A theory of the transition to critical period plasticity: inhibition selectively suppresses spontaneous activity

Abstract

What causes critical periods (CPs) to open? For the best-studied case, ocular dominance plasticity in primary visual cortex in response to monocular deprivation (MD), the maturation of inhibition is necessary and sufficient. How does inhibition open the CP? We present a theory: the transition from pre-CP to CP plasticity arises because inhibition preferentially suppresses responses to spontaneous relative to visually driven input activity, switching learning cues from internal to external sources. This differs from previous proposals in (1) arguing that the CP can open without changes in plasticity mechanisms when activity patterns become more sensitive to sensory experience through circuit development, and (2) explaining not simply a transition from no plasticity to plasticity, but a change in outcome of MD-induced plasticity from pre-CP to CP. More broadly, hierarchical organization of sensory-motor pathways may develop through a cascade of CPs induced as circuit maturation progresses from "lower" to "higher" cortical areas.

Fig. 1

(A) Model architecture: a cortical neuron receives excitatory input from 28×28 contralateral inputs and 20×20 ipsilateral inputs It also receives inhibitory input from nearby cortical neurons, which are not explicitly modeled. The excitatory synaptic strengths are subject to activity-dependent plasticity. (B) Hypothesis of how the maturation of inhibition initiates the CP for OD. Visual input is twice as strong as, but only 1/10 as frequent as, spontaneous input. During the pre-CP, the spontaneous input drives the cortical cell, so the contribution of the rare visual input to plasticity is relatively small. At CP onset, maturation of inhibition subtracts equally from all responses, causing greater proportional weakening of spontaneous than visual input and shifting many responses to spontaneous input below the threshold for Hebbian plasticity. This makes visual input the primary driver of plasticity. (C) Input correlations (covariance) in simulations are modeled as Gaussian functions of the distance between input retinotopic positions. At a given distance, inputs from the same eye (“contra” or “ipsi”) are more correlated than those from opposite eyes (“between-eye”). Under normal rearing (NR), the visually-evoked covariance (gray curves) is four times as strong as, but retinotopically as precise as, the spontaneous input (blue curves). During MD to the contralateral eye, the closed-eye and between-eye visual covariances (red curves) are reduced in amplitude and are retinotopically broadened because of the blurring of the visual stimulus through the closed eyelid.

Fig. 2

Model activity-dependent plasticity during the pre-CP: activity-dependent retinotopic refinement without OD plasticity. In Figs. 2–3, C and I stand for contralateral and ipsilateral eye, respectively. (A) Left panels: Development of synaptic strengths (color) over time during the pre-CP under NR (Top) and MD (Bottom; C closed from time 0). Final strengths are insensitive to initial conditions, which are identical in the two cases. Upper and lower halves show C and I strengths, respectively, over a one-dimensional (1-D) section of retinotopic positions through the center of the 2-D RFs (note, there are more C than I input neurons). Right panels: final 2-D RFs of both eyes. Under NR, retinotopic refinement occurs with little OD competition. Under MD, there is less retinotopic refinement and both eyes' strengths weaken relative to NR, but again there is little OD competition. (B) 1-D sections of the final synaptic strengths at time 100 under NR (black) and MD (red) (solid: C; dashed: I). (C) Contralateral bias index (CBI: see Supplemental Methods, S2d), which ranges from 0 to 1 for responses varying from purely driven by I to purely driven by C; the CBI settles to 0.75 under NR (black) and 0.68, only slightly lower, under MD (red), reflecting denser input from C than I. (D–E) RF width (D) and the response magnitude (E) (defined in Supplemental Methods, S2d) vs. time. Solid lines: C; dashed lines: I; black: NR; red: MD; green, monocular inactivation (MI) of C.

Fig. 3

Model activity-dependent plasticity during the CP: activity-dependent retinotopic refinement and OD plasticity. The simulation was continued from the final state of Fig. 2 under NR and CP was modeled as mature inhibition (increase of m from 0 to 5 at time 100). (A) Development under NR to time 150 and MD, initiated at time 150. Now, MD causes a strong shift to the open (I) eye. Panels are as in Fig. 2A. Note, homeostatic plasticity rapidly increased synaptic strengths after inhibitory strengthening. (B) One-dimensional sections of the synaptic strengths at time 150 under NR (black) and at time 200 under MD (red) (solid: contra; dashed: ipsi). Conventions as in Fig. 2B. (C) Comparison of the time courses of the contralateral bias index (CBI) under MD for CP (m=5, solid-line) and pre-CP (m=0, dotted-line) conditions. Maturation of inhibition yields about a five-fold greater change in CBI. Note that the final CBI for the CP under MD indicates only slight ipsilateral advantage (CBI=0.45): the summed synaptic strength for each eye is similar, although I has stronger strengths, because of denser input from C. (D) The RF width and the response magnitude vs. time as in Fig. 2D–E (solid:C; dashed I). (E–F) Final CBI v.s. the spontaneous-to-visual ratio of firing rates (E) and the ratio of gains (F) under various levels of the inhibitory strength (varied from 0 to 10) and the Hebbian threshold (varied from 1 to 3 Hz; shown with different colors). Black crosses indicates parameters used in Panel A–D.

Fig. 4

Tetrode recording from freely behaving adult mice. Spontaneous-to-visual activity ratio is a physiological correlate of OD plasticity as predicted by the model. (A) Experimental setup: periodic LED flashes (1 sec) were presented to freely behaving adult mice while putative pyramidal cells were recorded from V1. (B) Average firing response to LED flashes of typical multi-unit activity from WT (top) and GAD65-KO mice (bottom) before (left) and after (right) acute administration of diazepam. Green bars mark stimulus duration. (C) Summary of the spontaneous-to-visual ratio of firing rates before and after acute administration of diazepam (+Dz) in different mouse groups. Diazepam decreased the spontaneous-to-visual ratio of GAD65-KO mice but did not significantly change the ratios of adult WT mice or GAD65-KO mice chronically treated with diazepam earlier in life (KOc). Box-and-whisker plots show 25^th to 75^th percentiles (box), median (red line in box), the range of data falling outside the 25^th to 75^th percentiles by up to 1.5 the distance from 25^th to 75^th percentile (whiskers) and outliers outside this range (red crosses). (D) Cumulative distribution of spontaneous-to-visual ratios of firing rates of various mouse groups. Ratios of WT mice were smaller than those of GAD65-KO mice. Acute diazepam decreased the ratio of GAD65-KO mice far beyond adult WT levels. In KOc mice, the visual response ratio had returned to a level similar to naïve GAD65-KO mice, but no longer responded acutely to the drug as in adult WT mice (C).