At the macroscale, perfusion is influenced both by scaffold composition and architecture, and by the features of the chosen perfusion system.

In terms of scaffold composition and architecture, bone scaffolds must be highly porous since porosity up to 90% facilitates fluid perfusion inside the structure of the scaffold (Abbasi et al., 2020). It also influences mechanical properties, and biological features of bone tissue, by promoting cell adhesion and proliferation (Mccoy and O’brien, 2010; Yeatts and Fisher, 2011; Wittkowske et al., 2016; Zhao et al., 2016). The most long used techniques to obtain porosity in 3D bone scaffolds are: 1) solvent casting/particulate leaching technique, based on the use of substances (porogens) dispersed in a polymer solution and dissolved when the structure is set; 2) foaming gas, based on the use carbon dioxide at high pressure; 3) freeze-drying, based on the removal of water or other solvents under a vacuum, in a frozen sample; 4) phase separation technique, based on thermal separation of a polymer solution into a polymer-rich phase and a solvent-rich phase that is removed by extraction, evaporation, or sublimation. However, these techniques often lack a precise control of the microarchitecture, due to the scarce control over pore shape, size, and interconnectivity (Tu et al., 2003; Hollister, 2005; Janik et al., 2015; Fereshteh, 2018). Additive manufacturing/3D printing is a more recent scaffold fabrication technique which is more reproducible and accurate, and allows controlling geometry at the macro and microscale (Bose et al., 2013). The 3D structure is obtained by a computer-aided design-based model and is formed, layer-by-layer, through the deposition of a powder, liquid, or solid materials. Notably, among these techniques, bioprinting also allows the simultaneous deposition of biomaterials and cells, thus recapitulating both bone microarchitecture and cell distributions of native tissues. For a comprehensive overview of 3D printing techniques, please refer to Moroni et al. (2018).

Manufactured porous scaffolds can be then included in several types of macroscale bioreactors, like spinner flasks and rotating bioreactors, or in milliscale bioreactors, to reproduce interstitial perfusion. Spinner flasks and rotating wall vessels are two basic and inexpensive alternatives to static cultures and allow better nutrient transport and proliferation rate (Mccoy and O’brien, 2010; Yeatts and Fisher, 2011). They use convective flow to ensure the mixing of culture media around the 3D cellularized scaffolds. Spinner flasks are bioreactor systems made of a cylindrical container, with a stirring element at the bottom that ensures culture medium circulation and mixing (Figure 1A). These devices are suitable for mimicking native bone environment and study bone tissue formation and cellular function since they increase mass transport, shear stress, diffusion of nutrients. They also allow the removal of toxic products (Mccoy and O’brien, 2010; Yeatts and Fisher, 2011; Chen et al., 2013; Jin et al., 2014; Kedong et al., 2014; Song et al., 2016b; Zhang et al., 2017; Duan et al., 2018; Melke et al., 2018; He et al., 2019; Nadine et al., 2019; Tsai et al., 2019; Rubert et al., 2021; Van Beylen et al., 2021).

Rotating bioreactors offer major advantages over spinner flasks since they also control the supply of oxygen and exert low fluid shear stress and turbulence (Kimelman-Bleich et al., 2011; Weszl et al., 2012; Song et al., 2013; Shekaran et al., 2015; Song et al., 2016a; Demir et al., 2018; Westman et al., 2019; Daulbayev et al., 2020; Nokhbatolfoghahaei et al., 2020a; Nokhbatolfoghahaei et al., 2020b; Nokhbatolfoghahaei et al., 2020c). The most common rotating wall bioreactor is formed by two concentric cylinders: 1) the outer one is a culture chamber and accommodates the cellularized scaffold, submerged in culture medium, while 2) the inner cylinder is static and permits gas exchange (Figure 1B). However, in these rotation-based bioreactors, internal nutrient transport is limited to the outer compartment. This asymmetric spatial distribution of nutrients across the thickness of the scaffold may lead to the formation of a dense cell layer on the surface that, in turn, may cause an uneven distribution of fluid perfusion and shear stress and, ultimately, the formation of necrosis and impaired mineralization at the core of the construct (Mccoy and O’brien, 2010; Yeatts and Fisher, 2011). Active perfusion offered by perfusion bioreactors that directly pumps fluid through the cellularized structures overcomes these limitations.

