The adult hematopoietic system arises from the hematopoietic stem cell (HSC) that resides in and is supported by the HSC niche within the bone marrow. The HSC niche is a complex, interconnected network of cellular and molecular components within the bone marrow microenvironment (BMME) (Frisch, 2019). A major challenge that hampers progress in the field of the BMME is the lack of a reliable model system that can recapitulate the human bone marrow. Murine models have been used extensively because they allow for in vivo studies and powerful genetic tools. However, mouse models often fail to recapitulate the human environment. Conventional 2D cultures are another common approach used to investigate the human BMME for the study of normal hematopoiesis and hematopoietic malignancies (Congrains et al., 2021). However, 2D cultures lack the ability to completely elucidate bone marrow pathophysiology and are typically limited to 2–3 different cell types. Overall, the inadequacies of conventional preclinical models contribute to a relatively poor understanding of the human BMME and the large attrition rates of Phase II and III of clinical trials (Arrowsmith and Miller, 2013).

In recent years, in vitro 3D models of various organ systems have been developed to facilitate basic biology research and drug development and screening. The development of these preclinical disease models is beneficial in understanding complex microenvironmental interactions with the potential implications of discovering novel treatment strategies (Duarte et al., 2018). The three-dimensional culture strategy is advantageous as it can provide an avenue to recapitulate the complex tissue microstructures while reducing the exhaustive requirements of animal models (Huh et al., 2011; Low and Tagle, 2017a; Low and Tagle, 2017b; Osaki et al., 2018; Kelm et al., 2019; Low et al., 2020). Therefore, a 3D microphysiological system will advance the field of BMME research by providing a viable platform for applications in mechanistic studies and drug discovery.

Several groups have shown the use of fibrin-based hydrogels in microfluidics platforms and the incorporation of a dynamic vasculature component were able to sustain HSCs in vitro (Chou et al., 2020; Nelson et al., 2021). Synthetically derived scaffolds such as PEG crosslinked with MMP-degradable peptide or tethered with adhesive ligands show promising evidence of HSC sensitivity towards culture dimension and flow conditions (Bray et al., 2017; Rödling et al., 2017). Although these developments provide valuable insights into establishing niche incorporated models, the outcomes of these models are limited to studying myelotoxicity towards drugs and radiation rather than insights into the mechanisms of HSC/niche interactions. Most of the current models lack major BMME components such as osteoblastic cells. Osteoblasts are important in establishing the endosteal HSC niche compartment, regulating bone remodeling and promoting the survival of lymphoid progenitors (Duarte et al., 2018). Our group and others have shown that leukemic cell interactions with the BMME, specifically through osteoblastic crosstalk, can drive leukemic progression and development (Frisch et al., 2012; Staversky et al., 2018; Ackun-Farmmer et al., 2021). Therefore, it is crucial to incorporate the osteoblastic components in addition to the vascular niche to generate a reliable 3D model of the BMME.

In this study, we generated a 3D model of the murine BMME with functional osteoblastic, endothelial, and stromal cell compartments to recapitulate the endosteal niche that is favorable for HSC survival. We used a commercially available microfluidics system, the Human Emulation System from Emulate Bio, and fibrin hydrogel scaffolding to culture HSC and performed competitive transplant assays to evaluate the self-renewal property of BMME-on-chip grown HSCs. The development of this preclinical model is valuable for the study of BMME pathologies, BMME-hematopoietic cells crosstalk, and identifying novel therapeutic targets.

Our goal was to generate a BMME-on-chip with niche components that are reliable, cost-effective, and easily adopted across research laboratories (Figure 1). The development of in-house microfluidics models is beneficial in generating complex architecture for in vitro systems; however, it can be time-consuming and limited to a specific laboratory. In this study, we utilized Emulate Chip S-1™ to provide the geometry and mechanical support to culture cells. The chip contains two channels separated by a semi-porous membrane which allows for the physical separation of cellular compartments. The chip is assembled with a housing system, Pod™, that contains input and output reservoirs for both channels and is connected to a module that automatically regulates fluid flow in each channel. The validated setup allows for efficient culturing of cells without the need for frequent media changes, troubleshooting flow modules, or optimizing microfluidics architecture.

We have cultured bone marrow stromal cells (BMSC) isolated from wild type C57BL/6 J mice into the apical channel of Chip S-1 to create a supportive osteoblastic layer (Figure 2). Upon 14 days of culture under a continuous flow rate of 30 μL/h with osteoblast differentiation media, robust differentiation and mineralization were observed in chips, as shown by positive alkaline phosphatase and Von Kossa staining and quantification (Figure 2B). On D14, we introduced mouse endothelial C166 cells on the basal surface of the membrane to generate a supportive vascular endothelial monolayer (Figure 1C). C166 cells completely occupied the membrane of the channel at D2 and surrounded the walls of the channel by D7, as indicated by positive nuclei staining (DAPI, Figure 2D). On D7, we found robust endothelium formation as shown by cobblestone morphology and tight junction protein expression, zonula occludens-1 (ZO-1) (Figure 2E). On D14, a cell-laden fibrin hydrogel was introduced in the apical compartment containing BMSC and whole bone marrow cells. The chip was then subjected to a culture regime of supplemented endothelial media in the basal channel at 30 ul/hr. The chip was grown for 7 days before harvesting on D21 for analysis (Figure 1B).

