Page 358 - IJB-9-2

International Journal of Bioprinting Laser bioprinting of hiPSC-derived neural stem cells and neurons

identified activity events per neuron and minute, averaged interactions among neurons during the bursting episodes,
over all neurons in the network. which is the key dynamic difference across experimental
preparations. The GTE scores were finally set to 0 (absence
2.6.2. Global network activity (GNA) and distribution of connection) or 1 (connection present), shaping
of burst amplitudes directed yet unweighted connectivity matrices. For clarity
It quantified the capacity of the neurons to exhibit of language, the term “functional” was used instead of
coordinated activity and was computed by the fraction of “effective” throughout the description of results.
neurons in the network that were active together without
repetition in a sliding window of 2 s width and 0.5 s step. 2.6.5. Network measures
By construction, the GNA varied between 0 (no activity) The network characteristics of the inferred functional
and 1 (full network activation). Peaks in the GNA profile connectivity matrices were inferred using the “Brain
identified network bursts, i.e., neuronal coordinated Connectivity Toolbox” run in Matlab. Two main
[55]
activations that encompassed a significant part of the measures were considered, namely the Global Efficiency
network. Since neuronal sporadic activity was abundant, G eff [55,56] and the community statistic Q [55,57] . G varies
eff
a burst was deemed significant when its amplitude A between 0 and 1 and accounts for the capacity of a neuronal
b
verified A > µ bgnd + 3 SD bgnd , where µ bgnd and SD bgnd are the circuit to exchange information as a whole. G ≅ 0
b
eff
mean and standard deviation of background activity. All indicates that neurons tend to communicate locally and
significant burst amplitudes A , across realizations and for with few neighbors, while G ≅ 1 indicates that all neurons
eff
b
a given experimental condition, were pooled together to communicate among themselves across the entire network.
build the distribution of amplitudes. These distributions Q varies between 0 and 1 and describes the strength of
were finally compared among experimental conditions functional communities, i.e., groups of neurons that tend
to determine the tendency of a given preparation to show to be functionally more connected among themselves than
strong bursting. with the rest of the network. For Q ≅ 0, no structure is
detected, and the entire network shapes the only community.
2.6.3. Raster plots of collective activity A value of Q ≅ 0.3 typically indicates clear communities.
To prevent the abundance of neuronal sporadic activations The extreme case of Q = 1 corresponds to the situation in
from masking the computation of functional connectivity, which there are as many communities as neurons.
the original raster plots of each realization were processed
to generate new ones that contained only the neurons and 3. Results
corresponding activations of significantly strong network
bursts. Thus, these rasters excluded all uncorrelated Figure 1A provides a sketch of the cell printing technology,
background activity. The goodness of such a strategy was and Figure 1B shows the accuracy of our system to precisely
investigated through numerical simulations in a recent allocate cells in predefined locations.
study .
[51]
3.1. Change in the composition of cell cultures
2.6.4. Functional connectivity during differentiation
Causal relationships among pairs of neuronal spike NSCs were partially differentiated to neurons for 5, 10,
trains were inferred from the raster plots of collective or 20 days, or to astrocytes before being used for printing
activity, and using a modified version of the Generalized experiments. At any point in time, there is a mixture of
Transfer Entropy (GTE) [52-54] . Given a pair of spike already differentiated cells and undifferentiated NSCs. The
trains corresponding to neurons X and Y, an effective resulting cell composition was determined by fluorescence
connection was established between X and Y whenever imaging and Western blot analysis.
the information contained in X significantly increased Figure S1 (in Supplementary File) depicts phase
the capacity to predict future states of Y. To evaluate the contrast microscopic images of cells before dissociation for
effective connectivity among all pairs of neurons, binarized printing; while NSCs exhibited cortical rosettes, neurons
time series (“1” for the presence of a spike, “0” for absence) increasingly formed dendritic extensions during pre-
were constructed and computed in a fast implementation differentiation (d5, d10, d20 neurons). However, some cells
of GTE run in MatLab. Instant feedback was present, and with dendritic extensions could already be seen in the NSC
Markov Order was set as 2 . The actual GTE estimate was culture, which demonstrates spontaneous differentiation
[52]
then compared with those from the joint distribution of of NSCs to neurons. Thus, NSC and neuron cultures were
all inputs to Y and all outputs to X, setting a connection as not homogeneous.
significant whenever the GTE estimate exceeded the mean
+ 1.5 standard deviations of the joint distribution. This Fluorescence images with antibody staining for mature
threshold was considered optimal to capture the effective neuron marker MAP2, astrocytic marker GFAP, and

Volume 9 Issue 2 (2023) 350 https://doi.org/10.18063/ijb.v9i2.672