Fabrication materials.

The selection of the most appropriate materials for each application of OoC depends on some factors that shall be taken into account in BM mimicking devices.

The first one concerns the biocompatibility of materials, needed to guarantee an adequate cellular response in a specific application [72].

The second one is the maintenance of the long-term culture, which is vital in the studies of hematopoietic processes [73] and biological processes associated with mesenchymal cells (e. g. differentiation and ECM secretion).

Next factor is gas exchange governed by permeability. Standard cell cultures require stable and well-defined O₂ and CO₂ concentrations; any variations in these parameters can be harmful to the cells.

One more factor concerns chemical products that may form at the interface pharmaceutical product/material. These chemical products can be converted into signals that influence the cellular response to a given clinical treatment [74]. This is important, for instance, when BM chips are used in drug testing.

And one more factor is related to the optical properties of the material. The imaging techniques that are usually used for data collection (e.g., fluorescence microscopy, confocal microscopy, phase-contrast microscopy) require the use of transparent OoC devices that can be monitored without disassembling.

The present review showed that PDMS, PMMA, silicon, PC, and silicone were the main materials used to create BM channels in OoC devices. Polymeric materials were used more often, suggesting that they have the properties required for in BM mimicking OoC devices [75]. Compared to brittle glass and silicon materials that require non-trivial sealing protocols, thus making the fabrication process expensive and complex, polymers offer cost-effective manufacturing and allow creating high- quality low-cost products.

The review has shown that PDMS was the most commonly used polymer material for the construction of BM channels in OoC devices. PDMS is a biocompatible material that can be sterilized in an autoclave and is easy to use in microengineering, explaining its widespread use in the production of OoC devices [76]. Furthermore, PDMS has remarkable elasticity, low production costs [77, 78], and high oxygen permeability, which allows using it for long-term cultures. The translucency of the material facilitates the use of imaging technologies, e.g., confocal and fluorescence microscopy for the biological characterization of the environment [79].

In the articles that were analyzed in this review, PDMS was used in several applications for in vitro mimicking of BM, including a) assessment of normal and abnormal behavior of BM cells in drug presence [50, 52, 54, 55, 57, 60, 63, 68–70], and b) creation of long-term cultures (more than 20 days) [53, 58, 64, 69, 70].

At the same time, even though it is the first-choice material in the analyzed articles, some characteristics of the PDMS are disadvantageous for its use in BM mimicking OoC devices. For example a) the absorption of the medium culture and drug components; b) formation of bubbles; and c) distortion of the device’s geometry [77, 80].

Recently, efforts have been made to identify and test new materials able to replace or modify the PDMS for use in BM OoC devices [81–85]. To overcome the disadvantages mentioned above some authors proposed the use of coating methods and design adequation [84, 85].

Among the alternative materials identified in this systematic review, PMMA appears to be a good option and is cited in three publications. PMMA is the acrylate-family polymer known for fracture resistance, translucency, and elasticity [86], as well as high chemical resistance [87]. In addition, PMMA is a low-cost and easily manufactured material [88]; it is also less hydrophobic than PDMS or other polymers used in microfluidic devices, thus ensuring a lower absorption of hydrophobic molecules crucial characteristics for the preservation of cell culture.

Fabrication techniques.

The choice of fabrication methods is based on material properties. For the devices made of PDMS as the main material, the conventional soft-lithography was used [51–53, 55–58, 60, 62–64, 66, 67, 69, 70]. Soft-lithography is a technique based on conventional photolithography for fabrication or replication of structures made of elastomers. It is generally used to produce micro- and nanostructures [78, 89] and its low cost and versatility makes it the most common technique in the fabrication of microfluidics devices, since it allows building different planar and non-planar surfaces and 2D and 3D structures with a resolution of up to 30 nm and a minimum size of 10 to 100 nm [90, 91].

The main drawback of this technique is that it is time-consuming and requires highly trained and specialized personnel for the operation of the equipment [52].

