Long-term in vitro culture of 3D brain tissue model based on chitosan thermogel

Methods for studying brain function and disease heavily rely on in vivo animal models, ex-vivo tissue slices, and 2D cell culture platforms. These methods all have limitations that significantly impact the clinical translatability of results. Consequently, models able to better recapitulate some aspects of in vivo human brain are needed as additional preclinical tools. In this context, 3D hydrogel-based in vitro models of the brain are considered promising tools. To create a 3D brain-on-a-chip model, a hydrogel capable of sustaining neuronal maturation over extended culture periods is required. Among biopolymeric hydrogels, chitosan-β-glycerophosphate (CHITO-β-GP) thermogels have demonstrated their versatility and applicability in the biomedical field over the years. In this study, we investigated the ability of this thermogel to encapsulate neuronal cells and support the functional maturation of a 3D neuronal network in long-term cultures. To the best of our knowledge, we demonstrated for the first time that CHITO-β-GP thermogel possesses optimal characteristics for promoting neuronal growth and the development of an electrophysiologically functional neuronal network derived from both primary rat neurons and neurons differentiated from human induced pluripotent stem cells (h-iPSCs) co-cultured with astrocytes. Specifically, two different formulations were firstly characterized by rheological, mechanical and injectability tests. Primary nervous cells and neurons differentiated from h-iPSCs were embedded into the two thermogel formulations. The 3D cultures were then deeply characterized by immunocytochemistry, confocal microscopy, and electrophysiological recordings, employing both 2D and 3D micro-electrode arrays. The thermogels supported the long-term culture of neuronal networks for up to 100 d. In conclusion, CHITO-β-GP thermogels exhibit excellent mechanical properties, stability over time under culture conditions, and bioactivity toward nervous cells. Therefore, they are excellent candidates as artificial extracellular matrices in brain-on-a-chip models, with applications in neurodegenerative disease modeling, drug screening, and neurotoxicity evaluation.


Introduction
The study of neuronal development and plasticity, neurological disorders and drug discovery is a highly challenging field of research.This is primarily due to the intricate complexity of the central nervous system, particularly the brain.The brain can be viewed as a complex assembly of different tissues that work in coordination [1].Each brain tissue has a unique architecture with specific cellular phenotypes and extracellular matrices (ECMs) [2].Traditionally, the investigation of brain function and diseases has relied on imaging techniques in living humans, ex-vivo analysis of brain tissues, in vivo studies in animal models, and in vitro experiments using 2D cell cultures [3][4][5].While these techniques have contributed to key discoveries in neuroscience, they all have well-known limitations, including limited spatial and temporal resolution in imaging, cell death in brain slices due to axotomy, interspecies variation in animal models, and the oversimplification of 2D cell cultures [6][7][8].Consequently, these limitations have hindered the translatability of preclinical results to clinical studies in drug discovery [9].In this context, the advent of human induced pluripotent stem cells (h-iPSCs) represents an important step toward human in vitro models and has led to the development of brain organoids and scaffold-based 3D neural tissues [10,11].These models, although simplified in comparison with their in vivo counterparts, have been demonstrated to be able to recapitulate basic interactions between nervous cells and the surrounding microenvironment, both in physiological and pathophysiological conditions [12].Organoids exhibit selforganization and complex architecture that mimics neuronal development but suffer from limited nutrient and oxygen diffusion, resulting in necrosis and a limited lifespan [13].
On the other hand, scaffold-based 3D models offer advantages in terms of tunability, modularity, and engineering capabilities.However, a significant challenge in this field is identifying materials that can effectively mimic the brain's ECM.These materials need to possess mechanical properties and biochemical cues that support the functional maturation of 3D neuronal networks [14,15].Consequently, there is a need for a material that can support long-term cultures, especially those based on 3D neuronal networks derived from h-iPSCs, as achieving functional maturation of induced neurons (iNeurons) in 3D settings requires a relatively long time (6 weeks or more) [16,17], necessitating thus a stable biomaterial [18].Hydrogels are widely used biomaterials for mimicking the brain's ECM [19,20].They consist of crosslinked hydrophilic polymer chains that create 3D porous networks [21][22][23].Both synthetic and natural polymers can be used to form hydrogels through chemical or physical cross-linking strategies.Natural polymers offer advantages such as biocompatibility, biodegradability, and, to some extent, bioactivity [24,25].In terms of cross-linking strategies, physical cross-linking is preferred over chemical cross-linking, which often involves cytotoxic cross-linking agents [26].Furthermore, physical cross-linking allows the sol-to-gel transition to occur in the presence of cells, enabling their incorporation into the 3D matrix.
To date, Matrigel, a commercially available tumor-derived basement-membrane-like matrix, is the most used natural hydrogel for encapsulating neuronal cells [27].However, Matrigel exhibits batchto-batch variability, contains tumor-derived proteins, and lacks stability in long-term cultures.In the search for alternative natural hydrogel matrices, various approaches have been evaluated and proposed [28][29][30].One approach involves using decellularized brain ECM, which offers a physiologically relevant environment for 3D cell cultures [31][32][33].However, decellularized ECM has limitations, including low mechanical strength, rapid degradation, and ethical concerns due to its animal origin.Another approach focuses on using ECM-derived polysaccharides, proteins, or combinations thereof to reduce complexity and create matrices with well-defined and adjustable properties [34].
Recently, we have demonstrated that the polysaccharide chitosan (CHITO) by itself is able to support neuronal adhesion, growth and functional maturation in 2D cultures derived from both primary neurons and iNeurons co-cultured with astrocytes [35,36].We also demonstrated that the same cell phenotypes seeded onto a granular CHITO scaffold can develop 3D highly interconnected and functional networks, up to 21 and 42 d in vitro respectively [17,37,38].
CHITO, obtained through the deacetylation of chitin, is a cationic polysaccharide known for its favorable properties, including biocompatibility, biodegradability, as well as antimicrobial and antifungal properties [39].Additionally, CHITO shares structural similarities with various extracellular glycosaminoglycans and performs similar morphogenetic functions [40,41].
Most interestingly, Chenite et al [42] demonstrated that a CHITO solution in the presence of β-glycerophosphate salt (β-GP) remains in a liquid state at room temperature at physiological pH (7.0-7.4) and gels at 37 • C [43].Since then, CHITOβ-GP thermogels have been extensively studied for delivering therapeutic biomolecules and encapsulating living cells [43].In this respect, while thermogels have been widely explored for applications such as bone and cartilage regeneration [44,45].Their use in nervous tissue engineering has been relatively limited.Some studies have evaluated their potential for regenerating spinal cord injuries and peripheral nerves [46][47][48][49].For instance, Bhuiyan et al [48] recently investigated the application of CHI-β-GP thermogel for injectable delivery of stem cells or drug candidates aimed at enhancing brain tissue regeneration following injury.The in vitro tests were carried out using PC12 cells, an immortalized rat derived cell line with neuronal cell-like properties, which is poorly representative of healthy neuronal cell behavior.
In the present work, we investigated the ability of CHITO-β-GP thermogels to sustain the functional maturation and the long-term culture of 3D neuronal networks, derived both from co-cultures of primary rat cortical neurons and astrocytes and of cortical iNeurons and astrocytes.
To this purpose we selected [42,50] two thermogel formulations, differing in the gelling kinetics.The gelling kinetics were investigated by rheological characterization and the elastic modulus was evaluated for both formulations both for the plain hydrogels and for the hydrogels encapsulating cells.Injection tests were performed both without and with cells and the morphology of the plain hydrogels was characterized by scanning electron microscopy (SEM).The two formulations were then used to carry out the two cocultures mentioned above.Then immunocytochemistry and confocal imaging were used to characterize cellular distribution and vitality and to evaluate the formation of the 3D neuronal networks.Finally, the electrophysiological activity of the 3D networks derived from iNeurons was characterized by microelectrode arrays (MEAs), with both planar electrodes and 3D electrodes.
Overall, the results demonstrated that CHITO-β-GP thermogels can sustain neuronal growth, maturation and the development of dense 3D functional networks in long-term cultures.

