Computational modelling of schizophrenia-associated alterations of ion-channel-encoding gene expression predicts a decrease in delta power

Figure 1:. Fitting of reduced-morphology interneuron models. A-D: Complete and reduced morhologies of LBC(A-B) and NGC(C-D) neurons respectively. E-F: f-I curves of the complete morphology model versus the best-fit reduced-morphology models for LBC(E) and NGC(F). Inset: Absolute error of various reduced morphology dendrite lengths; in blue a grid search over a large range of lengths, with a large step length; in orange a fitting around the lowest error range with smaller step lengths to zoom in on the best fit. G-H: A fitting of various numbers of compartments per dendrite in the reduced morphologies of LBC(G) and NGC(H), where 19 compartments per dendrite was the value used when fitting lengths. Inset: Error values for 1, 3, 5, 7 and 9 compartments.

Figure 2:. Illustration of the connections, firing profiles and the spectral power of the model network. A: Connection diagram between populations, with connection probabilities derived from the literature [16] and scaled by a factor of 6. B: Population spike trains of all neurons in the control network model. C: Sample power spectrum for the PC population from 0 Hz to 20 Hz. D: Soma potentials of a single neuron from each population; NGC(blue), LBC(red) and PC(green).

Figure 3:. Predicted alterations in delta power based on expression data. A: Spectral power (left) and average delta power (right) of L5PC spiking activity from simulations of N=30 networks, smoothed by LOESS [39]. CommonMind data from ACC (purple) and PFC (gray) were used for modelling the effects of population-averaged SCZ-associated alterations of gene expression in the respective brain regions, lighter respective colors representing each random seed. The default network model was used for the HC (brown) case. A significant decrease in delta power was observed for both SCZ groups. B: Average spectral power for other frequency bands (theta: 5–8 Hz, alpha: 8–12 Hz, beta: 13–30 Hz, gamma: 30–150 Hz). A significant decrease is seen in the ACC region for all the frequency bands, and in the PFC region for the lower three frequency bands. C: Spectral power (left) and average delta power (right) of subject-wise simulations based on ACC gene-expression data. In these simulations, the ion-channel conductance coefficients (see Table 1) were calculated for each subject (both HC and SCZ) by normalizing the individual gene expressions by the mean expression levels in the HC population. The left panel…

Figure 4:. Ion-channel and neuronal population effects on delta power and f-I curves. A: Average delta power from simulations of N=30 networks where the SCZ-associated expression-level alterations were implemented in only one neuron population at the time (as opposed to Figure 3 where the ion channel conductances of all neurons were altered). Similar to Figure 3A, we used population-averaged ion-channel coefficients (Table 1). A significant decrease in delta power is found for CACNA1C and CACNA1D for both ACC and PFC. P-values: ACC CACNA1C and CACNA1D 0.015, ACC CACNA1I 0.515, ACC HCN1 0.894, ACC KCNB1 0.626, ACC KCNQ3 0.965, ACC GABBR2 0.450, PFC CACNA1C and CACNA1D 0.001, PFC CACNA1I 0.871, PFC HCN1 0.848, PFC KCNB1 0.408, PFC KCNQ3 0.337, PFC GABBR2 0.762. B: Summed delta power from simulations of N=30 random seeds per population for all population, performed with the average population ion-channel coefficients. A significant decrease in delta power is found in the PC population for both ACC and PFC. P-values: ACC PC 6.73·10−4, LBC 0.723, NGC 0.976, PFC PC 6.37·10−4, LBC 0.636, NGC 0.183.C,D,E: The effect of genetic variants on the f-I-curves of each simplified neuron; PC(top),…