified hydrogels with RGD and Col1 is not conclusive.

The loss of cells due to the large pores of the hydrogel can be alleviated by fabricating hydrogels with smaller pore sizes. Nair et al. (2015) found that a 60:40 combination of PU and PLLA created a hydrogel with dual nano and microporous property via thermally induced phased separation (TIPS) technique (24). This formulation showed the highest fibronectin adsorption compared to other combinations and thus was used to culture KG-1a cells. VLA-4 integrin expressed on AML cells can bind to stromal fibronectin and the downstream effects confer CAM-DR to chemotherapy (103, 104). The authors found that KG-1a cells cultured in fibronectin-coated PU/PLLA have increased cell adhesion and that the original CD34⁺/CD33^- phenotype was maintained after 7-day culture as compared to a reduction of the phenotype in fibronectin-coated tissue culture plate control (24). CAM-DR was observed by the lack of KG-1a cell death in the hydrogel after treatment of Ara-C as compared to profound cell loss in 2D conditions. When treated with DNR, KG-1a cells in the hydrogel remained drug-resistant for a longer time as compared to controls (24). These findings lead the authors to investigate the mechanism that contributes to chemoresistance in KG-1a cells cultured in the 3D culture. It was shown that the upregulation of the anti-apoptotic Bcl2 and p27^kip1 , which can result in cell growth arrest, was associated with chemoresistance. Subsequently, the inhibition of Bcl2 via the inhibitor ABT737 increased cell death of KG-1a cells in the presence of Ara-C and DNR (24).

Natural hydrogels can offer superior biocompatibility with cells, which explains why several research groups prefer to explore the utilization of these hydrogels for culturing HSPC and AML cells in vitro. In a paper published by Shin et al. (2016), the researchers used alginate hydrogel ionically crosslinked with RGD peptide to culture leukemic cells with distinct genetic mutations; MOLM-14 (MLL-AF9), U-937 (without MLL-AF9), and K-562 (CML cells with BCR-ABL) in a 96-well plate format. The authors investigated the proliferation of leukemic cell lines in 3D hydrogels with a range of physiological tissue stiffness relevant to the hematopoietic system stiffness (Young’s modulus, E = 0.075 kPa ~ 3 kPa) (37). These cells express α₅β₁ which binds to the RGD sequence in fibronectin (37). The authors found that leukemia cells formed amorphous aggregates in viscous matrix while forming single large spheroids in solid matrices. Shin et al. (2016) revealed that the MOLM-14 shows increased proliferation responses across matrix stiffness whereas K562 cells show a stiffness-dependent decrease in cell number (37). The authors found that RGD and a softer matrix confer resistance of AML cell lines to multiple drugs. The study advanced the field by showing that cell-extrinsic factors can have different impacts depending on the AML cell used and that genetic mutations and physical environments can have varying relationships (37).

Li et al. (2018) explored the use of human-derived decellularized Wharton’s Jelly matrix processed from umbilical cords (23). The rationale for using this matrix is that it contains similar components to bone marrow ECM, such as collagen, fibronectin, hyaluronic acid, and sulfated proteoglycans. The study used three different cell lines: HL60; Kasumi-1; and MV4-11 and tested the responses of the AML cell lines to Ara-C treatment. The results showed that AML cells in the Wharton’s Jelly matrix had reduced proliferation compared to cells in 2D suspension culture, but there was no difference in apoptosis. Additionally, the cells in the 3D matrix had characteristics of LSCs and a significant population of cells that were resistant to chemotherapy. The study also found that the AML cells in the Wharton’s Jelly matrix had an increased ability to self-renew and were more resistant to the chemotherapy drug Ara-C (23). The study further corroborated the association between elevated N-cadherin expression in 3D cells and heightened chemotherapeutic resistance of AML cells.

