Integrated functional neuronal network analysis of 3D silk-collagen scaffold-based mouse cortical culture

Summary Bioengineered 3D tunable neuronal constructs are a versatile platform for studying neuronal network functions, offering numerous advantages over existing technologies and providing for the discovery of new biological insights. Functional neural networks can be evaluated using calcium imaging and quantitatively described using network science. This protocol includes instructions for fabricating protein-based composite scaffolds, 3D in vitro culture of embryonic mouse cortical neurons, virally induced expression of GCaMP6f, wide-field calcium imaging, and computational analysis with open-source software and custom MATLAB code. For complete details on the use and execution of this protocol, please refer to Dingle et al. (2020).


SUMMARY
Bioengineered 3D tunable neuronal constructs are a versatile platform for studying neuronal network functions, offering numerous advantages over existing technologies and providing for the discovery of new biological insights. Functional neural networks can be evaluated using calcium imaging and quantitatively described using network science. This protocol includes instructions for fabricating protein-based composite scaffolds, 3D in vitro culture of embryonic mouse cortical neurons, virally induced expression of GCaMP6f, wide-field calcium imaging, and computational analysis with open-source software and custom MATLAB code. For complete details on the use and execution of this protocol, please refer to Dingle et al. (2020).

BEFORE YOU BEGIN Extraction of silk fibroin fibers
Timing: 3 h Silk fibroin is a natural polymer and a suitable material for both in vivo and in vitro biomedical applications because of its biocompatibility, tunable chemical and mechanical properties, and long-term stability in vitro yet with 100% degradability over time in vivo (Rockwood et al., 2011;Rouleau et al., 2020). We have previously reported several in vitro brain models, with both primary rodent neurons and human induced pluripotent stem cells (iPSCs), using porous scaffolds fabricated with silk fibroin and extracellular matrix composites. Such models recapitulated many key features of the brain tissue (Cairns et al., 2020;Cantley et al., 2018;Dingle et al., 2020;Liaudanskaya et al., 2020;Rouleau et al., 2020;Tang-Schomer et al., 2014).
Native Bombyx mori (silkworm) silk is composed of the core silk fibroin protein and the adhesive sericin proteins. This section describes the step-by-step instructions on the removal (degumming) of sericin and the extraction of fibroin fibers. See Figure 1.
Note: Each batch of 6.25 g cut cocoons yield approximately 1.5 pieces of 10 cm diameter 3 3-4 mm height silk sponge scaffold.
16. Dialyze to remove LiBr by changing 4 L RG or DI water six times in the next 48 h on a stir plate.
First two changes should be 1-2 h apart. Remaining four changes should be at least 4 h apart. Check conductivity of the final water with a handheld conductivity meter (see Key resources table), which should match fresh RG or DI water. 17. Collect silk fibroin solution in 50 mL conical tubes (does not need to be sterile). 18. Centrifuge silk fibroin solution at 4 C at 12,700 3 g for 20 min. Pour silk fibroin solution into new conical tubes. Repeat centrifugation and pass silk fibroin solution through a 100-mm strainer into new conical tubes. 19. Measure concentration by drying 500 mL of silk fibroin solution in a weigh boat (smaller boats with diameter < 4 cm typically work better) in a 60 C oven for a minimum of 8 h. a. Silk fibroin concentration % (w/v) = 100 3 dry weight of silk fibroin (g) O 0.5 mL.
Note: If the silk concentration is <6% (w/v), place silk fibroin solution back in new dialysis tubing and hang it in a fume hood for 8-12 h while exposed to air to allow evaporation.
Pause point: This solution can be used immediately or stored at 4 C for up to 2 weeks.

