Developmental sensory experience balances cortical excitation and inhibition

Synaptic development in rodent AI generally occurs over the first postnatal month^16,17. During this time, the nascent organization of AI can be extensively altered by passive exposure to structured auditory stimuli, such as repetitive sequences of pure tones at a given frequency^3,7. Recent studies indicate that inhibitory circuits play key roles in this process, first enabling and eventually limiting the extent of receptive field plasticity^6,18,19. However, it is unclear how inhibition at the cellular and network levels is developmentally coordinated to shape receptive field selectivity and control cortical plasticity. As sensory-evoked subthreshold responses cannot yet be directly measured optically, by extracellular recording in vivo, or by intracellular recording in vitro, here we used in vivo whole-cell recording to study synaptic organization and plasticity of developing AI.

We made 107 whole-cell voltage-clamp recordings from rat AI neurons in vivo. Recordings were obtained from postnatal day 12 (P12) to P30 and from adult animals. Frequency tuning profiles of AI neurons were measured with pure tones at different holding potentials to compute excitatory and inhibitory conductances^11-15.

We found that, in young animals, excitatory and inhibitory frequency tuning profiles were individually present but uncoupled, unlike the highly correlated and balanced excitation and inhibition measured in neurons from older animals. For example, at P14 (Fig. 1a, Supplementary Fig 1), although substantial tone-evoked excitatory and inhibitory responses were observed (Fig. 1a, top), magnitudes of excitation and inhibition were uncorrelated across frequencies (Fig. 1a, bottom; r: −0.01). This was in contrast to recordings from adults, in which excitatory inputs were balanced by a proportional amount of co-tuned inhibition (Fig. 1b; r: 0.87; Supplementary Fig. 2).

By P21, the correlation between excitatory and inhibitory inputs had increased (Fig. 1c). Initially after hearing onset (~P12), excitation and inhibition were generally mismatched. During P15-P21, excitatory and inhibitory responses were partially but not strongly correlated, relative to balanced tuning observed in adult AI. Moreover, stimuli evoking maximal excitation and inhibition in a given cell differed during development. In mature AI, excitatory and inhibitory best frequencies were approximately the same, while in young AI, peak excitatory and inhibitory responses were over an octave apart on average (Fig. 1d).

This early imbalance between synaptic inputs was due to low selectivity of inhibitory tuning. To measure sharpness of tuning, we normalized tuning curves and used linear slopes (Supplementary Fig. 3) and standard deviations of Gaussian fits (Fig. 2) as selectivity indices for excitation and inhibition. For the P14 cell shown in Figure 1a, inhibition was less tuned than excitation (Fig. 2a), in contrast to the sharper inhibitory tuning observed in older animals (Fig. 2b), as previously suggested¹⁹.

Development of inhibitory frequency tuning was delayed relative to excitation (Fig. 2c,d). Sharpness of excitatory tuning was close to adult levels by ~P15, but inhibitory tuning was slower to emerge, eventually maturing around P25-P30. Thus early in cortical development, inhibitory inputs are poorly tuned for spectral features of sensory stimuli, leading to imbalanced synaptic tuning profiles and accounting for high-threshold or less-structured spiking receptive fields and frequency maps reported in young animals^7,15,19,20.

A generally low inhibitory tonus at the onset of AI development might account for nonselective frequency tuning of inhibition. However, while many aspects of neuronal excitability and synaptic transmission change over development¹⁷ (Supplementary Figs. 4,5), average strengths of tone-evoked inhibitory conductances were similar in young and adults, in terms of absolute magnitude (Supplementary Fig. 3d) and relative to excitation (Fig. 2d). This suggests that the pace of the AI critical period is not determined solely at the level of individual synapses, e.g., by gross strengthening of GABAergic inhibition. Instead, changes of inhibitory transmission must be coordinated across the cortical network, to shape the tuning of intracortical inhibition throughout the entire receptive field in such a way as to match and balance the overall structure of excitatory inputs.

What factors contribute to refinement of excitatory-inhibitory balance during the AI critical period? Sensory experience is known to control development of cortical receptive fields and networks^{3,4,6,7,18,21}. We therefore examined the effect of patterned sensory stimulation on maturation of synaptic frequency tuning.

We first measured baseline tuning before repetitively playing a tone of a specific frequency for 3-5 minutes (patterned stimulation). After 5-10 minutes, we assessed changes to synaptic tuning. In young but not adult animals, patterned stimulation led to rapid increases in synaptic strength (Fig. 3), enhancing excitatory-inhibitory balance, shortening response latency (Supplementary Fig. 4f), and persisting for 30+ minutes.

