TL;DR
This paper presents a novel computational model of synaptic plasticity that unifies both presynaptic and postsynaptic changes. The model demonstrates that this dual-site plasticity leads to more precise sensory perception and a remarkable ability to rapidly relearn previously acquired information, offering a cellular basis for memory savings.
What Problem Does This Paper Address?
For decades, neuroscientists have known that synapses, the connections between neurons, can change their strength over time – a phenomenon called synaptic plasticity. This plasticity is crucial for learning and memory. While it's established that these changes can occur at either the presynaptic terminal (where the neuron releases neurotransmitters) or the postsynaptic neuron (which receives them), or both, the functional advantages of having plasticity at both sites simultaneously have remained unclear. This paper aims to bridge this gap by investigating the consequences of a unified model of synaptic plasticity that incorporates both presynaptic and postsynaptic modifications.
Methods
The authors developed a biologically-tuned mathematical model of spike-timing-dependent plasticity (STDP). This model uses exponentially decaying "traces" of past presynaptic and postsynaptic activity to govern changes in both a presynaptic factor (P, representing release probability) and a postsynaptic factor (q, representing quantal amplitude). The synaptic weight is then a product of these two factors (w = Pq). This phenomenological model was carefully tuned to experimental data from layer-5 pyramidal cells in the rat sensory cortex. They then used this model to simulate receptive field development in a simplified neural network and to explore the dynamics of learning and relearning.
Datasets
The model was fitted to existing experimental data from slice paired recordings of layer-5 pyramidal cells, specifically focusing on frequency- and timing-dependent STDP data (Sjöström et al., 2001). Additionally, the model's predictions for improved sensory perception were compared against in vivo experimental data (Froemke et al., 2013). No new datasets were introduced in this paper.
Key Insights and Ideas
- Unified Plasticity is Key for Precision: The model shows that combining presynaptic and postsynaptic plasticity significantly improves the signal-to-noise ratio (SNR) and discriminability of sensory inputs, leading to more precise receptive fields compared to plasticity at only one site.
- Presynaptic Changes Enable Reliability: While postsynaptic changes can improve reliability, the model reveals that presynaptic potentiation is crucial for counteracting the increased variance that arises from postsynaptic potentiation, thus ensuring reliable synaptic transmission.
- Hidden Memory Traces for Rapid Relearning: The model demonstrates that the distinct dynamics of presynaptic and postsynaptic plasticity leave behind a "hidden" postsynaptic memory trace. This trace allows for remarkably fast relearning of previously learned information, even after a long period of disuse.
- Cellular Basis for Memory Savings: This rapid relearning capability provides a potential cellular substrate for the psychological phenomenon of "memory savings," where relearning is faster than initial learning.
- Experience-Dependent Asymmetry: The findings suggest that the distribution of plasticity between pre- and postsynaptic sites is not fixed but can be experience-dependent, allowing for flexible adaptation to changing environments.
Key Results
- Simulations showed that unified pre- and postsynaptic STDP led to receptive fields with reduced variability and improved discriminability compared to postsynaptic plasticity alone.
- The model predicted that presynaptic potentiation (increasing P) is necessary to achieve high signal-to-noise ratios (SNR) for sensory inputs, complementing postsynaptic potentiation (increasing q).
- In learning simulations, the model demonstrated that relearning a previously learned stimulus was up to an order of magnitude faster than initial learning, attributed to the persistent postsynaptic memory trace.
- The model's predictions for improved sensory perception after learning were consistent with experimental observations.
Limitations and Future Work
- Model Simplification: While biologically tuned, the model is still a simplification of complex biological processes. The exact mechanisms and timescales of plasticity can vary across different brain regions and cell types.
- Generalizability: The current model is primarily fitted to neocortical data. The authors suggest that the framework could be extended to other brain regions, but acknowledge that different parameters might be needed to capture regional variations in plasticity expression.
- Future Directions: The authors propose that understanding the interplay between different timescales of pre- and postsynaptic plasticity could reveal general principles of neural computation. Further research could explore how this unified plasticity model applies to other brain areas and cognitive functions, and investigate the precise molecular mechanisms underlying the distinct temporal dynamics of presynaptic and postsynaptic changes.
Summary updated