Cortical spheroids derived from human IPSCs express CB1, DAGLα, MAGL, and FAAH

The ECS plays a dual role in the developing brain by modulating both growth cone directionality7,37,39 (Fig. 1a) and presynaptic feedback inhibition at mature synapses28,46 (Fig. 1b). Disruptions to this system during critical periods of development may have lasting impacts on neural circuitry building. We first examined ECS gene expression levels in cortical spheroids derived from control iPSCs and iPSCs from 3 children with the neurodevelopmental disorder ASD (Fig. 1c). Using quantitative RT-PCR, we found that DAGLα and MAGL, the principal synthetic and metabolic enzymes for 2-AG respectively, were expressed at significantly higher levels in cortical spheroids derived from autistic patients relative to controls (DAGLα: p = 0.007, MAGL: p = 0.012) (Fig. 1c). Across 3 independent cortical spheroid experiments, DAGLα expression increased from 0.102 ± 0.02 to 0.258 ± 0.05 -fold mRNA expression relative to TATA-BP. Even more dramatically, MAGL expression increased from 0.056 ± 0.01 to 0.351 ± 0.11 -fold mRNA expression relative to TATA-BP. Additionally, the enzyme responsible for anandamide metabolism, FAAH, had significantly higher expression (p = 0.014) in ASD spheroids relative to controls and increased from 0.050 ± 0.01 to 0.159 ± 0.04-fold mRNA expression relative to TATA-BP. Significant expression differences were not observed for CB1, DAGLβ or NAPE-PLD. We attempted to amplify CB2 receptor cDNA but no amplification was observed, leading us to conclude that this receptor is not expressed at detectable levels in our model. The lack of CB2 serves as an internal control for our model system because the use of dual-SMAD inhibition blocks mesodermal and endodermal differentiation and therefore the cortical spheroids do not generate microglia, the cell type which primarily expresses CB2 in the brain47,48.

Using immunofluorescent staining and imaging, we observed abundant cytosolic, synaptic, and neurite expression of CB1 in our 90-day old, control patient-derived cortical spheroids (Fig. 1d). We also observed the 2-AG enzymatic regulators, MAGL and DAGLα, in our 90-day old cortical spheroids, as well as FAAH. Interestingly, in concordance with our qRT-PCR results, we found that the area of DAGLα and MAGL was increased in cortical spheroids derived from 2 out of 3 of our ASD patient lines (Supplemental Fig. S1). DAGLα area significantly increased (p > 0.001) from 0.109 ± 0.01% in our control patient cell line to 0.795 ± 0.06% and 0.469 ± 0.05% in cortical spheroids derived from ASD patient 1 and 2, respectively. MAGL area was also significantly increased (p > 0.001), from 0.636 ± 0.06% in the control patient cell line to 1.34 ± 0.14% and 1.45 ± 0.05% in cortical spheroids from ASD patient 1 and 2, respectively. In the third patient, DAGLα was significantly increased (mean: 0.216 ± 0.02%, control patient vs ASD patient 3: p > 0.001) but MAGL was not significantly different from cortical spheroids derived from the control patient IPSCs (mean: 0.462 ± 0.04%) (Supplemental Fig. S1). These findings confirm the presence of the ECS in our cortical spheroid model and suggest increased expression of 2-AG enzymes MAGL and DAGLα may occur in concordance with previous observations of ECS alterations associated with ASD49,50.