Chitosan purification
CHITO was purified using a method based on the procedure outlined by Qian and Glanville [51], with several modifications.Specifically, 5 g of CHITO powder were dissolved in 500 ml of 0.1M hydrochloric acid through overnight stirring at 40 • C. The resulting acidic solution was then filtered under vacuum to eliminate insoluble particles.CHITO was subsequently precipitated by adding 5M NaOH while maintaining continuous stirring, maintaining the pH in the range of 8-9.Following the precipitation step, 5 ml of 10% (w/v) SDS was introduced into the slurry, and the mixture was heated at 95 • C for 5 min.After cooling to room temperature, the pH was adjusted to 10 using 0.5M NaOH.The resulting slurry was filtered under vacuum, and the hydrated CHITO was thoroughly washed five times with 500 ml of Milli-Q water, with the washes performed at 40 • C.

Preparation of CHITO thermogels
Two different CHITO formulations, namely ChiGel 1 and ChiGel 2, were prepared by dissolving 2.5% and 3.33% (w/v) purified CHITO powder in 0.1M acetic acid, respectively.The solutions were stirred at room temperature overnight and then sterilized by autoclaving at 120 • C for 20 min.Subsequently, they were stored at 4 • C. As for the gelling agent, β-GP was used at two different concentrations, 10% and 20% in water, and it was sterilized by filtration through 0.2 µm syringe filters, followed by storage at 4 • C. To create the thermogels, the sterilized CHITO solutions were mixed with the sterilized gelling agent solutions, added drop by drop, with a volume ratio of 0.8:0.2 for ChiGel 1 and 0.6:0.4 for ChiGel 2. All the hydrogels had a final concentration of 2% (w/v) CHITO.Subsequently, 30 µl of the mixed solution was dispensed into a circular polydimethylsiloxane (PDMS) mold.The PDMS mask served as confinement for the thermogel solution, giving it a circular shape after gelification.These confinement structures had a maximum height of 1 mm and an inner diameter of either 5 mm or 3 mm, and they were fabricated using a standard 1:10 ratio between curing agent and elastomer.

Rheological characterization
Rheological properties were assessed using an Anton Paar instrument (Physica MCR 301, Austria) equipped with a cone-plate geometry comprising a 50 mm diameter plate set at a 2 • cone angle with a truncation at 210 µm.The evolution of the storage modulus (G ′ ), loss modulus (G ′′ ), and complex viscosity of the samples was measured at a constant shear strain (5%) and constant frequency (1 Hz) immediately after the preparation of each hydrogel.The gelation kinetics were monitored isothermally at 37 • C for 40 min.Furthermore, to examine the thermosensitive characteristics of the hydrogels, rheological parameters were recorded during temperature ramps ranging from 4 • C to 40 • C at a rate of 1 • C min −1 and during isotherms maintained at 37 • C.

Dynamic mechanical analysis
Indentation tests were performed using a custombuilt dynamic mechanical analysis (DMA) apparatus.This apparatus consisted of a mini-shaker operating in a range between 1 Hz and 10 kHz with maximum force of 1.5 N, a laser vibrometer to measure the displacement, set at 80 µm V −1 and a force transducer.All the devices were connected to a PC through an NI acquisition card and the signals were elaborated by mean a LabView program.
The sample used had a cylindrical shape, measuring 5 mm in diameter and 3 mm in height.It was positioned vertically on a plate connected to the shaker.A cylindrical indenter, also with a diameter of 5 mm, was then brought into contact with the upper surface of the sample, followed by the application of a pre-strain.Specifically, a 10% strain was applied to assess the stiffness of the external layers and the system's response under low stress conditions, while a 20% strain was employed to evaluate the internal bulk properties without inducing damage to the internal structure.
During the test, oscillations occurred within the frequency range of 2 Hz to 100 Hz.The sinusoidal signal for displacement and force, representing the response and stimulus, respectively, was processed by the software to generate a spectrum containing the storage modulus (E ′ ) and the loss modulus (E ′′ ) vs. the frequency.