Perfusion bioreactors can be considered as milli fluidics devices, with millimetric channel/culture chamber dimension, ranging from 1 to 10 mm, containing fluids volumes from 1 to 100 ml (Freed and Vunjak-Novakovic, 2002). These devices accurately mimic the effect of interstitial fluid flow into cellularized constructs by ensuring better environmental control, good mass transport, and ultimately, physical cellular stimulation (Mccoy and O’brien, 2010; Yeatts and Fisher, 2011). By comparing static and dynamic culture conditions, Tocchio et al. clearly demonstrated that three days-perfusion in a customized perfused bioreactor better avoided the formation of a necrotic core within a hydrogel-based porous scaffold, when compared to a cylindrical bulk agarose hydrogel placed in a static cell culture flask (Tocchio et al., 2015). Of course, here, it must be highlighted that the different porosity and composition of the scaffolds also play an important role on the formation of the necrotic core. These bioreactors foresee the use of pumps that perfuse the media through the scaffolds, either continuously or non-continuously, and that can be fully automatized. Several types of perfusion bioreactors have been tested so far, both commercial (i.e., U-Cup bioreactors, CELLEC Biotek AG108) and custom. These systems often have in common a basic functional module: a culture media reservoir, peristaltic pumps, a tubing circuit, and culture chambers (Figure 1C). In perfused bioreactors, fluid shear stress is finely tuned and applied directly by perfusing the cell-laden scaffold, or indirectly by applying external deformations, which cause perturbation of the media in the culture chamber (Mccoy and O’brien, 2010; Tseng et al., 2014; Wang et al., 2014). External deformation is obtained through cyclic compressive or tensile loading, torsion or ultrasound. In bone bioreactors, external deformation in the range of 1–30% has been obtained by applying compressive stress, and in the range 3–10% by applying tensile stress, both up to 21 days (Mauck et al., 2002; Kavlock and Goldstein, 2011; Liu et al., 2012; Petersen et al., 2012; Petri et al., 2012; Hao et al., 2013; Ramani‐Mohan et al., 2018). Torsion and ultrasound have been rarely used (Drapal et al., 2021). To date, however, a systematic investigation of different routes to apply shear stress on cells and scaffolds have not yet been carried out, which hinders a clear selection of the optimal route for different applications.

In conclusion, perfusion bioreactors have overcome the limits of previous macroscale systems, as they can reproduce bone microenvironment in a more accurate and controllable manner. As a demonstration, from 2010 to date, 81% of papers on bioreactors focuses on the effectiveness of perfused bioreactors in inducing bone cell differentiation and mineralization, while all the other strategies combined account for 9% (Web of science database). As a result, in the recent years, perfused bioreactors are replacing spinner flask and rotating vessel for 3D in vitro models of bone: after 2010, average 58 research papers/year with “Bone Perfus* Bioreactors” as key words vs seven paper/years with “Spinner Flask bone” and “Rotating vessel bone” as keywords (Web of science as reference database).

Microfluidic systems, also called microfluidic bioreactors, are miniaturized bioreactor systems (Mestres et al., 2019), with precise micrometric design to simulate nutrient delivery, paracrine communication, and specific crosstalk between multiple cell types, in either a 2D or a 3D microenvironment. Most importantly, microfluidic systems permit the application of mechanical stress at the microscale level, by tuning physiological flow and fluid shear stress (Figure 2) and more closely mimic the physiological stimuli that occur during cell-cell and cell-ECM interactions. Furthermore, microscale models do not lead to volumetric displacement of fluid that causes transient normal forces that may alter cell function, as it occur in macro and milliscale systems (Moraes et al., 2011). Finally, microfluidic systems allow for simultaneous refinement of biomechanical and biochemical properties to form a chemical gradient along with application of the physical stimulus (Mestres et al., 2019). Other key advantages of microfluidic devices are: 1) a significant reduction of the amount of reagents and cells to be used and 2) the possibility to quickly analyse large samples arrays, with high levels of precision and resolution and with live imaging (Sackmann et al., 2014). Here’s why, although the principle behind flow-induced shear stress is the same, microfluidic systems are more advantageous for mechanotransduction studies and analysis of complex phenomena, such as osteogenesis and angiogenesis, in real-time and at a high-resolution, than perfusion bioreactors. On the other hand, it should be underlined that these systems are still at their infancy in the bone research field and do not allow to reproduce the phenomena at the macro/milliscale. In particular, they fail to reproduce the complex 3D porous microarchitecture, composition and mechanical properties (e.g., Elastic modulus) of bone tissue (Rauh et al., 2011). By further analyzing the limitations of microfluidics, the standard materials used as ECM in these devices are thermoresponsive organic hydrogels (matrigel, collagen, fibrin) that can mimic or not the organic phase of bone and are easily injectable, but do not resemble bone inorganic composition and microarchitecture. Recent studies have paved the way to functionalize hydrogels with an inorganic phase (Liang et al., 2020), such as calcium phosphate (HA microbeads, HA nanoparticles, tricalcium phosphate TCP), borosilicate glass-ceramics, that, however, are still far from reproducing bone tissue characteristics (Sharma et al., 2021). Moreover, in micro-bioreactor, cell culture surface is very small (around 0.5–0.8 mm²) and can be seeded only with a few thousand cells. This feature can be considered as a pro and cons (Mattei et al., 2014). On the one side, small volumes allow the insertion of patient-derived biopsies or cells inside the microfluidic chamber, which is crucial for the development of personalized therapeutic approaches (Arrigoni et al., 2017). On the other side, miniaturized organ-on-a-chip may be too simplistic in representing organ complexity (Mattei et al., 2014). In Table 1, we summarized pros and cons of perfusion device at the macro, milli and micro scale.