We wanted to further analyze whether ECS components exhibit synaptic localization characteristic of fetal autocrine signaling. However, confocal microscopy is limited by resolution and prevents us from determining whether ECS components such as CB1 correctly localizes to the presynaptic compartment in our system. To overcome this limitation, we used stochastic optical reconstruction microscopy (STORM) which has a resolution of up to ~ 20 nm in x,y, allowing us to resolve individual synapses which have a synaptic cleft distance of ~ 20 nm26,51. Examples of excitatory and inhibitory synapses can be found in Fig. 1e, f, respectively. Using STORM, we analyzed 126 excitatory synapses and confirmed presynaptic localization of CB1 in our cortical spheroid model, consistent with previous findings in other models7,52,53. The median distance between CB1 and presynaptic marker VGLUT-1 was 0.060 µm, which is significantly smaller (p < 0.0001) than the median distance between CB1 and postsynaptic marker PSD-95 (0.141 µm) (Fig. 1g). This indicates that CB1 is closer to the presynaptic marker than the postsynaptic marker and preferentially localizes to the presynapse. The median distance between VGLUT-1 and PSD-95 (0.131 µm) was similar to distance between CB1 and PSD-95 and is consistent with our previous STORM measurements of synaptic cleft size26,27. Additionally, we investigated CB1 localization at inhibitory synapses (n = 47) and observed a significantly shorter (p = 0.014) median distance between CB1/presynaptic marker VGAT (0.046 µm) compared to CB1/postsynaptic marker gephyrin (0.062 µm) (Fig. 1h). Further, we observed presynaptic localization of DAGLα (Fig. 1i) and postsynaptic localization of MAGL (Fig. 1j) at excitatory synapses. The median distance between DAGLα and VGLUT1 (0.052 µm) was significantly shorter (p = 0.006, n = 77 synapses) than the distance between DAGLα and PSD95 (0.086 µm). MAGL was postsynaptic, with a significantly shorter (p > 0.001, n = 105 synapses) median distance between MAGL and PSD95 (0.045 µm) compared to the distance between MAGL and VGLUT1 (0.072 µm). These localizations are consistent with a developmental autocrine CB1 signaling paradigm34,35.

Since the number of CB1 molecules at inhibitory versus excitatory synapses could impact the effect of pharmacological treatment, we used STORM microscopy to analyze the distribution of CB1 receptor count at excitatory and inhibitory synapses. We found that CB1 receptors are more abundant at excitatory synapses (530 ± 25 CB1 molecules/synapse) than at inhibitory synapses (262 ± 14 CB1 molecules/synapse) in the outer, 100 µm of the cortical spheroid (p > 0.001, unpaired t-test) (Supplemental Fig. S2). Association of the presynaptic terminal with a postsynaptic process is indicative of synapse formation, and increased postsynaptic area is indicative of synaptic strengthening54. Our model recapitulates synaptic scaling at both excitatory and inhibitory synapses as demonstrated by the positive relationship between the molecular count of pre- and postsynaptic markers at a given synapse (Supplemental Fig. S2). We therefore sought to determine whether the number of CB1 molecules scaled with increased postsynaptic association. We observed a positive relationship between CB1 receptor count and postsynaptic marker count at both excitatory (PSD95) and inhibitory (gephyrin) synapses (Sup Fig. S2), suggesting that CB1 receptors exhibit synaptic scaling.

Thus, we have determined that human cortical spheroids express ECS machinery, and that CB1, the predominant ECS receptor type in the brain, localizes to presynaptic compartments at both excitatory and inhibitory synapses. We also observed DAGLα localization to the presynapse and MAGL localization to the postsynapse in excitatory synapses using STORM microscopy. Our data supports cortical spheroids as a model of the fetal ECS system.

Treatment with CB1 antagonist SR141716A increases the number and total area of excitatory synapses

Having established the expression and presynaptic localization of CB1 within our system, we sought to determine how ECS disruption impacts synaptogenesis. In order to selectively perturb CB1 during synaptogenesis, we allowed cortical spheroids to develop for 90 days, so as to not disrupt neural differentiation and migration preceding synaptogenesis. At 90 days old, our cortical spheroids model the mid-gestational fetal brain25, a critical window of development during which the brain undergoes rapid synaptic proliferation11. Disruptions to the spatial and temporal regulation of synaptogenesis during this critical window is thought to drive developmental disorders such as ASD20,55. At 90 days of development, we have previously demonstrated that our cortical spheroids exhibit both excitatory and inhibitory synapses26,27. Furthermore, these synaptic connections exhibit a high level of plasticity, and are readily altered by acute perturbations to either the intracellular cytoskeleton or extracellular matrix26,27. Thus, we have established a window to selectively observe how CB1 signaling contributes to the initial formation of synaptic connections and subsequent development of synaptic activity. In order to selectively disrupt the process of synaptogenesis, we acutely treated 90-day old cortical spheroids with selective CB1 antagonist SR141716A (SR) for 24 h and observed the resulting effects on excitatory and inhibitory synapses.

