Highlights d Rats show stronger memory for highly rewarded locations d Disruption of hippocampal SWRs only affects memory for highly rewarded locations d Hippocampal replay events are biased to trajectories associated with high reward

We developed a dual-compartment neurofluidic system with inter-connecting microchannels to connect neurons from their respective compartments, placed on a planar microelectrode arrays. The design and development of the compartmented microfluidic device for neuronal cell culture, protocol for sustaining long-term cultures, and neurite growth through microchannels in such a closed compartment device are presented. Using electrophysiological measurements of spontaneous network activity in the compartments and selective pharmacological manipulation of cells in one compartment, the biological origin of network activity and the fluidic isolation between the compartments are demonstrated. The connectivity between neuronal populations via the microchannels and the crossing-over of neurites are verified using transfection experiments and immunofluorescence staining. In addition to the neurite cross-over to the adjacent compartment, functional connectivity between cells in both the compartments is verified using cross-correlation (CC) based techniques. Bidirectional signal propagation between the compartments is demonstrated using functional connectivity maps. CC analysis and connectivity maps demonstrate that the two neuronal populations are not only functionally connected within each compartment but also with each other and a well connected functional network was formed between the compartments despite the physical barrier introduced by the microchannels.

Falcon is a novel open-source software for closed-loop neuroscience. It has sub-millisecond intrinsic latency and gives the experimenter direct control of CPU resources. We envisage Falcon to be a useful tool to the neuroscientific community for implementing a wide variety of closed-loop experiments, including those requiring use of complex data structures and real-time execution of computationally intensive algorithms, such as population neural decoding/encoding from large cell assemblies.