Extended Data Figure 5

a, Expression of ChR2 throughout PAG in consecutive coronal brain sections (left), and fibre placements in Vglut2-ires-Cre mice of experimental and control groups (right). b, Light-evoked effect on freezing behaviour induced by activation of different glutamatergic subpopulations of PAG neurons in naive animals (n = 12 PAG, n = 10 vlPAG, n = 7 vlPAG-to-Mc, Kruskal–Wallis test, P < 0.001, Dunn’s multiple comparison post-hoc test). c, EnvA-ΔG–mCherry–rabies-mediated, Cre-dependent monosynaptic retrograde tracing of inputs to GAD2⁺ and vGluT2⁺ neurons in the vlPAG and dl/lPAG. d, Rabies-mediated labelling of presynaptic neurons within PMD (left panels) and CEA (right panels) of Gad2-ires-Cre and Vglut2-ires-Cre mice (scale bar, 100 μm). e, Statistical analysis reveals differential input to vGluT2⁺ (n = 3 mice for each vlPAG and dl/lPAG) and GAD2⁺ (n = 4 mice) neurons within vlPAG or dl/lPAG. While CEA preferentially targets GAD2⁺ neurons of the vlPAG (1 × 3 ANOVA, F_(2,6) = 21.67, P < 0.01, Tukey’s post-hoc test), vGluT2⁺ neurons of the dl/lPAG receive stronger inputs from PMD (1 × 3 ANOVA, F_(2,7) = 287, P < 0.0001, Tukey’s post-hoc test). f, Cell-type-specific monosynaptic rabies tracing of VMH and LH inputs to vGluT2⁺ neurons in the dl/lPAG, vlPAG and vlPAG GAD2⁺ neurons (scale bars in overview, 200 μm; in zoom-in, 25 μm). g, Quantification of starter cells in the PAG and presynaptic cells in CEA and hypothalamic subregions. Boxes indicate median and 25th–75th percentiles of the distribution. **P < 0.01; ***P < 0.001.