Timing: 5 days
This section provides step-by-step instructions for fabricating porous silk fibroin scaffolds with pore sizes of 425-500 mm. These scaffolds will be used for 3D cortical culture. We had previously determined that 6% (w/v, in water) silk fibroin solution resulted in scaffolds with desirable mechanical properties for neuronal culture (Tang-Schomer et al., 2014). To initiate the self-assembly of aqueous silk fibroin into a scaffold composed of insoluble b-sheet crystalline structure, sodium chloride (NaCl) particles are incorporated in the silk solution, incubated, and later washed out to create pores. The resulting sponge material is manually trimmed to the final dimensions. See Figure 2.
21. Stack stainless steel sieves in the top to bottom order of lid, 500-mm pore size sieve, 425-mm pore size sieve, and receiver tray (see Key resources table). Pour a small amount of NaCl at a time and sieve NaCl particles to size range of 425-500 mm. Shake until very little NaCl passes through the bottom 425-mm sieve. See Methods Video S6.
22. Prepare 60 g sieved NaCl particles in a large weigh boat (e.g., Fisherbrand 02-202-102, 110-mm diameter). 23. In a 10 cm plastic petri dish, first add 30 mL 6% (w/v) silk solution, and then slowly and evenly pour 60 g NaCl from the large weigh boat into silk solution. See Methods Video S7.
CRITICAL: Achieving homogenous sponge formation, even distribution of NaCl particles is important. Slowly pour NaCl in a back and forth motion to add thin layers and rotate the dish after each layer.
25. Fill a 4 L beaker with RG or DI water and rinse the cured sponge. Carefully break the edge of the plastic petri dish and slide a microspatula underneath the sponge to release the sponge from the dish. Soak and gently squeeze the sponge with DI water multiple times until the sponge is soft. See Methods Video S8.
26. Place the sponge in a clean petri dish lid and secure it to a float buoy using rubber bands. Wash the sponge in 4 L RG or DI water on a stir plate and change water every few hours over 1-2 days. Check conductivity of the final water, which should match fresh RG or DI water.
Pause point: The sponge can be stored in water in a petri dish, parafilm-sealed, at 4 C for up to 3 months. 28. Trim scaffold height to 1.5 mm with a razor blade. A useful method to keep height consistent is a customized, 3D printed scaffold trimmer. The SDL file of the scaffold trimmer can be downloaded from https://github.com/mattiabonzanni/Integrated-Functional-Neuronal-Network-Analysis-of-3D-Silk-Collagen-Scaffold-based-Cortical-Culture. 29. Place the scaffolds in an autoclavable container filled with RG or DI water. Autoclave-sterilize the scaffolds using a 120 C, 20-min liquid cycle.
Pause point: Store the sterile scaffolds at 4 C for up to 3 months.

OPEN ACCESS
Note: Store at 4 C for up to 1 month.
Note: Store at 4 C for up to 2 weeks.
Note: Store at 4 C for up to 1 month.
Note: Prepare 2 mL aliquots and store at À20 C for up to 6 months.

