*3.2. Characterization of 3D Bioprinted Neural Constructs*

Neural cells differentiated for about 4 weeks were dissociated, resuspended in the Matrigel/alginate solution and printed. We have performed several experiments in which cells were dissociated in the window of time between day 25 and day 35 of differentiation (indicated in red in the diagram of Figure 1B). During the printing process, the bioink and the crosslinking solution met at the ending tip of the coaxial extruder. Here, Ca2<sup>+</sup> ions triggered the gelation of alginate in the bioink. This gel adhered to the functionalized glass substrate so that, by moving the extruder, a micrometric cell-embedding gel fiber was spun out and deposited in pre-determined positions. In this work we printed the cells as a reticulum (Figure 1C; Movies S1 and S2). Such architecture was chosen as it allows optimal perfusion of culture medium, which can reach all the cells in the construct. Moreover, areas with lower and higher cell densities are formed along the fibers and at the crossing points, respectively, providing useful information on the behavior of the cells in the 3D construct under different density conditions. Alginate removal by enzymatic treatment 3 h after the printing process promoted the acquisition of neuronal morphology by the first day post printing (Figure S3). Notably, such mild enzymatic treatment did not affect the shape of the printed construct, which was stabilized by Matrigel polymerization. Immunostaining of neurofilaments showed that the structure of the reticulum was maintained over time and that neuronal cells projected their axons and dendrites both within and across the fibers (Figure 1D). Printed cells were then analyzed in terms of viability at different days post printing (DPP). Results shown in Figure 1E indicated that the great majority of the cells were viable at DPP1 (78 ± 3.8% live cells; average ± standard deviation; three constructs, nine fields each) and DPP7 (71 ± 3.5% live cells; average ± standard deviation; three constructs, nine fields each), suggesting that both physical parameters and bioink formulation did not harm neural cells during and immediately after the printing process. Moreover, viability was consistently maintained over time as assessed by live/dead staining up to DPP50 (68 ± 8% live cells; average ± standard deviation; one construct, nine fields). We noticed that the reticulum structure was to some extent maintained at this late time point.

We then assessed possible alterations in neuronal cell fate acquisition caused by either the printing process and/or subsequent cell differentiation within the 3D bioprinted construct. Bioprinted cells were compared with cells maintained in conventional 2D conditions for the same time and cells that were encapsulated in bioink droplets not subjected to printing process (3D bulk). Neuronal morphology was maintained intact in both 3D bulk and 3D bioprinted cells at DPP7 and DPP40 (Figure 2A). In the same samples, marker analysis by RT-PCR showed proper expression of: *PAX6*, *FOXG1* and *TBR2* as neuronal progenitor markers; *TBR1*, which reveals the presence of mature cortical neurons; and *GFAP*, a common astrocyte marker (Figure 2B and Figure S4). These results were further supported by immunostaining analyses of TBR1 and MAP2 at DPP7 (Figure 2C). Bioprinted neural cells were maintained in neuronal differentiation medium up to DPP70. At this late time point the reticulum structure was, to some extent, maintained and cells properly expressed neuronal and astroglial markers (Figure 2D,D').

**Figure 1.** 3D bioprinting method and analysis of viability post printing. (**A**) Schematic representation

of the outline of the bioprinting method. (**B**) Outline of the human induced pluripotent stem cell (iPSC) neural differentiation protocol in conventional 2D culture and representative images of differentiating cells in these conditions at the indicated time points. The window of time in which cells have been dissociated for bioprinting experiments in this work is indicated in red. (**C**) Image of the printed 3D construct. Scalebar: 2 mm. (**D**) Mosaic reconstruction of confocal images of bioprinted neural cells at DPP7, stained with a MAP2 antibody (green) and DAPI (blue). Scalebar: 200 μm. (**E**) Live (green) and dead (red) cell staining in the bioprinted construct at the indicated days post printing (DPP). Scalebar: 150 μm (left panels); 50 μm (right panels).

**Figure 2.** Analysis of neural marker expression in the 3D bioprinted construct. (**A**) Phase contrast images of cells within the 3D bioprinted construct ("3D printed" panels), at the indicated days post printing, and cells in conventional monolayer conditions ("2D" panels) or resuspended in the bioink ("3D bulk" panels) and maintained for the same time of differentiation. (**B**) RT-PCR analysis of neuronal progenitor markers (*PAX6*, *FOXG1*, *TBR2*), a cortical neuron marker (*TBR1*) and an astrocyte marker (*GFAP*). GAPDH was used as a housekeeping control. (**C**) Immunostaining analysis of bioprinted cells at DPP7. MAP2 (green), TBR1 (red) and DAPI (blue) signals are shown. Scalebar: 150 μm. (**D**) Mosaic reconstruction of confocal images of bioprinted neural cells at DPP70, showing the entire sample, stained with MAP2 (green), TBR1 (white) and GFAP (red) antibodies. Scalebar: 2 mm. (**D'**) Mosaic reconstruction of confocal images of the region inside the white box in panel D, acquired at higher resolution. Scalebar: 300 μm.

