Pairing Electrophysiology and Single Cell RNA Sequencing to Identify Mechanisms of Cellular pH/CO2 Sensitivity in Respiratory Chemoreceptor Neurons

Abstract

Respiratory chemoreceptors are specialized neurons that are excited by hypercapnic acidosis (HA) and encode brain pH and/or CO2 levels to drive the appropriate changes in breathing. A subset of serotonin‐producing (5‐HT) neurons develop an apparent intrinsic chemosensitivity to HA after ~P12, but the molecular mechanisms driving their cellular pH/CO2 sensitivity remains unknown. Given that Sudden Infant Death Syndrome (SIDS) in humans has been linked to developmental 5‐HT system defects and potential alterations in ventilatory chemoreflexes, we aimed to develop a new technique to identify transcriptional differences among pH‐sensitive and ‐insensitive 5‐HT neurons in order to ultimately define the genetic determinants of cellular pH/CO2 sensitivity. Acute brainstem slices (200 uM) were prepared from 18–23 day‐old transgenic Dahl Salt Sensitive rats that express eGFP in all 5‐HT neurons (SSeGFP). Firing rates were measured via cell‐attached patch in eGFP+ 5‐HT neurons superfused with an artificial CSF (aCSF) containing inhibitors for synaptic blockade (10 mM CNQX, 50 mM D‐AP5, 20 mM Gabazine) while cycling between control (5% CO2 bal. O2; pH= 7.36; 5 min) and HA (15% CO2 bal. O2; pH= 7.10; 10 min) conditions. Once the cellular phenotype was established for pHsensitive (n=6 cells) or ‐insensitive (n=5) cells, a second electrode filled with small amounts (<1ul) of intracellular solution spiked with RNAse inhibitor was used to isolate the cellular contents for subsequent single cell RNA Sequencing (scRNA‐Seq). Specificity of cellular isolation was confirmed by high levels of expression of several serotonin and neuronal genes, and a lack of glial gene expression by sc‐qRT‐PCR (n=7) or sc‐RNAseq. The scRNA‐Seq read summary indicated an average of ~45M reads, a quality score of ~33.5, and ~80% mapping rate per sample, from which >10,000 known genes were quantified. However, small sample sizes prevented obtaining statistical significance for differential expression (Q > 0.05). Using R software packages, unsupervised analyses (hierarchical clustering and principal component analyses) failed to clearly separate samples by phenotype, and a power analysis based on FPKM values from the 11 samples suggested sample sizes of 20–60 cells/group are required. However, a supervised analysis (pre‐defining experimental groups) using Support Vector Machine (SVM) learning coupled with Recursive Feature Elimination (i.e. creates a model with all genes and recursively reduces the number of genes to identify the fewest genes to explain the predefined phenotype) revealed a candidate molecular signature for CO2 sensitive 5‐HT neurons. Overall, we demonstrated we can obtain transcriptomic data from electrophysiologically‐characterized 5‐HT neurons at the single cell‐level, and analyses of these pilot data suggest unsupervised and supervised machine learning approaches may identify unique genetic profiles in pH‐sensitive 5‐HT neurons. These data provide the field of respiratory physiology with preliminary markers of 5‐HT neuron CO2/pH sensitivity and validate that this approach may be used to address questions pertaining to genetic determinants of a given cellular phenotype.Support or Funding InformationNIH R01 HL12358 04 & Parker B. Francis FoundationThis abstract is from the Experimental Biology 2019 Meeting. There is no full text article associated with this abstract published in The FASEB Journal.