Micropathological Chip Modeling the Neurovascular Unit Response to Inflammatory Bone Condition

Organ‐on‐a‐chip in vitro platforms accurately mimic complex microenvironments offering the ability to recapitulate and dissect mechanisms of physiological and pathological settings, revealing their major importance to develop new therapeutic targets. Bone diseases, such as osteoarthritis, are extremely complex, comprising of the action of inflammatory mediators leading to unbalanced bone homeostasis and de‐regulation of sensory innervation and angiogenesis. Although there are models to mimic bone vascularization or innervation, in vitro platforms merging the complexity of bone, vasculature, innervation, and inflammation are missing. Therefore, in this study a microfluidic‐based neuro‐vascularized bone chip (NVB chip) is proposed to 1) model the mechanistic interactions between innervation and angiogenesis in the inflammatory bone niche, and 2) explore, as a screening tool, novel strategies targeting inflammatory diseases, using a nano‐based drug delivery system. It is possible to set the design of the platform and achieve the optimized conditions to address the neurovascular network under inflammation. Moreover, this system is validated by delivering anti‐inflammatory drug‐loaded nanoparticles to counteract the neuronal growth associated with pain perception. This reliable in vitro tool will allow understanding the bone neurovascular system, enlightening novel mechanisms behind the inflammatory bone diseases, bone destruction, and pain opening new avenues for new therapies discovery.


Introduction
Blood vessels and nerve fibers are vital elements of the skeletal system [1][2][3][4] and important for the maintenance of bone or osteopontin. [20][21][22] Also, angiogenin released by osteoclasts is essential to maintain blood vessels growing in long bone through an angiogenin/plexin-B2-rRNA transcription signaling, resulting in endothelial cells proliferation. [8] We have demonstrated that the differentiation of human mesenchymal stem cells towards osteoblasts (bone forming cells) leads to marked impairment of their ability to promote axonal growth. [23] The mechanisms by which osteoblasts provide this nonpermissive environment for axons include paracrine-induced repulsion through Semaphorin 3A, Wnt4, and Sonic hedgehog and loss of neurotrophic factors expression (drastic reduction of nerve growth factor (NGF) and brain-derived neurotrophic factor (BDNF) production).
In addition, neuro-vasculature is actively modulated under inflammatory bone and joints affecting diseases, such as rheumatoid arthritis (RA) and osteoarthritis (OA). [20,[24][25][26] In these pathologies, a decrease in the total bone mass is observed recurrently connected to an increase of osteoclasts resorbing activity. [27][28][29] In inflammatory mouse model of OA, studies reported that osteoclasts are implicated, via netrin-1 signaling, in the pathological sensory innervation of the subchondral bone and endplates (intervertebral disc degeneration model). [30,31] Moreover, blood vessels are known to suffer changes in cell adhesion and permeability [32] and a re-organization of the sensory nerve fibers, characterized by an alteration in the morphology, density, and sprouting into areas of the joint-that are normally poorly innervated-is also re-arranged. [33][34][35][36] Several of the molecular mediators produced during inflammation act on peripheral nerve terminals of nociceptor neurons to produce pain sensitization. [37] It is evident that there is an established crosstalk between bone cells, vascularization, and innervation patterns relevant in physiological and inflammatory pathological settings. Nevertheless, the mechanisms underlying these intricate interactions remain unknown. A deeper understanding of the signaling pathways involved in the interactions across bone microenvironment, endothelial cells, and sensory neurons would accelerate the development of new therapies to improve bone cells balance, counteract pain and avoid osteonecrosis caused by lack of blood supply.
Microfluidic-based devices perform as promising tools to address complex crosstalks in appropriate in vitro microenvironments. Existing models of the bone microenvironment only mimic the presence of either blood vessels [38][39][40][41][42][43] or nerve fibers. [23,44,45] To our knowledge, current microfluidic-based models do not couple the neurovascular unit to the bone microenvironment to provide a complex and accurate system resembling the bone compartment. [46] To fill this gap, in this work, we designed, optimized, and validated a NeuroVascular Bone chip (NVBchip) to unravel the interconnection between these systems. Our goal was to model the neurovascular unit at the inflammatory bone microenvironment and provide a consistent tool to understand how nerve fibers and blood vessels self-organize under inflammatory stress. We demonstrated that this platform is suitable for testing drug delivery systems to counteract the inflammatory process by providing anti-inflammatory loaded nanoparticles (NPs) to the system.
We acknowledge that this platform has the potential to deliver robust information about the mechanistic events at the bone inflammatory niche-including nociception-/pain-related-and serve as a therapeutic screening tool.

Compartmentalized Microfluidic Chip to Enclose Neuronal, Vascular, and Bone Cells: Setting the Design
Compartmentalized microfluidic chips have been used to explore the neurovascular communication, mainly in the context of the central nervous system, to mimic the blood-brain barrier. [47][48][49][50][51][52][53][54][55][56][57] Concerning the in vitro replication of the bone microenvironment, vasculature has been integrated in microfluidic models given its importance in bone maintenance [58,59] or pathologies. [42,60] The lack of appropriate models to recapitulate the crosstalk of the neurovascular unit at the bone tissue was the motivation to establish a new microfluidic design to enclose neuronal, vascular, and bone cells into a single chip. [46,61] To fulfill this goal, we designed a new microfluidic platform-NVBchipcomprising three interconnected channels ( Figure 1A). Of note, the use of different compartmentalization features was critical given the anatomical connection between the systems in vivo. The NVBchip is designed to provide a structural organization that allows the distal contact of axonal terminals with endothelial compartment combined with a direct and close contact of endothelial and bone cells.
The first channel, separated from the remaining by microgrooves (3 (H) x 10 (W) x 450 (L) μm), allow the proper segregation of somal and axonal components of neuronal cells, [62] reproducing the in vivo setting where only the neuronal projections reach the bone microenvironment ( Figure 1B). A small punch (Ø 3 mm) centered in the neuronal channel ensures the open access to seed neuronal explants and to align them with the microgrooves and minimize the variability of neurite sprouting in the axonal side ( Figure 1A-C).
The two channels on the axonal side represent the subcompartments of the bone niche: vascular and bone units. They are divided by micropillars (100 μm long, 50 μm wide, and 50 μm interspaced) to provide a superficial tension that restrains the 3D seeding of the cells within the vascular unit (VU), while allowing the later interchange and migration of the cells between the two subunits ( Figure 1B). The loading of the hydrogel is performed by pipetting through the 5 mm reservoir linked to the central channel ( Figure 1C,D, Video S1, Supporting Information). Afterwards, the medium was added to the adjacent compartment ( Figure 1E, Video S1, Supporting Information). Finally, the bone unit presents an extra Ø 6 mm punch to create a larger culture area to the bone cells seeding. All the 3 channels are punched in the extremities to create media reservoirs from where cells are seeded and the medium is changed ( Figure 1C).
