Blocking NMDA, AMPA and GABAA receptors affected the firing patterns

To understand the effect of the drugs on the electrophysiological activity of the neuronal networks, we first considered the single-channel-level dynamics by analysing the normalized IFR during the spontaneous and chemically evoked activity. The 100E configuration—although not significant with respect to the spontaneous (i.e. not treated) initial condition—showed a slow decrease of the IFR trend when exposed to APV (Fig. 2a, Table S1, S2), while CNQX led to a rapid decrease of the 100E firing, noticeable within the first 5 min after that the drugs were added to the medium (Fig. 2d, Table S1, S2). On the other hand, BIC—as expected—maintained the firing pattern of this configuration constant (Fig. 2g, Tables S1, S2). Concerning the 75E25I configuration, a similar trend can be observed with respect to the 100E, when APV (Fig. 2b) and CNQX (Fig. 2e) were added to the medium: both drugs caused a decrease of the IFR with a faster timing when CNQX was used (Tables S1, S2). BIC, instead, resulted in an increased firing activity of the 75E25I (Fig. 2h, Tables S1, S2), noticeable from the first 5 min after the presence of the drug. Lastly, the 100I configuration, in the case of APV treatment (Fig. 2c, Tables S1, S2), showed a decrease—not statistically significant—in the IFR during the first 25 min, stabilizing again close to the values of the spontaneous activity after about 30 min, while exhibited an increase when treated with BIC, showing a similar timing of action with respect to the 75E25I configuration when treated with the same drug.

Fig. 2

Spiking activity characterization during the chemical modulation. a–i Normalized Instantaneous Firing Rate (IFR) of 100E (red), 75E25I (orange), and 100I (blue) configurations when APV (first row), CNQX (second row), or BIC (third row) were used. The timepoint “Spont” represents the normalized firing rate computed over the 10-min recording of the spontaneous activity. Data are represented with the mean (dot) and the standard error of the mean (whiskers) (∗ refers to p < 0.05)

Since the major changes in the firing patterns mostly occurred within the first 20 min after the drug delivery, we focused the analysis of the IFR in this range with a smaller bin size, i.e., 1 min (Fig. S1). The trends of the 1-min-binned IFR for each configuration and each drug showed a higher variability during the first 5–10 min due to the possible perturbation of the drug injection within the neuronal networks [9]. After this interval, the effect of the drugs on the IFR stabilized. For this reason, we chose to analyse more thoroughly the differences between the spontaneous and the chemically evoked activity between 10 to 15 min after the drug addition (see Sect. “Short-term effects of drug treatment”). Finally, concerning the bursting pattern, the 100I configuration exhibited a number of bursting units under the threshold for all the cultures (Fig. S2a)—as expected—since these neuronal networks were characterized by a tonic firing pattern (Fig. S3a) and uncorrelated activity (Fig. S3b). For this reason, the neuronal networks composed of exclusively GABAergic neurons (i.e., 100I) were excluded from further analysis related to network bursting events and functional connectivity investigations.

APV and CNQX abolished the network events

To understand the effect of the drugs on the network-level activity, we initially evaluated the electrophysiological activity from a qualitative point of view by observing the raster plots and the cumulative instantaneous firing rate profiles (Fig. 3a). Both configurations, i.e., 100E and 75E25I, showed sustained spontaneous activity characterized by repeated bursting and network bursting events (Fig. 3a). In the presence of APV or CNQX, the network bursting events were completely abolished in both configurations resulting in solely random activity (Fig. 3a). On the other hand, BIC appeared to slightly decrease the rate of network events in the 100E, while it increased the frequency of network bursts in the 75E25I. Quantifying the NBR, it revealed that both APV and CNQX drastically reduced the NBR in the 100E (Fig. 3b, c, Table S3, S4) and 75E25I (Fig. 3e, f, Table S3, S4) configurations. It is worth to highlight that both drugs affected the rate of the network bursting events of the physiological configuration (i.e., 75E25I) since the first 5 min after that the drugs were added to the medium, while the effects on the fully excitatory networks were appreciable after 10 and 5 min from the APV and CNQX presence, respectively. Finally, BIC halved the frequency of network events of the 100E networks (Fig. 3d, Table S3, S4), while, conversely, caused a significant increase in the network bursting rate of the 75E25I configuration (Fig. 3g, Table S3, S4).

