Continuous memory stabilization based on different adaptive processes in neural circuits

学习和创造记忆的能力使人类和动物能够在复杂和不断变化的环境中生存。感知我们的环境会引起神经活动，导致突触适应各种可塑性过程。我们对神经元和突触过程如何相互作用，选择一些特定的输入信号来记忆，而忽略其他信号知之甚少。特别是，目前还不清楚神经元网络如何稳定记忆并将其维持更长的时间，特别是考虑到神经元基质在记忆的生命周期中并不保持恒定，而是发生了很大的变化。记忆的形成、巩固和长期稳定是基于神经元和突触过程的动态机制，其在许多时间尺度上操作。这些过程只被部分理解，它们的相互作用仍然是未知的。因此，这项工作的目标是建立一个神经理论框架，以研究形成，巩固和长期稳定。我们将特别关注潜在的神经过程及其在不同时间尺度上的相互作用，这将有助于更好地理解记忆的动态。我们的中心假设是，记忆巩固或稳定是一个多步骤的过程：我们提出，通过突触和睡眠诱导的网络巩固过程的输入依赖性相互作用，首先，记忆表征或细胞组装是通过强烈互连的神经元组形成的，然后，这些细胞组装中的一些在更长的持续时间（天）内稳定下来，而另一些分裂导致遗忘（通过合并进行选择）。接下来，我们认为，细胞组装体的持续稳定提供了足够的时间，通过缓慢的结构可塑性过程（长期稳定）重建其内部连接，使其稳定更长的时间。这种方法的最终结果应该是一个跨不同时间尺度的动态记忆模型，它允许研究记忆的连续稳定性。

The ability to learn and to create memories enables humans and animals to survive in complex and changing environments. Sensing our environment causes neural activity resulting in synaptic adaptation by various plasticity processes. Little is known about how neuronal and synaptic processes interact to select some specific input signals to remember while disregarding others. In particular, it is unclear how the neuronal network stabilizes a memory and maintains it for longer periods of time, especially given that the neuronal substrate does not remain constant but changes substantially during the lifetime of a memory. Memory formation, consolidation, and long-term stabilization are dynamic mechanisms based on neuronal and synaptic processes, which operate across many time scales. These processes are only partially understood and their interactions are still widely unknown. Thus, the goal of this work is to develop a neuro-theoretical framework to investigate formation, consolidation, and long-term stabilization. We will specifically focus on the underlying neural processes and their interactions across different time scales, which should lead to a better understanding of the dynamics of memory. Our central hypothesis is that memory consolidation or stabilization is a multi-step process: We propose that, by the input-dependent interaction of the processes of synaptic and sleep-induced network consolidation, first, memory representations or cell assemblies are formed by strongly interconnecting groups of neurons and, then, some of these cell assemblies are stabilized for longer durations (days), while others disintegrate leading to forgetting (selection by consolidation). Next, we suggest that the days-lasting stabilization of cell assemblies provides enough time to stabilize them for longer times by restructuring their internal connectivity through the slow processes of structural plasticity (long-term stabilization). The final outcome of this approach should be a dynamic memory model across different time scales, which allows investigating the continuous stabilization of memories.