An early phase of instructive plasticity before the typical onset of sensory experience

Abstract

While early experience with moving stimuli is necessary for the development of direction selectivity in visual cortex of carnivores, it is unclear whether experience exerts a permissive or instructive influence. To test if the specific parameters of the experienced stimuli could instructively sculpt the emergent responses, visually naive ferrets were exposed to several hours of experience with unusual spatiotemporal patterns. In the most immature ferrets, cortical neurons developed selectivity to these patterns, indicating an instructive influence. In animals that were 1-10 days more mature, exposure to the same patterns led to a developmentally-typical increase in direction selectivity. We conclude that visual development progresses via an early phase of instructive plasticity, when the specific patterns of neural activity shape the specific parameters of the emerging response properties, followed by a late phase of permissive maturation, when sensory-driven activity merely serves to enhance the response properties already seeded in cortical circuits.

Fig. 1. Design of a spatiotemporal stimulus family.

a Left: X-T (space-time) view of vertical sinusoidal grating shifting to left at each step, termed forward stimulus (F). Each strip represents video frame. Yellow box indicates hypothetical receptive field. Right: Hypothetical X-T inputs (ON only) that depict positions, latencies of inputs that would drive optimal response to F. b Left: X-T view of vertical sinusoidal grating (stimulus S6) with scrambled phase steps [8 3 6 2 7 4 1 5]. Right: Hypothetical X-T inputs (ON only) that depict positions, latencies of inputs that would drive optimal response to S6. c Plot of best-aligned correlation with forward (F), backward (B) motion for all 5040 unique phase sequences (black x) and our selections for stimuli (blue circles). F, B represent forward and backward smooth motion; S1–S6 are phase-scrambled stimuli that deviate substantially from F, B; CP1, CP2 are counterphase stimuli. d Video frame strips of all stimuli. e Responses of hypothetical cells with input kernels optimized for indicated stimuli. Cell optimized for F (KF, orange) responds strongly to F, but not to B, S5, or S6. Cell optimized for B (KB, green) responds strongly to B but not F, S5, or S6. Cells optimized for either S4 (KS4, cyan) or S6 (KS6, purple) respond weakly to F and B and poorly to each other’s optimal stimulus. f Response Projection Index (RPI) indicates how tuning curve of a given cell matches those of cells optimized for 2 stimuli. Left: Each cell’s normalized response curve in 10-dimensional space. Response is compared to the responses expected from hypothetical reference neurons optimized for 2 stimuli (such as F and B). Distance in vector space between actual response (gray vector) and vector line defined by reference neurons is calculated (D1 and D2), and index is calculated RPI = (D1 − D2)/(D1 + D2). If cell’s responses match that of hypothetical neuron optimized for first (second) reference stimulus, then RPI is −1 (1). Right: Scatter plot of RPI index values for kernels optimized for stimuli indicated. X-axis is RPI relative to F, B and Y-axis is RPI relative to F, S4.

Fig. 2. In naive ferret, 6 h experience with phase-scrambled grating pattern caused an increase in pattern selectivity.

a GCaMP6s responses to the family of spatiotemporal stimuli before, after 6 h of training with pattern S4, indicating substantial increase in selectivity for S4. Animal’s eyes were opened prematurely on P30. b Single cell selectivity index (SI) values for different stimuli (F—a phase-regular, unscrambled direction stimulus, and S4/S6—phase-scrambled stimuli). Selectivity for F decreases, while selectivity for S4 increases.  All visually-responsive cells included. c Responses to example cells (in B) before, d after experience. Error bars are standard error of mean (SEM) across trials. Cells δ, ε exhibit strong responses to S4. e Orientation, direction tuning before, after training. Blue dots represent visually-responsive cells that do not exhibit significant variation across direction stimuli; green bars represent orientation-selective but not strongly direction-selective cells (DI < 0.5), cyan arrows indicate strongly direction-selective cells (DI ≥ 0.5). f Direction tuning of example cells (in e) before, g after the experience. Numbers indicate direction index values. Error bars are SEM. h Response Projection Index (RPI) for F vs B (X-axis) and F vs S4 (Y-axis) for cells measured before (green) and after (blue) 6 h of experience with S4. There is an upward shift on Y-axis, indicating cells exhibit responses more like cell optimized to S4. i Grand average of responses before and after 6 h exposure to S4. On average, there is an enhancement of the response to S4. Error bars are SEM across cells. j Difference in cell RPI (F vs S4) before (N = 138 cells), after (N = 82 cells) experience (error bars: 95% confidence intervals), indicating significant increase in selectivity to S4. * indicates 95% confidence interval does not include 0. Cells that exhibited significant variation across scrambled stimuli included.  k Direction index values before (N = 50 cells), after (N = 200 cells) exposure to S4. Direction index values decreased slightly after exposure to S4. Error bars are SEM across cells. * Indicates 95% confidence interval does not include 0 (see l). Cells that exhibited significant variation across direction stimuli included. l Estimated difference in DI of cells before, after the experience (error bars are 95% confidence intervals), indicating significant decrease in DI with S4 experience. * Indicates 95% confidence interval does not include 0.

Fig. 3. In a second visually naive ferret, 6 h of experience with a phase-scrambled grating pattern caused an increase in selectivity for that pattern.

The eyes were opened prematurely on P31. Panels a–l are as described in Fig. 2, except that S6 was used as the training stimulus. In this animal, some cells were tracked across time and Greek letters appear more than once. This animal exhibited increased selectivity to training stimulus S6, and no significant alterations to direction selectivity. N = 25 cells before and N = 62 cells after for measurements of F, B, S1–6, CP1–2, and N = 41 cells before and N = 145 cells after for direction tuning. Premature animals varied in the influence of phase-scrambled grating patterns on traditional direction selectivity.

