average of 66 ± 17 million 100 bp reads per sample. Reads were trimmed using Sickle software (https://github.com/najoshi/sickle) and only paired reads with a quality score >20 and a minimum length of 20 bp were pseudo-aligned and quantified to merged GENCODE (M9) and RefSeq gene models using Kallisto71. Samples were normalized and pairwise comparisons (cell body to neuropil) were made using Sleuth software (https://github.com/pachterlab/sleuth)72, with a false discovery rate (FDR, q value) of 0.05 (reported in Results). Scatter plots for Kncj6 transcripts corresponding to accession numbers NM_001025584.2 (Girk2a) and NM_010606.2 (Girk2c) were made using log2 normalized counts.