In the study by Xu et al. (2019), the authors explored the response signatures of Ara-C in bone marrow mononuclear cells (BMMCs) from both AML patients and healthy patients encapsulated in Matrigel, a commercial hydrogel rich in laminin and collagen IV (38). The researchers first optimized the growth conditions to recapitulate the BMME using BMMCs from healthy patients. They found that BMMSCs were capable of multilineage differentiation, including both hematopoietic lineages and stromal cells. Next, the researchers characterized the ex vivo growth and sensitivity to Ara-C of cryopreserved BMMCs from 49 AML patients. RNA-sequencing analysis was performed to determine differential gene expression patterns between 3D and 2D culture plates. The data showed unique gene expression signatures and novel genetic mutations associated with sensitivity to Ara-C treatment. This publication demonstrates the ability to integrate genomic approaches with 3D cultures of human cells and shows the feasibility of using these platforms to study human cell interactions with chemotherapies.

An alternative natural hydrogel that is commonly used for cell culture is fibrinogen hydrogels. Fibrinogen is derived from the polymerization of fibrin in the presence of thrombin which takes place during blood clotting response. Alhallak et al. (2021) proposed a fibrinogen hydrogel that is incorporated with plasma aspirate from a human donor. The combination of plasma culture with fibrinogen gel is rationalized to promote the proliferation and drug resistance of AML cells (39). Plasma contains a plethora of cytokines, chemokines, and growth factors that mimic the in vivo environment, whereas synthetic hydrogels lack these essential components. In this research, the fibrinogen and plasma culture hydrogels were used to culture both leukemic cell lines (KO52 and NOMO-1) and primary AML cells. The results showed that the AML cell line had a much higher expansion rate at days 3 and 7 in hydrogels compared to 2D cultures. Additionally, primary AML cells cultured in 2D showed no expansion on day 3, but when cultured in 3D hydrogel, the cells expanded 4.5-fold. The researchers also investigated the drug resistance of AML cells to Alisertib (Ali), Ara-C, and Venetoclax. In 2D culture, the drugs induced significant AML cell killing, exceeding 50%. However, in 3D culture, the drugs induced mild to no killing of cells. These findings suggest the chemoresistance of AML cells due to the protective fibrinogen and plasma culture hydrogel similar to native BMME.

Velliou et al. (2015) explored the effect of environmental factors on the proliferation and metabolic evolution of AML cells in 3D and 2D. In this model, a similar hydrogel was used as described before by Nair et al. (24). In this model, K-562 cell line (ATCC CCL-243), lymphoblast cells isolated from the bone marrow of a 53-year-old patient with chronic myelogenous leukemia (CML), were cultured in a PU-based 3D hydrogel coated with Col1 (101). The authors investigated the effects of different levels of oxidative stress and glucose on the phenotype of K-562 in 3D culture. The PU-based hydrogel showed great maintenance of K-562 cells in vitro as compared to other hydrogels tested (110). In this study, varying glucose concentrations were tested which were: (1) optimal tissue culture media concentration (4.3 g/L); (2) maximum physiological glucose concentration (1.3 g/L); and (3) minimum physiological glucose concentration (0.6 g/L). The glucose concentrations were tested against hypoxic (5% O₂) and normoxic (20% O₂) conditions. Understanding the impact of these factors on 3D culture systems is essential for future drug testing, as the microenvironment may synergistically interact with environmental factors, influencing drug response (101). All in all, the authors found varying effects of oxidative stress and glucose levels between 2D and 3D cultures. Specifically, under hypoxia-induced oxidative stress, glucose levels had a more profound impact on the proliferation of K-562 cells in 3D compared to 2D. In 3D under hypoxia, the highest level of glucose induces higher cell proliferation meanwhile the lowest level of glucose suppresses the expansion of K-562 cells (101). In addition, cells cultured in 2D showed higher metabolic activity indicated by increased metabolite levels in media such as lactate and glutamate. The findings suggest that the mode of culture can have a significant impact on the responses towards oxidative stress and glucose levels.