Injection tests
The injectability of both types of thermogels was initially assessed manually by passing them through an 18 G needle (1.27 mm in diameter) and then quantitatively by measuring the maximum force required to extrude the hydrogels using the ElectroForce 3200 instrument from Bose Corporation, USA, which was equipped with a 200 N load cell.
Immediately after mixing CHITO and β-GP, the resulting solution was loaded into a 1 ml syringe and allowed to sit for 5 min at 37 • C. Subsequently, the force needed to extrude the pre-gel was monitored, and the highest force recorded for the various formulations was compared.Additionally, the morphology of the extruded material, including aspects such as continuous threads, segments, droplets, etc, was observed.Different needle sizes were tested, including 20 G (910 µm), 21 G (810 µm), 23 G (610 µm), and 25 G (510 µm), to assess their impact on injectability.

Preliminary printability tests
3D bioprinting was carried out with the BIO X pneumatic-driven extrusion bioprinter from CELLINK.Each of the experiments were performed a 3 ml cartridge equipped with a standard sterile 25 G conical nozzle.The cartridge, loaded with ChiGel 2, was place in incubator at 37 • for 30 min to induce pre-gelification; the cartridge temperature was maintained at 37 • C, and the material flow was carefully controlled by maintaining a pressure between 20 and 30 kPa, while the printing speed was set at 5 mm s −1 .

Morphological analysis by SEM
The thermogels were prepared in a PDMS mold with a diameter of 5 mm and a thickness of 500 µm, and this process was carried out in duplicate.Subsequently, the gels were allowed to sit for 24 h at 37 • C to ensure complete gelation.Afterward, the thermogels were kept at −20 • C overnight and freeze-dried under vacuum for 24 h.Small sections of each sample were carefully cut, and then they were sputter-coated with a thin layer of gold, approximately 30 nm in thickness, using the Polaron sputter coater from Thermo VG Scientific, located in East Grinstead, UK.Following this preparation, images were captured at magnifications of ×200 and ×1.00k using a Hitachi S-2500 instrument from Hitachi in Chiyoda, Tokyo, Japan.These images were taken in the secondary electron mode at 10 KV.

Cell cultures 2.2.7.1. Primary cells cultures
Cortices were dissected and extracted from embryonic Sprague-Dawley rats on gestational day 18, following sterile procedures.The fetal cortical tissue was enzymatically digested using 0.125% Trypsin in Ca ++ and Mg ++ free Hank's buffer for a duration of 20 min at 37 • C. Subsequently, neurons were suspended in a plating medium composed of neurobasal medium with 2% B-27 supplement, 1% glutamax, and 1% penicillin-streptomycin. Cultures were maintained in an incubator at 37 • C in an atmosphere containing 5% CO 2 and 95% humidity for a duration of 4 weeks, with half of the medium being replaced once a week to sustain the cultures.The experimental protocol was conducted in accordance with the guidelines set forth by the European Animal Care Legislation (2010/63/EU), as well as the approval of the Italian Ministry of Health, following the provisions of DL 116/1992.Additionally, the research adhered to the guidelines of the University of Genova, with a focus on reducing the number of animals used for the project and minimizing any potential suffering.

h-iPSCs and neuronal differentiation
An rtTa/neurogenin 2 (NgN2) positive cell line of h-iPSCs was generated from fibroblasts obtained from a healthy donor, which were kindly provided by the Donders Institute for Brain, Cognition, and Behaviour at Radboudumc in Nijmegen, Netherlands.The stability of these cell lines was achieved through lentiviral transduction and selection using G418 and puromycin, as detailed in the study by Cooke and Rosenzweig [52].Upon thawing, the cells were immediately seeded into six-well plates that had been pre-coated with Matrigel.Essential 8 Flex medium supplemented with 1% penicillinstreptomycin, 50 µg ml −1 G418, and 0.5 µg ml −1 puromycin was used as the culture medium.The entire medium was replaced every 2 d to maintain cell health and pluripotency.Pluripotency was sustained by splitting the cultured h-iPSCs twice a week using ReLeSR, and the plates were stored in an incubator at 37 • C with a 5.5% CO 2 atmosphere.
Early-stage neurons were generated by inducing the overexpression of the neuronal determinant NgN2 in h-iPSCs through a 3 day treatment with doxycycline.These neurons were then co-cultured with astrocytes, a process previously detailed in the study conducted by Muzzi et al in 2021 [17].

Astrocytes
Cortical astrocytes were obtained from frozen tissue derived from rat embryos at embryonic day 18 (E18), and these samples were generously provided by the Department of Experimental Medicineat the University of Genoa, Italy [53].Upon thawing, the astrocytes were seeded into T-75 flasks containing Dulbecco ′ s Modified Eagle Medium (DMEM) highglucose supplemented with 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin.These flasks were maintained in an incubator at 37 • C with 5% CO 2 , and the medium was refreshed every 3 d.For co-culture, astrocytes were used in a 1:1 ratio with iNeurons.The experimental protocol adhered to the guidelines set forth by the European Animal Care Legislation (2010/63/EU) and was approved by the Italian Ministry of Health, in accordance with DL 116/1992.The research also followed the guidelines of the University of Genova, with a strong emphasis on reducing the number of animals used in the project and minimizing any potential suffering.

Preparation of CHITO thermogels for cell encapsulation
Primary cells and iNeurons/astrocytes were suspended and thoroughly mixed using a positivedisplacement pipette directly within the ChiGel 1 and ChiGel 2 solutions.The cell density employed was 6.6 × 10 6 cells ml −1 (as illustrated in figure 1).Subsequently, 30 µl of the cell/ChiGel solution mixture was carefully dispensed into PDMS molds, which had been previously positioned on standard petri dishes (35 mm) for subsequent immunocytochemistry characterization or directly onto the surface of the MEA for electrophysiological characterization.All the samples were placed in incubator at 37 • C and after 30 min, the complete neurobasal medium was added.All hydrogels had a final CHITO concentration of 2% (w/v).

