The RNAscope multiplex fluorescent kit was used according to the manufacturer’s instructions for fresh frozen tissue sections (Advanced Cell Diagnostics), with the exception that 16µm tissue sections were fixed with 4% PFA at 4°C for 60 minutes and the protease treatment step was shortened to 15 minutes at room temperature. Probes used to identify nuclear and cytoplasmic enriched transcripts were designed antisense to the following mouse genes: Calb1, Grik1, and Pvalb. Following hybridization and amplification, stained sections were imaged using a 60X oil immersion lens on a Nikon TiE epifluorescence microscope.

To determine if spots fell within the nucleus or cytoplasm, a boundary was drawn around the nucleus to delineate its border using measurement tools within Nikon Elements software.  To delineate the cytoplasmic boundary of each cell, a circle with a diameter of 15um was drawn and centered over the cell (Fig. 5). RNA spots in each channel were quantified manually using counting tools available in the Nikon Elements software. Spots that fell fully within the interior boundary of the nucleus were classified as nuclear transcripts. Spots that fell outside of the nucleus but within the circle that defined the cytoplasmic boundary were classified as cytoplasmic transcripts. Additionally, if spots intersected the exterior boundary of the nucleus they were classified as cytoplasmic transcripts. To prevent double counting of spots and ambiguities in assigning spots to particular cells, labeled cells whose boundaries intersected at any point along the circumference of the circle delineating their cytoplasmic boundary were excluded from the analysis. A linear regression was fit to nuclear versus soma probe counts, and the slope was used to estimate the nuclear proportion.

Coronal brain slices from Nr5a1-Cre;Ai14, Scnn1a-Tg3-Cre;Ai14, and Rbp4-Cre_KL100;Ai14 mice were stained with anti-dsRed (Clontech #632496) to enhance tdTomato signal in red channel and DAPI to label nuclei. Maximum intensity projections from six confocal stacks of 1-µm intervals were processed for analysis. Initial segmentation was performed by CellProfiler \cite{Lamprecht2007} to identify nuclei from the DAPI signal and soma from the tdTomato signal. Segmentation results were manually verified and any mis-segmented nuclei or somata were removed or re-segmented if appropriate. Area measurement of segmented nuclei and somata was performed in CellProfiler in Layer 4 from Nr5a1-Cre;Ai14 and Scnn1a-Tg3-Cre;Ai14 mice, and in Layer 5 from Rbp4-Cre_KL100;Ai14 mice. A linear regression was fit to nuclear versus soma area to highlight the differences between Cre-lines.

For measurements of nucleus and soma size agnostic to Cre driver, we used 16 µm-tissue sections from P56 mouse brain. To label nuclei, DAPI was applied to the tissue sections at a final concentration of 1mg/ml. To label cell somata, tissue sections were stained with Neurotrace 500/525 fluorescent Nissl stain (ThermoFisher Scientific) at a dilution of 1:100 in 1X PBS for 5 minutes, followed by brief washing in 1X PBS. Sections were coverslipped with Fluoromount-G (Southern Biotech) and visualized on a Nikon TiE epifluorescence microscope using a 40x oil objective. Soma and nuclei area measurements were taken by tracing the boundaries of the Nissl-stained soma or DAPI-stained nucleus, respectively, using cell measurement tools available in the Nikon TiE microscope software. All cells with a complete nucleus clearly present within the section were measured, except that we excluded glial cells which had very small nuclei and scant cytoplasm. Measurements were taken within a 40x field of view across an entire cortical column encompassing layers 1-6, and the laminar position of each cell (measured as depth from the pial surface) was tracked along with the nucleus and soma area measurements for each cell.

For each cell in the experiments above, the nuclear proportion was estimated as the ratio of nucleus and soma area raised to the 3/2 power. This transformation was required to convert area to volume measurements and assumed that the 3-dimensional geometries of soma and nuclei were reflected in their cross-sections. This is true for approximately symmetrical shapes such as many nuclei and some somata, but will lead to under- or over-estimates of nuclear proportions for asymmetrical cells. Therefore, the estimated nuclear proportion of any individual cell may be inaccurate, but the average nuclear proportion for many cells should be relatively unbiased.

Data and code to reproduce all figures are publicly available from GitHub at https://github.com/AllenInstitute/NucCellTypes.

The authors declare no competing interests.