Optogenetic stimulation was used on mESC-derived MEBs to implement training regimens during two important stages of neural development: neurogenesis (while still in suspension) and synaptogenesis (seeded on functionalized glass or MEAs) (Fig. 1a). Training regimens consisted of periodic stimulation with 5 ms pulses at 20 Hz in 1 s intervals for an hour (Supplementary Fig. 1a). This regimen has been shown to enhance axonal growth30, and thus would suggest that it could lead to a shift in structural potentiation in a neural network. The regimen was repeated every 24 h as differentiation occurred within the EBs, with an expectation that consistent repetition would enhance the potentiation and cause long-term changes in the firing patterns of the network. Following established differentiation protocols of mESC towards mature motor neurons31,32,33, the described training regimen was started at D2 of differentiation, at which point stem cells have been induced towards neuronal lineages, and specialization and maturation of motor neurons has been shown to take place in the subsequent 7 days (Fig. 1b). Since one of the transcription factors that drove differentiation, retinoic acid, is light sensitive, media was changed every single day immediately after stimulation to ensure that stimulation effects on MEBs were not artifacts (i.e. false positives) caused by photodegradation of factors (Supplementary Fig. 1b)34. Furthermore, since the differentiation was monitored with the expression of the motor neuronal marker Hb9 through a GFP reporter, we used the plateau of GFP expression between D8 and D9, as an indicator that D9 was an appropriate time point for seeding the MEBs on glass (Supplementary Fig. 1c). Thus, after these 7 days (D2-D9) of differentiation, stimulated (S) and non-stimulated (NS) cultures were seeded on MEA chips (Fig. 1c). Careful seeding practices were applied to ensure that ~ 20 MEBs were seeded within the sensing area of the MEAs for a ~ 50% coverage by the MEBs (Supplementary Fig. 2). Seeding in this manner ensured empty space between clusters for the extension of processes, even though some nearby clusters would start fusing into larger clusters. The resulting two groups of samples seeded on MEAs were further subdivided into two more experimental groups, referring to whether or not a training regimen was continued during network formation on chip for the consequent 15 days (D10-D25). For ease of discussion, S or NS prior to a colon (e.g. S:X or NS:X) will refer to the presence or lack thereof of stimulation, during neurogenesis, while S or NS written after a colon (e.g. X:S or X:NS), indicates the presence or absence of stimulation during synaptogenesis (Fig. 1a).

Figure 1

Approach to training mESC-derived motor neuronal embryoid body networks during neurogenesis and synaptogenesis. a Representative diagram of experimental setup combining differentiating ChR2 mESC’s and MEAs. b Representative diagram of ChR2 mESC differentiation toward motor neuronal embryoid bodies monitored by the expression of GFP guided by the motor neuronal specific Hb9 promoter (scale bar: 200 µm). c Representative image of fabricated MEA chip. d Representative spontaneous spike trains from MEA recordings of cultured embryoid body networks.

Figure 2

Intact MEBs indicate formation of internal networks and form active networks between them a (i) Scanning electron micrograph of two embryoid bodies. (scale bar: 200 µm) and (ii) confocal image showing dense clusters of synaptophysin between cultured embryoid bodies (scale bar: 50 µm). b (i) MEB cryosections showing usual internal structure. (Scale bar: 50 µm) with (ii) zoom in of internal structure of a sectioned embryoid body (scale bar: 15 µm). c Representative confocal image of MEB cryosection stained for GAD65/67 and vGlut. Triangles show GAD65/67 clusters d. Representative confocal image of entire field of view for neural culture grown on the MEA sensing area (scale bar: 200 µm) with scanning electron micrograph zoom in of embryoid bodies extending processes atop of sensing electrodes. e. Bar graph for average firing rate of 15 active electrodes for cultured embryoid body networks exposed to known neuronal signaling molecules at sequential addition of tonic baths of 10, 100 and 250 µM. Glut Glutamate, ACh Acetylcholine, cAMP cyclic AMP, cGMP cyclic GMP, NE norepinephrine, GABA gamma-aminobutyric acid) across 5 min of recording/exposure (n = 15; error bar represents SEM, * p < 0.05; ANOVA with Tukey post-hoc test).