Besides soft-lithography, two other fabrication techniques were used a) laser cutting [61, 68]; and b) injection molding [54, 59, 65]. Process automation makes the laser cutting and injection molding relatively simpler fabrication methods for the operators compared to soft-lithography. However, fabrication of the master mold is very expensive because of the high pressures and temperatures involved. The most commonly used polymers are the most rigid ones, such as PC and PMMA.

Active flow mechanisms.

The syringe and peristaltic pumps were the most used active flow mechanisms in the revised publications. These mechanisms comprise systems containing a pump, a culture medium reservoir, and the connecting pipes between the device and reservoir. There were identified some differences between these systems, e.g., syringe pumps produce unidirectional flow, not allowing fluid recirculation, while peristaltic pumps allow recirculation. Because of the recirculation of the culture medium, peristaltic pumps can better mimic the physiology of the cellular microenvironment and, hence, are more suitable for the studies of the pharmacokinetics and pharmacodynamics of a drug in the human body [50, 100].

Though the pumps are the most used approach for fluid flow induction, they have some disadvantages. One problem related to the use of peristaltic or syringe pumps or tube-connected devices is that these systems can generate the so-called “air bubbles”, which are harmful to cells and must be avoided to maintain the integrity of experiment [52, 101]. As reported in some studies included in this review [50, 66], to overcome this drawback and prevent the bubbles, a “bubble trap” is usually attached to the system. The “bubble traps” require the use of a vacuum system, hydrophobic-coated surfaces, membranes, and pressurized fluids [102]. The “trap” is based on the floating of air bubbles associated with their low density. The bubbles are collected in a compartment that separates them from the fluid, allowing regular flow and preventing the cells from damage [101].

Passive flow mechanism.

Taking the difficulties presented by active flow mechanisms into account, some studies included in the review used passive flow mechanisms, such as gravity-driven flow [52, 61, 65, 68]. This mechanism involves the use of fluid reservoirs located at different heights. In this case, gravity generates a flow proportional to the elevation and the density of the fluid [100]. This low-cost method prevents the formation of bubbles, is simple in operation, and generates flows with physiological shear stress on the cells [61]. Though having many advantages, the renewal of the culture medium in gravity-generated flow systems can be a problem. This can lead to recirculation of the culture medium with fewer substances and more waste from cell activity, which may lead to cell death and prevents long-term experiments [50].

It is noteworthy that through the development of OoC devices capable of growing cell culture under perfusion of culture medium, several studies were able to generate the dynamic BM microenvironment in vitro. Several in vivo characteristics of the BM were mimicked using different flow generation techniques: cellular stimuli created through physiological shear stress generated by the fluid [56, 58]; regulation of BM physiological parameters through fluid perfusion [55]; in vitro generation of flow within the medulla with physiological interstitial velocities [57, 66]; and creation of a physiological vascular system, including the supply of oxygen and nutrients and the removal of metabolic products [70]. This indicates that perfused cell culture in vitro can generate more realistic environments for cell culture, thus improving understanding of specific phenomena that occur in the BM.

BM microenvironment.

Based on the different studies included in this review, three models representing the BM were identified: hydrogel/gel; scaffolds; and ECM-secreted (Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Materials used for mimicking the bone marrow.

https://doi.org/10.1371/journal.pone.0243840.t002

The first approach used to mimic the BM microenvironment was the hydrogel model. A hydrogel can be defined as a solid lacking a macroporous design [70]. The hydrogel is a polymerizable material that permits forming transparent 3D structures. In the ECM formation, the polymerization process of the 3D network influences the interactions between the cells and the matrix in this microenvironment [108]. That is, the cells interact with the cellular microenvironment similarly to their in vivo counterparts. This interaction influences the cell's phenotype and the factors they secrete in ECM [109]. In addition, the transparency of the material permits obtaining high-resolution images of the culture [62].