Most microfluidic chips comprising endothelial cells either use micropillars or porous membranes. Both allow an open, less constrained, closer, interaction between the cells. [47,48,50,63,64] On top of that, micropillars offer the support structure to achieve 3D lumen-like vessels. [46,65] The microchannels features provide a restrained connection and can be observed in models targeting specific cellular components (e.g., axonal terminals targeting). For our model, we opted for the two modalities: narrow microchannels connection from neuronal to VU and more On the left side, the nerve unit (blue) includes embryonic DRG explants as source of neuronal cells. This unit was separated from the remaining by microgrooves that allow the crossing of axons but not of cell bodies. In the central channel, the VU (purple) contains the endothelial cells seeded within a hydrogel. Micropillars separate this unit from its adjacent one, creating a superficial tension that allows for the restrain of the hydrogel. A) On the right side, the bone unit (green) where osteoclasts will be seeded. Not to scale. B) Tile scan microscopy image of the device's microfeatures: microchannels and micropillars. Scale bar 500 μm. C) Low magnification microscopy tile scan image of the device's full structure showing the long compartments and the medium reservoirs. Scale bar 5 mm. D,E) Images showing the loading of hydrogel (light blue) to the central channel (t = 60 s) and the addition of aqueous solution (dark blue) to the lateral channel (t = 80 s). Scale bar 500 μm.
permissive micropillars separation between vascular and bone units.
The rationale behind the sequential compartmentalized configuration is based on the anatomical aspects to reproduce the projection of sensory neurons into the neurovascular unit, where only the axonal terminals interact directly with forming capillaries, followed by cell-cell direct contact within the bone cell in the later compartment. By merging previous microfluidic-based features, we were able to design a new model that emerges as a new platform to recapitulate the complex neurovascular unit, adjacent to the bone compartment, representative of the bone innervation/vascularization processes.

Generation of the Vascular Unit (VU): 3D Substrates and Microfluidic Conditions
Vascular system is critical to supply nutrients and clear metabolites from tissues. In parallel, it is the main route for the circulation of immune cells as well as therapeutic drugs or their delivery systems. [66] As such, the first optimization step of the proposed platform was to establish the hydrogel that would sustain endothelial cells culture and allow the formation of a capillary-like structure microvascular network. [67,68] This was performed first outside the microfluidic devices using the ibidi μ-slides angiogenesis to assess the tube formation (in 2D and 3D). The angiogenic capacity of two different hydrogels, fibrin and collagen/fibrin, was evaluated and compared to the Matrigel control.
Endothelial cells start to self-organize on top of all substrates after 24 h. Circular structures were observed on collagen/fibrin and fibrin hydrogels while the Matrigel control showed a higher arrangement of the tubular-like structures. All tested substrates were able to support the maintenance of the structures for 48 h ( Figure S1A-C, Supporting Information). Endothelial cells remained viable, and no differences were observed across the different hydrogels ( Figure S1D-F, Supporting Information).
As collagen I is the main component of the organic fraction of bone extracellular matrix, [69] the formulation composed of collagen and fibrin was further tested for the 3D endothelial cell culture. Four cell densities were tested: 2.25 × 10 6 , 5 × 10 6 , 10 × 10 6 , and 20 × 10 6 cells mL −1 . The lowest cell densities, although being previously reported in other bone in vitro models, [58] were insufficient to ensure cell viability (data not shown). Regarding the highest cell densities, 20 million cells mL −1 showed a greater endothelial cell arrangement within the fibrin/collagen and Figure 2. VU optimization. Endothelial cells (HUVEC) were seeded in 3D within A) Matrigel, B) Collagen/Fibrin hydrogels. C) Metabolic activity of the 20 million cells mL −1 density over 96 h of culture within the different hydrogels was assessed by the resazurin assay. Within the microfluidic channels, the cells formed a lumen structure covering the entire channel, represented by its D) orthogonal projections and a E) 3D reconstruction using IMARIS software. F) The structure permeability was assessed at day 1 and day 6 of culture. CD31 expression at G) day 1 and H) day 6 was assessed by immunostaining. F-actin is depicted in green and nuclei in blue (A, B, D, and E); CD31 staining in (G) and (H) is represented in green and nuclei in blue. Scale bars: 100 μm.
Matrigel substrates (Figure 2A,B). Moreover, cell metabolic activity was kept unchanged up to 96 h of 3D culture, regardless of the substrate used ( Figure 2C).
The endothelial cells arrangement was more evident on Matrigel substrate, a commercially available basement membrane extract. [68] Despite its potent angiogenic effect, its composition is undefined and dependent on batch variability, and for this reason there is a need to move forward with other suitable alternatives to incorporate into the model. It is described that capillary growth rates are usually faster in fibrin gels, when compared to collagen, as vasculogenesis is favored in softer matrices and collagen/fibrin matrices are usually stiffer when compared to fibrin. [70,71] Herein we show that the cells embedded in the collagen/fibrin hydrogel were still able to organize themselves and maintain their metabolic activity. In fact, the collagen/fibrin hydrogel is a suitable candidate for the assemble of vascularized bone models, and has been previously used in the establishment of 3D bone organotypic cultures. [38,58,72,73] Paired with the above-mentioned collagen/fibrin results, we proceeded with this substrate to follow the optimization steps within the microfluidic devices. 20 million cells mL −1 were seeded within the central channel of the microfluidic devices to establish the VU. After 24 h, a compact layer of endothelial cells covering the walls of the microchannel was observed ( Figure 2D). A lumen-structure was obtained with full coverage of the total area of the channel walls ( Figure 2E, Video S2, Supporting Information). This technique, called "viscous finger patterning", is used to create channel geometries and lumens surrounded by hydrogel layers. It is based on the displacement of a more viscous fluid by a less viscous one. [74] To confirm the integrity of capillarylike structure, a permeability assay was performed using a fluorescent tracer (40 kDa fluorescein isothiocyanate (FITC-dextran).
www.advancedsciencenews.com www.advhealthmat.de The tracer was added to the VU at a concentration of 200 μg mL −1 and aliquots were collected from the adjacent compartment at different timepoints, over the course of 60 min. The results showed a tendency for a decrease in the permeability from day 1 to day 6 of endothelial cells culture inside the microfluidic chip (Figure 2F). The reduction in the permeability of the endothelial layer is associated with the increased organization and maturation of the capillary structure and higher connection between the cells. Studies reporting the formation of vessel-like structures in microfluidics described the capacity to retain the integrity of the membrane, consequently low permeability coefficient, up to day 7 of culture. [75][76][77][78] To further characterize the maturation of the endothelized channel, we assessed the expression of CD31 (plateletendothelial cell adhesion molecule, PECAM-1), known to be expressed by hematopoietic and endothelial cells, important to sustain the vascular barrier integrity. [79,80] Immunostaining was performed on endothelial cells fixed at days 1 and 6. It was observed that the endothelial cells express CD31 marker at days 1 and 6 inside the microfluidic chips ( Figure 2G,H). It was possible to detect CD31 protein predominantly localized at the cell membrane more evident at day 6 ( Figure 2H), reinforcing the stabilization of the cell-cell adhesion in the lumen-like structure.