Using confocal image analysis (Fig. 2a), we determined the effects of SR treatment on excitatory and inhibitory synaptogenesis by independently measuring pre- and post-synaptic marker area. Cortical spheroids were stained with antibodies against excitatory synaptic markers [vesicular glutamate transporter 1 (VGLUT-1) and postsynaptic density protein 95 (PSD-95)] or inhibitory synaptic markers [vesicular GABA transporter (VGAT) and gephyrin (GEPHRYIN)]. We defined the area of overlap between presynaptic marker (VGLUT-1 or VGAT) and their respective postsynaptic marker (PSD-95 and GEPHRYIN) as a “synapse”. We determined the effect of SR on the number of synapses and size of synapses in the outer 100 µm of the spheroid using this method. Example confocal images used for analysis are given in Fig. 2b, c. Under basal conditions, our cortical spheroids have more excitatory synapses than inhibitory synapses27. However, we found that SR treatment impacts both excitatory and inhibitory synapses. SR treatment increased expression of excitatory synapses markers in a dose-dependent fashion, whereas increased inhibitory synapses were only observed at the lower dose of 30 nM SR. To compare the area of synaptic markers across treatment groups, we normalized the area of the synaptic marker to the endogenous β-actin-GFP expression in our cortical spheroids.

The area of excitatory presynaptic marker VGLUT-1 significantly increased from 32.4 ± 4.6% in the vehicle control to 76.7 ± 6.7% and 74.0 ± 7.8% in the 30 nM and 300 nM SR dose groups, respectively (0 vs 30: p < 0.001, 0 vs 300: p = 0.001) (Fig. 2d). The area of the excitatory postsynaptic scaffold PSD-95 also significantly increased from 14.9 ± 1.4% in the vehicle control to 27.8 ± 2.4% and 49.5 ± 3.0% in the 30 nM and 300 nM SR dose groups, respectively (0 vs 30: p < 0.001, 0 vs 300: p < 0.001) (Fig. 2d). Additionally, there was a significant, dose dependent relationship between the low and high doses of SR on PSD-95 expression (30 vs 300: p < 0.001) (Fig. 2d). Using the colocalization of excitatory presynaptic marker VGLUT-1 and postsynaptic marker PSD-95, we determined that SR treatment significantly increased both the total area (30 vs 300: p < 0.001) (Fig. 2f) and number (30 vs 300: p = 0.003) (Fig. 2g) of excitatory synapses in a dose dependent manner. Excitatory synapse area significantly increased from 1.6 ± 0.3% in the vehicle control to 5.7 ± 0.6% and 10.7 ± 1.2% in the 30 nM and 300 nM SR dose groups, respectively (0 vs 30: p < 0.001, 0 vs 300: p < 0.001) (Fig. 2f). Additionally, the number of excitatory synapses per area of actin significantly increased from 0.03 ± 0.005 synapses/µm2 in the vehicle control group to 0.05 ± 0.005 synapses/µm2 and 0.09 ± 0.007 synapses/µm2 in the 30 nM and 300 nM SR dose groups, respectively (0 vs 30: p = 0.009, 0 vs 300: p < 0.001) (Fig. 2g). The size of individual excitatory synapses trended towards an increase but was found not significant by Kolmogorov–Smirnov test, despite a rightward shift in the cumulative distribution plot (Fig. 2i). Having observed that CB1 antagonism increases excitatory synaptogenesis, we also sought to determine whether synaptic CB1 distribution was altered in response to SR treatment. We therefore examined CB1 localization to excitatory and inhibitory synapses. CB1 area as a percent of total excitatory synapse area significantly increased from 25.6 ± 1.5% of total excitatory synapses to 37.5 ± 1.6% percent of total excitatory synapses after application of 30 nM SR (0 vs 30: p < 0.001) (Fig. 2h). Surprisingly, the percent of CB1-positive excitatory synapses returned to a value similar to the vehicle control after application of 300 nM SR (28.4 ± 2.6% of total excitatory synapses) (0 vs 300: p = 0.719, 30 vs 300: p = 0.014) (Fig. 2h). Thus, the 30 nM SR treatment captures a window of dynamic ECS alterations at excitatory synapses.