Scaffold coating
Timing: 1 h each on day À1 and day 0 This step coats the silk scaffolds with PDL and laminin, which is essential for cell adhesion to the silk scaffolds. This procedure should be done in a biosafety hood under sterile conditions. See Figure 3. Cortical neuron isolation and cell seeding Timing: 5-6 h on day 0 This section describes the isolation of primary neural cells from mouse cerebral cortices and the subsequent seeding of cells onto 3D silk scaffolds. See Figure 3.
Note: Proper institutional approval per animal protocols must be attained prior to embarking on this method.
Note: Animal dissection should be done in a biosafety hood under sterile conditions. If the lab is not equipped with a dissection microscope inside a biosafety hood, it is acceptable to perform dissection on the bench. Tissue dissociation and cell culture should be done in a biosafety hood under sterile conditions. Protocol Optional: Refer to ''Quality control assay #1, steps 84-89'' and ''Quality control assay #3, steps 94-97'' to examine the viability of the neurons.
5. Chill 150-200 mL HBSS on ice. 6. Fill three 10 cm dishes with ice-cold HBSS. Fill several 60 mm dishes with ice-cold HBSS and keep on ice. 7. Sacrifice timed-pregnant (E16.5) mouse. Shave abdomen and clean with 70% ethanol. Hold abdomen with tissue forceps and cut open skin with surgical scissors.
Note: To avoid contamination from the skin, avoid reusing these tools.
8. With separate scissors and forceps, remove and place the uterus in the first 10 cm dish of ice-cold HBSS to remove excess blood.
Note: Blood will adversely affect the health of the cells.
9. Transfer the uterus to the second dish of HBSS, extract embryos, and place embryos in the third dish of HBSS. 10. Decapitate the embryos with scissors and place the heads in the third 60 mm dish of HBSS on ice.
Note: Keep tissues on ice as much as possible to maintain viability.
11. Under a dissection microscope, stabilize the head by inserting angled forceps into the eye sockets and hold down. Gently make a midline cut with a scalpel blade and peel back the skin and skull with Dumont #5 forceps. Carefully remove the brain with a spatula and place the brain in ice-cold HBSS in a 60 mm dish. 12. Separate the cerebellum and the hemispheres with forceps. 13. Peel and remove the meninges with forceps. 14. With the medial side facing up, remove the midbrain and separate the striatum and cortex with forceps. Collect cortices in a 50 mL conical tube with HBSS on ice. 15. Remove HBSS from the conical tube containing cortices. Rinse the cortices with 30 mL PBS. 16. Carefully remove the PBS and add 2 mL of 0.25% Trypsin + 0.3 mg/mL DNase solution. 17. Transfer tissues/Trypsin/DNase to a 35-mm dish and incubate at 37 C for 20 min. 18. Add 2 mL Neuro Medium Plus 10% FBS to the tissue/Trypsin/DNase and transfer the 4 mL mixture to a 15 mL conical tube. 19. Add an additional 4 mL of Neuro Medium Plus 10% FBS. Triturate with a 5 mL pipette tip 20-25 times to mechanically dissociate the tissues. 20. Pass the solution through a 100-mm cell strainer into a new 50 mL conical tube. Rinse the strainer with an additional 4 mL Neuro Medium Plus 10% FBS. 21. Count the cells (recommended at 1:10 dilution) using Trypan Blue dye.
22. Centrifuge at 210 3 g for 5 min. 23. Remove supernatant. 24. The final seeding cell number and volume per scaffold is 2 3 10 6 cells in 5 mL (i.e., 400 3 10 6 cells/mL) of Neuro Medium Plus 5% PBS. Because the cell pellet accounts for a significant volume at this density, only add 3.5 mL per 2 3 10 6 cells of Neuro Medium Plus 5% PBS to resuspend the cell pellet. 25. Aspirate dry the PDL/laminin-coated scaffolds and transfer each scaffold to the center of a well in a 96-well plate.
CRITICAL: Do not dry the scaffolds until the cell suspension is ready for seeding and it is recommended to dry no more than 30 scaffolds at a time. Dry scaffolds quickly become fragile, so cell suspension needs to be added as soon as possible.
CRITICAL: Shear force can affect cell viability. Once the suspension seems homogeneous, stop pipetting.
27. Add 5 mL of cell suspension directly onto the scaffold.
Optional: A repeater pipette can be used to speed up the seeding process.
Note: If scaffold gets stuck on the pipette tip, use another pipette tip or forceps to gently dislodge.

Adeno-associated virus (AAV) infection
Timing: 1 h on day 1 This section describes the protocol for infecting 3D cultures with AAV-Syn1-GCaMP6f (Chen et al., 2013). Cells adhere onto the 3D silk scaffold surfaces overnight before neurons are infected with AAV-Syn1-GCaMP6f, an adenoviral system for expressing the genetically encoded calcium indicator (GECI) under the human synapsin 1 promotor. This procedure should be done in a biosafety hood under sterile conditions. See Figure 3.
32. Dilute AAV in pre-warmed Neuro Medium (serum-free) to a final concentration of 2 3 10 10 GC/ mL. 33. Add 500 mL AAV/Neuro Medium to individual wells of 48-well plate for infection (1 well for each scaffold to be infected). 34. Using forceps, transfer each scaffold into one of these wells containing AAV/Neuro Medium.
Incubate at 37 C for 24 h.
Note: It is normal to have many unadhered cells on the bottom of the 96-well plate from day 0. Seeding efficiency is estimated to be around 60%-70%.