Generation of the Bone Unit (BU): Physiological and Inflammatory Conditions
When the bone homeostasis is disrupted, there is an unbalance between bone formation and resorption leading to loss of mechanical strength. [69] Osteoclasts, the cells responsible for bone degradation, represent a higher challenge in the optimization of in-chip microfluidics and, so far, no models were presented comprising fully differentiated primary osteoclasts in microfluidic platforms. Therefore, we focused our efforts on optimizing the culture of differentiated osteoclasts as the key players of the bone unit compartment.

Osteoclasts Maturation at the Bone Unit
Prior to starting the microfluidic chips optimization, we have tested the differentiation of the primary mouse osteoclasts, derived from bone marrow hematopoietic lineage, on two different coatings: poly-D-lysine (PDL), widely used to promote cell adhesion, and collagen I, due to its resemblance with the natural bone matrix. [69] The different coatings were performed in glass coverslips and standard tissue culture plates (TCPS) were used as positive control ( Figure 3A). Multinucleated osteoclasts were observed in both PDL ( Figure 3B) and collagen-coated conditions ( Figure 3C), despite the lower number when compared to control ( Figure 3A). The quantification of multinucleated cells with more than 3 and 10 nuclei, showed a tendency of increased osteoclast differentiation, and percentage of area occupied by the osteoclasts, in presence of collagen coating ( Figure 3D,E). The modification of surfaces with collagen type I was reported to improve monocyte proliferation and differentiation to osteoclasts [81] and collagen-based materials are also used to culture bone cells when establishing organotypic models. [38] The first approach on the optimization inside the microfluidic chips was performed using the initial closed microfluidic channel ( Figure S2A, Supporting Information), with a density of 7.5 × 10 4 cells per channel. Under this experimental setup monocytes kept their mononucleated state without fusing and reaching the formation of multinucleated cells ( Figure S2B,C, Supporting Information). Therefore, we hypothesized that either the reduced culture area or the hypoxic conditions could be preventing the full maturation of the cells and their fusion to form multinucleated osteoclasts. To overcome this limitation a larger area for the bone compartment was provided for the seeding of osteoclasts. As depicted in Figure 3F, this upgraded design included a 6 mm punch uniting the already existing 5 mm punches. This modification allowed the differentiation of the osteoclast precursors, as indicated by the presence of multinucleated cells (Figure 3G). To increase the population of multinucleated osteoclasts within the microchip, the cytokines concentration was adjusted. Cytokines concentrations were increased from 30 to 50 ng mL −1 of macrophage colony-stimulating factor (M-CSF) and from 100 to 120 ng mL −1 of receptor activator of nuclear factor kappa-B ligand (RANKL), as previously shown to lead to a higher homogeneous osteoclasts culture in vitro, [82] resulting in a higher number of multinucleated osteoclast within the microfluidic chip ( Figure 3H). The quantification of number of cells with more than 3 and 10 nuclei revealed an enhanced osteoclast differentiation when exposed to higher cytokines concentration ( Figure 3I). Also, larger osteoclasts were obtained under this stimulus (Figure 3J). The maturation of the osteoclasts in the microfluidic chips was also supported by the expression of Cathepsin K (Figure 3K-K'') and by the staining of Tartrate Resistant Acid Phosphatase (TRAP) ( Figure 3L), golden standard markers of osteoclasts differentiation. [83,84] Overall, the collagen coating, the larger culture area (6 mm), and the adjusted cytokines' concentrations (50 ng mL −1 of M-CSF and 120 ng mL −1 of RANKL) added daily, were the established conditions set to the bone unit. The combination of these methods allowed to obtain consistent cultures with multinucleated osteoclasts present inside the microfluidic chips.

Osteoclasts Angiogenic Profile under Inflammatory Conditions
In inflammatory bone pathologies, the disruption of bone mechanical and physiological proprieties is mostly related to the high osteoclast activity on bone degradation front. [27][28][29] Given the significant impact that osteoclast activity has in musculoskeletal inflammatory disorders, these cells are a target for the current and new development of bone-driven therapies. [20] In addition, osteoclasts sustain inflammatory microenvironments by secreting molecules with a major impact on angiogenesis and innervation processes. [20,30] Secretion of interleukin 1 (IL-1 ) pro-inflammatory cytokine is implicated in the pathophysiological changes that occur during different disease states, such as RA, neuropathic pain, OA, or vascular disease. [85] Therefore, it is experimentally and clinically relevant to set the conditions for the creation of pro-inflammatory osteoclasts' phenotype within the in vitro models. In this study, we followed previous data to generate pathologically activated osteoclasts under IL-1 cytokine (0.5 ng mL −1 ) stimulation. [86] In the control conditions, Figure 3. Testing of protein coatings and cytokines concentrations to achieve fully differentiated osteoclasts. Primary murine osteoclast progenitors were seeded in multiwell plates directly on A) TCPS or in glass coverslips coated with B) poly-D-lysine (PDL) or C) collagen I. For each coating, D) the percentages of cells which were multinucleated with more than 3 and more than 10 nuclei, as well as the E) percentage of area occupied by mature osteoclasts were quantified. F) An upgrade of the initial design of the microfluidic chamber was tested to improve osteoclast differentiation, increasing the total culture area by adding an extra 6 mm punch connecting the already existing media reservoirs. Different concentrations of cytokines were tested: osteoclasts' morphology cultured under G) 30 ng mL −1 of M-CSF and 100 ng mL −1 of RANKL and H) 50 ng mL −1 of M-CSF and 120 ng mL −1 of RANKL. For the different concentrations, I) the percentages of cells which were multinucleated with more than 3 and more than 10 nuclei, as well as the J) percentage of area occupied by mature osteoclasts were quantified. To confirm differentiation of osteoclasts with the defined culture conditions, immunostaining against cathepsin K (K; detailed view of multinucleated osteoclast K'(actin) and K''(cathepsinK)) and L) TRAP staining were performed. Actin is stained in green and nuclei in blue (A, B, C, G, H, K); K) cathepsin is stained in red. Scale bars: 100 μm.
