Untangling Basal Ganglia Network Dynamics and Function: Role of Dopamine Depletion and Inhibition Investigated in a Spiking Network Model

Model validation using GPe as a readout. A, Firing rate and CV of GPe neurons with cortical beta activity used as input for the model, and comparisons with experiments (from Mallet et al, 2008). i, iii, Firing rate and CV of randomly selected GPe neurons in the control and lesioned (dopamine set to 0) network. ii, iv, Firing rate and CV for GPe TI and GPe TA neurons in the lesioned network. B, Coherence of GPe neurons in the control and lesioned network when activated with cortical beta input. i–iv, Coherence for random GPe versus GPe, specifically GPe TI versus GPe TI, GPe TI versus GPe TA, and GPe TA versus GPe TA. Black dotted line shows significance of p = 0.05. C, Phase relationship of GPe neurons in the control and lesioned network when activated with cortical beta input. i–iv, Phase relationship for random GPe versus GPe, GPe TI versus GPe TI, GPe TI versus GPe TA, and GPe TA versus GPe TA. D, E, and F, same as A, B, and C, but for cortical slow-wave input.

Model validation using STN as a readout. A, Firing rate and CV of STN neurons with cortical beta input for model and experiment (Mallet, et al, 2008). i, ii, Firing rate and CV of STN neurons in control and lesioned (dopamine set to 0) network. B, Coherence of STN and GPe neurons in control and lesioned network with cortical beta input. i–iv, Coherence for STN versus STN, STN versus GPe, STN versus GPe TI, and STN versus GPe TA. Dotted black line shows significance of p = 0.05. C, Phase relationship of STN and GPe neurons in control and lesioned network with cortical beta input. i–iv, Phase relationship for STN versus STN, STN versus GPe, STN versus GPe TI, and STN versus GPe TA. D, E, and F, same as A, B, and C, for cortical slow-wave input.

Local network dynamics and the role of dopamine-dependent perturbations. A, Synchrony and oscillations in control and lesioned network components. Upper panel shows amount of synchrony in MSN D1, MSN D2, FSN, GPe TA, GPe TI, SNr, and STN during control (black bars) and after dopamine depletion (white bars). Lower panel shows the same but for oscillations instead of synchrony. B, Upper panel shows the relative change in synchrony in MSN D1, MSN D2, FSN, GPe TA, GPe TI, SNr, and STN (y axis) compared with the lesioned network when restoring the parameter on the x axis to the value it had in the control network (no dopamine depletion). Lower panel shows the same but for oscillations instead of synchrony. C, Phase relationship of STN and GPe TI when decreasing MSN D2 firing rate in the lesioned model. Cartoons illustrating effect of no change, medium decrease, and large decrease of MSN D2 firing rate. Signals represent instantaneous population firing rate in each nuclei.

Control of firing rates of MSN neurons due to striatal inhibition. A, Illustration how firing of MSN D1 and MSN D2 neurons are affected by inhibition from FSN, GPe TA, and MSN collaterals. Left top panel shows the firing rate of MSN D1 neurons with cortical input successively changed by a factor of 0.8 to 1.5 relative control, when keeping all inhibition (dark blue), only MSN–MSN inhibition (light blue), only FSN–MSN inhibition (cyan), only FSN–MSN static synapse inhibition (light green), only GPe TA–MSN inhibition (orange), and without any inhibition (dark red). Right top panel shows the relative contribution of the different sources of inhibition with successively increased cortical input and for MSN–MSN inhibition (light blue), FSN–MSN inhibition (cyan), FSN–MSN static synapse inhibition (light green), and GPe TA–MSN (orange) inhibition. Left and right bottom panels show the same as left and right top panels, respectively, except for MSN D2. B, Illustration of the role of lateral inhibition when a subpopulation, 10%, of MSNs are activated. MSNs D1 and D2 are enumerated and given an index (id) between 1-1500. Left top panel shows the result when randomly selected MSN D1 neurons are activated; middle top panel shows the result when only nonconnected MSN D1 neurons are activated. Left and middle bottom panels show the same as left and middle top panels, except for MSN D2. Right panel shows the connection diagram for MSNs, in which a dot indicates a connection between MSNs. It illustrates how groups of MSNs, here those with low id, are not connected. C, Left top panel shows the firing rate of the 10% of MSN D1 activated with a 100-ms-long input burst for the case when randomly connected MSNs (blue) and specifically nonconnected MSNs (red) are targeted. Right top panel shows the mean difference between the firing rate of the two differently selected populations of MSN D1 neurons when varying the activated MSN population size between 2% and 20%. Left and right bottom panels show the same thing as left and right top panels, except for MSN D2.