In vivo, oxygen gradients occur naturally due to limitations in oxygen transport and metabolic consumption of oxygen by cells (Rexius-Hall et al., 2014) and different level of oxygen in the microenvironment may change cell behaviour and response. Furthermore, oxygen levels are specific to different cell types or components of bone (Hirao et al., 2007; Volkmer et al., 2008; Volkmer et al., 2010; Volkmer et al., 2012; Inagaki et al., 2017; Lee et al., 2017; Stegen et al., 2018; Urdeitx et al., 2020). For instance, in bone tissue, different compartments are characterized by different levels of oxygen tension (pO₂), which is high in the periosteum, low in cortical bone, and even lower in bone marrow, despite its very high vascular density (Spencer et al., 2014).

In vitro, perfusion ensures the transport of oxygen across the cellularized construct, both at the milli- and microscale. However, obtaining a fine control of the oxygen level in in vitro models is not trivial. In milliscale systems, optical oxygen micro sensors can be inserted in the cell-laden scaffold to send pO₂ data, through an oxygen-triggered feedback mechanism. Data are collected by a computer that controls a syringe pump that, according to the type of input received by the computer, adjusts the perfusion rate by activating or stopping the fluid flow (Volkmer et al., 2008; Volkmer et al., 2010; Volkmer et al., 2012). Therefore, these systems are suitable for setting a predefined pO₂ in the culture chamber, but are unsuitable to create oxygen gradients (Figure 3). As for microfluidic systems, different devices have been developed to study the effect of O₂ tension on cellular behaviour, including the response to drug treatments. They can be engineered to allow a tight control of cell exposure to given oxygen levels, either one single oxygen level or multi-condition oxygen levels. This is obtained by the use of off-chip computer-controlled gas mixers, flow of oxygen scavenging chemicals, or on-chip gas mixer layouts (Rexius-Hall et al., 2014). Therefore, microfluidics appear more promising to achieve controlled oxygen gradients than the milliscale devices. These technologies are now well established and prospectively will be more extensively applied for tissue modelling. To date, however, very few papers are available.

For applications of bone tissue constructs in tissue engineering, it is already clear that the lack of vascularization may result in functional and physical failure upon implantation due to cell necrosis (Bandaru et al., 2020; Harvestine et al., 2020; Hann et al., 2021). Bone vascularization is critical to ensure bone cell survival since it allows the perfusion and delivery of fundamental cell nutrients and the removal of waste products of cell metabolism and this is the reason why inclusion of vascularized structures also in perfused bioreactors is a very active field of investigation.

Direct axial loading on bone and cartilage in vivo results in compressive forces that, in turn, generate pressure gradients and interstitial fluid flow. It has been already well established that osteoblasts and MSCs directly respond to shear stress by increasing the expression of the early and late osteoblastic markers and calcium deposition. The use of perfused bioreactors is extremely suitable to study these types of mechanic-biological stimulations. The most commonly applied mechanical stimuli to obtain shear stress in vitro in perfused bone models is compression; also tension, torsion and ultrasound are widely investigated although with aims other than studying osteogenic differentiation. In this chapter, we will discuss on how milli and microbioreactors have been used to study the effect of fluid perfusion on osteogenic differentiation.