under M-CSF and RANKL stimuli, multi-nucleated osteoclasts presented a round shape and cytoplasmatic filopodia feature related to the high mobility of these cells (Figure 4A,B). Under pro-inflammatory conditions, we observed that osteoclasts presented an irregular cytoplasmatic protrusion ( Figure 4C,D). These morphological features have been associated with higher aggressiveness and resorptive activity in accordance with previous studies. [86] Under inflammatory settings, osteoclasts were shown to release cytokines as IL-1 and tumor necrosis factor-alpha (TNF-). [87,88] To confirm the contribution of osteoclasts to sustain the inflammatory response and their angiogenic capacity, the secretome conditioned for 24 h was screened to detect a panel of over 20 secreted cytokines and chemokines. The data indicate that the osteoclasts exposed to IL-1 present a positive balance towards a pro-angiogenic potential ( Figure 4E), in line with the literature since angiogenesis is potentiated in inflammatory conditions. [89] It is important to note that the increase of IL-1 detected by the ar-ray is not due to the addition of exogenous cytokine during stimulation, since the recombinant protein used and the detected by the array are from different species.
The upregulated factors have been implicated in inflammatory signaling pathways, [90][91][92][93][94][95] and described to be mainly secreted by macrophages progenitors. [96][97][98][99][100] Of note, among those, there are molecules described to directly impact bone metabolism or to be involved in inflammatory bone diseases. Granulocyte colony-stimulating factor (G-CSF) was one of the first cytokines detected in human synovial fluid from inflamed joints. G-CSF was reported to drive inflammatory joint disease by increasing the production and mobilization of myeloid lineage cells from the bone marrow and inducing the trafficking and local activation of these cells in peripheral tissues. [92,93] Fibroblast growth factor (FGF) signaling showed to be important for joint function and the impairment of the signaling pathway leads to progression of OA. [92] C-C Motif Chemokine Ligand 11 (CCL11) was reported to mediate inflammatory bone resorption, playing a role in osteoclast migration and resorption activity. [95,101] Interferon-gamma (IFN-) and monokine induced by interferon-gamma (MIG, or CXCL9), have been related with different inflammatory diseases. IFN-has been shown to induce vascular permeability, in an animal model of inflammatory bowel disease, [102] through a VE-cadherin mediated vascular barrier disruption. [103] Considering MIG, the inhibition of its signaling axis has been shown to suppress the progression of periodontitis. [104] In this in vitro conditions, we observed that the osteoclasts profile shifts towards a pro-angiogenic one. Still, the increased secretion of anti-angiogenic factors as IFN-and MIG might be important to support the increased permeability of endothelial barrier under pro-inflammatory stimuli.
Our data indicate that the osteoclasts support the bone inflammatory microenvironment. In fact, different subsets of inflammatory osteoclasts contribute to the regulation of inflammatory processes but also to maintain an immunosuppressive bone microenvironment. [20,105] These insights make osteoclasts an important target for anti-inflammatory therapies of chronic inflammatory diseases as well as for influencing the bone environment.

Assembly of Vascular and Bone Units under Inflammatory Conditions
Besides its function to provide nutrients, clear metabolites, and facilitate cellular exchange, the vasculature is also highly affected in pathological settings. Increased vascular permeabil-ity of the endothelium is often associated with inflammatory conditions, [32,106] along with the rearrangement of the tubular structures.
The relationship between vasculature and bone-resorbing osteoclasts is not fully understood yet. Data suggest that suppression of osteoclast formation with osteoprotegerin impairs angiogenesis. [21] Additionally, other study reported that osteoclasts regulate anastomoses of type H vessels, at the growth plate resorption, during bone growth. [107] Still, no in vitro studies have clearly addressed how osteoclasts specifically impact vascular network formation in vitro. [108] This is surprising given that several therapeutics typically target osteoclasts, with potential impact on angiogenesis and consequently osteonecrosis. [109] Therefore, after the optimization of each unit separately, the next step included the assembling of both vascular and bone units ( Figure S3, Supporting Information) and validation of the vascularized bone model under the inflammatory settings.
The VU function was characterized by assessing the integrity and cohesion of the tubular-like structures upon exposure to inflammatory osteoclasts. Permeability assay, using 40 kDa FITCdextran, was evaluated at the VU adjacent to the bone inflammatory unit (V-IU)-osteoclasts under IL-1 exposure-and compared to experimental setup of VU adjacent to the physiological bone unit (V-BU)-osteoclasts in standard culture conditions. Our results show that the inflammatory osteoclasts in the adjacent unit increased the permeability of the endothelial barrier ( Figure 5A). This paracrine effect is observed in the direct crosstalk between the two cell types and not in the presence of the secretome alone or medium supplemented with pro-inflammatory cytokines ( Figure S4A-C, Supporting Information).
Inflammatory setting increases vessels permeability, reduces the number of tight junctions, and alters the endothelial cells' metabolomics profile. [110] Stimulation by a cytokine mix comprising TNF, IL-1 , IFN-, and lipopolysaccharides was reported to result in the loss of vascular endothelial (VE)-cadherin and zonula occludens-1 adhesion molecules expression, indicating disruption of the endothelial barrier. [111] A model stimulated by TNF-exhibited a tenfold decrease in transendothelial electrical resistance. [47] Given the increased permeability of the endothelial barrier in our model under inflammatory conditions, the expression of CD31 adhesion molecule was evaluated ( Figure 5B-D). In fact, we could observe that cellular distribution of CD31 was mainly localized at membrane level in control conditions ( Figure 5B), while in presence of bone unit (physiological and inflammatory conditions) CD31 was found predominantly localized in the cytosolic space ( Figure 5C,D). The translocation of CD31 adhesion molecules from membrane to cytosol can disrupt the endothelial junction complex at the site of cell contact, which can be associated with the increase in membrane permeability. [79,112,113] Microfluidic models addressing the endothelial barrier permeability and the expression of adhesion molecules have been reported. [57,114] These models have been used to study the bloodbrain barrier function, screen drug candidates, and predict the clearance of pharmaceuticals when exposed to pro-inflammatory cocktails. [57] Our NVBchip model can fit similar purposes in the bone pathologies context as we were able to recapitulate critical events that occur under inflammatory microenvironment.

Assessment of In Chip Nerve Sprouting in the Vascularized and Inflammatory Bone Unit
Sensory neurons are described to be affected by the presence of an inflammatory environment. [37,115] Along with the modulation of the vasculature network, also the neuronal connections are rearranged, changing the concentrations of neuromodulators within the microenvironment. In some pathologies, the increase in neuronal density is related to higher nociception at the periphery, culminating with higher pain sensation. [116,117] Within the inflammatory NVBchip, we proposed to address the response of sensory neurons sprouting to the vascular bone unit schematized in Figure 6A. For this reason, the final step in the establishment of the micropathological chip was to include sensory neurons and validate the interaction of the nerve unit with the previously established and characterized vascular and bone units.