OPEN ACCESS
Collagen I gel casting and culture maintenance

Timing: 2-3 h on day 2 and 30 min every 3-4 days
This section describes casting of Collagen Type I onto the seeded 3D scaffolds. By this day, cells have adhered to the surfaces throughout the 3D scaffold. Collagen gel will fill the pores of the scaffold to provide an additional substrate for 3D neurite extension. This procedure should be done in a biosafety hood under sterile conditions. See Figure 3. The functional measurement of the 3D system is achieved using the genetically encoded calcium indicator (GCaMP) to detect changes in the cytoplasmic Ca 2+ concentrations in the neural culture (Chen et al., 2013). A large field of view is needed to investigate functional networks at mesoscale, i.e., activity of populations of neurons across hundreds of mm (Betzel and Bassett, 2017). This section describes time-lapse recordings of Ca 2+ transients in neurons expressing GCaMP at a low magnification/large field of view to capture the entire 3D culture projection. The image data are used for network analysis in the subsequent sections. See Figure 4.
Note: Ca 2+ imaging is recommended at 3 weeks of culture. GCaMP expression is sufficient at 2 weeks of culture for Ca 2+ imaging but higher sample-to-sample variability has been observed at an earlier time point.
Note: For terminal time points, this procedure does not need to be done under sterile conditions. For longitudinal studies, conduct this procedure under sterile conditions as much as possible.
44. Set the microscope stage-top incubator to 37 C and 5% CO 2 . 45. Remove enough medium (approximately 1.5 mL per well) so that the 3D culture is not floating, but still submerged in medium.
CRITICAL: 3D cultures must not be moving during time-lapse imaging.  49. Choose the microscope objective based on the required field size and resolution. In this study, a 43 objective, with 4.15 mm 3 3.51 mm field size was used to capture the full 3D culture projection. 50. Set the acquisition rate, time-lapse duration, FITC channel exposure time, gain, and binning. A minimum 1 Hz is required (higher frequency recommended) to capture Ca 2+ events, and the acquisition parameters should be optimized according to user's microscope and camera setup. Table 1 describes the factors to consider for each acquisition parameter. 51. Move the stage so that the 3D culture is centered in the image field of view. 52. Record time-lapse series of green fluorescence signal.
Note: If there is no GCaMP signal or change in fluorescence intensity, see Troubleshooting 5 and Troubleshooting 6.
53. Export the time-lapse files as OEM .tif image stacks.

Generating region-of-interest (ROI) mask for network analysis
Timing: <1 h This section describes how to design and convert to binary mask using Adobe Illustrator and Adobe Photoshop. The goal is to create a mask containing 37 hexagonal ROIs, with 500 mm center-to-center distancing ( Figure 5). The mask must be binary (black/white) in Bitmap format in order to be used in the subsequent Image Processing Step.
Alternatives: Most microscope software has built-in ROI mask functions and can be utilized if it generates a mask in black/white binary and bitmap format. We also provide a pre-made mask with FOV size 4.15 mm 3 3.51 mm (852 px 3 720 px), which can be downloaded at https:// github.com/mattiabonzanni/Integrated-Functional-Neuronal-Network-Analysis-of-3D-Silk-Collagen-Scaffold-based-Cortical-Culture. Note: LZW compression is fine. This generates a binary mask that can be read by FluoroSN-NAP in the subsequent steps. An RGB or grayscale format mask is incompatible. This section describes using open-source software (FluoroSNNAP) to extract fluorescence intensity versus time traces and identifying Ca 2+ events in each ROI (Patel et al., 2015). A global synchronization index is calculated in this step. Cross-correlation of Ca 2+ traces from each ROI pair is calculated and a matrix of cross-correlation coefficients of all the paired ROIs is generated.  (1), number of times to perform surrogate resampling = 20, minimum size of synchronization cluster = 3 75. Select Analysis / Preferences / Functional Connectivity. In this study, we use Cross-Correlation for functional connectivity analysis. See Troubleshooting 8 for recommendations on the choices of the functional connectivity. Save preference. 76. Select Analysis/ Process single file. At the end of the analysis run, a processed_analysis.mat file is generated for each time-lapse and will be used for further analysis.
77. Rename the processed_analysis.mat file to processed_analysis_(your_file_name).mat file in the folder, but do not rename it in MATLAB (the RunNetworkAnalysis.m script described in the next section works under the assumption that the name of the file loaded in the Workspace is proc-essed_analysis.mat). CRITICAL: Rename the processed-analysis.mat file in the folder after it is generated, or it will get overwritten by FluoroSNNAP when the next time-lapse stack is processed.
Optional: To check the output file is in the correct format, open the processed_analysi-s_(your_file_name).mat in MATLAB. It should load a 1 3 1 processed_analysis.mat structure, containing 21 fields. To view the cross-correlation coefficient matrix, open FC/CC/C.