Morphological characterization of 3D neuronal networks 2.2.9.1. Immunocytochemistry
Immunocytochemistry was carried out to validate both CHITO thermogels for primary and humanderived cells, with the aim of assessing cell morphology, spatial distribution, and the development of 3D networks.For both cell populations (primary cells and iNeurons), 3D cell cultures were fixed in 4% paraformaldehyde in phosphate buffer solution (PBS), pH 7.4 for 30 min at room temperature.Permeabilization was achieved with PBS containing 0.3% Triton-X100 for 15 min at room temperature and non-specific binding of antibodies was blocked with an incubation of 1 h 30 min in a blocking buffer solution consisting of PBS, 0.3% BSA and 0.5% FBS.Cultures were incubated with primary antibody diluted in PBS blocking buffer at 4 • C overnight in a humidified atmosphere.Cultures were rinsed three times with PBS and finally exposed to secondary antibodies.The following primary antibodies were used: microtubule-associated protein-2 1:500 (MAP-2, polyclonal Synaptic System), glial fibrillary acidic protein 1:500 monoclonal antibodies (GFAP, Sigma) and 4 ′ ,6-diamidino-2-phenylindole 1:10 000 (DAPI, Sigma).Cultures were rinsed twice with PBS and finally exposed to the secondary antibodies: Alexa Fluor 488, Alexa Fluor 549, Goat anti-mouse or goat anti-rabbit, diluted 1:700 and 1:1000 (Invitrogen Life Technologies S. Donato Milanese).For live neuron imaging, 3D cell cultures were exposed for 3 h to NeuroFluor™ NeuO (0.25 µM).

Optical microscopy and confocal imaging
Immunofluorescence evaluation was performed using an Olympus BX-51 upright microscope, and image acquisition was carried out with a Hamamatsu Orca ER II digital cooled chargecoupled device(CCD) camera, which was controlled using Image ProPlus software developed by Media Cybernetic.
Confocal imaging was conducted on Leica TCS SP5 AOBS Tandem DMI6000 Inverted coupled with a Leica IRAPO 25X objective, 0.95 NA.The Leica TCS SP5 AOBS Tandem microscope is provided by Leica Microsystems, located in Mannheim, Germany.Data were analyzed using LASX V2.0 software, developed by Leica Microsystems Srl in Italy.Additionally, the image-processing package Fiji was used for further analysis of the acquired data.These microscopy and imaging systems enabled the comprehensive evaluation of immunofluorescence and confocal imaging, facilitating the study's objectives in visualizing and analyzing cellular structures and interactions.

Functional characterization of 3D neuronal networks 2.2.10.1. MEA recording and analysis
Standard MEA60 and 3D MEA60 (MultiChannel System) were used to record the electrophysiological spontaneous activity of neuronal network at different days in vitro (DIVs).MEA60 are glass devices in which 60 titanium nitride electrodes are embedded at the center of the culture well.They are arranged in an 8 × 8 grid without electrodes at the corner, spaced 200 µm among them with 30 µm diameter, generating an active area of 1.6 mm × 1.6 mm.The 3D MEA60 have the same spatial organization of the electrodes, but in this case the electrodes are pyramidal (they are 100 µm high and have a tip with a diameter of 12 µm) and are 250 µm spaced among them.Devices were sterilized in the hoven at 120 • C for 2 h.Sterilized PDMS mask (inner diameter = 3 mm; thickness = 1 mm) were placed on the MEA to confine the 3D neuronal network.
The electrophysiological activity was acquired with the 2100 System (MEA 2100-System, MCS) and signals were sampled at 10 kHz.During the recording, sterility was maintained using an autoclavable cap on top of the MEA chamber while normal conditions were maintained by pre-heating the head stage at 37 • C and by delivering humidified air with 5.5% CO 2 over the MEA thanks to a plastic box placed on top of the recording system.Recordings lasted 10 min and data were analyzed offline using MATLAB (The MathWorks) scripts.

Data and statistical analysis
Raw data were first subjected to high-pass filtering (cutoff frequency = 200 Hz) then to spike detection and finally analyzed as point processes.Precise time spike detection algorithm was implemented in MATLAB as detailed in Maccione et al [51].Spike were extracted if the peak was five times higher than the noise's standard deviation and the peak-to-peak amplitude was six times the noise's standard deviation, considering 1 ms of peak lifetime and refractory period.Array wide firing rate was evaluated by dividing the recording in 100 ms bin and counting all the spikes detected in each bin.The mean firing rate (MFR) was evaluated as the sum of all spikes recorded in one electrode divided by the recording time.We defined as active an electrode that has MFR > 0.1 spk s −1 burst duration (BD) was obtained by calculating the duration (ms) of each burst.The percentage of random spikes was evaluated as the total number of spikes in the culture that does not belong to a burst.We defined network burst (NB) as synchronous activity among channels [17].We imposed the condition that at least 10% of the active electrode must fire with an ISI ⩽ 100 ms.

Statistical analysis
The data are presented as mean ± SD.The comparison of means by one-way ANOVA followed by Tukey test was done using Graph-Pad InStat software ( * p < 0.05).Electrophysiological statistical analysis was carried out using OriginPro 8 (OriginLab Corporation, Northampton, MA, USA).All data are presented as mean ± standard error of the mean.Statistical analysis was performed using a non-parametric Kruskal-Wallis's test, since data do not follow a normal distribution (evaluated by the Kolmogorov-Smirnov normality test).Differences were considered statistically significant when p < 10 −3 .In order to determine which of the sample pairs are significantly different, post-hoc test, using Dunn's test, has been applied.