Fig. 3

a Representative raster plots and respective cumulative instantaneous firing rate profiles (bin = 10 ms, overlapped) of the electrophysiological activity of the 100E (red) and 75E25I (orange) configurations in spontaneous conditions (first column) and after APV (second column), CNQX (third column), and BIC (fourth column) administration. A black dot represents a detected spike, while a dense black band indicates a network burst event. b-g Network bursting activity characterization during the chemical modulation. Normalized Network Bursting Rate (NBR) of 100E (red) and 75E25I (orange) configurations when APV (b, e), CNQX (c, f), or BIC (d, g) were used. The timepoint “Spont” represents the normalized NBR computed over the 10-min recording of the spontaneous activity. Data are represented with the mean (dot) and the standard error of the mean (whiskers) (∗ refers to p < 0.05 and ∗ ∗ to p < 0.01)

Short-term effects of drug treatment

To better analyse the impact of the drugs on network bursting patterns, we compared the spontaneous activity with the chemically evoked activity between 10 and 15 min after drug treatment (see Sect. “Blocking NMDA, AMPA and GABAA receptors affected the firing patterns”). Major changes were exhibited by the network bursting activity, in which both APV and CNQX induced a reduction of the network bursting units (Fig. 4a, Tables S5, S6) in both 100E and 75E25I, but significant only for the 100E. On the other hand, BIC induced the generation of network events characterized by a comparable number of units for both 100E and 75E25I (Table S5, S6). The NBR resulted to be highly affected by the drugs (Fig. 4b). In particular, both APV and CNQX caused a strong, statistical decrease with respect to the spontaneous activity in the NBR of both 100E (pAPV = 0.0463, pCNQX = 0.0150) and 75E25I (pAPV = 0.0463) configurations, leading to values close to zero (NBR100E_APV = 0.06 ± 0.03, NBR75E25I_APV = 0.01 ± 0.01, NBR100E_CNQX = 0.13 ± 0.10, NBR75E25I_CNQX = 0.09 ± 0.08) proving the abolishment of the network bursting activity, as hinted by the raster plots of Fig. 3a. Conversely, BIC led to a statistical increase in the rate of the network events in the 75E25I (NBR75E25I_BIC = 3.09 ± 0.40), and a slight—but statistical—decrease in the 100E configuration (NBR100E_BIC = 0.44 ± 0.15), as qualitatively suggested by the raster plots in Fig. 3a. Since APV and CNQX completely suppressed the network events of the cultures, we then investigated changes in the average network burst shapes and in the network burst duration (NBD) in the neuronal networks treated with BIC (Fig. 4c, d, Table S7). The network burst shape showed an increased amplitude, both in 100E and 75E25I when treated with BIC. Moreover, an increase in the network burst duration can be observed in both configurations (NBD100E_BIC = 1.90 ± 0.49, NBD75E25I_BIC = 1.48 ± 0.20), albeit not statistical.

Fig. 4

Drugs' effect in the 10–15 min after the administration. a Plot of the normalized number of network bursting (NB) units for 100E (red) and 75E25I (orange) configurations. b Box plots of the normalized Network Bursting Rate (NBR) for each configuration. c, d Average network burst shapes (Spike Time Histogram, STH) for the spontaneous (grey) and the BIC-modulated activity for the 100E (c, red) and 75E25I (d, orange) configurations. Inset: box plots of the Network Burst Duration (NBD) for each configuration in spontaneous (grey) and BIC modulated (red or orange) conditions. Scatter plots are represented with the mean (dot) and the standard error of the mean (whiskers). Box plots are represented with the percentile 25–75 (box), the standard deviation (whiskers), the median (line), the mean (square), and the minimum and maximum (crosses) values (∗ refers to p < 0.05)

Drug-driven changes in the functional connectivity

The neuronal connectivity showed to be affected by the drug treatments as well as the neuronal dynamics. Specifically, the use of APV resulted in a drastic reduction in the strength of the connections in both the 100E and 75E25I configurations, as shown in the connectivity maps and graphs (Fig. 5). On the other hand, it was difficult to evaluate qualitative changes in the number of links and the number of nodes from such intricate structures, although a reduction of both features was suggested. For what concerns CNQX, the AMPA receptor antagonist caused a reduction of the connection strength and the number of links and nodes as well as in the case of APV usage in both 100E and 75E25I (Fig. 5). Finally, BIC did not induce observable changes in the connectivity of the 100E configuration, while a substantial increase in the strength and in the number of links can be observed in the connectivity map and graph of the 75E25I configuration (Fig. 5). As a step forward, we computed the number of nodes and links during the chemical stimulated activity, by distinguishing the new ones, those which remained stable those which disappeared when treated with the drugs (Fig. 6). APV led to a reduction of both nodes and links, especially in the 100E configuration (Fig. 6a, d), as well as CNQX, in which a high number of extinguished nodes and links can be observed (Fig. 6b, e). On the other hand, BIC usage confirms to not affect particularly the 100E configuration (Fig. 6c), while in the 75E25I configuration led to a noticeable increase in both the number of new nodes (28%) and the number of new links (70%) (Fig. 6f). Finally, we evaluated the Cpeak values (Table S8), indicating the strength of the connections. The APV usage showed a decrease in the Cpeak values in both 100E and 75E25I configurations (Fig. 6g, h, Cpeak_100E_spont = 0.23 ± 0.03, Cpeak_100E_APV = 0.05 ± 0.03, Cpeak_75E25I_spont = 0.22 ± 0.06, Cpeak_75E25I_APV = 0.13 ± 0.06), although statistically significant only in the case of 100E (pAPV = 0.018). On the other hand, CNQX showed a slight decrease in the Cpeak value exclusively for the 75E25I configuration (Fig. 6h, Cpeak_75E25I_CNQX = 0.13 ± 0.06). Lastly, BIC did not cause any changes in the strength of the connections of the 100E (Fig. 6g), leading to an increase of such value only in the heterogeneous configuration (Fig. 6h, Cpeak_75E25I_BIC = 0.37 ± 0.08).