The seeding of DRG explants (containing the cell soma of sensory neurons) in the microfluidic platforms has been previously optimized by our group. [44] DRG was seeded into a 3 mm punch, used to allow open access to the neuronal cell compartment and to align the explants with the microgrooves.
The setup of the complete model, with the incorporation of sensory neurons, allowed to fully establish the assembly timeline. After microfluidic replication, sterilization, and glass attachment, endothelial cells are the first to be seeded in 3D collagen-fibrin hydrogel. The dried device, along with the micropillars feature contributes to the hydrogel restrain at the central channel. Neuronal DRG explants are the following cells to be seeded in the left compartment after laminin coating. Finally, osteoclasts are seeded, after collagen coating of the large opened  To generate the inflammatory environment, pro-inflammatory cytokine IL-1 was added to the osteoclasts culture medium (orange). B) The spatial interaction of HUVEC and DRG extensions was addressed. C-E) Images from the same z-stack show that axons grow on top of the endothelial monolayer in the bottom of the channel (z = 0). F) A 3D reconstruction performed using IMARIS software shows that neurites appear in the central channel through the microgrooves. Actin staining in green, NF200 in red, and nuclei in blue. The vascular-nerve unit was established adjacent to bone (VN-BU; F) or G) inflammatory (VN-IU; G) units. Inverted images showing NF200 staining in black. Axonal growth was quantified using AxoFluidic. [44] H) This software returns two values: the value, which is related to the extension of growth, and the I) A value, which represents the fraction of saxons that effectively cross the microgrooves. Scale bars: 100 μm (B-F), 500 μm (G, H). Data represented as mean ± SD; *P < 0.05 and **P < 0.01. area, in the right compartment. At day 6 of maturation, the axonal extension from neuronal to vascular-bone units was evaluated and quantified under inflammatory and physiological conditions. Nerve fibers positive for neurofilament 200 (NF200) were observed interconnected with endothelial cells at the VU (Figure 6B). As depicted in 3D reconstruction, neurites appear into the vascularized compartment through the microchannels (Figure 6F). From there, the fibers grew around the endothelial cells at the top and bottom layers ( Figure 6F, Video S3, Supporting Information). Higher detail presented in Figure 6C-E (z-stack images) shows that the axonal growth occurs aligned with endothelial cells layer that covers the channel. Previous reports demonstrated that, in vivo, endothelial and neuronal cells can connect directly [118] providing evidence that these interactions are not an in vitro artifact.
Axonal sprouting was quantified using AxoFluidic, a software previously developed by our group. [44] Images of axonal growth were obtained by immunostaining of NF200 in the physiological and inflammatory models ( Figure 6G,H). The algorithm quantifies the axonal growth from the microgrooves towards the vascular-bone unit. It returns the A constant value representing the number of axons that reach the VU and the value related to the length of the axons in the VU. The analysis shows that, under inflammatory settings, there is a significant increase in the sprouting and extension of axons towards the VU (Figure 6I), without affecting the fraction of axons that effectively cross the microgrooves ( Figure 6J). The axonal growth remained unchanged for the neurons exposed to conditioned medium and respective controls ( Figure S5, Supporting Information).
Our observations are in agreement with the literature as HUVEC are acknowledged to secrete neurotrophic factors, namely BDNF, that stimulate axonal growth from different DRG species. [9,119] Moreover, it was shown that pro-inflammatory activation of endothelial cells enhances their neurotrophic ability through the secretion of exosomes containing functional miRNA. [120] In vivo experiments demonstrated that angiocrine factors derived from brain endothelial cells regulate the homeostasis and regeneration of neural stem cells by neurotrophin-3 expression. In stress conditions, after hypoxic injury, upregulation of vascular endothelial growth factor (VEGF)-A through the activation of VEGF receptor-2 enhances the production of nitric oxide, which induces BDNF in brain capillary endothelial cells to drive the expansion and maturation of neuronal cells. [9] Osteoclasts might play a direct role in axonal sprouting as these cells are recognized to induce neurite growth in OA mouse models through netrin-1 secretion. [30] Within our NVBchip model, it will be interesting to dissect indirect pathways of interaction as inflammatory osteoclasts might induce also the secretion of neurotrophic factor by endothelial cells.

In-Chip Neurovascular Modulation under Micropathological Bone Inflammation: Testing Anti-Inflammatory Nano-Based Delivery Systems
The proposed platform has not only the great potential to provide new insights into the poorly known crosstalk between innervation, angiogenesis, and bone in a pro-inflammatory setting, but also to be applied as a reliable and biologically relevant in vitro model, able to screen potential novel treatments targeting bone inflammation and pain. As proof of concept and to demonstrate the responsiveness of our model, we have tested the potential of a well-known non-steroidal anti-inflammatory drug (NSAIDs), ibuprofen, either encapsulated into poly-lactic-co-glycolic acid (PLGA) NPs or in its free form. Despite the controversy in the use of NSAIDs in bone healing, [121] ibuprofen has been shown to have beneficial effects in bone and cartilage [122,123] healing by reducing inflammatory cytokines and nerve inflammation. Taking advantage of the multiple features that nanomaterials can offer, [124] PLGA NPs encapsulating ibuprofen were produced, aiming at surpassing the short half-life of the drug, improving drug half-life and stability by providing a protective shield, [125] and consequently increasing its retention and local effectiveness. PLGA NPs were prepared by the precipitation method, as described, [126] and presented a mean size of 206±4 nm, polydispersity index of 0.05±0.02, and almost neutral zeta potential of −0.98±0.59 mV (Figure 7A). An association efficiency of 71.9±4.24% was achieved and the release profile showed that 32.3±18.5% of ibuprofen was released in the first 24 h at physiological pH of 7.4 ( Figure 7B).
Treatment with ibuprofen-loaded NPs was performed on day 6 of the triculture, in the presence of differentiated inflammatory osteoclasts. NPs were added to the VU at a final concentration of ibuprofen of 15 μg mL −1 and allowed to diffuse through the channels for 24 h. Soluble ibuprofen, at equal concentration, was used as control. Finished the incubation, the axonal growth was measured. A decrease of axonal sprouting was observed (Figure 7F,G), upon 24 h when compared to control ( Figure 7D,G), without affecting the fraction of axons that effectively cross the microgrooves ( Figure 7H). This time point corresponded to 30% of drug release from the particles. The soluble form of ibuprofen did not reproduce the same effect reinforcing the relevance of the PLGA nanocarrier ( Figure 7E,G).