Chemical and mechanical properties of hydrogels depend on the type of hydrogel used in cell cultures [110]. Type I collagen, alginate, Matrigel [56], hyaluronic acid [67], fibronectin [69], and fibrin [51] are some of the protein types and ECM components that compose natural hydrogels [8]. The use of natural hydrogels provides a suitable microenvironment for the cultivation of BM cells, keeping their functions similar to those in vivo [57]. However, different production protocols of these components lead to variation in the compositions, making it difficult to predict and standardize the effects of the hydrogel matrix on cell behavior [110]. In the gel/hydrogel models, the most used components mimicking the BM environment were a) type I collagen-based hydrogel [54, 55, 57, 60, 62, 65, 69, 70] and b) alginate-based hydrogel [50, 52, 56, 68]

Type I collagen is fibrillar collagen. Probably it is the most studied and abundant collagen type. Its Young’s modulus depends on a density. Different studies have reported the production of 3D collagen structures with Young’s modulus values around 10 Pa (0.5 mg/ml) [111], 300 Pa (2 mg/ml) [111, 112], and 1200 Pa (4 mg/ml) [111]. Type I collagen with different densities can be used to mimic BM niches with different stiffness because its mechanical properties and their control methods are well known. Nevertheless, the low mechanical strength of collagen gels results in their significant contraction caused by the cells, and this is an important drawback of collagen use in systems without geometric constraints [113].

On the other hand, alginate hydrogel Young’s modulus ranges from 100 Pa (relatively elastic) to 10 kPa (moderately stiff) [114, 115] making it more suitable for mimicking the BM because BM Young’s modulus ranges from 0.1 to 1 kPa [1]. For example, in cultures of non-adherent abnormal cells, such as myeloblastic lineage cells, alginate provides a more realistic environment than Matrigel, retaining leukemic cells and improving their response [50]. In addition, when used as an ECM, alginate hydrogel provides greater cell growth compared to standard cell culture protocols [1].

Another way of representing ECM is by structures referred to as scaffolds; the scaffolds can be made of hydroxyapatite, polystyrene, and bony structure (Table 2). Scaffolds are well-defined macroporous 3D structures with properties similar to those found in vivo [116]. Scaffolds are characterized by their geometry, mechanical properties, and composition. The pores, the size, architecture, fabrication methods, and used material govern the properties of the scaffold [117]. In the reviewed publications, the scaffolds were made of synthetic materials [64, 66] or bony structures [63, 66].

Scaffolds can mimic trabecular bone, especially its structural and mechanical properties, allowing, therefore, to represent a spongy bone and BM niches [64]. This permits modelling the endosteal niche located near the trabecular bone that regulates the behavior of human hematopoietic cells [6].

Unlike the hydrogel-formed structure, the use of the scaffolds has some technical drawbacks. The materials commonly used to create scaffolds, such as hydroxyapatite, polypropylene, and polystyrene, may not be transparent [63, 64, 66]. Since a large part of the analysis techniques is based on light microscopy, opacity makes it difficult or impossible to view and record images, affecting the assessment of the device's effectiveness.

Finally, in addition to studies pointing out the hydrogel/gel and scaffolds, some studies have used living cells to produce the ECM by cultivating the cells on the surface of the device itself, creating specific niches of BM microenvironment [53, 58]. Using niche cells to generate the microenvironment through the secretion of ECM, growth factors and cytokines produce a matrix with chemical and mechanical properties that maintains the cellular behavior similar to in vivo.

For cell cultures on the device surfaces [56, 59, 69, 70], specific coatings were used to mimic the surface properties of the cellular environment of organs and tissues. The coating improves the cell adhesion and generates a phenotypic representation as expected in vivo, ensuring a realistic cell behavior in a given environment [118].

BM cell types and cellular culture.