We expect that the BMME-on-chip can provide a conducive environment for marrow cell culture due to the multifaceted culture system. In particular, we are interested in investigating the fate of hematopoietic stem cell and progenitor cells (HSPC) within the marrow compartment after 7 days of chip culture. Therefore, on D21, we harvested cells from BMME-on-chip and control mice BM and stained them for lineage (Gr1, CD3e, B220, and Ter119) and HSPC (Sca1 and cKit) markers⁷ (Figure 3A). HSPC (LSK, Lin−Sca+cKit+) population was preserved in healthy BMME-on-chip after 7 days of culture at ∼0.3% of total cells. This is two-fold greater than levels to in vivo of ∼0.15% (Figure 3B). The data strongly suggest that BMME-on-chip is a reliable surrogate for HSPC maintenance in vitro.

HSPC are self-renewing and multipotent cells able to reconstitute the BM and differentiate into mature blood cells. Therefore, the standard functional analysis of HSPC is via a mouse BM transplantation model, whereby experimental HSPC are transplanted into myeloablated mice with the addition of competitor cells as an internal control (Figure 4). Sorted LSK cells from C57BL/6 J mice (CD45.2+) were grown in chips with CD45.2+ BMSCs for 14 days (Figure 4A). The whole content of each chip was mixed with whole BM cells from CD45.1+ PepBoy B6 mouse as a competitor and transplanted into a lethally irradiated CD45.1+ PepBoy B6 mice (Figures 4B–C). Equal numbers of CD45.2+ freshly isolated bone marrow cells were also mixed with CD45.1+ BM cells and transplanted into lethally irradiated CD45.1+ recipients as controls. We analyzed peripheral blood (PB) cells via flow cytometry at multiple timepoints after the transplant. We found that CD45.2+ chip LSK were able to successfully engraft in recipient mice at similar levels compared to control HSPC from whole bone marrow (Figure 5). In particular, chip LSK gave rise to ∼40–60% of total PB cells, which include myeloid populations, B lymphocytes, and T lymphocytes (Figure 5B). To confirm multilineage differentiation capacity, we compared the frequency of myeloid and lymphoid cells that were derived from CD45.2+ cells in primary recipients. CD45.2+ cells harvested from the BMME-on-chip generated myeloid cells at similar frequencies compared to cells harvested from freshly isolated bone marrow (C). At 28-weeks post-transplant, myeloid frequencies were at ∼20% in peripheral blood of both chip and mouse recipients (D), while lymphoid populations were at ∼80% (D-E). Thus, our findings suggest that chip-derived cells have the equivalent potential for generating myeloid and lymphoid cells to the mouse control group.

Bone marrow engraftment was analyzed by detecting chimerism of CD45+ HSPC populations at a very late timepoint, 28-weeks post-transplant. Flow cytometry was used to quantify chimerism in LSK population and lineage-committed myeloid progenitors (Lin−cKit+) and lymphoid precursors (Lin−Sca1+) (Figures 6A–E). Chimerism was observed in all three populations in the majority of the mice transplanted with HSPC from bone marrow and chip (Figures 6H–J). In particular, CD45.2+ donor cells were observed at a high percentage in LSK populations of control recipients M1, M4, and M5 and in chip recipients C1, C4, and C5 (Figure 6H). All of the five control recipients showed engraftment of CD45.2+ donor cells, with M2 and M3 having a majority of the repopulated LSK derived from donor cells. On the other hand, two out of the five chip recipients showed low to no repopulation of LSK by donor cells. Myeloid and lymphoid committed cells also showed chimerism as indicated by CD45.2+ cells in control and chip recipients (Figures 6I and J).

HSPC subpopulations were further analyzed within the LSK cells in recipient mice with high chimerism (M1, M4, M5, C1, C4, and C5) (Figure 7). Our findings showed CD45.2+ repopulation of myeloid-biased (MPP2 and MPP3) and lymphoid-biased (MPP4) multipotent progenitor populations in chip recipients at similar levels to control recipients (Figures 7B–D and G–I). In addition, CD45.2+ short-term and long-term repopulating HSC (ST-HSC) and (LT-HSC) were also observed at high levels, 40% for both groups, in chip recipients. Overall, the majority of mice transplanted with HSPCs obtained from BMME-on-chip cultures maintained similar distribution of HSPC subpopulations even through this very long-term engraftment.