The electrical activity of the resulting neuronal cultures was measured with the MEA system and the raw data was filtered to remove low frequencies (< 200 Hz), to remove undesired voltage artifacts (e.g. stimulation artifacts), and extract action potentials recorded as spiking events (Fig. 1d). A two-step procedure was used to remove false positives from the analyzed data: (1) the detection threshold was set at a value at which no positives would be detected from the ground electrode, then (2) the recorded spikes at each electrode were inspected to ensure that the detected spikes had the appropriate voltage phases relating to action potentials: depolarization, repolarization and refractory period.

MEB cultures form active neural networks with excitatory and inhibitory populations

In this work, neural networks were cultured from intact MEBs, in contrast to growing them as a monolayer after dissociation. The long-term goal of our study is the modulation of electrical activity of the MEBs towards downstream implantation in in-vivo or in-vitro experimental systems and modulating the functionality of such systems through the resulting interaction. When cultured in their intact form, MEBs tend to keep their spheroid shape, while extending processes which contain neurites that form networks as they undergo synaptogenesis (Fig. 2a). Furthermore, dense web-like neurite structures form within the spheroid itself (Fig. 2b) and both excitatory (vGlut) and inhibitory (GAD65/67) receptors stain positively (Fig. 2c).

Network formation was validated by exposing MEB cultures grown on MEAs (Fig. 2d) to varying concentrations of commonly used exciting and inhibiting signaling molecules for 5 min: L-glutamate, acetylcholine, cyclic AMP, cyclic GMP, norepinephrine and GABA. (Fig. 2e). As expected, L-glutamate evoked a statistically significant (repeated measures ANOVA with a Greenhouse–Geisser correction, n = 15; F(1.28,17.89) = 18.78, p = 1.88E-4) response in the network. A post hoc Tukey test showed a statistically significant positive difference at p < 0.05 between 0 µM to 10 µM, while higher concentrations, 100 µM and 250 µM, showed a decrease in firing rate with the latter showing a statistically significant negative difference to the spontaneous firing rate, most likely related to excitotoxicity35. Other excitatory signaling molecules, acetylcholine and cyclic AMP, evoked a continuously excitatory response (repeated measures ANOVA; ACh (with Greenhouse–Geisser correction), n = 15: F(2.13,29.78) = 16.14, p = 1.31E-5 and cAMP: F(3,42) = 125.49,p = 4.20E-15) continued a gradual increase in firing rate with increasing concentrations. Cyclic GMP, another cyclic nucleotide similar in function as cAMP, failed to evoke any statistically significant effect on firing rate (repeated measures ANOVA with a Greenhouse–Geisser correction, n = 15; F(2.08,29.18) = 2.86, p = 0.07). On the other hand, the inhibitory neurotransmitters evoked statistically significant effects on the MEB-derived networks, with norepinephrine (repeated measures ANOVA, n = 15; F(3,42) = 81.43, p = 1.53E-17), showing a statistically significant decrease at p < 0.05 in a post hoc Tukey test from 0 µM to 10 µM, and 100 µM to 250 µM, while GABA (repeated measures ANOVA, n = 15; F(3,42) = 191.55, p = 1.60E-24) showed a statistically significant decrease in firing rate at p < 0.05 in post hoc Tukey test at each concentration. The responses corroborated the development of endogenously active neural networks expressing different kinds of receptors. The observations that MEBs extend processes within the body itself while responding to both excitatory and inhibitory signaling molecules would lead to the hypothesis that these MEBs could be forming intrabody circuits which could be “trained” during differentiation and have these changes last after network formation.