Afterwards, we tracked the fluorescent-labeled NPs to understand if they were internalized, by which cells, and how they were distributed in the system. NPs were able to diffuse passively through the device with no dynamic flow applied. Given the high resistance presented by the microchannels leading to the somal compartment, is more likely for the NPs-containing medium to diffuse into the osteoclasts compartment. Interestingly, although nerve terminals were present at the vascular compartment, no axons or endothelial cells were detected with internalized NPs (Figure S6A,B, Supporting Information) and no alterations in axons morphology were observed ( Figure S6C,D, Supporting Information). The increased permeability of the endothelial barrier under inflammatory conditions enabled the diffusion of the NPs towards adjacent inflammatory bone compartment. Fluorescent NPs were detected at the inflammatory bone unit and internalized by the cells (Figure 7C).
PLGA NPs were reported to be internalized by neurons in different settings. [127] Neuronal human cell line showed a wide distribution of the fluorescent PLGA NPs in the cytoplasm and within the nucleus promoting a curcumin-mediated neuroprotective effect. [128] In vivo, anesthetic-loaded PLGA NPs were reported to be internalized by ≈50% of DRG neurons and to reduce pain in mouse models of chronic DRG compression. [129] In these reported studies, the PLGA NPs were directly administered to the cell soma while, in our system, the nerve terminals are exposed to the NPs at their axonal terminal. In our inflammatory NVBchip model, the effect on axonal sprouting decrease might be mediated indirectly through the reduction of inflammation at the cells internalizing the NPs bone compartment ( Figure 7C).
PLGA NPs are widely used as drug delivery systems, due to the properties of this synthetic polymer, namely its biocompatibility and biodegradability, and to the possibility to encapsulate a wide variety of hydrophobic and hydrophilic drugs. [130][131][132] These polymeric NPs have been previously described as a delivery system for ibuprofen. This strategy has proven to increase the effects of ibuprofen on human gastric [133] and lung cancer cells. [134] Additionally, in vivo, the NP-mediated delivery allows for a controlled, pH-dependent, release of ibuprofen in the colon of Wistar rats. [126] Our results show that the effect of ibuprofen in the axonal growth of DRG neurons is more pronounced when the drug is delivered by PLGA NPs, which is in line with the conclusions from these studies.
The potential of PLGA NPs for drug delivery for treating neuropathic pain has been explored. The recent advances in nanomedical options include siRNA, miRNA, plasmid vector, or drugencapsulated PLGA NPs, which have shown significant analgesic effects. [127] Herein, we could appreciate that the ibuprofen-loaded PLGA NPs treatment reduced the inflammatory-derived axonal sprouting, possibly related to pain-associated behaviors, [25,135,136] by counteracting the inflammatory process in the bone unit sustained by the osteoclasts.
Microfluidics has been demonstrated as one of the most promising techniques to fabricate high-performance drug delivery systems with uniform morphology, size and distribution index, reduced variability, and controllable drug-delivering capacity. [137,138] On the other hand, microfluidics are extremely useful to test the drug delivery systems' functionality under physiologically/ pathologically relevant in vitro setups. [57] Blood-brain barrier microfluidic models were developed to assess the NPs transport across the endothelial cell layer. It was reported that NPs pass more easily through the barrier when combined with the peptides that enhance the adhesion of NPs to the endothelial Figure 7. Ibuprofen-loaded PLGA NP treatment to revert inflammatory profile at the micropathological chip. A) Physical characterization of size, polydispersion index, and NPs charge. B) Kinetics of ibuprofen drug release at pH 7.4 over 120 h. Ibuprofen-loaded NPs internalized by osteoclasts at the bone unit after being added to the system through the endothelial tubular-like structure. C) NPs in green, actin in red, and nuclei in blue. Scale bar: 100 μm. D) Axonal growth was assessed in the control condition and following treatment with E) free ibuprofen or F) ibuprofen-loaded PLGA NPs. NF200 in black. Scale bars: 500 μm. Axonal growth was quantified using AxoFluidic. [44] G) This software returns two values: the value, which is related to the extension of growth, and H) the A value, which represents the fraction of axons that effectively cross the microgrooves. Data represented as mean ± SD. layer, especially in flow conditions. [139] Using dynamic conditions Papademetriou et al. reported that shear flow impacted the binding and internalization of liposomes NPs by brain endothelial cells in microfluidic models. [140] Endothelial cells respond to increased flow shear with decreased ability to uptake the NPs. Therefore, modifying NPs surfaces with endothelial-cell-binding ligands partially restores uptake, suggesting that functionalizing NPs to bind to endothelial cells enables NPs to resist flow effects. [141] Tunable endothelial permeability was also evaluated using microfluidic devices under static conditions. It has been reported that diffusing of different sizes fluorescent polystyrene NPs across the HUVECs monolayer, grown in microfluidic devices, showed to be dependent on Angiopoietin 1 and cyclic adenosine monophosphate (cAMP) treatments, [142] and TNF-treatment enhanced NP translocation across the endothelial cell monolayer and polyester membrane. [142] NPs transport across endothelial barrier was evaluated in sophisticated microenvironments as tumor area bordered by branched endothelial vasculature to identify parameters beneficial for extravasation, tumor penetration, and cancer therapy. [141] The authors confirmed that NP size is a critical factor for passage across endothelial barrier and tumor accumulation, by comparing different diameter carboxylic acid-terminated NPs to reveal a higher degree of extravasation and tumor penetration with 40 nm NPs, relative to the larger counterparts (70 and 130 nm). [141] In our study, we observed extravasation through the endothelial cell barrier for 200 nm PLGA NPs. It is important to notice that our system was under inflammatory conditions that enhance the passage of the NPs to adjacent compartments by decreasing the permeability of the endothelial layer. Moreover, it is relevant to reinforce that our data were retrieved from a complex microfluidic system, comprising several cell types and secreted molecules, which so far has no similar models described.

Conclusion
We designed, optimized, and validated an advanced NVBchip model to address the intricate relationship between sensory neurons, endothelial cells, and osteoclasts, in bone physiological and inflammatory pathological conditions. For the first time, a microfluidic-based model was presented comprising fully differentiated mouse primary osteoclasts. Moreover, this is the first bone microenvironment model, including, not only the important vasculature contribution, but also the sensory innervation intimately connected to the pain perception, which will substantially impact the bone research field.