To investigate long-term HSPC function, a secondary transplant assay was performed by injecting bone marrow cells isolated from primary recipient mice into myeloablated secondary recipients (Figure 8A). PB was analyzed at 4- and 8-weeks post-transplant. Consistent with engraftment data in the bone marrow, only C1, C4, and C5 samples from the chip group showed engraftment of CD45.2+ cells at 4- and 8-weeks (Figures 8B and C). Overall, CD45.2+ chip-derived cells comprised approximately 20% of total live cells at 4- and 8-weeks post-transplant (Figure 8D). Although the mouse group showed higher trends of engraftment, the comparison showed no statistical significance between the groups. Importantly, long-term differentiation of HSPC into mature blood populations was maintained as indicated by monocyte and lymphocyte populations that were CD45.2+ post-transplant (Figures 8E and F).

Our findings currently showed that chip grown HSPC are self-renewing and capable of regenerating the hematopoietic system, which current 3D systems have not been able to demonstrate. This reliability of the chip to provide a permissive microenvironment for HSC maintenance makes it a valuable platform for future work in studying HSPC interactions with the BMME and BM pathologies. Here, we present step by step methodologies to generate BMME-on-chip, including protocols for Emulate Chip preparation, cell culture, MSC isolation, chip characterization, and competitive transplant assay.

We anticipate that the modular nature of this BMME-on-chip will allow for rigorous analysis of individual cellular components of the bone marrow microenvironment in a model that more closely recapitulates the in situ HSC niche. The modular nature of this model and the ability to use different cell populations from unique sources and genetic backgrounds is a powerful tool. We anticipate that this system will be used to clarify the roles of individual cell populations of the BMME during hematopoietic homeostasis, as well as during aging, and in pathological states such as leukemia and multiple bone marrow failure syndromes.

The described murine BMME-on-chip provides validation that an in vitro system is capable of maintaining multi-potent and self-renewing HSCs for at least 14 days. Competitive transplantation of HSCs into an immune competent recipient is the gold standard for quantifying HSC function. As this analysis is not possible with the use of human cells, this represents a strength of the murine system in validation prior to the use of a fully human BMME-on-chip. However, it also represents a limitation as mouse models do not always recapitulate the human environment. Previously published BMME-on-chip devices that use human tissue have demonstrated the ability to maintain phenotypic HSPCs (Chou et al., 2020). Therefore, it is likely that the functional ability to self-renew and repopulate the hematopoietic system of a myeloablated recipient that we report here in the murine model is translatable to a human environment.

While our murine BMME-on-chip includes multiple key components of the bone marrow HSC niche, it is not fully representative of the physiologic bone marrow. For example, osteoclastic bone resorption is an ongoing process in vivo and is a component not included in our system. In addition, osteocytes embedded in bone are the most abundant bone cell in vivo and are not present in our BMME-on-chip (Frisch, 2019). The relative importance of these cells and processes for the regulation of HSCs are not well-established, and thus the implication of their absence in our system is not known (Soto et al., 2021). However, the ability of our BMME-on-chip to maintain a fully functional HSC population suggests that they are not required for HSC maintenance.

In leukemia such as acute myeloid leukemia (AML) and chronic myeloid leukemia (CML), leukemic cells can interact with the surrounding stroma to create a favorable environment for leukemia stem cell LSC maintenance and expansion of leukemic cells (Frisch et al., 2012; Shafat et al., 2017; Behrmann et al., 2018; Staversky et al., 2018; Xu et al., 2021). The BMME is made up of cellular and noncellular components such as BMSC, endothelial cells, osteoblasts, osteoclasts, fibroblasts, adipocytes, macrophages, and extracellular matrix that provide crucial signals for the homeostatic regulation of hematopoietic systems. Leukemia has been shown to interact with the BM niche by modulating endothelial activation, establishing crosstalk with fibroblasts, and inhibiting osteoblasts (Ryningen et al., 2005; Frisch et al., 2012; Pezeshkian et al., 2013; Krevvata et al., 2014; Shafat et al., 2017; Staversky et al., 2018). In particular, our group and others have shown that leukemic cell interaction with osteoblasts is an important component in sustaining leukemic burden. However, the interactions of the HSC niche with bone marrow pathologies are poorly understood (Soto et al., 2021). AML is the most common form of leukemia that arises in the bone marrow (Soto et al., 2021). The standard therapy for AML of combined daunorubicin and cytarabine administration, termed 7 + 3 induction therapy, which was first employed in the 1970s, is still the predominant upfront therapy for most AML patients and gives rise to poor outcomes with an overall 5-years survival rate of <30% (Roussel et al., 2020; Kantarjian et al., 2021). Therefore, a better understanding of the role of the BMME in AML is critical for the improvement in therapy and patient outcomes.

The use of a BMME-on-chip that recapitulates many of the cellular interactions that occur in the BMME in vivo will allow for rigorous analysis of the BMME during hematologic malignancies. A murine environment, like the one in the BMME-on-chip described herein, will validate that phenotypes identified in vivo murine leukemic models can be recreated in an in vitro environment, and suggest that the in vitro environment of a similar human BMME-on-chip will recreate human bone marrow phenotypes that are not possible to directly observe in vivo. Therefore, this BMME-on-chip microphysiological platform will serve as a critical pre-clinical model for the identification of novel therapeutic targets.