Stimulation during neurogenesis results in morphological changes in MEB cultures

The effects of stimulation during differentiation were initially observed in neurite extension and presynaptic protein clustering. While it has been reported that neurite outgrowth could be enhanced if neural populations simultaneously underwent optogenetic stimulation30, it was not clear if effects of the stimulation on MEBs done in suspension would still result in an increase of neurite extension when later seeded on chips, as this would indicate some stable long-term changes in the neuronal system. To quantify this, S:NS and NS:NS MEBs were seeded at low confluence on gridded coverslips and imaged 6 times every two hours on D10 (1 DIV) to quantify the number of extending neurites (Fig. 3a). Observations showed a consistently statistically significant positive difference (ANOVA, n = 20; 14hrs: F(1,38) = 215.44, p = 0.0; 16hrs: F(1,38) = 148.40, p = 1.08E-2; 18hrs: F(1,38) = 257.32, p = 0.0; 20hrs: F(1,38) = 199.14,p = 1.11E-2; 22hrs: F(1,38) = 221.35, p = 0.0; 24hrs: F(1,38) = 76.11,p = 1.31E-2) of number of neurites extended for S:NS samples, compared to NS:NS, for each of the six hours the two groups were measured and compared. This indicates an increased rate of neurite extension as a result of the stimulation during neurogenesis (Fig. 3b). Next, we wanted to observe the effect of stimulation during differentiation on the propensity of the network to form synapses. To quantify this, the clustering of presynaptic synaptophysin stained with anti-SY38, was counted along individual neurites as well as per unit area between the groups NS:NS and S:S (Fig. 3c). By D11 (2 DIV) S:S samples showed a statistically significant ~ twofold increase (ANOVA, n = 10; F(1,18) = 24.58, p = 1.02E-4) of synaptophysin clusters per neurite than NS:NS samples (Fig. 3d). This increase of pre-synaptic clusters per neurite combined with the increase in neurite extension resulted in S:S samples presenting a statistically significant higher synaptophysin clusters per unit area than NS:NS counterparts at D11 (ANOVA, n = 10; F(1,18) = 40.18, p = 5.68), D13 (ANOVA, n = 10; F(1,18) = 131.58, p = 1.04E-9) and D15 (ANOVA, n = 10; F(1,18) = 74.87, p = 7.88E-8) (Fig. 3e). When monitoring the difference of pre-synaptic clusters per unit area at D13 and D15, the statistically significant difference indicated that optogenetic stimulation during neurogenesis evoked physiological responses on two important aspects of neural network development: neurite extension and presynaptic clustering (Fig. 3e).

Figure 3

Stimulation during neurogenesis affects key morphological parameters of network formation. a. Representative phase contrast images of neurite extension along the periphery of embryoid bodies between non-stimulated (NS) and stimulated during neurogenesis (S) samples (scale bar: 50 µm). b. Bar graphs representing the average number of neurites protruding from the periphery of embryoid body normalized by the perimeter of the embryoid body at a given time after seeding. Each point signifies the number of extending neurites normalized by the perimeter of an individual embryoid body (n = 20; error bar represents SEM, *p < 0.05, ANOVA with Tukey post-hoc test). c. Representative fluorescence images of synaptic puncta stained against SY38 at D11 along a neurite. Arrow denote presynaptic puncta. (scale bar: 5 µm). d. Bar graphs representing the average number of presynaptic puncta along the length of neurites for D11. Each point corresponds to the average number of synaptic puncta along a neurite normalized the length of the neurite per field of view (n = 10; error bar represents SEM, *p < 0.05, ANOVA with Tukey post-hoc test). e. Bar graphs representing the average number of presynaptic puncta per unit area for D11-D15. Each point corresponds to the average number of synaptic puncta per unit area in an individual field of view (n = 10; error bar represents SEM, *p < 0.05, ANOVA with Tukey post-hoc test).

MEB network synchronicity is amplified by stimulation during neurogenesis and synaptogenesis

Network synchrony is a common parameter used to characterize a developing neural network, as it gives information on the network’s plasticity and connectivity. Various studies have successfully shown that the presence of chronic stimulation results in improved network synchrony36,37,38. In our study, we wanted to observe the long-term effects of stimulation regimens on the network synchrony and determine if these effects were amplified or shifted when the training regimen during neurogenesis was extended during synaptogenesis. From the raster plots of the spontaneous activity recorded at D21, the increased level of synchronous activity was notable between NS:S and S:S samples versus S:NS and NS:NS (Fig. 4a). This can be appreciated by the peaks above the raster plots, which correspond to a summation of the activity across all electrodes, where synchronous networks would result in discrete peaks whereas in samples that lacked coordinated firing, the resulting line plot seemed to lack any peaks.