Thus the major features of AI synaptic receptive field development are a progressive balancing of excitation and inhibition, driven by sharpening of initially poorly-tuned inhibitory inputs, and susceptibility to exposure-induced synaptic modifications that enhance excitatory-inhibitory coupling. Given their similar age dependence, we hypothesized that there was a relationship between formation of excitatory-inhibitory balance and the age window for synaptic plasticity. We noticed that patterned stimulation led to changes not only at the presented input, but also across other frequencies. Consequently, excitatory and inhibitory tuning profiles became more similar. Although these synaptic responses exhibited complex spectrotemporal evolution, there were three general principles of developmental synaptic receptive field modification induced by patterned stimulation.

First, potentiation of excitation and inhibition was not limited to the presented frequency alone, but spread to frequencies within one octave (Fig. 4a). Second, excitation and inhibition depressed at their respective best frequencies (Fig. 4b). Third, these positive and negative changes to multiple inputs cooperated to increase excitatory-inhibitory balance (Fig. 4c).

Changes at stimuli other than the presented frequency were predominant in enhancing excitatory-inhibitory coupling. Considered separately, changes to the presented frequency did not greatly increase correlation (Fig. 4d, “presented only”), while changes to other inputs could entirely account for increased correlation (Fig. 4d, “unpresented only”).

Acceleration of frequency tuning was specific for pure tones. Stimulation with white noise (Supplementary Fig. 6) did not affect excitatory-inhibitory balance, regardless of the temporal structure of noise presentation, i.e., short bursts or continuously. Interestingly, noise bursts enhanced excitation and inhibition, but irrespective of frequency, leaving excitatory-inhibitory correlation unaltered.

Patterned stimulation dramatically changed spike output, in terms of spike timing precision and tuning curve structure. We made 22 whole-cell current-clamp or cell-attached recordings from P12-P21 or adult AI neurons in vivo. Spike firing was imprecise during P12-P21 (Supplementary Fig. 7a), such that latency was highly variable from trial to trial. Spike timing was much more precise in adults (Supplementary Fig. 7b), tightly locked to tone onset. Patterned stimulation increased spike output and enhanced temporal precision in young animals but not adults (Supplementary Figs. 7c-e), as predicted from simulations¹² (Supplementary Fig. 7f). Surprisingly, while maturation and plasticity of excitatory and inhibitory responses predicted much of action potential generation and spiking receptive field development, synaptic inputs by themselves did not seem to fully predict the absolute values of spike latency and variability. Thus development of other factors (e.g., ion channel expression²²) play important roles in coupling cortical synaptic input to spiking output.

Lastly, we asked whether accelerated balancing of excitatory and inhibitory inputs through patterned stimulation was sufficient to prevent further modifications of synaptic receptive fields. We performed three final experiments to probe the limits and duration of cortical plasticity in relation to excitatory-inhibitory balance.

We first measured synaptic tuning before and after an initial episode of patterned stimulation, followed by another round of patterned stimulation ~30 minutes later to determine if additional modifications were induced. An example P18 cell is shown in Figure 5. Initially, excitation and inhibition were weakly correlated due to low selectivity of inhibition (Fig. 5a; r_pre: 0.27). Patterned stimulation with 4 kHz tones increased excitation, inhibition, and excitatory-inhibitory correlation (Fig. 5b; r_post1: 0.77). Thirty minutes later, a second episode of 4 kHz patterned stimulation was presented, but tuning was not significantly changed (Fig. 5c; r_post2: 0.82).

Therefore, relatively brief episodes of structured sensory experience during development reorganize AI synaptic receptive fields, first limiting and eventually preventing subsequent patterned stimulation from triggering further modifications. After first modifying excitation and inhibition, no additional changes were induced by a second period of patterned stimulation (Fig. 5d), as by then the coupling between excitation and inhibition had reached mature levels.

To more completely determine the temporal dynamics of these effects, we made consecutive recordings from the same animals for hours after one round of patterned stimulation (Supplementary Fig. 8). For example, recordings in Supplementary Figure 8a-c were made in the same P17 animal. Initially, excitatory-inhibitory correlation was low (Supplementary Fig. 8a; r_pre: 0.22). Twenty-five minutes after 1 kHz patterned stimulation, correlation in a second neuron was much higher (Supplementary Fig. 8b; r_post1: 0.98); but two hours later, correlation in a third neuron was reduced (Supplementary Fig. 8c; r_post2: 0.54). Seemingly regardless of location within AI, patterned stimulation increased excitatory-inhibitory correlation for hours before returning to initial lower levels (Supplementary Fig. 8d).