Collectively, these results demonstrate that iPSC-derived cortical neuronal cells can be bioprinted and further cultured in 3D constructs without causing major survival and differentiation issues.

#### *3.3. Functional Analysis*

Single-cell patch-clamp and time-lapse calcium imaging recordings were then performed to assess the degree of maturation achieved by the 3D bioprinted construct. Even though the 3D construct was 300 μm thick, the selected bioink displayed sufficient transparency to visible light and softness to patch pipette insertion (Figure 3A). Using patch clamp recordings, we investigated the expression of the passive and active membrane properties on 3D bioprinted cortical neurons at day 7 after printing. As expected at this experimental point, resting membrane potential (−17.7 ± 1.5 mV; *n* = 36), cell capacitance (14.8 ± 0.89 pF; *n* = 45) and membrane resistance values (1.97 ± 0.23 MΩ; *n* = 44) were typical of neuronal progenitors [21] and similar to those observed in parallel 2D cultures (Figure S5), indicating that the printing process did not impair neuronal viability. We then characterized the ability of cortical neurons to generate action potentials. Neurons in the 3D construct displayed large inward voltage-dependent Na<sup>+</sup> currents (−777.31 <sup>±</sup> 73.16 pA at 0 mV; *<sup>n</sup>* <sup>=</sup> 43; Figure 3B,C and Figure S5) which activated near −40 mV and peaked at 0 mV, and voltage-dependent K+ currents (865.75 ± 63.28 pA at +40 mV; *n* = 43; Figure 3B,D and Figure S5). Current pulses were able to induce action potentials in almost all tested cells. The mean threshold for first action potential generation was −32.85 ± 2.86 mV (*n* = 15; 20 pA of current injection). However, the minimum current required to elicit firing in some of the tested cells was 10 pA (Figure 3E,F). As expected, no synaptic activity was recorded at 7 days post printing (data not shown).

Given the optical transparency of the 3D bioprinted constructs at DPP7, fluorescence time-lapse recordings lasting 5 min each were performed, thus preserving a good signal-to-noise ratio (Figure 3G). Fluorescence time-lapse analysis of spontaneous calcium oscillation in Fluo4-AM loaded 3D neuronal network indicated the presence of individual calcium activity (mean firing frequency = 0.015 ± 0.001 Hz; FOVs = 38; mean firing amplitude = 0.083 ± 0.002 A.U.) with little synchronized firing (syncro index = 0.223 ± 0.020; FOVs = 38). The small degree of synchronous activity was confirmed by the low correlation coefficient value between each pair of neurons in the field as displayed by the heatmap in Figure 3G (average correlation coefficient value = 0.006; max correlation coefficient value = 0.046 ± 0.005; FOV = 38), thus indicating the establishment of early and immature neuronal networks.

**Figure 3.** *Cont.*

**Figure 3.** Functional analysis of the 3D bioprinted construct. (**A**) Single-cell patch-clamp recording of an iPSC-derived neuronal cell encapsulated in the 3D bioprinted construct at DPP7. (**B**) Representative scheme of the recording protocol is shown. The inward sodium currents are highlighted in the purple box and the permanent outward potassium currents are highlighted in the green box. (**C**) Average trace of the large inward voltage-dependent Na<sup>+</sup> currents. (**D**) Average trace of the outward voltage-dependent K<sup>+</sup> currents. (**E**) A single action potential evoked in current clamp recording is shown. The minimum current required to elicit firing was 10 pA, however more of the 50% of tested cells (*n* = 9 out of 15) responded to 20 pA (**F**). (**G**) Calcium traces as a function of ΔF/F0 of cortical neurons isolated within the 3D network shown on the left at DPP7. On the right, a representation of the firing pattern and a relative heatmap of the Pearson correlation coefficients within the cells of the same network are shown.

#### **4. Discussion**

In this paper we describe a method to obtain 3D cortical constructs in which human cortical neurons and glial cells survive in the long term, holding their cellular characteristics and functional properties. Moving from conventional neuronal cultures, in 2D, to more realistic 3D models is considered a crucial advancement in neurobiology [22]. The recent discovery that differentiating hPSCs have the ability to self-assemble into brain organoids, which recapitulate to some extent the brain structure in 3D [23], has given a twist in the way neurodevelopmental and neurodegenerative diseases are modeled and approached [24]. 3D bioprinting could provide important advantages, in terms of automation and reproducibility, over self-assembled brain organoids [25]. Recent reports showed that undifferentiated human iPSCs and ESCs can be bioprinted and then converted, post-printing, into cell types of interest [4]. This approach will not likely generate useful artificial tissues, as it does not allow control on the position of individual cell types, generated during differentiation, within the construct. The complementary approach, used in this work and in [13,14], and consisting in bioprinting specific cell types obtained by pre-printing hPSCs differentiation, would be more advantageous, allowing better control of the resulting bioprinted construct.

