Single neurons act as a memory buffer for space

The ability of the brain to briefly retain information is believed to depend on persistent neural firing. Historically, persistent firing has been attributed to recurrent synaptic networks, in which neural activity reverberates. This view aligns with the widely accepted principle that neurons generate action potentials only in response to sufficiently strong input, and therefore do not serve as memory components. Here, we present evidence challenging this view, demonstrating in behaving mice that individual hippocampal neurons can sustain persistent firing, acting as a memory buffer for spatial representation. We first demonstrate that TRPC4 ion channel silencing selectively disrupts the capability of individual neurons to persistently fire using in vitro recordings in mice. This manipulation impairs spatial working memory performance, and significantly and selectively reduces persistent firing in vivo during the maintenance (delay) period of the task. Despite general belief that persistent firing maintains working memory contents, we find persistent firing to reflect continuous firing of spatial cells coding for spatial information, suggesting that maintenance of spatial representation in the hippocampus may depend on intrinsic cellular mechanism. We further demonstrate that TRPC4 silencing selectively affect hippocampal spatial coding where mice stayed in a place for an extended time (start and goal areas), pointing out the specific role of intrinsic cellular mechanisms of persistent firing in the retention of spatial information and potential coexistence of intrinsic cellular mechanism with attractor dynamics. Finally, task performance correlated with the strength of persistent firing in the goal area, suggesting that maintained goal representation in the hippocampus is crucial for the task performance. These findings redefine neurons as active contributors to information retention beyond their conventional role as passive input-output units, potentially reshaping our general understanding of brain computation.