Previous studies indicate that prolonged exposure to random stimuli can undo or extinguish prior modifications to neural circuits, potentially leaving maps and receptive fields in unrefined yet plastic states^4,19. Conversely, a few days of exposure to tonal stimuli early in life can profoundly affect AI topography and receptive fields, lasting days to months afterward^3,7. In our final experiment, we therefore used much longer periods of patterned stimulation, to examine conditions for AI critical period closure.

Rat pups (P9-P11) were repetitively exposed to 2 kHz or 7 kHz tones for one to three days in their home cages⁷. We then made recordings in anesthetized P12-P16 animals, comparing synaptic receptive fields (as in Fig. 1) and degree of exposure-induced synaptic modifications (as in Fig. 3) to age-matched controls.

Early exposure to pure tones accelerated development of excitatory-inhibitory balance and prevented subsequent patterned stimulation from inducing synaptic modifications (Supplementary Fig. 9). The cell shown in Supplementary Figure 9a was recorded from a P16 animal exposed to repetitive 2 kHz tones from P11-P14. Correlation between excitation and inhibition (r_pre: 0.66) was high, presumably due to the prior structured sensory exposure. After measuring synaptic tuning, we repetitively presented 16 kHz tones for five minutes, but patterned stimulation did not significantly change synaptic strength or excitatory-inhibitory balance (r_post: 0.64).

Early exposure to repetitive tonal stimuli for a few days enhanced average excitatory-inhibitory correlation at P12-P16 compared to similarly aged animals raised under normal laboratory acoustic conditions (Supplementary Fig. 9b). This precocious maturation of synaptic receptive fields occurred in absence to changes to cell-intrinsic properties such as input resistance and synaptic kinetics (Supplementary Fig. 9c). Furthermore, this accelerated development prevented changes in synaptic strength and balance by additional periods of patterned stimulation during recording (Supplementary Fig. 9d). These data indicate that accumulation of sensory experience early in development, when excitation and inhibition are still uncorrelated, rapidly improves the balance and structure of synaptic receptive fields, preventing additional exposure-induced changes from occurring; in essence, ending the critical period of heightened cortical plasticity.

In conclusion, we have defined here the AI critical period as that developmental stage when brief episodes of structured sensory experience dramatically alter synaptic receptive fields. During this period, patterned tonal stimulation sets in motion a coordinated set of bidirectional modifications to specific elements of synaptic receptive fields which endure for hours (after minutes of exposure) or weeks (after days of exposure). Synaptic modifications could be forms of spike-timing-dependent plasticity^23-25, heterosynaptic plasticity^26-28 or homeostatic changes such as synaptic scaling²⁹. Together, these maturational adjustments dynamically fine-tune cortical networks, balancing excitatory and inhibitory inputs to reduce the efficacy of sensory stimuli in modifying synaptic strength and receptive fields later in life.

Our results demonstrate that maturation of inhibitory circuitry is not simply expressed as an overall increase in inhibitory strength with age. Rather, inhibition is strong early in life, but is initially poorly tuned for spectral features, unmatched to excitatory inputs. Another recent study³⁰ also found that inhibition was high shortly after hearing onset, but that many neurons in layer 4 of AI seemed to have pre-balanced excitatory and inhibitory synaptic receptive fields. Here we found that, while in some neurons, excitation and inhibition were similarly balanced at young ages (Fig. 1c), in different cells even from the same animals, excitatory-inhibitory co-tuning and correlation could be much lower. Various components of cortical receptive fields may develop at different rates, such as frequency tuning curves at other intensity levels. Likewise, different cortical layers, sectors, or microcircuits within AI might also have differential developmental trajectories or sensitivity to the acoustic environment. Thus, overall, the progressive balancing of excitation and inhibition is apparently a network-level phenomenon. Consistent experience with reliable, patterned sensory stimulation refines intracortical inhibition precisely in proportion to excitation. The progressive remodeling of inhibitory receptive fields by sensory experience leads to balanced synaptic activity in cortical networks and limits further exposure-induced modifications, closing the developmental critical period. Given the importance of balanced excitation and inhibition for the temporal precision of activity in the auditory cortex^12,13, failure of this maturational adjustment to occur could be catastrophic for development of speech and language skills, with lasting consequences such as dyslexia or other development-associated language impairments^{5, 10}.