Coupling, membrane conductance, and ion channel mRNA profiles in the establishment and maintenance of network activity in the crustacean cardiac ganglion

Neural networks produce critical rhythmic behaviors throughout an animal's lifespan, despite growth, differing environments, and changes in physiological state. This requires networks which balance stability in their properties with the plasticity necessary to respond to altered demands or perturbations. Studying the mechanisms which confer these properties requires a well characterized system with a known network topology and identifiable neurons that are amenable to both electrophysiological and molecular characterization and manipulation. Here, we use two networks from Cancer borealis to explore activity dependent regulation of cell connectivity, changes in cell properties with prolonged perturbation, and reliability of gene expression as a means for cell identification. For the first two topics we use the cardiac ganglion alone. The cardiac ganglion consists of a kernel of four interneurons that drive five motor neurons (termed large cells, LCs) which innervate the heart musculature. LCs burst synchronously due to simultaneous stimulation and electrical coupling through gap junctions. Depolarizing pharmacological perturbations have been shown to result in hyperexcitability (Ransdell et al., 2012a) and disrupt synchrony between LCs (Lane et al., 2016) eliciting rapid plasticity in ionic currents and electrical coupling which restores synchrony and excitability (Ransdell et al., 2012a; Lane et al., 2016). The salient electrophysiological signal which elicits coupling plasticity has not been identified. Using voltage clamp we directly control LC depolarizations to vary amplitude and timing of activity between LCs. We find that timing between cells, rather than depolarization elicits plasticity with the direction, i.e., potentiation or depression, being determined by the degree of desynchronization. With dynamic clamp we artificially couple networks from two animals and show that strong coupling with sufficient desynchronization can compromise a cell's output. These results suggest that coupling strength is tuned promoting synchrony or baseline cellular activity in a degree dependent manner. While rapid compensatory plasticity to hyperexcitability has been shown, it is unknown whether the changes are solely post-transcriptional and whether the short-term changes persist over longer time scales. We perturb networks for one or twenty-four hours and compare LCs' excitability, membrane properties, and abundances of ion channel and gap junction transcripts. We find evidence of rapid transcriptional changes at one hour, which may be maintained or regress at twenty-four hours. Additionally, we find that membrane properties and excitability are not maintained from one to twenty-four hours, suggesting a failure to maintain homeostasis or that additional compensatory changes are occurring at the network level. To address our third topic, we use LCs in addition to neurons collected form the stomatogastric ganglion which coordinates mastication and filtering in the digestive track. Both systems allow for unambiguous identification of cells based on anatomy or neuronal projections. We use this to evaluate the efficacy of cluster estimation procedures, clustering methods, and classification algorithms to determine the number of cell types present, group like cells together, and identify cells based on gene expression alone. We use single cell RNA-seq and single cell qRT-PCR to measure all contigs or a select set of ion channel, receptor, and gap junction mRNAs. We find these methods do not reproduce the known number of cell types present. Furthermore, although clustering and classification both outperform chance, we are unable to recapitulate cell type with complete accuracy from these data. These results indicate that, while promising, determining cell type by molecular profiling should not be relied on as the sole metric of cell type determination.