Network analysis with RunNetworkAnalysis.m
Timing: variable (typically 1-2 min per file) This section describes a step-by-step walkthrough of using RunNetworkAnalysis.m which can be accessed at https://github.com/mattiabonzanni/Integrated-Functional-Neuronal-Network-Analysisof-3D-Silk-Collagen-Scaffold-based-Cortical-Culture. This custom MATLAB code loads the proces-sed_analysis.mat file generated from FluoroSNNAP in the previous step, extracts and organizes the variables and computes parameters that describe the network properties (described in detail after the steps). This code runs on MATLAB (version 2019 or later), and Bioinformatics Toolbox Add-on is required.
In brief, the analysis is based on graph-theory descriptors. A graph (network) is composed of nodes interconnected by edges (Bullmore and Sporns, 2009). In our analysis, each ROI is a node and the statistical dependence (assessed using the cross-correlation method in this manuscript) between two ROI activities is an edge. The computed parameters describe several features of the reconstructed network topology.
The protocol here focuses on mesoscale activity, namely the statistical dependence of the activities of clusters of neurons, to capture the population, rather than single-cell activities, of the entire 3D culture. The user can however customize the size and shape of the ROIs to address different questions. It follows that different spatial parcellations lead to different interpretations of the resulting network (Messé , 2020).
Note: Descriptions of Network Parameters -Node. Each ROI is represented in the network as a node. Overall, a node can be defined from a single neuron to entire circuits; considering the multiscale organization of neuronal networks, there is no single, privileged scale (Hallquist and Hillary, 2019).
Note: Descriptions of Network Parameters -Edge. The edge between two nodes represents their functional connectivity, e.g., the statistical dependence between their activities. The value of functional connectivity between two nodes must not be necessarily interpreted as an indication of a physical connection (structural connectivity).

Note: Descriptions of Network Parameters -Binary versus Weighted networks.
A weighted network is a network whose edges have values between 0 and 1. An edge weight represents the weighted statistically dependence of the activity of the two nodes. On the other hand, such estimates are associated with measurement noise (small nonzero and/or negative weights). A binary network could be thus derived through a cutoff procedure with a userdefined threshold to the edge weights. An edge is preserved if its weight is higher than the threshold and removed otherwise. The preserved edges assume a value of 1 in the matrix; the eliminated edges assume a value of 0 instead. 83. The code outputs the following variables that can be viewed in MATLAB and in the exported Excel spreadsheet. These variables are summarized in Table 2 and Figure 6 and described in greater detail below:

Download and unzip
Note: In this analysis, each ROI is defined as a node, and the terms ROI and node therefore can be used interchangeably. See Troubleshooting 12 for error message regarding negative weighted edge. Table) -A calcium event is a transient and synchronous change in fluorescence of a cluster of neurons in response to their intracellular elevation of Ca 2+ associated with an action potential. In the Calcium Events Table, each column represents a ROI, and the data represents the time of peak (s) of the Ca 2+ events per ROI (a peak is defined as the time at which a calcium event reaches its maximum value).

Note: CalciumEventsTable (Calcium Events
Note: ISITable (Inter-spike Interval Table) -The inter-spike interval (ISI) is the average time (s) between two consecutive Ca 2+ events ( Figure 6A). It is computed for each ROI and organized in a tabular format. The ISI is calculated as: where Peak is the time of peak of the n th event. Note: AverageFrequency -The average frequency of Ca 2+ events of all ROIs (events/min). Interpretation: A greater average frequency indicates faster neuronal activity.
Note: PercentageActiveROI -The percentage of ROIs with at least one recorded calcium event. It represents the fraction of the entire sample active during the measurement ( Figure 6B -Green hexagon: Active ROI; Gray Hexagon: Inactive ROI). Interpretation: A greater value suggests a larger proportion of active surface area of the sample.
Note: SyncIndex (Synchronization Index) -The global synchronization index of the sample. It quantifies the global degree of synchronization of multiple time series recorded simultaneously from multiple ROIs. It is important to notice that the entire traces and not the firing patterns are used to compute the synchronization index. Interpretation: A greater global synchronization index suggests that the sample activity is more synchronous ( Figure 6B).
Note: N-The number of ROIs/nodes. It is customized by the user while creating the ROI mask (ex. Figure 6 -seven ROIs).
Note: NodeDegree (Node Degree) -The node degree ( Figure 6C -the numbers refer to the central green ROI), normalized to the maximum number of edges (N-1), calculated as: In a weighted graph, degree i is the sum of the edge weights for edges directly connected to node i . In a binary graph, degree i is the number of edges directly connected to node i . Interpretation: A greater node degree suggests a stronger connectivity of the i th node to the other nodes.
Note: Node degree distribution -A Kernel density estimation graph showing probability distribution versus. node degree value. Interpretation: A distribution curve with a long tail may suggest the existence of hub nodes -a small number of nodes with node degree greatly  Note: Modules and Modularity -A module is a group of nodes that are densely/strongly connected within modules and weakly/sparsely between nodes belonging to other modules (Figure 6D -green, red, and blue modules). Modularity is a value that measures the strength by which the network can be partitioned into modules. Number of modules and module assignment of nodes in a network are identified by maximizing the network's modularity during computation. Interpretation: Real world networks tend to divide naturally into communities or modules. A greater modularity indicates stronger community structure of the network (Newman, 2006).