A wide variety of cell types was used to constitute the bone marrow microenvironment. Mesenchymal stem cells, osteoblasts, and endothelial cells are some of the reported cell types, in addition to hematopoietic [59, 65, 67, 69, 70] and leukemic cells [50, 52, 57, 61, 62, 65, 67, 68]. Other studied cell types are given in Table 1.

The studies have evaluated both co-cultures and single-cell cultures. As BM is a very complex biological organ, it is essential to understand the functioning of its regions to choose the best cell culture approach [6]. The culture of one cell type provides a more manageable environment for in vitro testing, making it possible to analyze the collective behavior of specific cell types under well-defined physiological conditions. However, the in vivo cell environment is composed of several cell types and the cell-cell and cell-ECM interaction are determinant for cellular behavior. For example, cells that are present in the endosteal, central, and perivascular niches play key roles in stem cell behavior, regulating differentiation into red and white blood cells that will migrate into the bloodstream [119, 120]. A simulation of this complex environment in vitro is challenging for researchers, as regions with different mechanical and biological properties need to be reproduced in the same device [3], requiring the adjustment of the culture medium, oxygen supply, fluid flow, and mechanical properties in a complex microfluidic system. Nevertheless, most of the publications in the review use the co-culture approach to represent the cellular microenvironment of BM. This suggests the evolution of technologies and methods of cell culture that allow implementing this model with the in vitro systems. However, some screened publications used single-cell cultures to represent in a simplified way a phenomenon that in reality occurs in a complex BM environment.

Though BM environment is a dynamic system governed by cell-cell and cell-ECM interactions, integrating the ECM and cells properties in the OoC devices analyzed in this systematic review allowed detecting various cellular behaviors and phenomena associated with BM: the regulation of HSP and HPC cells in the hematopoietic microenvironment by interaction with niche cells, such as osteoblasts, perivascular endothelial, and perivascular stromal cells [55]; the preservation of MM cells in a 3D ossified tissue constructed by a human osteoblast cell line (endosteal niche) [53, 58]; the effect of chemotherapeutic drugs on abnormal cells in a 3D environment with co-cultivation of BM cells [57, 63]; the hematopoiesis process in a 3D BM model based on collagen hydrogel and bone powder [60]; the long-term culture of HSPC in co-culture with MSC in a 3D scaffold that mimics trabecular bone [64]; the regulation of normal and abnormal HSPC behavior in a 3D multi-niche system [67]; and the mimicking of hematopoiesis processes in a 3D matrix in co-culture with MSC and HUVEC [70].

Static cultures of multiple cell types.

The first approach, static cultures of multiple cell types, was based in a 3D co-culture in vitro system without the presence of fluid flow. BM mimicking using this approach allowed reproducing the phenomena of vessel formation in the healthy BM [51] or the vessel formation induced by leukemic cell lines [62]. It is known that the presence of blood vessels in BM physiology is vital for the supply of nutrients, oxygen, and growth factors necessary for niche cell survival and regulation of hematopoietic cell behavior [5].

The region where these vessels are located in a healthy BM is called the perivascular region, or perivascular niche of the BM. With an objective to mimic this niche specifically, publications considered endothelial and stromal cells as basic components, since these are the cell types present in the native composition of the perivascular niche of the BM [8].

The simulated behavior of cell migration [51, 62] and adhesion [51], vasculogenesis [51], and angiogenesis [62] of niche cells were reported to be similar to in vivo; in addition, there was also the formation of vessels through the association between multicellular aggregates and fibroblasts sprouting in a primitive capillary plexus as it is detected in the microenvironment of BM.

Besides the study of cell behavior around the vascular region, the 3D co-culture approach also provides useful tools for understanding the response of tumor cells to the action of drugs. In a 3D matrix based on type I collagen, an experiment carried out to evaluate the chemotherapeutic effects on plasma cell myeloma reported a reduced action of drugs when the abnormal cells were cultured with bone stromal cells [54], resulting in increased viability (live/dead) of the cells. This phenomenon occurs because of a mechanism known as cell adhesion–mediated drug resistance caused by a) the direct MM–BM stroma cell adhesion; b) by MM-ECM adhesion; or c) by the secretion of a soluble factor [54]. Cell adhesion-mediated drug resistance is related to the increase of drug resistance mediators and avoiding the apoptosis process [122].