Interestingly, the effect of SR on inhibitory synapses was greatest at the lower, 30 nM dose of SR. The area of presynaptic inhibitory marker VGAT significantly increased from 23.2 ± 2.3% in the vehicle control to 38.5 ± 3.8% in the 30 nM SR dose group (0 vs 30: p = 0.003) (Fig. 2e). A decrease in VGAT was observed in the high dose group (19.4 ± 1.5%) when compared to the low dose group (30 vs 300: p < 0.001) (Fig. 2e). When compared to the controls, the area of the postsynaptic inhibitory scaffold, gephyrin, was not significantly altered, however, there was a decrease between low and high doses (30 vs 300: p = 0.011) (Fig. 2e). There was a significant increase in the area (0 vs 30: p = 0.012) (Fig. 2f) and number (0 vs 30: p = 0.001) (Fig. 2g) of inhibitory synapses in the low dose SR group compared to the control group. The area of inhibitory synapses increased from 1.02 ± 0.1% in the control group to 1.63 ± 0.2% and 2.41 ± 0.5% in the 30 nM and 300 nM dose groups, respectively (0 vs 30: p = 0.012, 0 vs 300: p = 0.030) (Fig. 2f). The number of inhibitory synapses per area of actin changed from 0.015 ± 0.002 synapses/µm2 in the control group to 0.035 ± 0.005 synapses/µm2 and 0.021 ± 0.003 synapses/µm2 in the low and high dose groups, respectively (0 vs 30: p = 0.001, 30 vs 300: p = 0.047) (Fig. 2g). Unlike the results we observed earlier where there was a redistribution of synaptic CB1 at excitatory synapses, we did not observe significant differences in CB1 localization to inhibitory synapses and the percent of CB1-positive inhibitory synapses remained steady at approximately 40% (Fig. 2h).

In order to investigate mechanisms of increased synaptogenesis by CB1 antagonism, we used image analysis to measure the ratio of active RhoA to total RhoA (Fig. 3a). CB1 activation is associated with rapid growth cone retraction through the GTPase RhoA system39; additionally, antagonizing RhoA through ROCK inhibition increases excitatory synapse formation26. We therefore sought to determine if CB1 antagonism changed RhoA activation through ratiometric image analysis at VGLUT1-positive synapses. Activated RhoA was distinguished from total RhoA by an antibody targeting the GTP-bound form of RhoA compared to an antibody that distinguished total RhoA levels. Treatment of cortical spheroids with 30 nM and 300 nM SR141716A decreased the relative intensity of RhoA activation at excitatory synapses (0 nM vs 30 nM: p > 0.001, 0 nM vs 300 nM: p > 0.001) (Fig. 3b), consistent with the observed increase in excitatory synapses at these doses.

Thus, using the CB1 selective antagonist SR141716A, we successfully manipulated the cortical spheroid system, resulting in increased excitatory synaptogenesis. This increased excitatory synaptogenesis corresponded with increased inhibitory synaptogenesis and CB1 expression at excitatory synapses selectively at the lower, 30 nM SR treatment. These results demonstrate the functionality of the ECS in our cortical spheroids and suggest that 30 nM SR treatment could potentially restore excitatory and inhibitory synaptic balance in disrupted systems.