Rheological and injectability characterization
Rheological analysis was carried out to investigate the sol-gel transition behaviors of ChiGel 1 and ChiGel 2. ChiGel 1 exhibited slower gelling kinetics compared to ChiGel 2. This difference was evident in the trend of the storage modulus (G ′ ) over time during the isotherm test at 37 • C, particularly for samples stored at 4 • C for different timepoints until 30 d (figures 2(A) and (B)).In contrast, ChiGel 2 showed higher initial slopes in the curves at different time points (0 min, 5 min, and 24 h), and there were no significant differences in G ′ values between samples stored at 4 • C for up to 24 h (figure 2(B)).This behavior was further confirmed by the gradual increase in G ′ (storage modulus) for the samples stored at 4 • C (figure 2(A)).Furthermore, the stiffness of ChiGel 1 after 30 d was slightly higher (8 kPa) than that of ChiGel 2 (5 kPa).
To further investigate the thermosensitive properties, temperature ramps were performed.Both Then the injectability of both ChiGel 1 and ChiGel 2 pre-gels was evaluated.The results indicated that only ChiGel 2 pre-gel was suitable for proper extrusion, while ChiGel 1 pre-gel exhibited noncontinuous flow and did not gel during the extrusion process.Specifically, the injectability of pre-gels (ChiGel 1 and ChiGel 2) was further characterized based on injection speed (figure 2(D)) and needle size (figure 2(E)).By increasing the injection speed, the maximal force required for extrusion also increased.The required maximal force fell within a range of 1.8-3N, depending on the injection speed, figure 2(D).This behavior suggests that the rate of injection has a direct impact on the force required to extrude the pregels.Moreover, by decreasing the size of the needle, the maximal force required for extrusion increased.For both formulations, the increase in maximal force was observed over a range of 2.5-6N, depending on the needle size.This indicates that the diameter of the needle has a significant influence on the force needed for extrusion, figure 2(E).ChiGel 2 pre-gels exhibited cohesive behavior, forming a linear and continuous filament during injection into a saline bath at 37 • C.This indicates good extrudability and cohesiveness of ChiGel 2 (supplementary movie 1).In contrast, ChiGel 1 pre-gels produced discontinuous segments during extrusion, which may suggest a less cohesive network within ChiGel 1.These results highlight the importance of injection speed and needle size in the extrusion of pre-gels, with ChiGel 2 demonstrating better extrudability and cohesiveness compared to ChiGel 1.The cohesiveness of the pre-gel is a critical factor to ensure the successful delivery of materials in various biomedical applications.Nevertheless, it is important to note that injectability does not always guarantee the material's suitability for the printing process by extrusion [52].To evaluate the printability of ChiGel 2, we printed different geometries as shown in supplementary figure S1.
These findings highlight the distinct rheological and thermosensitive characteristics of ChiGel 1 and ChiGel 2, with ChiGel 2 demonstrating faster gelling kinetics, greater responsiveness to temperature changes, and suitability for injectability.These properties are crucial considerations for their application in various biomedical and tissue engineering contexts such as in vivo applications or as bioink.

DMA and SEM characterization
The results of the rheological tests were also supported by DMA characterization (figures 3(A)-(D)).As a first step, the mechanical properties of the thermogels without cells were evaluated (gray bars).From one hand, the elastic modulus E ′ was found to be around 6 kPa, with a pre-strain of 10%, for both the thermogels stored for 24 h at 37 • C, (figures 3(A) and (B), gray bars).From the other, E ′ was found to be 11.3 ± 2.6 kPa for ChiGel 1 and 9 ± 2,8 kPa for ChiGel 2 with a pre-strain of 20%, (figures 3(C) and (D), gray bars).The 20% pre-strain was applied in order to take into account the contribution of the internal stiffness, that relates to the response of the inner layers of the network.After 20 d of storage at 37 • C, E ′ for ChiGel 1 increased up to 17.7 ± 2.7 kPa and 28.8 ± 2.40 kPa with a prestrain of 10% and 20% respectively (figures 3(A) and (C), gray bars), whereas E ′ remained constant for ChiGel 2, (figures 3(B) and (D), gray bars).As second step, the mechanical properties of the thermogels loaded with cells were evaluated (violet bars).As relates to ChiGel 1, E ′ was found to be lower respect to the one without cells both at DIV 1 and DIV 20.Namely, at DIV 1 E ′ was 3.7 ± 0.9 kPa and 6.7 ± 0.9 kPa with a pre-strain of 10% and 20% respectively and at DIV 20 E ′ was 14.9 ± 1.5 kPa and 19.3 ± 0.6 kPa with a pre-strain of 10% and 20% respectively, (figures 3(A) and (C), violet bars).As relates to ChiGel 2, at DIV 1 no difference respect to the thermogels without cells was observed in E ′ , being 6 ± 2 kPa and 10.7 ± 1.9 kPa with a pre-strain of 10% and 20% respectively, (figures 3(B) and (D), violet bars).Instead, at DIV 20 E' was positively affected by the presence of cells, being 23.07 ± 1.5 kPa and 37.68 ± 2.3 kPa with a pre-strain of 10% and 20% respectively, (figures 3(B) and (D), violet bars).
From the morphological point of view, the inner microstructures of the two thermogels under SEM observation (figures 3(E) and (F)) were found to be slightly different.ChiGel 1 showed less dense matrix with pores ranging in diameter from 5 to 20 µm (figure 3(E)), whereas ChiGel 2 exhibited a denser network and pores with diameters ranging from 2 and 10 µm (figure (F)).