Note: NumberofModules (Number of Modules) -
The number of modules in which the network has been partitioned ( Figure 6D -green, red, and blue modules).
Note: ModulesComposition (Modules Composition) -A table summarizing the module index (first column), the number of ROIs in each module (second column) and the node indices in each module (third column).
Note: AverageEdgeWeight (Average Edge Weight) -The average value of the normalized node degree of the sample. Interpretation: In a weighted graph, it represents the average edge weight of a node. In a binary graph, it represents the average probability that a node is connected to another one. Larger values correspond to greater average connectivity of the network.
Note: AverageClustCoeff (Average Clustering Coefficient) -The average clustering coefficient of a network ( Figure 6E) is calculated as the average of the local clustering coefficients CC i of all the nodes as follows: In a weighted network, CC i represents the strength of triangle networks ( Figure 6E, left) formed by three neighboring nodes and is calculated as follows (Onnela et al., 2005): In a binary graph, CC i represents the probability that two nodes j and k are connected to each other while they are both connected to a node i ( Figure 6E, right). Interpretation: A greater average clustering coefficient suggests stronger tendency of the network to form local clusters. Real world networks tend to have a high clustering coefficient (Muldoon et al., 2016).
Note: AveragePathLength (Average Path Length) -The average path length of a network ( Figure 6F) is calculated as: where N is the total number of nodes and d(i,j) is the shortest path between node i and j ( Figure 6F shortest path is indicated in red). For a weighted network, the shortest path is calculated as dði;jÞ = 1=w ij , where w ij is the edge weight between the node i and j. For a binary network, the shortest path represents the smallest number of edges to connect node i and j. Interpretation: A greater average path length suggests that a longer distance or more nodes must be crossed to transmit the information between two random nodes. Average path length is also inversely correlated to the probability of shortcuts in the network. Real world networks tend to have a small value of the path length (Muldoon et al., 2016). It is also important to mention the debate on its interpretation in correlation-ll OPEN ACCESS based functional networks. In short, paths in those networks represents sequences of statistical association but may not necessarily characterize the presence of information routes (Fornito et al., 2010).

Quality control assays (optional)
The quality control assays provide additional procedures to examine the viability, development, and functionality of the neurons and 3D cultures. We recommend conducting the quality control assays at the same time as the experiments using the same batch of materials and cells, even though they are not required for the functional analysis.

Quality control assay #1 -2D
Timing: Add 0.5-1 h to each day of the 3D culture steps This step describes conventional 2D culture of cortical neurons. We recommend conducting it using the same batch of neurons and for the same duration as in the 3D culture to ensure the viability of the neurons from the isolation procedure.

Quality control assay #2 -drug response
Timing: Add 15-30 min per sample on Ca 2+ imaging day This section describes using neurotransmitter receptor antagonists, namely bicuculline (GABA A receptor blocker), picrotoxin (GABA A and GABAr receptor blocker), NBQX (AMPA receptor blocker), AP5 (NMDA receptor blocker), and tetrodotoxin (sodium channel blocker) on the 3D cultures to elicit known responses. We recommend conducting it after 2-3 weeks of culture and at the same time point as the control 3D cultures.  This section describes infecting the 3D culture with AAV to express RFP in the neurons, in order to visualize the structural neuronal network in the 3D culture using confocal microscopy.