Perfused cultures of single cell types.

Using the second approach, perfused cultures of single cell types, which is based on a perfused 3D system without co-culture, publications reported better results for toxicity response and cellular proliferation than those reported in conventional 2D control groups. Evaluating viability (toxicity and proliferation), myeloblastic cell lines showed greater sensitivity to drug exposure [50, 52] and megakaryoblastic cell lines were maintained in long-term culture [61].

The two previously mentioned approaches showed that a simplified BM mimicking with the in vitro system was able to represent the phenomenon that occurs in vivo. In other words, a system that uses the minimum characteristics essential to mimic a specific phenomenon can replicate it in vitro. The studies that use a 3D co-culture [51, 54, 62] or perfused 3D single-cell culture [50, 52, 61] consider the influences of two specific features of microenvironment on cell’s response, besides the ECM interactions, namely: a) cell-cell interactions; and b) fluid flow action.

Cell-cell and cell-ECM interactions play a key role in regulating cell functions and the microenvironment. These interactions occur through the so-called cell junctions and involve mediator proteins including cadherins and integrins that sense and respond to the external signals of the microenvironment [123].

The dynamic flow of the physiological fluid supplies nutrients, oxygen, and growth factors, and removes metabolic products from the cells. In addition, cellular behavior is influenced by flow-generated shear, which promotes a biological response of cells according to the mechanisms involved in cellular mechanotransduction [124, 125].

It is important to point out that the two approaches discussed so far have considered isolating the co-cultured cells in a static system or single-cell culture in a perfused system. Despite the important achievement in obtaining similar biological results as in vivo, the individual evaluation of these mechanisms is a simplified representation of the BM microenvironment and does not consider the dynamics of interactions and stimuli.

Perfused cultures of multiple cell types.

The third group of studies that attempted to overcome these issues and took into account all the three basics conditions for representation of BM environment in vitro was clustered: a 3D environment, formed by the proper matrix; a co-culture with specific bone marrow cell; and a perfused culture medium, as a source of oxygen, nutrients, and shear stress stimulation. In this systematic review, this group of studies was further subcategorized by specific applications and approaches: a) focusing in mimicry specific regions of BM; b) different approaches to BM ECM representation; c) studies of physiological process; and d) studies of pathological process. The organization and content of the subcategories in this subsection is presented below (Fig 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Organization and content diagram.

Scope of Perfused cultures of multiple cell types section.

https://doi.org/10.1371/journal.pone.0243840.g002

Considering a BM as an environment composed of specific and well-established regions, the use of perfused 3D co-culture approach allows a more integrated mimicking of bone marrow niches, specifically perivascular, endosteal, and their deconstructed combination.

Kotha et al. developed a 3D matrix based on hydrogel co-cultured with normal and abnormal cells that allowed the study of some important cellular processes around the perivascular niche: transmigration and the adhesion of hematopoietic cells in the healthy and pathological marrow [65]. Cell migration, cell-cell interaction, phenotypic analysis, and gene expression analysis were evaluated in this study. The results showed the device was able to represent the behavior of hematopoietic cells around the perivascular region and is a useful tool for the study of the dynamics of these cells in the niche.