Figure 4

MEB network synchronicity is amplified by stimulation during neurogenesis and synaptogenesis. a. Representative raster plots of MEB cultures at D25 showing network synchrony by line plots of the sum of active electrodes for each time point. b. The average correlation value (χ) was calculated for active electrodes across time for an average value for each electrode, then mapped to their respective spatial position on the MEA array. c. Bar graphs representing the mean correlation value across the culture for the MEA cultures at the different days of recording. The correlation value for the culture was calculated using active electrodes during spontaneous time of each culture for each day of recording. Each point corresponds to the correlation value across electrodes for each MEA culture. (n = 3; error bar represents SEM, *p < 0.05, ANOVA with Tukey post-hoc test).

Similarity between electrode recordings was quantified with cross-correlation in order to quantify synchronous behavior. Values for the similarity across the network were obtained by calculating cross-correlation for all electrode combinations (Supplementary Fig. 3). For this analysis, only spontaneous recordings of active electrodes (electrodes detecting at least 10 spikes/min) were used to quantify the long-term effects of the training regimen on steady state synchrony. When average correlation values per electrodes were mapped to their position on the chip, NS:S and S:S samples showed high synchrony level ((stackrel{-}{chi }) > 0.5) across the entire network for spontaneous recordings at D21 (Fig. 4b). This showed that synchronous behavior extended across the entire network and was markedly higher for networks that were stimulated during synaptogenesis.

Interestingly, when the network wide mean synchronicity was calculated for each recording day, a trend of higher synchrony was observed for samples that had been exposed to some form of training regimen (NS:S, S:NS or S:S) but no statistical significance was observed at D11 (ANOVA, n = 3; F(3,8) = 3.42, p = 0.073) and D13 (ANOVA, n = 3; F(3,8) = 1.77, p = 0.23). At D15, a statistically significant difference (ANOVA, n = 3; F(3,8) = 7.47, p = 0.010) was observed, with a post hoc Tukey test performed at p < 0.05 showing statistical significance between NS:S and S:NS (stackrel{-}{chi }) values. Subsequently, while no statistical significance was observed for D17 (ANOVA, n = 3; F(3,8) = 3.88, p = 0.055), D19 (ANOVA, n = 3; F(3,8) = 3.58, p = 0.066) and D21 (ANOVA, n = 3; F(3,8) = 3.61, p = 0.065), a gradual trend was observed for the synchronicity of networks undergoing training during synaptogenesis (NS:S and S:S) being larger than their counterparts (NS:NS and S:NS). At D23, there was a statistically significant difference among the experimental groups (ANOVA, n = 3; F(3,8) = 8.73, p = 6.6E-3). Post hoc comparisons using Tukey test at p < 0.05 indicated that the (stackrel{-}{chi }) value for NS:S and S:S were higher than both NS:NS and S:NS groups. This statistically significance was sustained for D25 (ANOVA, n = 3; F(3,8) = 6.46, p = 0.016), with the post hoc Tukey test showing significant difference between (stackrel{-}{chi }) for S:S and (stackrel{-}{chi }) for NS:NS as well as S:NS. (Fig. 4c).

Spectral density elucidates changes in steady state firing

Conventionally, electrophysiological behavior is characterized by firing rate during set epochs and burst parameters (Supplementary Fig. 4). However, when analyzing these parameters during spontaneous firing, there was no discernable trend in the change of long-term firing rate or burst parameters between experimental groups. However, when observing the spike data during steady state of a more mature neural network (D25), there were deviations on how the spike firing clustered into bursts, despite the fact that no clear change in the number of spikes was observed (Fig. 5a). We accredited this seeming conflict between the quantitative and qualitative data to the selection method of the burst detection parameters (See Quantification and statistical analysis). In order to avoid arbitrariness in the selection of these parameters, we decided to characterize the data in the frequency domain. For this reason, we focused on characterizing spontaneous firing recorded on MEAs by comparing changes in the power spectrums of recorded signals calculated through Fourier transforms (Fig. 5b). To obtain spectral profiles, binned spike counts were divided into 10-s-long contiguous windows and transformed to the frequency domain, thus representing the power spectrum as a function of time (Fig. 5b). When initially calculating the power spectral density (PSD) and observing between the DC frequency and the Nyquist frequency, we noticed that most of the components appeared below 7 Hz for all samples. For this reason, we compared samples between 0.1 Hz (to remove DC component) and 5 Hz. Focusing between 0.1–5 Hz, all samples except S:S, showed frequency profiles of their respective firing patterns with components across the entire bandwidth of interest. This spontaneous heterogeneous firing patterns can be expected from these cultures formed from MEBs, as they are a super-network composed of individual networks from within each MEB. On the other hand, S:S samples show a clear change in their frequency profile, where most of the spectral power fell within 0.1-1 Hz.