Quality control assay #5 -gene expression
Timing: Variable (5-6 h per 24 samples of RNA isolation, 3 h for cDNA synthesis, 3 h per 96well plate of qRT-PCR) This section describes qRT-PCR analysis of gene expression of proteins associated with neuronal maturation and synaptic neurotransmission (See Key resources table and Table 3) of the 3D cultures shortly after cells seeding and as the neuronal networks develop.
CRITICAL: Use nuclease-free supplies. Keep mRNA samples on ice as much as possible.
101. Collect uninfected 3D cultures on day 1, 14, and 21 each in 600 mL RLT/bME buffer in a 1.5 mL nuclease-free microcentrifuge tube. Store at À80 C.Thaw samples on ice. Use a homogenizer pestle to manually grind 3D culture into fine pieces. 102. Use a motorized tissue grinder to further grind 3D culture for 20 s.
Pause point: Homogenized samples can be stored at À80 C.
Note: Long-term storage at this step can significantly affect mRNA concentration. Perform mRNA isolation as soon as possible.
103. Transfer sample to a QIAshredder column and centrifuge at 21,000 3 g for 2 min. 104. Isolate mRNA from the flow through using a Qiagen RNeasy Mini Kit according to the manufacturer's protocol (https://www.qiagen.com/us/products/discovery-and-translationalresearch/dna-rna-purification/rna-purification/total-rna/rneasy-mini-kit/#resources) with the exception to elute the mRNA using only 22-24 mL of nuclease-free H 2 O. 105. Measure mRNA concentrations using a NanoDrop spectrophotometer.
Pause point: mRNA can be temporarily stored at À80 C, but it is recommended to reverse transcribe to cDNA as soon as possible.
Pause point: cDNA can be stored long-term at À80 C.

EXPECTED OUTCOMES
AAV infection and Ca 2+ imaging By two weeks, a large number of neurons in 3D cultures infected with AAV-hSyn1-GCaMP6f-P2A-nls-dTomato should express strong nuclear dTomato signal ( Figure 7A). GCaMP6f has less prominent fluorescence signal ( Figure 7B) and is not as clearly visible as dTomato. Spontaneous neuronal activities, observable as changes in GCaMP6f signal intensity, should be seen in clusters (> 30 mm) of neurons (See Methods Video S10). It should be noted that silk scaffolds are autofluorescent and do produce background in both the red and green channels. Non-homogenous distribution of cells is not unexpected.

Network activity and analysis
At 3 weeks, the 3D cultures are expected to show spontaneous and sustained activity in most of the regions ( Figure 8A). Two figures will be generated: The first figure is Kernel Density Distribution of the node degree, with the average degree indicated as well. The second figure is a circular graph showing the functional connectivity graph. For weighted graphs, the edge weight is directly proportional to the thickness of the lines.
In addition, an Excel file is automatically saved containing the outputs divided in different sheets (as detailed in the code comments section).
Example results from real samples are provided in Figure 8, using weighted network analysis.
A representative raster plot (each dot is the time of peak of a single calcium event) is shown in Figure 8A (left). The number of events/min and synchronization index are then graphed for 7 samples ( Figure 8A -right). Upon network reconstruction (weighted network), the probability density function (pdf) of the normalized node degree is generated, as shown in the representative distribution of Figure 8B (top,left). The curve had a Gaussian-like profile suggesting the absence of hub nodes, namely nodes with a high node degree. Two topological descriptors, the average clustering coefficient and average path length, are computed from the network ( Figure 8B -top, right). Finally, the network is partitioned into modules ( Figure 8B -bottom, left; red nodes and edges) based on the computation of the modularity ( Figure 8B -right, bottom). Quality controls QC #1 -In 2D cultures, short neurites emerged on day 1. Visual networks with increasing density form between day 3 and 7 ( Figure 9A).
QC #2 -When 3D cultures are treated with picrotoxin or bicuculline, Ca2+ event frequency increases. When 3D cultures are treated with NBQX or AP5, Ca2+ event frequency decreases or ceases. When 3D cultures are treated with tetrodotoxin, all or almost all activities are eliminated.
QC #3 -High density of live calcein-positive cells throughout the 3D culture. 3D cultures are expected to have lower density of viable cells at 2-3 weeks of culture than at <1 week ( Figure 9B).
QC #4 -In AAV-syn1-TurboRFP infected 3D cultures, strong TurboRFP expression in neurons at 2-3 weeks of culture is observed, demonstrating an extensive neuronal network with long neurites (hundreds of mm). When visualized using confocal microscopy with z-stack imaging, observations of many small and large round objects are expected ( Figure 9C).