We demonstrated the functionality of the NVBchip model by characterizing each cell type compartment by the evaluation of cell-specific morphology and cellular markers. Bone unit was successfully established with the presence of differentiated multinucleated osteoclasts, sustaining a high pro-inflammatory response and pro-angiogenic profile when exposed to IL-1 . The VU was effectively achieved displaying the 3D tubular-like structure expressing CD31 during the culture time, while permeability studies allowed to confirm the integrity of the endothelial barrier. Last the functionality of the neuronal unit was accomplished by the axonal growth response to the inflammatory stimuli. Our model also proved to be a valuable tool to screen the effect of compounds or drug delivery systems on bone neurovascular system, supported by robust data retrieved from the quantitative measurements as permeability and axonal growth quantification in this platform.
There are some limitations associated to this particular setup, namely the use of cells from different species. The ideal is to obtain a fully-humanized NVBchip. The main challenge to overcome is related to the culture of human sensory neurons. One possibility would be to use human-induced pluripotent stem cells. Still, their culture in miniaturized platforms, alone or in coculture with non-neuronal cells, will require tight monitorization of their viability and phenotype stability.
We consider that we have transposed into a single chip the key cellular players and the cascade of events that are likely to occur in bone inflammatory pathologies. OA or RA, where inflammation and pain are chronically present, and where osteoclasts play major roles in bone degradation and pain triggered by medium acidification, can now be methodically addressed using this in vitro tool. This neurovascular bone model will be a useful platform to carry out studies such as validation and comparative therapeutic tests with the opportunity to perform the administration in the vascular compartment and assess the drug effects on axonal and bone cells functions.

Experimental Section
Production of Microfluidic Devices: The SU-8 master molds were produced at INESC following the described protocols [143,144] from which the microfluidic devices were reproduced with poly(dimethylsiloxane) (PDMS; SYLGARD 784 Silicone Elastomer). The cured PDMS molds were cut and separated from the master, and the media reservoirs were punched out with a 5 mm biopsy punch (KAI medical, BP-50F), while the compartment for the DRG explant alignment with the microgrooves was punched out with a 3 mm punch (KAI medical, BP-30F). To create the compartment for the seeding of osteoclasts, the two 5 mm punches were further united using an extra 6 mm punch (KAI medical, BP-60F). Following sterilization, the PDMS layer was mounted on top of the clean coverslip, slightly pressing both surfaces.
Cell Culture: All animal procedures were approved by the i3S ethics committee and by the Portuguese Agency for Animal Welfare (Direção Geral de Alimentação e Veterinária) in accordance with the EU Directive (2010/63/EU) and Portuguese Law (DL113/2013).
To evaluate the angiogenic potential of the substrates, cells were seeded on top of three different hydrogels (Fibrin, Collagen/Fibrin, and Matrigel) on Slide Angiogenesis (Ibidi, #81 506). Fibrin hydrogels were prepared by mixing a fibrinogen (12 mg mL −1 in tris-buffered saline (TBS); Sigma-Aldrich) solution with an equal volume of a solution of thrombin (4 U mL −1 ; Sigma-Aldrich), calcium chloride (CaCl 2 , 2.5 mm; Sigma-Aldrich), and aprotinin (10 μg mL −1 ; Sigma-Aldrich) in TBS. Collagen/Fibrin (Col/Fib) [38,72] hydrogels were prepared by mixing collagen type I (diluted according to the manufacturer instructions, to a final concentration of 3 mg mL −1 ; CORNING) and fibrinogen (4.5 mg mL −1 in TBS) at a mass ratio of 40:60, reaching a final concentration of 1.875 mg mL −1 . The collagen/fibrinogen solution was then mixed 1:1 with a 4 U mL −1 solution of thrombin in M199 medium Matrigel Matrix (CORNING) was mixed with an equal volume of M199 medium.
Prior to cell seeding, hydrogels were plated on the Slide and polymerized for 30 min at 37°C. Cells were seeded on top of the hydrogels with a density of 8 × 10 4 cells cm −2 and the culture was maintained for 48 h. Tube formation was assessed at 24 h and at 48 h by immunostaining against filamentous actin (F-actin).
3D Cell Culture: Cells were harvested as previously described, and resuspended in the thrombin solutions used for the preparation of each hydrogel, reaching a final density of 10 × 10 6 or 20 × 10 6 cells mL −1 . Fibrin and Col/Fib hydrogels were prepared as described previously using the cell-laden thrombin solutions and seeded in the Slide Angiogenesis. After polymerization, EGM-2 medium was added, and cells were incubated at 37°C and 5% CO 2 . Medium was changed after 24 h of culture and at 36 h a viability assay was performed, by staining the cells with Calcein AM (Invitrogen) and propidium iodide (PI, Sigma).
For the highest cell density, viability and morphology were assessed at 24, 48, 72, and 96 h; metabolic activity was also assessed by the resazurin assay. Briefly, cells were incubated for 4 h at 37°C and 5% CO 2 with complete EGM-2 medium supplemented with 10% resazurin (Sigma) and fluorescence was measured with a SynergyTM Mx Microplate Reader (BioTek), at an excitation wavelength of 530 nm and an emission wavelength of 590 nm.
Seeding on the Microfluidic Channels (Vascular Unit -VU): Upon reaching confluency, HUVEC were harvested and embedded within Col/Fib hydrogels, by resuspending the cell pellet in the thrombin solution prior to its mixing with the collagen/fibrinogen, reaching a final density of 20 × 10 6 cells mL −1 . The central channel was filled with the hydrogel and incubated for 10 min at 37°C before adding EGM-2 to the medium reservoirs.
Bone Unit Optimization: Coverslips were coated with either poly-D-Lysine (PDL, Sigma) or Collagen I (Sigma). PDL was diluted to 0.1 mg mL −1 in filtered dH 2 O and collagen I was diluted to 0.01% w/v in acetic acid (CH 3 COOH) 0.1m and 10 μg cm −2 were used to perform the coatings, for 5 h at 37°C, and then removed. Previously to cell seeding, the wells or the coverslips coated with PDL were washed with abundant dH 2 O; collagen-coated ones were let dry at RT.
On day 4 after isolation of bone marrow stromal cells, murine osteoclasts (mOC) were detached using a cell scraper. After centrifugation for 5 min at 1600 rpm, cells were resuspended in -MEM complete medium, supplemented with normal (30 ng mL −1 M-CSF and 100 ng mL −1 receptor activator of nuclear factor kappa-B ligand (RANKL, PeproTech)), or increased (50 ng mL −1 M-CSF and 120 ng mL −1 RANKL) concentrations of cytokines and cells were seeded with a density of 5 × 10 4 cells cm −2 (open devices) or of 7.5 × 10 4 cells per channel (closed microfluidic). To establish pro-inflammatory conditions, culture medium was supplemented with 0.5 ng mL −1 interleukin 1 (IL-1 , ImmunoTools). [86] On day 1 and day 2 of culture, fresh M-CSF and RANKL were added in the same concentrations as day 0, without exchanging the media. On day 3, medium was replenished, and the culture was maintained until day 4.