Still focusing on the perivascular niche mimicking, Marturano et al. presented an OoC based on a 3D scaffold of decellularized bone co-cultured with endothelial cells and mesenchymal stem [66]. The generated physiological flow was regulated in order to generate an oxygen gradient along the scaffold. The authors evaluated the formation of vessels, gene expression and histology [66] and compared the results with two control groups: static 2D, to evaluate the influence of flow and its control over oxygen and nutrient supply; and dynamic 3D in a polystyrene scaffold, to evaluate the influence of surface properties. In both cases, the device’s performance was better than the control group. In the first case, this is associated with the two-dimensional arrangement and the non-physiological design of the supply of nutrients and oxygen in the conventional method. In the second case, the composition of the scaffold surface is made of a synthetic polymer, that does not provide the cells with the necessary structure for adhesion. In addition, the synthetic polymer modified the natural cell phenotype, and consequentially altered the physiological processes of the microenvironment [66].

Furthermore, the perfused 3D co-culture approach was able to mimic another vital niche in bone marrow: the endosteal niche. The endosteal niche is a region near the spongy bone composed basically of the bone itself and the stromal cells. This niche regulates several hematopoietic processes including the quiescence [126]. From the studies that aim to mimic the endosteal niche, it was observed that this approach also allowed the development of an OoC device capable of representing the interaction between abnormal cells and the niche microenvironment. In the work that used bone cells to generate an ossified 3D matrix, Zhang et al. studied how multiple myeloma cells behaved in this environment [53]. Based on the evaluation of cell proliferation and cell-tissue interactions, this study reported the formation of a 3D arrangement between abnormal cells and bone tissue. This arrangement is responsible for a phenomenon known as “cell adhesion-mediated drug resistance”, already studied in a work that used an approach without fluid flow for the representation of BM microenvironment [54].

In another study, using the same device to assess cell viability and proliferation compared to that reported by a 3D static culture, Zhang et al. demonstrated that perfusion in the endosteal niche cultivated with multiple myeloma cells is an important factor of the microenvironment that improves the maintenance of the long-term co-culture [58].

To conclude the discussion about the use of complete systems in niche mimicking, it shall be mentioned that Aleman et al. developed an innovative deconstructed model of BM niches. Using a perfused 3D gelatin-based matrix, the authors simulated the three niches of bone marrow in a single system, identifying them as perivascular (periarterial and perisinusoidal), mesenchymal, and osteoblastic. The niches were placed in separate chambers interconnected by channels and the study reported the interaction of normal and abnormal hematopoietic cells with the niche cells and the grow factors that they release; this device created a system for the study of the physiological and pathological BM [67].

It shall be noted that despite the importance of representing BM with niche formation, most publications that have developed 'perfused 3D co-culture' devices have modeled the microenvironment as 'a general bone marrow niche' [55–57, 60, 63, 64, 68–70].

In an in vitro BM system, there are three aspects that define whether the microenvironment model is suitable for the expression of the native cell phenotype: cellular culture as in vivo; cellular processes as in vivo; and cellular response as in vivo. The most basic goal of this system is to provide cells with a microenvironment that makes them behave as in vivo.

From this perspective, one way to evaluate the in vitro cell culture and to study how the environment stimulates the cells is through the modification of the arrangement of 3D BM models. As already discussed, the marrow microenvironment can be built considering specific biochemistry related to the matrix composition (e. g. proteins and biomolecules), or a specific structure related to the scaffold design (e. g. pore size and organization).

Concerning the compositional approach for the representation of ECM, Wuchter et al. developed an OoC device that generates a microenvironment that stimulates hematopoietic cells to express their phenotype as in vivo [59]. The cell co-culture was composed of the progenitor mesenchymal stem cells and the ECM was generated by the mesenchymal stem cells themselves, in a cellular network that provided the complete integration of progenitor stem cells [59].

Torisawa et al. created a biomimetic hematopoietic environment based on a 3D matrix composed of type I collagen and bone powder to study the cell maintenance and culture in a mimicked native environment. This 3D matrix was inserted (subcutaneous implantation) into an animal (rat), and after 8 weeks a complete BM niche (histological analysis) was generated [55]. All niche cell types were found in the BM 3D matrix, including hematopoietic cells, and cell behavior was maintained in long-term culture. The results demonstrated this system can generate a physiological response of the BM to radiation and anti-radiation drugs (cell viability) [55].