Osteocytes are the most differentiated form of osteoblasts and the most common type (90–95%) of bone cells with a fundamental role in the regulation of bone and mineral homeostasis. In vivo, osteocytes reside in lacunae, which are interconnected microscale spaces 20–30 µm apart from each other (mouse bone). The intercellular dimension is important in cell–cell signalling for osteocyte process growth and mechanotransduction sensitivity (Gu et al., 2015). Nonetheless, osteocytes are quite rarely included in preclinical models, and even less in perfusion devices (average of three published papers per year since 2015, Scopus Database) (Mullen et al., 2013; Bellido, 2014; Choudhary et al., 2018; Sun et al., 2018). The main reason is possibly the difficulty of isolating osteocytes from the mineralized bone and recreating a 3D cell microenvironment that can mimic the lacunar-canalicular structure of bone tissue and the interstitial fluid flow (Webster et al., 2013; Sun et al., 2018), and the formation of an osteocytic network, all crucial for the osteocyte physiology. Thus, the use of 3D perfused devices may be particularly advantageous for the study of this specific bone cell. Form one side, 3D microfluidic device may allow the use of multicellular models, like osteocytes cultured with other cells of the bone microenvironment (e.g., osteoblasts and/or osteoclasts), to study the role of the osteocytic network on bone homeostasis and analyse its ability to induce osteoblastogenesis and osteoclastogenesis (Middleton et al., 2017; George et al., 2018). On the other side, osteocytes are mechanosensors, sensing to different kinds and extents of mechanical load, and reacting through the regulation of bone homeostasis (Brown et al., 2013; Bellido, 2014; Florencio-Silva et al., 2015; Terpos, 2015; Bellido et al., 2019), and thus, microscale bioreactors are very useful for study osteocyte-mediated mechanotransduction.

As an example, by the use of milli scaled devices, co-culture experiments showed that fluid shear stress induces the release of factors by osteocytes that affect osteoblasts and osteoclasts activity and modulate osteoblasts proliferation and differentiation, also through the release of nitric oxide (Vezeridis et al., 2006; Hoey et al., 2011; Wittkowske et al., 2016). Similarly, the application of compressive stress (5 min, 10 Hz, 2.5 N) on a 3D co-culture model, including osteocytes that were embedded in a Type I collagen gel, and osteoblasts that were cultured on top of the hydrogel, induced the formation of an osteocytic network within the collagen and increased collagen production by osteoblasts (Vazquez et al., 2014).

At the microscale, a larger number of studies has been published. Microfluidic devices incorporating osteocytes are usually focused onto three main aspects: 3D cell distribution, biomaterials, and mechanical stimuli (LA and Alam; Li et al., 2008; Gu et al., 2015; Wei et al., 2015; Xu et al., 2016; Middleton et al., 2017; George et al., 2018; Sun et al., 2018). To recapitulate this crucial structure in vitro, Gu et al. developed a cell construct in which osteocyte cell bodies were located into the interstitial spaces between BCP microbeads (ø 20–30 µm) (Gu et al., 2015). To recreate a bone-like microenvironment, other authors used collagen-based hydrogels and ceramics fillers that favoured the upregulation of osteocyte specific genes, such as Sost gene, a key osteocyte-specific marker for mechanotransduction (Wei et al., 2015; Middleton et al., 2017; Sun et al., 2018), or collagen-coating chamber that was effective in the maintenance of the osteocyte phenotype (Wei et al., 2015). Besides the composition of the ECM, the key feature that makes the model particularly suitable for reproducing the in vivo counterpart is the mechanical-induced shear stress that, as explained above, is mandatory for the characterization of osteocytes behaviour and features in bone homeostasis and pathology (LA and Alam; Li et al., 2008; Middleton et al., 2017; George et al., 2018). The most representative example is the device developed by Qiaoling Sun et al. who assembled an osteo-like structure mixing collagen-coated biphasic calcium phosphate microbeads (68% of HA and 32% of β-TCP) with MLO-A5 cells. Furthermore, cells were exposed to a cyclic compression-induced shear stress, obtained by cell chamber pressurization (Sun et al., 2018).

In summary, in fluidic bioreactors, osteocytes can form a 3D network to communicate with the other bone cells, and are extremely sensitive to mechanical loading, and respond to such external stimuli by releasing soluble factors which, in turn, control bone homeostasis. As a result, contrary to static models, microfluidics have the potential to mimic the dynamic osteocyte microenvironment (i.e., shear stress, ECM), paving the way to the development of in vitro reliable bone model, to fully understand the function of osteocytes in physiological and altered states.