Fig 1.
Modelling framework for TMS-induced long-term synaptic plasticity.
The Neuron Modeling for TMS (NeMo-TMS) toolbox integrates detailed neuronal models with TMS-induced electric fields, allowing the simulation of cellular and subcellular voltage and calcium responses during single and repetitive TMS pulses [34]. The direction of the electric field is represented by the vector E. We implement in NeMo-TMS a validated model of a CA1 pyramidal cell with detailed biophysics and reduced morphology, capable of generating realistic dendritic and somatic spikes [35,43]. Relative dendritic diameters are depicted. We introduced a unified voltage-dependent 4-pathway (pre- and postsynaptic; LTP and LTD) model of long-term synaptic plasticity (yellow circle with pre- and postsynaptic LTP (+) and LTD(-)), capable of reproducing the frequency-, timing- and location-dependence of synaptic changes [30], into the existing NeMo-TMS framework. 128 of plastic excitatory synapses were placed in the morphology.
Table 1.
Ion channels and membrane properties.
Changes from the original [35] are shown in bold italics.
Table 2.
Parameter values for four-pathway synaptic plasticity model.
Fig 2.
Unifying four-pathway model of long-term synaptic plasticity reproduces frequency dependence of synaptic changes in hippocampal CA1 pyramidal cells evoked by local electrical Schaffer collateral stimulation
(a) Local electrical stimulation (i.e. Schaffer collateral synaptic stimulation) regions (blue purple and magenta) shown on schematic. There is no somatic stimulation. (b) Schaffer collateral weight change for 900 pulses delivered at 1, 10, 30, and 100 Hz with inhibition active. Model (red) and data (black). (c) Same as b but with inhibition blocked. Model (red) and bicuculline data (black).
Fig 3.
The modelling framework for synaptic plasticity replicates proximal LTP induced by 10 Hz rMS protocol and predicts increasing LTP with higher stimulation frequency.
(a) Schematic of stimulation with electric field vector. (b) Mean LTP in the proximal stratum radiatum decreases with lower stimulation frequency; the induced LTP for 10 Hz rMS is in agreement with experimental data from Lenz (2015). (c) - (h) Induction of LTP from 900 pulse rMS protocol at various frequencies. Strong LTP is only seen in the proximal dendrites, with LTP in the stratum radiatum rapidly decreasing with lower frequency.
Fig 4.
Presynaptic and postsynaptic weights change monotonically linearly over stimulus duration.
(a) Time course of average presynaptic weights over stimulus duration for 900 pulse 10 Hz stimulus. (b) Time course of average postsynaptic weights over stimulus duration 900 pulse 10 Hz stimulus. Weight change in the stratum radiatum is far more postsynaptically dominated than in other strata.
Fig 5.
Pharmacology in silico: rMS-induced LTP depends on dendritic sodium and calcium through voltage-gated calcium channels, as well as synaptic NMDA channels.
(a): Schematic of electric field for all panels in this figure. Electric field vector direction shown by E. In addition, all regions of the CA1 pyramidal cell are synaptically stimulated because we assume that electric field not only depolarises the postsynaptic neuron but also elicits spikes in presynaptic neurons. (b): Elimination of LTP when dendritic sodium channels are suppressed, simulating TTX application. (c): Elimination of LTP when NMDAR (NMDA receptor)-like processes within the synaptic model are disabled in silico. (d): Large reduction in LTP induction (as compared to control, i.e. no perturbation condition) when L-type voltage-gated calcium channels are suppressed in silico. (e): Elimination of LTP observed when calcium-free solution is modelled.
Fig 6.
The synaptic plasticity model reproduces distal LTP for local electric TBS and predicts proximal and distal LTP for rMS TBS.
Induction of LTP by local electrical TBS to the perforant path. (a): Schematic of local electric stimulation for panels b and c. L-M: lacunosum-moleculare (perforant path target). (b): Negligible plasticity observed when inhibition is present. (c): Distal LTP observed when inhibition is blocked. (d): Schematic of rMS simulation with electric field vector for panels e and f. (e): LTP in both distal and proximal synapses when rMS is used. Unlike local electrical TBS, rMS TBS is able to induce large LTP when inhibition is present. (f): LTP in both distal and proximal synapses when rMS is simulated and inhibition is blocked.
Fig 7.
Voltage traces from local electrical and rMS TBS, recorded at the soma and distal apical tuft, account for the outcomes of synaptic plasticity.
(a) 100 Hz local electrical TBS only causes small somatic voltage depolarisations when inhibition is blocked. (b) 100 Hz local electrical TBS induces dendritic spikes when inhibition is blocked. (c) 100 Hz rMS TBS causes multiple somatic spikes. (d) 100 Hz rMS TBS induces dendritic large-amplitude voltage depolarisations when inhibition is blocked. (e) 100 Hz local electrical TBS causes almost no somatic voltage depolarisation when inhibition is enabled. (f) 100 Hz local electrical TBS does not induce dendritic spikes when inhibition is enabled. (g) 100 Hz rMS TBS causes multiple somatic spikes with inhibition enabled. (h) 100 Hz rMS TBS induces dendritic large-amplitude depolarisations even when inhibition is enabled.
Fig 8.
Distal LTP induced by simulated rMS is sustained by sodium dendritic spikes but less affected by inhibition than local electrical stimulation.
rMS TBS (red) generates LTP even when local electrical TBS (blue) does not. In the presence of simulated GABAergic inhibition, rMS induces LTP comparable to local electrical stimulation without inhibition (i.e. bicuculline simulation). When dendritic sodium spikes are blocked (i.e. when local dendritic TTX application is simulated), neither of the two stimulation protocols we tested was able to induce LTP in the dendritic tuft. We also tested blocking somatic as well as dendritic sodium channels, and found effectively identical results (not pictured). Sample of 10 models with different randomised synapse locations.