From the structural perspective, the maintenance of the behavior of cells and their long-term culture was achieved using a well-structured 3D scaffold formed by hydroxyapatite [64]. Hydroxyapatite mimics BM very well because it is analogous to the native bone structure. Using phenotypic and gene expression analysis in HSPC and MSC in co-culture, the authors reported that it was possible to replicate cellular behavior in an experiment performed on perfusion for 28 days. The comparison with a control group confirmed that modelling the BM ECM as a 2D monolayer is a poor and very simplified representation of a complex in vivo microenvironment.

Thus, the developed OoC-type devices mimic the BM microenvironment better and allow the maintenance of cell culture and replication of the in vivo cellular behavior. Furthermore, these systems also allow studying fundamental characteristics of cellular behavior: physiological processes.

Hematopoiesis is one of the most important physiological processes and is responsible for the development and maturation of blood cell elements including erythrocytes, leucocytes, and platelets.

Thon et al. developed a 3D system based on a hydrogel with cells stimulated by a flow-generated shear that improves platelet formation (cell differentiation analysis) compared to static 3D culture [56]. In this study, the vascular shear in microfluidic channels was the main contributor to the increase of platelet formation and for the reproduction of cell phenotype in vitro (cell morphology analysis)

In addition to the devices that evaluate the production process of specific cell lines in the BM, some devices seek to represent more broadly the entire hematopoietic process, mimicking what occurs in the medullary environment.

Using the previously developed device and experimental setup [55], Torisawa et al. evaluated the hematopoiesis processes in a BM-on-a-chip [60]. The results showed that after 13 days the blood cells production (histology analysis) and the hematopoietic cell behavior were maintained. This biomimetic environment simulated a physiological response of the BM to radiation and anti-radiation drugs [60].

The possibilities and results achieved with the BM OoC developed by Torisawa et al. characterize this device as an alternative to animal models [55, 60]. However, the results were based on animal cells; therefore, the translation into a human clinical trial would require the change of a protocol and a new approach to cell culture.

To advance in the study of BM for clinical applications, Chou et al., in an experiment cultivating one microfluidic channel with endothelial cells and the other with hematopoietic and mesenchymal cells in a hydrogel matrix formed by type I collagen and fibrinogen, developed an in vitro system with dynamic fluid flow reproducing hematopoietic processes [70]. This BM device improved the expansion and maintenance of differentiated hematopoietic cell functions when compared to static cultures. As reported by the authors [70], the improvement compared to conventional models has also been described for the toxicity response and drug treatment, thus, showing the advances that this device offers for the study of BM pathologies.

The study of pathologies in systems that mimic the three main aspects of a general BM niche can provide essential information for understanding how abnormal cells modify their surroundings and how it affects their response.

In human subjects, these studies can be made in two contexts: BM as a part of a complex system (multi-organ-on-a-chip); and BM as a complex system (bone-marrow-on-a-chip).

In the first context, McAleer et al. and Herland et al. developed a multi-system that improves the cells drug response [68] and provides a platform for drug screening [69]. Evaluating cell viability, cell density, chemotherapeutic assay, cell proliferation, cell quantification [68] and cell quantification, phenotypes analysis [69], both studies contributed to global understanding of the therapeutic effects of drug treatment in clinical trials.

In the second context, Bruce et al. mimicked a pathological BM environment with a tri-culture system (abnormal lymphoblastic cell, mesenchymal stem cell, and osteoblast) in a perfused type I collagen-based 3D matrix and recreated the intricate network of interactions (cell-cell and cell-ECM) that actually occur in vivo microenvironments. The reported conditions are correlated with the enhanced abnormal cell survival during the drug treatment [57] and underline that the leukemic cell lines show a higher resistance to chemotherapeutic drugs than cells in 2D models.