Figure 5

Stimulating training regimens modulates firing patterns in the frequency domain. a. Fifteen second representation of spontaneous voltage recording from NS:NS, NS:S, S:NS and S:S samples for D25. b. Smoothened (3 point moving average) and normalized (AUC) power spectra was calculated for contiguous 10 s windows across the 4 min of spontaneous recording NS:NS, NS:S, S:NS and S:S. Resulting matrices were averaged across samples. c. Bar graph for the sum of power spectral density magnitude from (b) across the spontaneous recording time between 0.1 Hz and 1 Hz (n = 3; error bar represents SEM, *p < 0.05, ANOVA with Tukey post-hoc test).

Moreover, if the signal power is summed between the frequency range of 0.1-1 Hz, the training regimen pattern had a statistically significant effect at p < 0.05 on the power magnitude within this frequency interval (ANOVA, n = 3; F(3,8) = 20.15, p = 4.37E-4). Post hoc comparisons using Tukey test at p < 0.05 showed a statistically significant difference between power magnitude withing 0.1-1 Hz of samples non stimulated during synaptogenesis (NS:NS, S:NS) and samples stimulated throughout development (S:S) (Fig. 5c). Moreover, the post hoc Tukey test indicated a statistically significant difference between power spectra values between NS:S and S:S, implying that combined stimulation of both neurogenesis and synaptogenesis had an amplified effect on modulating the power spectra of the networks than just stimulation during synaptogenesis. This statistical significance was not observed in the mature networks (D25: ANOVA, n = 3; F(3,8) = 0.063, p = 0.98) if the power was summed for the whole frequency interval of interest (0.1-5 Hz) (Supplementary Fig. 5).

Neurogenetic stimulation changes the opto-response of MEB networks

Another aspect of consideration on the effect of training MEBs during neurogenesis was whether the early stage perturbation had some effects on how the later-stage network would respond to the same perturbation. To study this, we recorded responses to optogenetic stimulation from sets of samples that had not undergone the training regimen during neurogenesis (Fig. 6a) and compared them to those set that had undergone such regimen (Fig. 6b). Initial observation showed a difference between how the networks responded when stimulated early in the network development (D11) versus more mature networks (D25). For example, when early networks, which had a low spontaneous firing rate (D11) were stimulated, there would be a very notable evoked response during stimulation followed by a quiescent state, where the network would barely fire before returning to the baseline spontaneous firing rate. In contrast, more mature networks (D25), would still show an evoked response during stimulation but would automatically return to baseline firing rate right after stimulation ceased. What was interesting was that the quiescent time after stimulation for early S:S networks were notably shorter than those from the NS:S samples (Fig. 6a-b). Moreover, at D25, while NS:S samples would return to the same baseline firing rate right after stimulation stopped, S:S samples showed a transient change in firing rate for several seconds after the stimulation stopped (Fig. 6a-b).