LIMITATIONS
Network activity depends on the health of the primary neurons and the distribution of cells within the scaffold. Therefore, the cultures are prone to sample-to-sample, batch-to-batch, and/or user-touser variability. The network analysis is based on the projection imaging of the 3D culture, and therefore the z-direction network information flow cannot be inferred from this protocol. Ultra-speed confocal microscopy, such as light sheet confocal microscopy, will be required to obtain x,y,z information at the speed necessary to detect Ca 2+ transients. This protocol focuses on wide-field, mesoscale network analysis among populations of neurons. It relies on synchronized Ca 2+ transients in local clusters of neurons for a Ca 2+ event to be detectable per ROI. If the neuronal activity is mainly asynchronous, Ca 2+ event detection becomes unreliable and cannot be distinguished from the background signal. When 3D cultures are under conditions where there will be low or no activity, global synchronization index may become artificially high due to the similarly of background noise, and network analysis cannot be reliably performed.
This protocol has the potential to be adapted to analyze neural networks of human iPSC-derived or neural stem cell-derived neurons. Key issues to consider include: 1) the AAV-GCaMP6f infection step needs to be optimized for human neuronal cultures and may require repeated AAV infections to maintain adequate GCaMP levels, 2) human neurons mature at a slower rate than rodent neurons and synapsin I expression is required for GCaMP6f expression, and 3) functional neural network formation may take longer in human cultures than in rodent cultures and therefore Ca 2+ imaging time points may need to be adjusted accordingly.

TROUBLESHOOTING Problem 1
Silk fibroin not fully dissolved after 4 h in 60 C (Before you begin step 12).

Potential solution
Inaccurate LiBr solution concentration due to the stock LiBr powder absorbing too much moisture from air. Use new LiBr and keep the container closed and parafilm-sealed as much as possible.

Potential solution
Incubate the samples on the stage at 37 C for a longer duration after placing on the stage and before imaging. Check the actual temperature of the culture on the stage to ensure that it is within the range of 36 C-38 C. Overheating can quickly and irreversibly damage the culture. Check multiple 3D cultures. Treat 3D cultures with picrotoxin or bicuculline to induce activity (See Quality Control #2).

Potential solution
Run as an administrator. In PC, right click the software, select Run as Administrator.

Problem 8
How to choose the functional connectivity method (major step 75).

Potential solution
Given a dataset, different functional connectivity approaches result in different networks. It is important to notice that there is no privileged technique. Each method comes with pros/cons that must be considered by the user, whom should choose based on the type of question to answer and the mathematical meaning of each approach. For further discussion, please refer to ''Fundamentals of brain network analysis'' (Fornito et al., 2016).

Problem 9
FluoroSNNAP freezes at the end of functional analysis (major step 76).

Potential solution
FluoroSNNAP can crash if there is little or no activity to perform network analysis. Check the original time-lapse images to visually determine whether there is enough activity. Run only the ''Calcium Events Detection'' module in FluoroSNNAP and check the Ca 2+ peaks in the resulting .txt file. Restart the software and/or restart the computer.

Potential solution
Make sure Bioinformatics Toolbox is installed. Check folder directory.

Problem 11
How to choose a threshold for binarization (major step 82).

Potential solution
The choice of the binarization threshold is a crucial step since it shapes the resulting network. We can use prior knowledges of the system to determine the noise level of the weight edges between two nodes, thus filtering out irrelevant connections. If unsure about the choice of a threshold, analyze the weighted matrix.

Problem 12
Negative edges in the functional connectivity matrix. The computation of path length does not admit negative values. In addition, the circularGraph.m will not accept negative entries (major step 83).

Potential solution
The meaning of negative edge values in the computation of the path length is under debate. The first option is to choose an alternative functional connectivity method to avoid negative edges. In ll OPEN ACCESS addition, the user can rescale the functional matrix between zero and one. If the code detects negative edges, it will display a dialog box reporting how many there were; moreover, it automatically substitutes them with zero entries (to allow the program to run). Finally, we can take the absolute value of each edge. For a recent survey, please refer to (Hallquist and Hillary, 2019).

RESOURCE AVAILABILITY
Lead contact Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, David L. Kaplan (david.kaplan@tufts.edu).

Materials availability
This study did not generate new unique reagents.