SR141716A increased variability of synaptic activity as measured by microelectrode array (MEA)

The effects of cannabinoid modulation on neural activity are complex due to CB1 localization at both glutamatergic and GABAergic synapses30.To address whether the complex changes in synaptogenesis altered the development of spontaneous activity in neural circuits, we used MEA to measure the extracellular field potential which corresponds to action potential. After 90 days of development, we dissociated cortical spheroids directly onto microelectrodes (Fig. 4a, b). In order to observe consistent activity in our control spheroids, we allow neurons to re-establish connections for an additional month after dissociation, resulting in reproducible activity measurements. We then measure spontaneous neural activity with or without SR treatment. Spontaneous extracellular activity caused by multiple, local action potentials is measured by the electrodes in units called “spikes” (Fig. 4c). Thus, a spike represents an increase in activity across a small area of multiple cells. Multiple spikes of 5 or more in quick succession (< 100 ms between spikes) are defined as “bursts” and represent rapid communication between populations of cells. Synchronous bursting between multiple electrodes is characteristic of mature communication patterns.

Figure 4
CB1 antagonism increases excitatory synaptogenesis in a
cortical spheroid model of fetal brain developmentCB1 antagonism increases excitatory synaptogenesis in a
cortical spheroid model of fetal brain development

SR141716A did not significantly increase WMFR or bursting frequency but increased variability (a) Schematic illustrating the process of cortical spheroid culture, dissociation, and plating. (b) Image of dissociated cortical spheroids on top of 16 microelectrodes. Image taken 19 days after dissociation. (c) Dissociated spheroids adjust to the MEA plate for 30 days. They are then treated with SR and recorded for 24 h. Raster plots of extracellular activity are analyzed for spiking and bursting activity. (d) The weighted mean firing rate (WMFR) of the vehicle control significantly decreased by 50% after 15 h (0 h vs 15 h: p = 0.0423, 0 h vs 18 h: p = 0.0131, 0 h vs 21 h p = 0.0094, 0 h vs 24 h p = 0.0099). There was no significant change to WMFR over 24 h of SR treatment. (e) Bursting frequency was significantly reduced after 6 h of vehicle treatment (0 h vs 6 h: p = 0.0442). (f) The average WMFR IQR of the control group over 24 h was 39.3 ± 4.7 compared to 61 ± 4.6 in the 3 nM group, 124 ± 7.3 in the 30 nM group, and 87 ± 7.7 in the 300 nM group. SR treatment significantly increased the WMFR IQR when compared to vehicle treated controls (VEH vs 3 nM: p = 0.026, VEH vs 30 nM: p < 0.001, VEH vs 300 nM: p = 0.001) (f). The variability of the 30 nM group was significantly higher than the 3 nM group (3 nM vs 30 nM: p < 0.0001) but the variability of the 300 nM group was significantly lower than the 30 nM group (30 nM vs 300 nM: p = 0.0166) (f) (g) Bursting frequency IQR within the control group over 24 h was 32.5 ± 3.5. SR increased the mean IQR to 110 ± 20 in the 3 nM group, 647 ± 88 in the 30 nM group, and 137 ± 27 in the 300 nM group. SR significantly increased bursting frequency variability (VEH vs 3 nM: p = 0.0221, VEH vs 30 nM: p = 0.0005, VEH vs 300 nM: p = 0.0267). Variability was greatest at 30 nM (3 nM vs 30 nM: p = 0.0008, 30 nM vs 300 nM: p = 0.0009). Effects in panels (d) and (e) were compared using one-way ANOVA with multiple comparisons. Effects in panels (f) and (g) were compared using Dunnett’s T3 multiple comparisons test. Data represented as mean ± SEM. Significance (*) defined by p < 0.05. Schematics in panel (a) and (c) were created with BioRender (https://biorender.com/).

