Rethinking how neural activity sculpts critical periods

A fundamental goal of neuroscience is to understand how neural activity shapes brain wiring during development. This process occurs during critical periods, temporal windows of heightened neural plasticity when activity dramatically influences circuit formation. David Hubel and Torsten Wiesel’s pioneering work in the 1960s began defining the biological mechanisms involved, showing that depriving cats of visual input during development irreversibly alters wiring in the visual cortex.

In the decades that followed, an explosion of studies focused on the activity-dependent mechanisms that govern critical periods. Research in the early 2000s, for example, showed that activity early in development is important for retinal ganglion cell axons to wire up properly and form eye-specific connections. These findings in the visual cortex, along with research in other areas of the brain, helped establish the notion that neural activity directs synaptic wiring during critical periods.

Around the same time, however, a handful of studies challenged this basic idea, showing that depriving activity—by removing input from one eye, for example—does not result in changes in axonal projections or the number of presynaptic terminals in the visual system. Nor does a lack of activity alter the formation of ocular dominance columns, a collection of neurons in the visual cortex that respond to visual inputs to the left or right eye. Based on these findings, the role of neural activity during development appeared to be not as well defined as it previously seemed.

The inconsistencies in the literature highlight a long-standing and under-recognized issue in the field: Visualizing activity-dependent changes during development is challenging, and in the past, most studies relied on indirect measures rather than direct empirical evidence of synaptic remodeling. This limitation constrained our understanding of critical periods.

But this is beginning to change. Thanks to advances in imaging techniques and molecular tools over the past 15 years, we can now visualize changes in the physical wiring of the brain during development with higher precision than ever before. New results using these methods are prompting researchers to rethink how neural activity sculpts the formation of circuits during critical periods.

T

ake, for example, synaptic pruning, an activity-dependent process in which the brain eliminates excess synapses during critical periods. Studies in mice established that neural activity triggers the accumulation of synapses, strengthening some circuits. Lack of activity weakens other synapses, which are then pruned. But these conceptual rules were largely inferred from functional studies, in which researchers measured changes in neuronal activity—such as excitatory postsynaptic potentials, for example—rather than observing direct structural evidence of pruning events. Structural evidence for these events is necessary because a lack of neuronal activity can sometimes trigger changes that resemble synaptic pruning but in fact reflect other influences. For example, differences in the size of a synapse or axon diameter can result in changes to neuronal activity that look like synapses are being pruned.

Our group used new connectomic approaches across cortical areas and layers to reveal there is little evidence of pruning after synapse overaccumulation. Instead, we show that in mice, synapses gradually and monotonically increase over development. Other teams have shown a similar monotonic increase in synapses over the lifespan of a mouse, using a marker for excitatory synapses called PSD95, and other experimental models have also shown a lack of activity-dependent pruning. Work in zebrafish, for instance, suggests that normal visual system development can proceed without activity-dependent synaptic pruning.

These findings force us to reconsider the precise role of neural activity in circuit formation. Part of the challenge of understanding the role of activity-dependent mechanisms in critical periods stems from species-specific differences. In my work, we have found that the number of synapses in the visual cortex increases during the critical period in mice but decreases in primates. One interpretation of these data is that during development, activity instructs neurons on which synapses to make in mice, but similar changes in activity tell primate neurons which synapses to prune. More broadly, these results point to important differences in the cellular mechanisms of how brains develop and experience critical periods that require consideration when designing experiments.

To reveal fundamental principles of activity-dependent circuit development, we need more comparative studies of mice, primates and humans. Comparative approaches have already shown success outside of neuroscience—in cancer research, for example. Studying tumor suppression mechanisms in long-lived mammals, such as elephants and bats, and researching naturally occurring cancers across various animal models, has improved the translation of cancer therapies to humans. That different species’ brains show mechanistic differences is not a reason to stop mouse work, but rather an opportunity to dive into comparative neuroscience.

To truly understand critical periods, as a field we need to actively test and challenge existing models rather than building on potentially flawed assumptions, and we need to remain open to revising our understanding of established fundamental principles. As Thomas Kuhn, a philosopher of science, argued, “to assess the relative merits of competing ideas, we must remain patient and constantly focused on assessing competing ideas until there is overwhelming evidence to tip the scales.” Novel approaches will be essential for testing new hypotheses. Artificial-intelligence tools, such as AlphaFold, could integrate complex datasets and help generate new and, most importantly, testable hypotheses for how activity alters structure during development. And we must test these hypotheses across molecular, computational and behavioral neuroscience. Only then will we capture the full complexity of critical-period plasticity.