Bioprinting neurons and glial cells represents a challenge. Neurons are vulnerable cells in vitro and environmental stress due to the printing process may affect neural cell viability and influence further differentiation and maturation. Our work is the result of an extensive effort in the optimization of the bioprinting process and bioink composition, with the goal to define proper conditions for generating human artificial 3D cortical neural tissues from hiPSCs. The generation of a bioprinted constructs, by combining hiPSC-derived spinal neuronal progenitor cells and mouse oligodendrocyte progenitor cells, has been recently reported by Joung et al. [13]. Moreover, spinal cord neural progenitors from hiPSCs have been successfully bioprinted by using a commercial lab-on-a-printer platform [14,26]. Here, for the first time, we describe the generation of constructs made of cortical neurons and glial cells by a custom extrusion bioprinter. In this work, we have obtained the best results with a bioink made of Matrigel and alginate. The selection of the bioink most suitable for the viability of the 3D construct remains a controversial issue. Indeed, both fibrin-based and Matrigel-based bioinks have been previously used for bioprinting hiPSC-derived spinal neural cells [13,14]. Matrigel, which is a matrix preparation extracted from the Engelbreth-Holm-Swarm mouse sarcoma, had been successfully used as a bioink component for the generation of hiPSC-derived cardiac and spinal cord bioprinted constructs [3,13]. However, its composition is rather undefined. This could represent an important limitation for basic and translational applications of bioprinted models, including those of the nervous system. Future studies are necessary for identifying more physiological, standardized and defined alternatives to Matrigel. To this direction, promising results have been recently obtained with decellularized extracellular matrix, used for the bioprinting of liver and hearth constructs [9].

Due to the vulnerability of neurons, their viability post printing is a major concern. In this regard, our results (70–80% of live cells) are comparable to previous works using hiPSC-derived neural progenitors [13,14] or an immortalized human neural stem cell line [27]. Moreover, our method allows long term survival of human neurons, up to 70 days post printing. To the best of our knowledge, this is the longest time of maintenance of hiPSC-derived neurons in 3D bioprinted constructs (14 days in [13], 30 days in [14], 40 days in [4] and 41 days in [26]). Further, this work suggests a possible approach to overcome some practical challenges associated with the bioprinting of 3D in vitro models containing cells of limited availability. In order to be able to produce constructs with arbitrary, high cellular density, without affecting the number of samples obtainable from each experiment, we adjusted the amount of bioink necessary for each construct to a few microliters (3 to 5 μL per sample). The dimension, visibility and weight of these samples are very limited, and their handling and maintenance in floating culture condition can be very challenging. To overcome this, we used functionalized micro-slides as receiving substrate during the printing step that guaranteed a prolonged adhesion of the samples to a flat, clear glass surface.

We here report that cortical 3D bioprinted constructs, as well as parallel 2D cultures, display functional properties typical of immature neuronal networks. Indeed, calcium imaging experiments showed sustained calcium spontaneous activity already at DPP7, in line with that reported for 3D bioprinted iPSCs [4,27], and spinal neural progenitors [13], thus suggesting that the printing process does not prevent the development of a functional network. However, passive and active neuronal properties, analyzed at single cell level by means of patch clamp recordings, were typical of immature neurons, and the absence of spontaneous synaptic activity indicated that network activity was mainly not dependent on action potential firing. This result is in line with data on 2D culture at the same time point.

This study opens the possibility for generating more complex human neural 3D constructs, for instance by printing mixed populations with precise ratios of neuronal and glial cells and/or printing iPSCs carrying pathogenic mutations associated to neurological diseases. Notably, the bioprinting approach used herein can be further implemented with more sophisticated microfluidic platforms that might allow the deposition of multiple materials and/or multicellular bioink within a single scaffold, by simultaneously extruding different bioinks or by rapidly switching between one bioink and another, as previously described [16], with the aim of controlling the localization

of individual cell types in predetermined positions of the 3D construct. Different specific neuronal subtypes, which can be obtained by iPSC differentiation, might be used as the cellular components of 3D constructs for disease models and drug screening. In the case of complex diseases with clear non-cell autonomous contribution, neural and non-neural cells could be printed together. In the long term, further development of this technology could provide bioprinted cortical neural constructs that can be exploited as customized, standardized and scalable pre-clinical models for drug safety and toxicity studies.