Illustration of the action selection network tested. A, Illustration of a simplified action selection network with two action channels. Light gray lines from cortex to MSNs indicate that these connections are weak (e.g., due to learning), whereas black indicates strong connections. Cortical state S1 connects to one striatal population in the direct (Go) pathway (denoted Go_A1), which inhibits the action 1 (A1) population in SNr. S1 also activates a striatal population in the indirect (NoGo) pathway (denoted NoGo_A2), hypothesized to disinhibit SNr populations not compatible with action 1 (i.e. in this example action 2), and vice versa for cortical state S2. Green box indicates the network in which only the direct (Go) pathway is activated, and the larger gray box indicates the network in which both the direct (Go) and indirect (NoGo) pathways are activated. SNr A1 and A2 populations are in turn assumed to control targets in the brain stem and/or thalamus via disinhibition. B, Illustration of two types of action selection scenarios. Top graph shows the firing rates of two hypothesized cortical states, state 1 (black) and state 2 (copper) that have been associated with two different actions after learning. Bottom two graphs show the resulting SNr firing rates for the co-activation of S1 and S2 representing action 1 (black) and 2 (copper). White indicates the interval on the x axis in which the SNr firing rates controlling action 1 and 2 are both above the selection threshold (i.e., no selection); blue indicates the interval in which the firing rate for action 1 is below the selection threshold, but action 2 is not (i.e., selection of action 1); and black indicates the interval in which both actions are selected (dual selection). In the beginning, good contrast can be obtained in SNr between the two actions (black and copper lines in middle panel), but at one point both will lead to completely inhibited populations in BG output nuclei, resulting in dual selection. Such a selection agent would work over a very limited range of inputs, and action selection would be possible only during a narrow window (middle panel, blue line). Instead, the capability to select only one action should preferably be obtained over a larger interval (bottom panel) with a mechanism ensuring that the contrast between the two actions is maintained over a larger range of inputs. C, Illustration of how assumed cortical input combinations could activate striatal populations, in turn facilitating action 1 and action 2, and how illustrations of action selection outcomes are made. First and second panels show examples how D1 (Go) or D2 (NoGo) populations are activated for consecutive trials by cortical state 1 and 2. For example, state 2 are stepwise increased over five trials and state 1 is kept constant, and this is then repeated, e.g., five times but with state 1 increased each time. Third panel shows the resulting firing rate of SNr neurons representing action 1 (black) and action 2 (copper). Black, gray, and light-gray boxes are three examples of input combinations, a, b, c, as illustrated with the arrows and corresponding letters. Bottom panel illustrates how these types of simulations are classified into a selection pie chart. In each pie (circle) the relative proportions of action selection outcomes are represented with the respective colors: white for no selection, black for dual selection, blue for selection of action 1, and red for selection of action 2.

Action selection performance of the basal ganglia network. A, Top left panel shows the action selection outcome with only the direct pathway activated, with respectively cortical input (i.e., S1 or S2) from 1× the base level to 3× the base level leading to selection of actions 1 and 2. Top right panel shows the action selection when both direct and indirect pathways are activated (similar action inputs as in top left panel). Middle left panel shows the action selection when both direct pathways are activated and MSN collaterals have been removed (similar inputs as in top right panel). Middle right panel shows the result when both direct and indirect pathway are activated and FSN–MSN inhibition has been removed (similar action inputs as in top right panel). Bottom left panel shows the action selection when both direct and indirect pathways are activated and in addition the hyperdirect pathway is co-activated through a burst in STN (similar action inputs as in top right panel). Bottom right panel shows the outcome when direct, indirect, and hyperdirect pathways are activated and GPe TA–MSN inhibition has been removed. For all six plots, 20% of the total MSN D1 or D2 pools were activated (the maximal MSN D1 or D2 action pool equaled each half of the total striatal MSNs). B, Action selection performance when scaling the size of the activated MSN pool. First row shows the action selection with only direct pathway activated, same as top left figure in A, but varying the relative size of the activated MSN population between 10% and 100%. Second row shows the same as the first row, except for the scenario when both direct and indirect pathways are activated. Third and fourth rows show the result for the scenario when both direct and indirect pathways are activated (as in row 2) and when in addition MSN collaterals or the FSN–MSN inhibition is removed. Fifth and sixth rows show the effect when the hyperdirect pathway is activated during action selection with and without TA–MSN present.

Restoring network dynamics and function after dopamine lesion. A, Action selection performance of control and lesioned networks. Left upper panel shows an action selection heat map of the control network with direct and indirect pathways both activated. Second upper panel shows the same but for the lesioned network. The remaining panels show the result for the lesioned network but with the dopamine parameter displayed in the heading restored to the control network parameter value. (compare with Fig. 4B). B, Change in synchrony and oscillations in MSN D1, MSN D2, FSN, GPe TA, GPe TI, SNr, and STN relative to the lesioned network model when silencing/increasing each connection. C, Change in synchrony and oscillations in MSN D1, MSN D2, FSN, GPe TA, GPe TI, SNr, and STN relative to lesioned network model when inhibiting/exciting each nucleus separately. A current was injected with a negative or positive value in each nucleus. The hyperpolarizing and depolarizing current steps were equally large, where the amplitude of the hyperpolarizing current was adjusted to silence the specific nucleus (mimicking a lesion of the nucleus). D, Action selection performance for a number of manipulations either to neuron excitability or to synaptic efficacy.