Figure 6

Stimulation during neurogenesis alters response to stimulation during network formation. Summed spike counts per each 100 ms for all active electrodes across the 20 min of recording were graphed for D11 and D25 for one representative sample from NS:S (a) and S:S (b). c. Zoom-in of a for 1 min, centered around the 20 s of stimulation at D25 for sample NS:S, the arrows represent the firing rate interval prior to stimulation (FRpre), the firing rate during stimulation (FRstim) and the firing rate after stimulation (FRpost). d. Bar graphs showing the mean firing rate increase between Frstim/Frpre for D11-D25. (n = 9; error bar represents SEM, *p < 0.05, ANOVA with Tukey post-hoc test)). e. Bar graphs showing the firing rate increase between Frpost/Frpre for D11-D25. (n = 9: error bar represents SEM, *p < 0.05, ANOVA with Tukey post-hoc test)). f. Raster plot of average correlation value for each electrode during 10 s bins across the entire recording time. g. Ratio of average correlation value prior to stimulation during recording and correlation value post stimulation (χpost/ χpre). (n = 3; error bar represents SEM, *p < 0.05, ANOVA with Tukey post-hoc test).

To quantify this behavior, the evoked firing rate during stimulation (FRstim) and the post-response firing rate (FRpost) were compared to the firing rate prior to stimulation (FRpre) for the three instances of stimulation within recording for each of the three MEA networks for both experimental groups (Fig. 6c). While the fold-change increase of firing rate FRpre to FRstim decreased with time for both NS:S (repeated measures ANOVA with Greenhouse–Geisser correction, n = 3; F(1.48, 11.83) = 14.79, p = 1.12E-3) and S:S (repeated measures ANOVA with Greenhouse–Geisser correction, n = 3; F(1.88, 15.02) = 11.02, p = 1.31E-3 (because more mature networks would have a higher baseline firing rate), when comparing the amount of evoked action potentials during stimulation (FRstim/FRpre), S:S samples seemed to respond more strongly to stimulation than NS:S samples (Fig. 6d). One-way ANOVA determined a statistically significant difference between NS:S and S:S FRstim/FRpre values for D13 (n = 9; F(1, 16) = 5.55, p = 0.031), D15 (n = 9; F(1,16) = 5.90, p = 0.027), D17 (n = 9; F(1,16) = 11.30, p = 4E-3), D19 (n = 9; F(1,16) = 8.78, p = 9.2E-3), D23 (n = 9; F(1,16) = 10.81, p = 4.6E-3) and D25 (n = 9; F(1,16) = 9.94, p = 6.2E-3), while only showing a trend (not statistically significant) of higher S:S FRstim/FRpre values for D11 (n = 9; F(1,16) = 4.48, p = 0.05) and D21 (n = 9; F(1,16) = 1.1, p = 0.31).

Additionally, the quiescent state response post-stimulation observed in early days (D11, D13 and D15), reflected itself in FRpost being less than FRpre, resulting in FRpost/FRpre < 1 for NS:S and S:S samples. We observed that this transient decrease in firing rate was statistically significantly shorter for the S:S samples than the NS:S for D11 (ANOVA, n = 9; F(1,16) = 19.95, p = 3.9E-4) and D13 (ANOVA, n = 9; F(1,16) = 9.49, p = 7.2E-3) (Fig. 6e). Repeated measured ANOVA indicated that FRpost/FRpre ratios increased for both NS:S (Greenhouse–Geisser corrected, n = 9; F(3.06, 24.48) = 36.92, p = 2.69E-9) and S:S (n = 9; F(7,56) = 5.66, p = 5.63E-5). Furthermore, at later days of network development, it was notable that FRpost/FRpre was ~ 1 for NS:S, meaning that the steady state firing rate was indistinguishable from that immediately following the termination of stimulation. On the other hand, S:S samples showed FRpost/FRpre values above 1 from D17 forward, indicating that the network would transiently increase in firing rate right after stimulation. One-way ANOVA showed that this increase between FRpost/FRpre values for S:S and NS:S was statistically significant for D17 (n = 9; F(1,16) = 12.19, p = 3E-3), D21 (n = 9; F(1,16) = 6.94, p = 0.018) and D23 (n = 9; F(1,16) = 9.91, p = 6.23E-3), while only showing a non-statistically significant trend for D19 (n = 9; F(1,16) = 2.16, p = 0.16) and D25 (n = 9; F(1,16) = 3.76, p = 0.071). It is relevant to mention that these effects were observed while there was no perceivable change in efficiency of the blue light to activate the ChR2 ion channels and evoke a response in the networks (Supplementary Fig. 6). These observations were corroborated by repeated measures ANOVA performed at p < 0.05, which showed no statistically significance change in efficiency (repeated measures ANOVA, n = 12; F(2,22) = 1.25, p = 0.31).