Cytokines Array: Osteoclasts were cultured in the presence or absence of IL-1 in a 48 well plate. At day 3 of culture, medium was exchanged and at day 4 (after 24 h) conditioned media from both conditions were collected. Conditioned media from 5 independent experiments (n = 5) were pooled and the Mouse Angiogenesis Array C1 (Ray Biotech) was used to detect a panel of 24 secreted cytokines and chemokines, according to the manufacturer's protocol. The spot signal densities of array membrane were scanned using the ChemiDoc Imaging System (BioRad). The raw numerical densitometry data were extracted from background. The data of samples were normalized to the positive control signals. The signal intensity for each antigen-specific antibody spot was proportional to the relative concentration of the antigen in that sample.

Co-Culture Establishment -Vascular and Bone Units (Physiological V-BU and Inflammatory V-IU):
To establish co-cultures of endothelial cells and osteoclasts within the microfluidic platform, the first step was to seal the central channel with the cell-laden hydrogel, as previously described. The EGM-2 medium was replenished after 24 h. On day 2, the collagen coating was performed on the osteoclast compartment for 5 h at 37°C, prior to seeding of the osteoclasts in the presence (V-IU) or absence (V-BU) of 0.5 ng mL −1 IL-1 . On days 3 and 4, cytokines were added to the osteoclast compartment, in the same concentrations and without changing the media. On day 5, both EGM-2 and -MEM were exchanged. On day 6, co-cultures were used for the permeability assays or fixed for immunostaining.
Permeability Assay: Permeability studies were performed on day 6 of the co-cultures, corresponding to day 4 of osteoclast differentiation. BioTek), with an excitation wavelength of 480 nm and an emission wavelength of 520 nm.

Triculture Establishment -Vascular-Nerve and Bone Units (Physiological VN-BU and Inflammatory VN-IU):
The triculture was assembled in a step-by-step manner, establishing timings that would allow its maintenance until full development of all the cell types involved.
The first step was to seed the endothelial cells in the central channel, as previously described. Following this, in the somal side, laminin (Sigma) 5 μg mL −1 in complete neurobasal medium was added and incubated overnight at 37°C. On the next day, the neurobasal medium was changed to a new medium also supplemented with 50 ng mL −1 7S NGF (Calbiochem, Merck Millipore) and the EGM-2 medium was exchanged. Embryonic DRG were seeded in the respective compartment, and their adhesion to the bottom of the well was confirmed. The next day, collagen coating and osteoclast's seeding were performed, as previously described, and EGM-2 medium was exchanged. Fresh M-CSF and RANKL were added on the two days following osteoclasts' seeding. The next day, all media were replenished and the triculture was maintained for 24 more hours.
Ibuprofen-Loaded PLGA Nanoparticles Treatment: Using PLGA-NPs as nanocarriers, ibuprofen (Sigma) was delivered to the inflammatory triculture model to assess their anti-inflammatory potential. On day 6 of the triculture, when inflammatory osteoclasts were already fully differentiated, ibuprofen-loaded PLGA-NPs were added to the VU to a final concentration of ibuprofen of 15 μg mL −1 . Free ibuprofen at the same concentration was used as a control. On day 7, the culture was fixed and stained against NF200 to assess the axonal growth.
PLGA NPs with a 20% theoretical drug loading were produced by the nanoprecipitation method, as previously described. [126] Briefly, the organic phase (PLGA in acetone) was added to the aqueous phase (0.5% Poloxamer 407 in water) under stirring and the NPs were afterwards purified through subsequent washes using Amicon filters. The physicochemical characterization regarding PLGA NPs size, polydispersity index (PdI), and zeta potential, using a Dynamic Light Scattering (DLS, Zetasizer, Malvern Instruments) was performed in 10 mm NaCl, at pH 6.0. Association efficiency and drug loading were quantified using the direct and indirect methods. Ibuprofen payload from fresh PLGA NPs (suspension) was evaluated indirectly from free ibuprofen isolated by centrifugation with Amicon filters. For the direct method, Ibuprofen-loaded PLGA NPs were lyophilized and then reconstituted by the addition of dimethylformamide for total NPs disintegration. Free ibuprofen from both methods was quantified by a validated High-Performance Liquid Chromatography (HPLC) method. Ibuprofen release from the PLGA NPs was addressed in PBS, at pH 7.4 under agitation (100 rpm), 37°C. At different time points (0, 4, 24, 48, 72, and 168 h), samples were collected and centrifuged for 20 min at 10 000 rpm (9600g), the free drug in the supernatant was quantified by HPLC.
Tartrate Resistant Acid Phosphatase (TRAP) Staining: Osteoclasts were stained for TRAP using Acid Phosphatase, Leukocyte (TRAP) Kit (Sigma-Aldrich), according to the manufacturer's instructions. Briefly, cells were fixed with citrate/acetone for 30 s, washed with dH 2 O, and allowed to dry. Then, cells were incubated for one hour at 37°C with TRAP solution (Acetate, Naphtol AS-BI Phosphoric Acid, Tartrate Solution, and Fast Garner GBC Salt in water), and afterwards washed with dH 2 O for 3 min and allowed to dry. Images were acquired using a Axiovert 200M (Zeiss) inverted fluorescence microscope.
Quantification of Osteoclast Differentiation: Images were acquired using an Axiovert 200M (Zeiss) inverted fluorescence microscope (cells seeded in well-plates) or a laser scanning confocal microscope Leica SP5 (cells seeded in microfluidic devices). Using the Cell Counter plugin of ImageJ, cells in each image were counted into three different groups (less than 3 nuclei, 3 to 9 nuclei, and 10 or more nuclei) and the percentage of cells in each one was calculated. Additionally, the mature osteoclasts were manually selected using the freehand selection tool and the area was measured, to calculate the percentage of total area occupied by these cells.
Quantification of Axonal Growth: Mosaic images were acquired using a laser scanning confocal microscope Leica SP5. The 8-bit images of the channel corresponding to the neuronal marker were used to quantify axonal growth using AxoFluidic, a software previously developed by the group, [44] which creates a spatial profile of axonal outgrowth and analyzes neurite length in the axonal side of the microfluidic platform.
Statistical Analysis: Data were presented as mean ± S.D. (standard deviation). Analysis of data was performed by one-way ANOVA followed by Kruskal-Wallis test, or two-way ANOVA followed by Geisser-Greenhouse correction, using GraphPad Prism 8.4.3 for Windows. Differences between groups were considered statistically significant when p<0.05.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.