Validation of clinical and physiological relevance of in vitro findings from MPS models are still a major obstacle. Currently, there is still a lack of conclusive evidence that 3D cell-based systems can accurately recapitulate disease pathophysiology (117). Sufficient quantitative and reproducible data must be provided to replace current 2D models already used in academic and industry laboratories (117). Indeed, the majority of 3D culture models discussed in this review have shown that when AML cells, whether derived from cell lines or primary AML cells, are cultured within hydrogel systems, they exhibit chemoresistance. While the findings from these studies validate a crucial phenotype within the leukemic BMME, there remain additional phenotypes that require validation to ensure the achievement of a comprehensive physiological recapitulation of the BMME. The majority of the articles reviewed in the previous section have been mainly focused on validating CAM-DR that is seen in vivo, one that conventional 2D cultures struggle to replicate. Particularly, the findings supported the idea that BMSCs, osteoblastic cells, endothelial cells, and matrix interactions can provide a protective niche for leukemic cells to evade chemotherapies. In addition, the goal of the 3D culture systems was focused on maintaining AML cell numbers and stemness in the presence of an artificial matrix or BMME niche components. AML drug research that uses established cell lines often lacks proper healthy cell controls such as healthy HSPCs (118). For drug discovery purposes, the drugs must show efficacy to AML cells and at the same time the absence of toxicity to the BMME cells and HSPCs (118).

More importantly, we believe that there needs to be an emphasis on the validation of key phenotypes of AML interactions with BMME components that were shown to promote disease progression. For example, we and others have shown that AML cells can modulate the BMME by functionally inhibiting osteoblastic cell activity and BMSC differentiation into mature osteoblasts (14, 92, 119–122). Leukemic cells can disrupt osteoblastic cell activity through secreted factors such as TPO and CCL3 (14, 21). As osteoblastic cells and BMSCs modulate HSPC quiescence, the modulation of their function negatively impacts the hematopoietic compartment in the bone marrow. In addition, in AML, the endosteal vascular niche is remodeled from anti-angiogenic and inflammatory cytokines secreted by leukemic cells such as TNF and CXCL12 (21). The downstream effects of these interactions include damage to BMSCs, endothelial cells, and osteoblasts which lead to impaired ability of the endosteal niche to support HSC maintenance (21). Findings such as these, which were shown in vivo experiments, must be validated in 3D culture systems to ensure that the system can mimic not only the AML maintenance but also the leukemic cell interactions with the BMME components in vitro. Validation of works performed in vitro and in vivo must be done to ensure a faithful recapitulation of the AML BMME. Findings from such experiments can provide evidence to drive the adoption of 3D culture systems in AML research.

The underlying genomic and molecular complexity of AML makes it challenging to find therapies that will benefit the majority of AML patients. As evident from previous studies, responses to hydrogel materials and chemotherapies in 3D culture can vary depending on the AML cell source (35, 39, 102, 109). These works suggest the need for exploring multiple cell types, either primary cells or cell lines to capture the heterogenous responses associated with cytogenetic abnormalities in AML. Multiple immortalized cell lines have been established throughout the years to study AML cell mechanisms, chemoresistance, and drug discovery which have provided valuable findings to the AML field. Skopek et al. (2023) provided a comprehensive review of the AML cell types commonly used in research (118). A summary of the cell lines used in the 3D culture systems is listed in Table 2 . MPS developers must carefully select AML cell lines that take into account the genetic mutations. For example, the NPM1 mutations are the most common genetic alteration in adult AML, but the occurrence of the mutation is only in about 30% of total cases (118, 125). For instance, researchers exploring phenotypes linked to NPM1 mutations may find it beneficial to select cell lines like OCI-AML3 to narrow down the focus of their 3D culture investigations.

The American Cancer Society estimated that there will be 62,770 new cases and 23,670 deaths of leukemia in 2024 based on the NCI SEER cancer statistics (126). Specifically in AML, the estimated new cases and deaths for males are higher as compared to females. In most invasive cancers, men are at higher risk of being diagnosed with cancer (41.6%) compared to women (39.6%). This is believed to be caused by greater exposure to carcinogenic environmental and lifestyle factors (126, 127). However, male predominance is still largely unexplained by risk factors and therefore the role of sex-related biological factors such as endogenous hormone exposure and immune function must be further investigated (126–128). Most AML cell lines were derived from male patients of white or Asian descent potentially due to the predominance of AML incidence in males ( Table 2 ). In B-cell chronic lymphocytic leukemia (CLL), in which the predominance is also in males, varying DNA methylation patterns between men and women may contribute to the sex-related difference in CLL risk (129). Although results in intrinsic sex differences are still limited, 3D culture systems could potentially serve as a tool to study cellular sex differences in the BMME and malignancy.