Morphological characterization of 3D neuronal networks
To evaluate the distribution of primary neuronal cells and iNeurons within the thermogels, both 3D cell cultures were labeled with NeuroFluor™ NeuO (supplementary materials figure S2).Real-time imaging of neurons encapsulated in the thermogels was conducted at various DIV.At DIV 4, both cell populations appeared to be uniformly distributed throughout the entire thermogel volume.By DIV 7, primary cells began displaying neuritic processes (as depicted in supplementary materials figure S2(A)), whereas iNeurons showed no processes, with only neuronal stomata marked (supplementary materials figure S2(B)).After 14 DIV, both cell populations exhibited the development of a complex neuronal network.Afterward, to assess the long-term culture sustainability of the thermogels, primary neurons were co-encapsulated with astrocytes.The neuronal networks were fixed at DIV 21 and immunolabeled for MAP-2 (green) and GFAP (red) to highlight the morphology of neuronal and glial cells within the 3D thermogel.Nuclei were stained with DAPI (blue).For both ChiGel 1 (figure 4(A)) and ChiGel 2 (figure 4(B)), neuronal cells presented spherical somas resembling those found in brain tissue [53] They exhibited a dense arborization of neuritic processes, which appeared partially fragmented due to their penetration into the microporous structure of the thermogel, contributing to the formation of a highly interconnected 3D network.Moreover, neurons and glial cells were observed to be distributed on different levels, as indicated by the distribution of their nuclei.
Moreover, it was observed that ChiGel 2 thermogels displayed no signs of degradation at any time point, and stable cultures could be maintained for more than 4 weeks.Conversely, ChiGel 1 thermogels proved to be unstable and exhibited a tendency to fragment during in vitro testing.The same experiments were subsequently carried out with iNeurons co-encapsulated with astrocytes, as illustrated in figures 5(A) and (B).
Also in this case, both thermogels successfully supported neuronal and glial adhesion and growth, (figures 5(A) and (B)).However, the network formed within ChiGel 1 (figure 5(A)) appeared less dense and homogeneous compared to the one formed within ChiGel 2. Immunofluorescence analysis revealed that ChiGel 2 provided a suitable 3D microenvironment for iNeurons and astrocytes (figure 5(B), MAP-2 and GFAP) supporting the formation and long-term viability of neural networks (figure 5(B), MERGE).Specifically, ChiGel 2 sustained an in vivo like morphology during the long-term in vitro culture over 103 d [54], as shown in figure 5(C).For that reason, further characterizations for both cell types (primary neurons and iNeurons) were performed onto ChiGel 2. The thickness of the 3D neuronal network encapsulated into ChiGel 2 was evaluated to be approximately 500-700 µm.The formation of a 3D network was further investigated using confocal microscopy, which revealed that ChiGel 2 allowed the development of a dense network for both cell populations, primary (figure 6) and human derived (figure 7).Both neuronal networks are visualized by z-stack reconstruction (supplementary movies 2 and 3).In the case of 3D primary neuronal networks (figure 6(A)), spherical neuronal stomata and extensive neuritic processes were observed.The 3D network was composed of a dense interlacing of neuritic processes and glial cell extensions, as illustrated in figure 6(B).Additionally, a comprehensive view of the neuronal network within ChiGel 2 was provided by the 3D reconstruction of a 150 µm z-stack of the cortical network, as depicted in figure 5(C).
Similar observations can be extended to iNeurons co-cultured with astrocytes, as both exhibited an in vivo-like morphology (figure 7(A)) and formed a rich, robust, and well-interconnected neuronal network (figure 7(B)).A max intensity projection of the orthogonal view is represented in figure 7: XYZ projection (figure 7(C)) shows the network developed within a region of ChiGel 2. Notably, the GFAPpositive cells within the thermogels exhibited a thin and elongated shape, which is a characteristic morphology often observed in a 3D culture system [38].

Functional characterization of 3D neuronal network
The spontaneous electrophysiological activity of 3D neuronal networks cultured on MEA60 was recorded at different DIV (47-52-56-63). Figure 8(A) shows the developing global electrophysiological behavior of these 3D networks over time.As qualitatively demonstrated in the raster plot, at DIV 47, the activity appeared mainly random, while with the progression of culture time, the activity became more synchronized.Moreover, it is possible to observe that the number of active electrodes increased with time.These observations were supported by the quantitative analysis presented in figure 8(B).Parameters extracted from the spike data analysis revealed that 3D neuronal networks exhibited an increase in the MFR values, rising from 0.59 ± 0.76 spikes s −1 at the beginning of the culture period to 1.013 ± 1.24 spikes s −1 .
Regarding the bursting behavior, the percentage of bursting electrodes remained consistent throughout the culture period, but there was an increase in the mean bursting rate (MBR) values.Specifically, the MBR values were 2.92 ± 4.67 bursts min    Following the electrophysiological recordings, 3D networks were fixed and immunolabeled for morphological characterization.Figure 8(C) shows 3D networks developed within ChiGel 2 onto MEA.Neuronal and glial cells were well interconnected and homogenously dispersed across several layers on top of the electrodes, indicating that the ChiGel 2 allowed a correct development of a three-dimensional neuronal network.
Spontaneous electrophysiological activity was also recorded from 3D network coupled with 3D MEA60, as depicted in supplementary figure S3.A preliminary and qualitatively comparison was carried out between MEA60 and 3D MEA60, supplementary figures S3(A) and (B).Supplementary figures S3(C) and (D) show 1 min of spontaneous activity as recorded by MEA (C) and 3D MEA (D), while supplementary figures S3(E) and (F) show a zoom in of 1 s of the raw activity, revealing bursting phenomena and functional neuronal maturation.Furthermore, in supplementary figure S3, panels (G) and (H) exhibited waveforms recorded by the same electrode at DIV 21 and DIV 42, both from MEA60 (supplementary figures S3(G) and S2(I)) and 3D MEA60 (supplementary figures S3(H) and S3(J)).These waveforms highlighted how, over time, firing activity and the peak of the average action potential waveform increased, indicating the development of functional synapses on the electrode.A preliminary and qualitative electrophysiological characterization of 3D networks onto 3D MEA was carried out, which is reported in supplementary figure S4.Supplementary figure S4(A) shows the spike waveforms from DIV 74 to DIV 106.Supplementary figure S4(B) provides corresponding raster plots representing spontaneous activity over 10 min recording.