In other publications, the same phenomenon of drug resistance was reported, but with different approaches. Houshmand et al., using a macroporous model based on perfused demineralized bone 3D matrix with abnormal erythroblast cell and MSC co-cultured, found that leukemic 3D microenvironment protected the cell from drugs cytotoxicity [63]. Abnormal cells embedded in demineralized bone 3D matrix showed greater resistance to drug treatment than in 2D models.

From a clinical point of view, these results contribute to understanding why some therapies for treating hematological disorders have not yet achieved the expected efficiency. Drug screening on conventional 2D platforms predicts cell response in a non-physiological environment, disregarding the dynamic environment of interactions and stimuli that exist in the human body. Here, the effect of drugs is overestimated because the model does not consider a physicochemical barrier that protects the cells from drugs [53, 54, 57, 58].

An overall conclusion from the analysis of all articles included in this revision is that the devices developed so far for mimicking the BM are effective. All publications reported results that demonstrated the effectiveness of the devices, showing an improvement in relation to conventional models and shedding new light on the knowledge about cellular phenomena in BM. However, despite the improvements and considerable success in recreating in vitro BM environment, some limitations and challenges remain.

The main limitations identified in OoC devices in this systematic review were related to two vital features of BM that regulate and maintain the niche cells and hematopoietic cells: niches integration; and oxygen supply.

Regarding niches integration, it is already known that BM is formed by integrating three main regions vital for BM cell behavior: endosteal, central, and perivascular.

To develop an integrated system that incorporates the mentioned niches, cells-cells and ECM-cells interactions shall be considered, whether in the same niche or not. In this systematic review only one publication implemented the concept of several niches in the same OoC device [67]. However, these niches were not considered as an integrated system, since they were not physically interacting with each other and only the biochemical interaction was considered. Hence, even considering this interaction, which is also important for regulating BM cells, the lack of physical interactions and other cellular processes, e. g., cellular migration, poorly represent the dynamically complex BM microenvironment.

Here, two relevant papers shall be mentioned that are still under the peer review process and were not formally included in this revision. Nelson et al. developed a device where co-culture is done with main cells that compose the three niches: osteoblasts (specific endosteal cell), endothelial cells (specific perivascular cell), mesenchymal cells, and hematopoietic cells (niches cells) [127]. The culture was implemented in a hydrogel matrix based on collagen types I and IV, fibronectin, and lamina, all components existing in the BM microenvironment [8]. However, this study does not completely consider distinct niches of the BM, failing to represent the physiological and structural/spatial aspects.

Glaser et al. developed a system with distinct and interconnected chambers for the representation of the endosteal and perivascular niches. In the endosteal niche, culture was performed with osteoblasts and BM stromal cells. The perivascular niche was done with endothelial cells and BM stromal cells. Then, hematopoietic cells were grown in these niches [128]. The cells cultured in each mimicked niche were analogous to the main cell composition of the BM niches in vivo [129]. The cultures were implemented under perfusion, in a 3D matrix based on fibrin, with a hydrogel used to represent hematopoietic niches in the BM [8].

Despite the evolution in mimicking the BM niches shown in the studies by Nelson et al. and Glaser et al., they did not evaluate the hypoxia on the cellular response, failing to represent the physiological condition of oxygen concentration within the BM.

Oxygen conditions in BM have a key role in cells regulation, yet it was considered only in one study included in this review [66]. The BM is a system in a state of hypoxia or low oxygen concentration. It is also important to note that there is an oxygen gradient in the niches and that the oxygen concentration is higher in the perivascular niche and lower in the endosteal niche. It was reported [15] that this parameter regulates the behavior of hematopoietic cells, and its representation in a device mimicking the BM is essential to obtain a more realistic model. This issue was raised by Chou et al., who pointed out the possibility of changing oxygen concentration to study the behavior of HSC as a function of oxygen availability [70], and also by Houshmand et al., who reported that the absence of the hypoxia condition is a limitation in mimicking the BM niche [63].