On the Analogy between Human Brain and LLMs: Spotting Key Neurons in Grammar Perception