Figure 5. Lower mismatch accelerates fear reduction and decreases renewal in a neural network model.

(A) Cue inputs presented to the network during training (shock memory), reexposure (with or without footshock) and test sessions (consisting of the tone and either context B (test) or A (renewal)). Color scale shows the cue received by each of the 100 neurons (B) Extinction over multiple sessions using the no-footshock (red bars) or footshock (blue bars) cue. Bars represent freezing, expressed as the activity ratio between shock neurons and the sum of shock and non-shock neurons in response to the test cue, at reexposure days 1 to 4. After each test, memory is updated according to the activity reached in response to the full reexposure pattern. (C) Effect of LVGCC blockade (i.e. setting the mismatch-induced degradation term D to 0). Removing the degradation term blocks deconditioning-update, but not regular extinction. (D) Network activity in retrieval tests during tone presentations (e.g. cued with the tone alone) and at the end of reexposure (e.g. cued with the full reactivation pattern), as well as on test and retrieval sessions. Lower mismatch (i.e. weak footshock) leads to retrieval of the original pattern on the first days, leading to memory updating through mismatch-induced degradation and lower retrieval on subsequent tests. (E) Mean synaptic weights between different neuronal groups after training and at the end of each extinction session. Heat map represents the connection from neuronal populations in the Y axis to those in the X axis in the no-footshock and footshock groups. Deconditioning-update leads to weakening of connections between context and shock neurons and of their inhibitory connections to other neurons. On no-footshock extinction, an extinction memory is formed with sparing of the shock representation.

Figure 5—figure supplement 1. Blockade of Hebbian plasticity impairs reconsolidation, blocks standard extinction and interacts with deconditioning-update.

(A) Effect of protein synthesis inhibition (modeled by setting the Hebbian plasticity term S to 0 in the model) during reexposure with a low degree of mismatch that does not induce extinction (i.e. reconsolidation condition). Bars represent freezing, expressed as the activity ratio between shock neurons and the sum of shock and non-shock neurons in response to the test cue, before and after a reexposure with either vehicle (dark bars, Hebbian plasticity on) or a protein synthesis inhibitor (light bars, Hebbian plasticity off). Unlike in deconditioning-update, which requires several sessions, freezing decreases after a single reexposure session under protein synthesis blockade. (B) Effect of protein synthesis inhibition on regular extinction (red bars) or deconditioning-update (blue bars), as defined by the patterns in Figure 5A. Blockade of Hebbian plasticity inhibits extinction with a pure extinction pattern, which is fully based on learning a new attractor; however, it causes deconditioning-update to initially progress faster due to greater weakening of the original memory. On the long run, however, extinction is impaired in this group as well due to blockade of new learning.