Discussion
The development of a reliable 3D in vitro model able to mimic features of the native tissue goes through the development of an artificial ECM able to recapitulate the mechanical characteristics, porosity, cell phenotype, and bioactivity of the in vivo counterpart [55].In the field of neural engineering, natural hydrogels, having a low modulus of elasticity and high-water content, are ideal 3D matrices mimicking human brain ECM [56,57].Based on previous work [17,35,38], the natural polysaccharide CHITO has demonstrated its bioactivity toward nervous cells, including both primary neurons and iNeurons, emerging thus as a suitable candidate for the development of 2D and 3D in vitro neuronal networks.Among all the properties of CHITO, one of the most interesting is its ability to form thermogels at physiological temperature and pH in the presence of a weak base such as β-GP.The temperature-dependent sol-to-gel transition is attributed to the heat-induced transfer of protons from CHITO to β-GP, reducing the repulsive forces among positively charged ammonium groups and allowing thus the interaction of CHITO chains [50].Even if CHITO thermogels have been investigated for a wide range of applications in the field of drug delivery and tissue engineering, very limited work has been carried out on their use for brain tissue engineering.In this study, we aimed to assess CHITO thermogels potential as an artificial ECM for the brain, recapitulating essential native ECM characteristics.Two thermosensitive CHITO formulations, as described in the literature [50][51][52][53][54][55][56][57][58][59], were characterized in terms of rheological properties, gelling kinetics, mechanical properties, morphology and bioactivity toward nervous cells.These two formulations differ in the gelling kinetics and consequently in their mechanical properties.Both formulations had a final CHITO concentration of 2% w/v and a β-GP concentration of 4% w/v.
The thermogels were produced by using PDMS molds with dimensions compatible with the active area of the MEAs used in this study, resulting in a thermogel thickness of approximately 700 µm and a diameter of 5 mm [17,38].Rheological and mechanical characterizations were carried out as a first step for both formulations since these properties greatly rely on the intrinsic properties of the used CHITO, in terms of molecular weight and degree of deacetylation [60].Mechanical characterization was carried out by DMA on both thermogels unloaded and loaded with cells.Initial gelation of both solutions occurred around 20 • C-25 • C and took about less than 5 min.The complete gelation was observed around 25 min for ChiGel 1 and 15 min for ChiGel 2 at 37 • C. The results indicated that both formulations followed a gelation kinetic compatible with cell culture, which is crucial to prevent cell stress and death [61].The storage modulus of the formulations varied linearly with CHITO initial concentration ranging from 0.1 to 0.4 kPa.These values are similar to the brain tissue storage modulus, indicating good mechanical compatibility [62][63][64][65].
Thermogels are also considered excellent injectable biomaterials, limiting cell damage from mechanical stress during syringe injection [64].Injectable thermogels should enable easy cell encapsulation, minimally invasive administration, and rapid gelation after injection [61].The viscoelastic properties are important features for injectability because the injection may strongly change physical stability and properties [65].The low storage modulus (G ′ ) at room temperature and the high storage modulus at physiological temperature (37 • C) of both the formulations suggested their injectability.Injectability characterization was carried out at physiological temperature in a physiological fluid, using needles with different sizes and by varying the injection speed.To reproduce the physiological environment, the injection study was performed in PBS at 37 • C to assess the forces required to manually extrude pre-gels from syringes and to assess self-healing after injection.Data demonstrated that by decreasing the inner needle diameter, the maximum force required to extrude the pre-gel increased (figure 2(D)) proportionally to the injection speed (figure 2(E)).Injection tests in physiological conditions confirmed the injectability of these thermogels, with forces required for injection falling within the range suitable for clinical in vivo applications.Typically, clinically relevant injection forces are less than 20 N [66,67].For in vivo injections, it is preferable to use thinner needles to enhance the patient's comfort.We selected needles with inner diameters ranging from 21 G to 25 G, which are commonly used for subcutaneous injections.Figure 2(D) clearly showed a correlation between the solution's viscosity and the maximum force required, and this correlation was more pronounced for ChiGel 2. Additionally, injection forces for both formulations were also assessed using needles with inner diameters ranging from 18 G to 20 G, which are also utilized in 3D bioprinting [68].However, the preliminary bioprinting test demonstrated a low 'shape fidelity' , indicating that when the material is deposited onto the flat print bed, it does not recover quickly enough the solid-like properties and as a consequence some fibers fuse together and close the pores.Further investigation is required to increase the shape fidelity and is actually ongoing in our group.
The mechanical properties of the native ECM are a critical parameter that should be considered when designing an artificial ECM, exerting significant effects on cell morphologies and behavior.In the field of brain tissue engineering, the mechanical properties of the substrate play a significant role in influencing neuronal behavior, affecting aspects such as morphology, the rate of neurite extension [69] the direction of axonal outgrowth, and overall functionality [66,67].The elastic modulus of both thermogels was characterized with and without cells, at DIV 1 and after 21 d.This evaluation aimed to understand the effects of cells, the development of neuronal networks, and the potential fragility of the thermogels.Interestingly, ChiGel 1 exhibited different mechanical properties with and without cells at DIV 1, indicating that the presence of the cells interferes with the strength of the matrix.Instead, no significant differences were observed at DIV 1 with and without cells for ChiGel 2. However, an increase in mechanical properties was observed at DIV 21 in the presence of cells for ChiGel 2. This suggests that the stiffness of the 3D thermogel is affected by the formation of the neuronal network and, possibly, by the natural ECM produced by the cultured neurons [38].It is well-documented in the literature that cortical neurons and astrocytes are highly responsive to changes in matrix rigidity when cultured on flexible substrates [70].
The elastic modulus of both thermogels was evaluated at 10 kHz and the values fall in the same range of brain tissue [71,72].This result suggests the suitability of these thermogels for brain tissue engineering, as they closely mimic the mechanical properties of the native brain ECM.
The inner structure of both thermogels was found to be porous, which is crucial for ensuring cell viability and promoting cell growth.Porosity is a significant property of the ECM and has an impact on various cellular processes, including viability, migration, proliferation, differentiation [73] and neurite outgrowth efficiency.Furthermore, ChiGel 2 exhibited a denser porosity compared to ChiGel 1.This increased porosity in ChiGel 2 can play a role in guiding the formation of highly interconnected networks [74], further enhancing its suitability for supporting neuronal growth and network formation.
An important aspect to be considered when developing 3D brain models is the ability to support longterm cultures, allowing for complete cell maturation and differentiation.This becomes particularly important when working with iPSCs or precursors cells for studying neurodevelopment or neurodegenerative diseases.The morphological characterization by immunocytochemistry revealed that CHITO thermogels provided a suitable 3D microenvironment for the maturation of both primary neurons and iNeurons, allowing the formation of highly interconnected neuronal networks and their long-term viability for up to 25 and 100 d, respectively.As a first step, we characterized both formulations with primary neurons from embryonic rodent.These primary neurons are regarded as the gold standard in neurobiology and neuroengineering due to their well-established properties and advantages, including cost-effectiveness, reproducibility, and the ability to differentiate into distinct axons, dendrites, and synapses, which can be clearly distinguished [75,76].
Then, CHITO formulations were tested with iNeurons.For iPSCs-derived model, iNeurons were co-cultured with astrocytes.It is well-established that astrocytes play a crucial role in controlling synapse formation and function, and they are essential for proper neural [77].iPSC-derived cells have a huge potential to model human diseases and to uncover disease mechanisms and novel therapeutic targets [78].Current 3D culture techniques are based on various approaches, including self-organizing organoids [79], neural tissue [80], and hydrogel based artificial ECMs [81,82].The literature reports several examples of 3D brain-like tissue models that rely on the use of the polysaccharide alginate with iPSCs, showing its ability to sustain differentiation into dopaminergic/glutamatergic neurons in long-term cultures representing an alternative model for applications in disease modeling (Parkinson's disease) and drug or chemical evaluation for mechanistic studies [80,83].Additionally, several studies have been focused on hyaluronic acid (HA) or HA-chondroitinbased hydrogels in their efforts to develop 3D neural models.These studies have aimed to identify the essential chemical cues required for creating a neuronal matrix that can trap the cell-produced ECM and neurotrophic factors, remodel the matrix, and support neurite outgrowth [84].Human neurons and glial cells have also been successfully cultured in Matrigel and PEG-heparin-based hydrogels [80,84].However, some of these biomaterials require a biofunctionalization to support neuronal cell adhesion and growth, and they often have limitations, including high batch-to-batch variability and low stability over time.It is well known that CHITO can promote neuronal adhesion, growth and differentiation of neural stem cells in vitro [85].CHITO thermogels were mainly used for in-situ bone and cartilage regeneration [38,61].To the best of our knowledge, this is the first work reporting its application in the brain field.Confocal microscopy confirmed the development of a 3D network of neurons co-cultured with astrocytes in ChiGel 2. For both cell populations, neuronal and glial cells developed an extensive and interconnected 3D network.Primary neurons and iNeurons were homogeneously distributed in the whole volume of the thermogel, and single cell bodies could be resolved well, showing a spherical soma; GFAP+ cells were distributed throughout the entire thermogel and showed a thin morphology.Both types of cells showed morphologies comparable to the ones found in vivo.The consistent morphologies observed in both types of cells align with prior results in the literature and suggest the suitability of CHITO thermogels for modeling in vitro neural systems [38].The 3D networks developed within the thermogel were demonstrated to be functional for at least DIV 106 onto 3D MEA.The 3D networks were characterized with respect to their spontaneous electrophysiological activity starting from DIV 47.The recording started at DIV 47 because we expected an already mature electrophysiological dynamic [17].The number of active electrodes progressively increased over the days in culture, suggesting that neurons are within a favorable environment for synaptic formation, as more neurons are involved in network communication as the days go by.However, by DIV 63, only half the number of the electrodes (37) remained active.This observation aligns with existing literature on 3D networks embedded in hydrogels and it can be explained by considering that neurons are distributed in all three dimensions within the culture, resulting in fewer cells directly interfacing with the electrodes.
Long-term cultures hold the potential to expand the utility of in vitro neuronal networks in disease modeling and in drug screening [83,86].To date, only a few studies have documented long durations of in vitro cultures [16,80,87].However, it is important to underline that typically hydrogels, in long term cultures, demonstrated to be able to support the 3D culture of neuronal cell clusters/spheroids.Focusing on polysaccharidic hydrogels, as an example, functionalized alginate and modified HA hydrogels have been tested for the long-term culture of h-iPSCs the form of spheroids.Indeed, in the present work, we have demonstrated that a CHITO-based hydrogel, without any further functionalization, is able to sustain the maturation of a highly interconnected neuronal network of single nervous cells.The use of CHITO to support neuronal adhesion, growth and maturation by using same cell phenotypes was already validated in our previous works, both in 2D and in 3D by using CHITO microbeads for the fabrication of a granular scaffold [17].In this respect, the use of microbeads was challenging due to the low stability of the 3D scaffold in the first DIV.The present work was then carried out to respond to the need of having a 3D material more easily workable but still bioactive toward neurons.From the functional point of view, as previously observed by Muzzi et al in the case of CHITO microbeads, the percentage of random spiking in 3D networks was higher than that observed in 2D cultures.Furthermore, we observed that 3D ChiGel networks exhibit more synchronous bursts (MBR, NBR) compared to 3D CHITO microbeads, along with longer durations for both bursts and network bursts.This activity indicates the formation of a very dense network with a high degree of connectivity, as also suggested by immunostaining for MAP-2 (figure 7).Therefore, the 3D thermogel model offers potential advantages such as enhanced hydrogel stability, stiffness similar to living brain tissue, no requirement for pre-treatment with adhesion proteins, a high level of connectivity, and in vivo-like electrophysiological behavior.

