Imaging, counts, and figures.

All slides were imaged with a 10x or 20x objective on an Olympus VS120 slide-scanning microscope. For brightfield images of DAB / NiDAB and Nissl-counterstained sections, we used extended focal imaging (EFI) to collect and combine in-focus light from all focal planes through the tissue. After reviewing whole-slide images in OlyVIA or VS-ASW software (Olympus), we collected additional EFI or multifocal image stacks with the 20x or 40x objective in some regions of interest. For dual-retrograde tracing (CTb + Fg) experiments, we counted double- and triple-labeled cells by scrolling up and down through 40x focal planes (12 μm stacks; 0.84 μm Z-spacing between each image) to confirm co-localization.

For mRNA co-localization we used 20x-40x image stacks to count HSD2-immunoreactive or Hsd11b2 mRNA-expressing cells (denominator), only those containing a nucleus (DAPI) in the plane of section, and to score each neuron for the presence or absence of the co-labeled mRNA of interest. We set a conservative counting criterion of at least 10 distinct fluorescent dots per cell (presumptive mRNA transcripts) for scoring individual Hsd11b2/HSD2 neurons, meaning that an individual neuron was scored as positive of gene expression of interest if and only if it contained at least 10 distinct transcripts for the mRNA in question, within the same cytoplasm containing Hsd11b2 mRNA or HSD2-ir surrounding that neuron’s DAPI+ nucleus and, in cases with ubiquitin mRNA labeling, also within the bounds of that neuron’s dense cytoplasmic mRNA for Ubc. For all mRNA expression, the vast majority of HSD2 neurons contained either zero or many more than 10 fluorescent dots (mRNA transcripts).

To prevent overcounting objects that span two tissue sections, we Abercrombie-corrected our HSD2 neuron counts using the average diameter of our criterion object, the nuclear diameter of HSD2 neurons (long-axis measurement), which ranged from 9.1 ± 0.1 μm (n=170) in immuno-labeled tissue or 9.4 ± 1.5 μm (n= 42) in mRNA-labeled tissue.

In every axon tracing case, we immunolabeled, imaged, Nissl counterstained, reimaged, and examined all sections from least one 1-in-3 series through the full brain, from the olfactory bulbs to the cervical spinal cord. To construct the whole-brain map of axonal projections, we manually traced axons and boutons with NiDAB immunolabeling for dsRed (Syp-mCherry) in Adobe Illustrator, then overlaid the Nissl-counterstained, re-scanned images of each section, which aided in labeling and outlining a minimal set of landmarks at each level for neuroanatomical clarity. We used neuroanatomical terminology from the peer-reviewed scientific literature (Geerling and Loewy, 2006b; Shin et al., 2008; Geerling et al., 2010; Verstegen et al., 2017) and, for some well-established regions and neuron populations, from commercially-available mouse brain atlases (Dong, 2008; Franklin and Paxinos, 2013). We used Adobe Photoshop to import raw fluorescence (grayscale) data into color channels for multicolor combinations, crop bitmap images, and adjust brightness or contrast. We added lettering and made drawings in Illustrator. Scalebars were traced atop calibrated lines in OlyVIA/VS-ASW to produce a simple white or black line.