Although the role of race and ethnicity is not fully understood in cell culture, the potential implications should be considered especially in the context of preclinical drug testing. The lack of racial and ethnic diversity can potentially contribute to cancer health disparities. A review in lung cancer recently published found over 800 lung cancer cell lines came mostly from white and Asian patients although lung cancer disproportionately impacts Black individuals in the US (130). In addition, the authors highlighted that no cell lines were identified from other groups such as Hispanic/latin(x), American Indian/American Native, or Native Hawaiian or other Pacific Islander (130). The review paper highlighted the need to establish additional cell lines to ensure the representation of all population groups in critical pre-clinical research. In the context of AML, the disparities in clinical trial enrollment, cancer cell biobanks, and in-hospital death rates in AML, highlight the need to acknowledge the lack of representation in research and care settings, and preclinical studies (131). In terms of cytogenetic landscape, multiple findings have associated ancestry-associated differences that contribute to molecular features in AML (131). For example, studies have reported that Black AML patients have lower frequencies of NPM1 and WT1 mutations and higher frequencies of IDH1/2-mutation (131, 132). As the understanding of the influence of race and ethnicity emerges, the field should anticipate the integration of these factors into basic and preclinical research.

MPS and 3D culture platforms are advantageous in drug discovery as they can recapitulate the complex tissue microstructures while reducing the exhaustive requirements of animal models (49–54). The field is still at its early stages with new models developed with modified approaches such as composite hydrogels with varying stiffness and bioprinted HTS systems (109, 133). However, the adoption of MPS and 3D culture approaches in AML research remains limited. MPS is still an emerging field facing challenges with validation, standardization, and acceptance (134). Standardization will be necessary to ensure the adoptability of 3D culture systems into preclinical studies. One of the primary hurdles in embracing 3D culture techniques, particularly concerning hydrogel construction in AML research, stems from the need to adopt the material fabrication techniques and equipment associated. As evident in this review, researchers must also select from a wide range of biomaterials as there is no standard hydrogel. This is a significant challenge for basic science or clinical labs lacking backgrounds in tissue engineering and biomaterials. Most AML 3D culture systems detailed in literature are custom-built, leading to significant challenges in seamlessly transferring these systems to other research laboratories. In this regard, using a commercially available device may be beneficial; however, cost must be considered, as these devices are newer to the market and can be expensive. Consideration must also be given to the cost of animal models, including the development of new syngeneic or genetic models to study specific BMME and AML cell interactions. Additionally, there is a cost associated with untranslatable findings from 2D culture models to clinical trials.

So far, this review has covered various 3D models encompassing static hydrogels, perfusable hydrogels, and microfluidics-based bone marrow on-a-chip. These models have provided findings that showed recapitulation of chemotherapy resistance of AML cells from matrix and niche cell interactions, and establishment of functional niches that support HSPC maintenance. In addition, the models have also shown distinct responses dependent on AML cell sources, whether primary or cell lines, that may reflect patient-specific responses to chemotherapies and thus provide more relevance to human studies compared to 2D or animal studies. To facilitate the adoption of 3D culture systems into preclinical studies, developers perhaps should consider aligning their approach to a specific context of use (CoU). Specific CoUs for MPS platforms currently include toxicology pharmacokinetics (ADME), pharmacodynamics, efficacy, and drug safety. By defining the CoU or particular goals of the devices, developers can define the strengths and limitations of their platform to convince stakeholders and researchers to adopt their systems. Examples of CoU or goals that are ready to be investigated in the 3D culture of AML BMME include investigating: 1) gender disparities using different cell sources; 2) aging BMME and AML cell progression; 3) initiation of disease due to perturbation of BMME in human cells; 4) development of therapy-related AML (t-AML); and 5) polypharmacy effects in BMME.