We determined that our vehicle decreased the weighted mean firing rate (WMFR) of dissociated cortical spheroids over a period of 24 h. Immediately after dosing, the mean WMFR was 7.64 ± 7.0% above the pre-treatment average, however, after 3 h of treatment the WMFR started to decrease (-9.11 ± 6.95% of pre-treatment) and continued to decrease. The WMFR of the vehicle control group significantly decreased from the initial recording after 15 h of vehicle treatment (0 h vs 15 h: p = 0.0423, 0 h vs 18 h: p = 0.0131, 0 h vs 21 h p = 0.0094, 0 h vs 24 h p = 0.0099) (Fig. 4d). We believe that this decrease is attributable to nutrient depletion over time. Interestingly, SR treatment prevented this decrease in WMFR. More strikingly, we observed highly variable activity in response to SR treatment which was not observed in the DMSO vehicle treatment group. WMFR means of the 30 nM and 300 nM group over time were approximately 50% greater than their respective pre-treatment values but there was no significant effect found over time.

The bursting frequency data showed the same trends as the WMFR data. Bursting frequency of the vehicle control group was initially 48.31 ± 10.7% greater than pre-treatment bursting frequency at hour 0, but later decreased to 50% of pre-treatment values after 12 h. Specifically, the vehicle treated wells had significantly decreased bursting frequency continuing after 6 h of treatment (0 h vs 6 h: p = 0.0442, 0 h vs 9 h: p = 0.0062, 0 h vs 12 h p < 0.0001, 0 h vs 15 h: p < 0.0001, 0 h vs 18 h: p = 0.0010, 0 h vs 21 h: p < 0.000, 0 h vs 24 h: p = 0.0001) (Fig. 4e). While the low dose of SR showed a significant decrease in bursting frequency after 18 h (0 h vs 18 h: p = 0.0440, 0 h vs 21 h: p = 0.0308), this decrease took a longer time to manifest when compared to the vehicle treated group (Fig. 4e). Treatment with 30 nM of SR increased bursting frequency by over 200% for every timepoint, but this increase did not vary significantly across time. The high dose of 300 nM SR also increased overall bursting frequency by about 75% but there was no significant effect over time.

While we did not observe significant increases in WMFR and bursting frequency, we did observe a high degree of variability across our SR dose groups. We measured the effects of variability by utilizing the interquartile ranges (IQR) of each timepoint within dose groups and then compared dose groups. The average WMFR IQR of the control group over 24 h was 39.3 ± 4.7 compared to 61 ± 4.6 in the 3 nM group, 124 ± 7.3 in the 30 nM group, and 87 ± 7.7 in the 300 nM group. SR treatment significantly increased the WMFR IQR when compared to vehicle treated controls (VEH vs 3 nM: p = 0.026, VEH vs 30 nM: p < 0.001, VEH vs 300 nM: p = 0.001) (Fig. 4f). The variability of the 30 nM group was significantly higher than the 3 nM group (3 nM vs 30 nM: p < 0.0001) but the variability of the 300 nM group was significantly lower than the 30 nM group (30 nM vs 300 nM: p = 0.0166) (Fig. 4f). Our bursting frequency results parallel the results of the WMFR, where SR caused significantly more variability compared to the vehicle control (VEH vs 3 nM: p = 0.0221, VEH vs 30 nM: p = 0.0005, VEH vs 300 nM: p = 0.0267) (Fig. 4g). Bursting frequency IQR within the control group over 24 h was 32.5 ± 3.5. SR increased the mean IQR to 110 ± 20 in the 3 nM group, 647 ± 88 in the 30 nM group, and 137 ± 27 in the 300 nM group. Similar to WMFR measurements, we observed a biphasic dose response that displayed significantly more IQR variability in our 30 nM group than in our 3 nM group (3 nM vs 30 nM: p = 0.001) and less variability in our 300 nM group when compared to the 30 nM group (30 nM vs 300 nM: p = 0.001) (Fig. 4g). Increased variability of synaptic activity, particularly at the 30 nM dose of SR, parallels the complex and differential changes to excitatory and inhibitory synapse formation we observed in our confocal analysis of synaptic area.

Read original article here.