Promotion of Structured Motor Program Diversity Through Since-last-state Memory in Drosophila Larvae

Abstract

Motor systems controlling locomotion must generate repetitive rhythmic activity, while also still retaining the ability to generate a diverse range of outputs. How motor systems monitor, regulate, and promote diversity of their own outputs is not well understood. Here, we perform single-step, variable-order and hidden-state Markov modelling (HSMM) on spontaneous fictive locomotor activity in the isolated Drosophila larval nervous system to examine how a motor system balances constraint and promotion of diversity amongst competing motor programs. We show that spontaneous fictive activity is structured by interacting mechanisms operating at multiple levels of sequence organisation. Analysis of one-step transition rules revealed a bias in activity towards activity states underlying exploration that in turn, promote transition to diverse outputs. In contrast, higher-order Markov, N-gram, and HSMM analysis indicated a memory biased towards revisiting recently executed motor programs. These mechanisms together suggest that the Drosophila larval locomotor system maintains a dynamic repertoire of possible motor outputs by monitoring recent activity and biasing future transitions accordingly. In this sense, fictive rhythmogenesis reflects a diversity-generating process: the larval locomotor network does not simply repeat a fixed motor programme or randomly transition from one state to another, but rather continually regulates access to rhythmic states based on recent experience. Together, these findings suggest that fictive locomotor dynamics are consistent with adaptive winner-takes-all competition between central pattern generating (CPG) modules that balance constraint and promotion of motor program diversity.