Imaging and cell counting

Images were captured through a 10x dry objective (NA 0.70), a 40x oil-immersed objective (NA 1.30) and a 63x oil-immersed objective (NA 1.40) from a Leica SP8 confocal microscope (Leica Microscopy). First, all processed brain slices, as well as all slices containing an injection site were imaged using 10x or 20x objectives. For cell counting of NeuN, GAD-67, PV and insula projectors, z-stacks of ROIs (z step: 0.5 μm, 3–4 frame average, 1024 × 1024 pixels) were scanned using the 40x objective (Fig. ​(Fig.1e-f1e-f and ​and3c3c and Supplementary Fig. 2a). Injection of retrograde tracers in two different targets is not an appropriate technique to quantify collateralization to two downstream regions due to three main limitations: (1) tracers have limited efficiency in retrograde transport (2) different CTB retrograde tracers have different efficiencies and (3) injections are performed to prioritize specificity over coverage of the target regions [45]. However, the overlap of CTB-555 and CTB-647 can instruct us whether collateralization exists. Additionally, we detected few, but some cell bodies containing both CTB-555 and CTB-647. Specifically, we found that 3.84 ± 1.35% of insula-amygdala labeled neurons were both IC-BLA and IC-CeA projectors (n = 5 mice, data not shown), and that 3.70 ± 0.14% of insula-LH labeled neurons were both IC-rLH and IC-cLH projectors (n = 4 mice, data not shown). Therefore, neurons projecting to both BLA and CeA, or to both rLH and cLH exist, but their exact proportion should be studied using different methods, such as synaptophysin expression using viral vectors [45]. Example immunofluorescent pictures of 5-HT1A+ or 5-HT2A+ cells were captured with the 63x objective (Fig. ​(Fig.1g-h,1g-h, ​g-h,3d-e,3d-e, ​d-e,5c-d5c-d and ​and5g-h5g-h and Supplementary Fig. 2b-c). In order to detect synaptophysin-eYFP or synaptophysin-mCherry signal, z-stacks of BLA, CeA, rLH and cLH have been scanned with 63x objective in a picture format of 1024 × 1024 pixels (Number of z-stacks: 30, z-steps: ~ 1 ± 0.10 μm, 2–3 frame average). All images have been processed using the open source Fiji software (ImageJ, NIH), and cell counting has been performed manually.