Conclusion
In this work, two different formulations of CHITOβ-GP thermogels were characterized with the aim of evaluating their potential as artificial ECMs in brainon-a-chip models.Both formulations exhibited rheological, mechanical, and morphological properties that render them suitable for use with nervous cells.Moreover, injectability tests demonstrated the potential applicability of ChiGel 2 for minimally invasive in vivo procedures, and they also hold promise as an alternative bioink for 3D bioprinting.Both ChiGel 1 and ChiGel 2 proved to be compatible with encapsulating neuronal cells, supporting the viability and growth of both animal and human-derived cell populations.In particular, ChiGel 2 thermogel showed the ability to sustain the maturation of a functional 3D human neuronal network over 100 DIV, representing an excellent system for long-term cultures, overcoming the limits introduced by organoids and spheroids.Importantly, this study represents the first instance of CHITO thermogels being employed with primary neurons and iPSCs to develop brain-on-achip models.

Figure 2 .
Figure 2. Rheological properties and injection test of CHITO thermogels.Storage modulus (G ′ ) and loss modulus (G ′′ ) variation as a function of time at 37 • of ChiGel 1 (A) and ChiGel 2 (B).(C) Temperature dependence of storage and loss moduli (G ′ , G ′′ ) of both formulation upon heating from 4 to 40 • C at a rate of 1 • C min −1 .Injectability of CHITO pre-gel: variability of injection force as a function of injection speed (D) and needle size (E) immersed for in PBS solution at 37 • .

Figure 8 .
Figure 8. Spontaneous electrophysiological activity characterization of 3D iNeuronal network in ChiGel 2 at different time in culture (47, 52, 56, 63 DIV): (A) raster plot showing 120 s of spontaneous activity; (B) mean electrophysiological parameters: number of active electrodes, mean firing rate, mean bursting rate, mean burst duration, network burst rate, network burst duration, perceptual random spikes, perceptual bursting electrodes.Data are shown as mean and standard deviation of the mean.(C) Fluorescence images of 3D iNeuronal networks fixed after the recording at DIV 63, labeled for MAP-2 (green), GFAP (red), DAPI (blue).Scale bar: 20 µm.