To further study how the training regimens affected network response, we also quantified the evoked response reflected in the network’s synchronicity for the initial stimulation done on the initial spontaneous interval of recording. For this purpose, raster-plots of the average values of cross-correlation (as calculated for the analysis in Fig. 4) were calculated using 10 s bins across the entire 20 min of recording (Fig. 6f). When quantifying the short term effect of stimulation during recording had on network synchronicity, by comparing (stackrel{-}{chi }) post to (stackrel{-}{chi }) pre, a trend was observed where the presence of a training regimen during neurogenesis seemed to cause the correlation fold-change ((stackrel{-}{chi }) post/(stackrel{-}{chi }) pre) for S:S samples to be higher than NS:S samples. One-way ANOVA detected a statistically significant difference between (stackrel{-}{chi }) post/(stackrel{-}{chi }) pre for S:S and NS:S for days D19 (n = 3; F(1,4) = 16.49, p = 0.015) and D23 (n = 3; F(1,4) = 11.12, p = 0.029) (Fig. 6g).

Changes evoked by stimulation during neurogenesis result in genetic changes

Given the effects on neurite extension, presynaptic clustering, frequency profiles and network response to stimulation that were observed as a result of the presence of training regimens on MEBs during neurogenesis, we proceeded to determine genetic changes that could provide possible mechanistic explanations. Total messenger RNA sequencing was performed and analyzed for stimulated (S) and non-stimulated (NS) MEBs at D9, as well as EBs at D2. The differentially expressed genes in MEBs that underwent training regimens during neurogenesis were compared to those that did not, both with respect to the genetic expression of EBs sampled prior to differentiation (at D2). A total of 749 differentially expressed genes between S and NS with p < 0.05 were detected and clustered and color coded with respect to the differential expression of D2 (Fig. 7a). There were 200 genes that were upregulated during control differentiation, but this upregulation was lessened for samples that underwent training regimen (black bar), while the upregulation of 172 genes was amplified for those same samples (red bar). On the other hand, there were 202 genes whose downregulation was stagnated for samples with training regimen (yellow bar). For 173 genes, the control downregulation was further amplified after stimulation (blue bar). Something important to note was that this observed differential expression did not include changes in phenotype populations, matching the immunostaining observations (Supplementary Fig. 7). This indicated that training regimen during differentiation did not seem to noticeably disrupt the rate of phenotype specification or generation of the neural populations that generally result from the differentiation protocol (Table 1). This suggests that training regimens affected other functional pathways rather than altering the differentiation of populations. For further analysis, a more stringent threshold (p < 0.0005) was set to detect the most promising genes as key factors for the behavioral changes seen in stimulated MEB cultures. This threshold resulted in 97 differentially expressed genes for the black cluster (Fig. 7b), 63 differentially expressed genes for the red cluster (Fig. 7c), 77 differentially expressed genes for the yellow cluster (Fig. 7d) and 71 differentially expressed genes for the blue cluster (Fig. 7e). From this pool, a thorough literature study was used to identify gene targets that had been reported to be related to known neural development and function (Table 2, Supplementary Fig. 9).

Figure 7

RNA Sequencing shows differential expression as a result of optical stimulation during neurogenesis. a. Heat map of standard deviation of differential expression for genes with p < 0.05 (n = 749). Genes were primarily clustered for: (1) genes that would overexpress during differentiation and underexpressed due to stimulation, (2) genes that would overexpress during control differentiation and overexpressed further due to stimulation, (3) genes that would underexpress during control differentiation and stimulation minimized that underexpression and (4) genes that would underexpress during control differentiation and stimulation amplified that underexpression. (first color column in order: black, red, yellow, blue). Significantly differentially regulated genes, with p < 0.0005 (n = 307) were extracted as column plots for: b. black, c. red, d. yellow and e. blue clusters.

Table 1 Expression comparisons for phenotypic gene targets.
Table 2 Significantly (p < 0.0005) differentially expressed genes reported in literature as regulators of neural development.

Source link

Leave a Reply

Your email address will not be published. Required fields are marked *