Fig. 5

Functional connectivity maps and graphs. Representative connectivity maps and connectivity graphs of 100E (red) and 75E25I (orange) configurations. In each pair, the functional connectivity map/graph of the spontaneous activity (on the left) is compared with the one after the drug treatment (on the right). In the connectivity maps, the connection between two units is represented by a pixel which colour represents its strength. In the connectivity graphs, each node is represented by dots and functional connections are represented with edges

Fig. 6

Characterization of the functional connectivity. a–f Pie charts of the number of edges and number of links new, unchanged, and extinguished with respect to the spontaneous activity for each configuration and for each administrated drug. g–h Box plots of the Cpeak values of the 100E g and 75E25I h configurations. Box plots are represented with the percentile 25–75 (box), the standard deviation (whiskers), the median (line), the mean (square), and the minimum and maximum (crosses) values (∗ refers to p < 0.05)

The presence of PTX and PTZ showed comparable effects with respect to BIC treatment

To conclude our work, we investigated the effects of two further drugs, i.e., PTX and PTZ, antagonist of GABAA receptors. Retracing the main steps highlighted for APV, CNQX, and BIC, we first evaluated the IFR of the neuronal networks after the drug treatment. PTX did not affect the firing pattern of the 100E configuration (Fig. 7a, Table S9, S10), while PTZ caused a slight decrease during the first 10 min, rapidly restored after 15 min (Fig. 7d, Table S9, S10). For what concern the effect of PTX and PTZ on the 75E25I configuration, the presence of such drugs showed similar effect with respect to BIC usage, that is an increase of the IFR (Fig. 7b, e, Table S9, S10). Nevertheless, a different time of action of the drugs can be observed: while BIC effect was quite instantaneous (noticeable within the first 5 min), PTX and PTZ showed a slower increase of the IFR, appreciable after about 15/20 min after the drug treatment. Alongside, the 100I configuration was affected exclusively by the usage of PTZ, which induced an increase of the IFR (Fig. 7f, Table S9, S10), while PTX did not provoke any changes in the firing pattern (Fig. 7c, Table S9, S10). Concerning the neuronal network events, the NBR of the 100E configuration after both PTX and PTZ treatment showed a decrease as noticed for the BIC usage (Fig. 7g, j, Table S11, S12). On the other hand, the 75E25I configuration showed different trends when treated with PTX or PTZ (Table S11, S12): the first one caused an initial slight decrease of the NBR, rapidly restored within 15 min (Fig. 7h); the second one provoked a slight increase of the NBR, which continued throughout the recording time (Fig. 7k). Finally, both PTX and PTZ led to an increase of the Network Burst Duration (NBD; Fig. 7i, l; NBD100E_PTX = 1.43 ± 0.26, NBD75E25I_PTX = 2.01 ± 0.42, NBD100E_PTZ = 1.83 ± 0.56, NBD75E25I_PTZ = 1.31 ± 0.30, Table S15), although without significant differences (Table S16), in both configurations, as well as in the case of BIC.

Fig. 7

Characterization of PTX e PTZ effects. a–c Normalized Instantaneous Firing Rate (IFR) over time of the 100E (red), 75E25I (orange), and 100I (blue) configurations when PTX was used. d–f Normalized IFR over time of the 100E (red), 75E25I (orange), and 100I (blue) configurations when PTZ was used. g–h Normalized Network Bursting Rate (NBR) over time of the 100E (red) and 75E25I (orange) configurations in the presence of PTX. Insets: Normalized number of network bursting units of the configurations in the spontaneous and PTX evoked conditions. i Normalized Network Burst Duration (NBD) of the 100E (left) and 75E25I (right) configurations in spontaneous (grey) and PTX evoked conditions (red and orange, respectively). j–k Normalized Network Bursting Rate (NBR) over time of the 100E (red) and 75E25I (orange) configurations in the presence of PTZ. Insets: Normalized number of network bursting units of the configurations in the spontaneous and PTZ evoked conditions. l Normalized Network Burst Duration (NBD) of the 100E (left) and 75E25I (right) configurations in spontaneous (grey) and PTZ evoked conditions (red and orange, respectively). Scatter plots are represented with the mean (dot) and the standard error of the mean (whiskers). Box plots are represented with the percentile 25–75 (box), the standard deviation (whiskers), the median (line), the mean (square), and the minimum and maximum (crosses) values (∗ refers to p < 0.05)