Fig. 4. In a ferret with 3 days of visual experience, 6 h of experience with a phase-scrambled grating pattern caused an increase in direction selectivity rather than selectivity for the phase-scrambled pattern.

The eyes opened naturally on day P33, and the experiment was performed at P36. Panels a–l are as described in Figs. 2, 3. S4 was used as the training stimulus. In this animal, some cells were tracked across time and Greek letters appear more than once. N = 170 cells before and N = 267 cells after for measurements of F, B, S1–6, CP1–2, and N = 124 cells before and N = 371 cells after for direction tuning. This animal exhibited decreased selectivity to training stimulus S4, and exhibited a significant increase in direction selectivity.

Fig. 5. In a ferret with 5 days of visual experience, 6 h of experience with a phase-scrambled grating pattern caused an increase in direction selectivity rather than selectivity for the phase-scrambled pattern.

The eyes opened naturally on day P33, and the experiment was performed at P38. Panels a–l are as described in Figs. 2–4. S6 was used as the training stimulus. In this animal, some cells were tracked across some trials and Greek letters appear more than once. N = 13 cells before and N = 41 cells after for measurements of F, B, S1–6, CP1–2, and N = 36 cells before and N = 67 cells after for direction tuning. This animal exhibited no change in selectivity to training stimulus S6, and exhibited a significant increase in direction selectivity.

Fig. 6. Relationship between changes in visual selectivity and parameters related to animal maturity.

a Animal age, days after eye-opening. ST indicates animals trained with S4, S6. Triangles indicate animals from ref. trained with bidirectional moving stimuli. Filled circle is single animal beyond critical period for direction selectivity development. b Animal age, initial orientation selectivity (1-CV). On average, orientation selectivity becomes stronger with age, but there is range of initial selectivity in youngest animals, likely reflecting range of cortical maturity achieved. c Difference in RPI for F vs trained stimulus (denoted ST; S4 or S6) before and after training (error bars 95% confidence intervals) plotted against days after eye-opening (ρ = −0.43, p < 0.165, DF = 12-2). d Same, but difference in direction index values (ρ = 0.54, p < 0.068, DF = 12-2). e Difference in RPI vs initial orientation selectivity that was measured at beginning of the experiment (before training stimulus exposure). Animals with lowest orientation selectivity exhibit strong changes in RPI, become more selective for scrambled training stimulus. ρ < 0* indicates significant negative correlation (ρ = −0.71, p < 0.009, DF = 12-2). f Same, but for DI (ρ = 0.44, p < 0.152, DF = 12-2). g Left: Changes in RPI (F vs training stimulus ST) for animals whose eyes were opened by experimenter (EO < 1) and animals whose eyes opened naturally before experiment (EO ≥ 1). * Indicates significant difference, t-test (mean ± SEM EO < 1: 0.16 ± 0.05, EO ≥ 1: 0.00 ± 0.03, p < 0.020241, DF = 12-2). Right: Changes in RPI (F vs ST) for animals that exhibited low initial orientation selectivity (1-CV < 0.3) and animals that exhibited higher initial orientation selectivity (1-CV ≥ 0.3) * indicates significant difference, t-Test (mean ± SEM 1-CV < 0.3: 0.04 ± 0.10, 1-CV ≥ 0.3:: −0.01 ± 0.03, p < 0.0045866, DF = 12-2). h Same as g, but change in direction index values indicated. Left: Difference is not significant (mean ± SEM EO < 1: 0.04 ± 0.10, EO ≥ 1: 0.18 ± 0.06, T-test, p < 0.25972, DF = 12-2). Right: Difference is not significant (mean ± SEM 1-CV < 0.3: 0.02 ± 0.11, 1-CV ≥ 0.3: 0.19 ± 0.06, T-test, p < 0.14439, DF = 12-2). Ns are animals (averages across all significantly-responding cells in each animal). The post-critical period animal was excluded in this analysis. RPI exhibited increases in inexperienced animals and in animals with low initial orientation index values.

Fig. 7. Responses became more similar to those of ideal neurons selective to the training stimulus.

a Principle component projection from 10-dimensional space to 2-dimensional space of mean responses (for each animal) to the chosen set of 10 stimuli, before and after training, with vectors indicating the transition from the mean response before training to after training (arrow points at mean state after training). Responses of hypothetical neurons optimized for each stimulus (KF, S4, CP2, etc.) shown. Animals that exhibited significant ∆RPI (F vs ST) are indicated (trained with S4 green, S6 blue). In this reduced view, average responses of significant animals moved closer to KST, while animals (8/8) that exhibited no significant effect moved to be near to KF, KB, KCP1, KCP2 (typical V1 receptive fields). b Change in RPI for significant animals with each stimulus used as a reference with the training stimulus (Sn vs ST). For animals trained with S4 or S6, values of RPI (S4 vs S4) or RPI (S6 vs S6) were excluded from the average as it is 0 by definition. Error bars indicate SEM of the mean. * or ** indicates one-tailed T-test (*p < 0.05, **p < 0.005, DF = 6, DF = 3 when X is S4/S6) with mean > 0. Comparison for each RPI (X vs ST) is single comparison evaluating only stimulus X. Changes in responses became more like a hypothetical neuron tuned to the training stimulus KST than stimuli F, B, S1, S3, S4, CP1, and CP2, and changes in responses remained equally close to stimuli S2, S5, and S6 (when S6 was not the training stimulus) on average. As responses changed in 10-dimensional space, they moved closer to KST for most stimuli while moving no closer or farther from KS2, KS5, and KS6. This is consistent with the idea that the training stimulus provided an instructive influence on receptive field properties in this early developmental period.