dPAG single unit recordings during risky foraging.

(A) Rats underwent pre-robot, robot, and post-robot sessions, successfully securing pellets in pre- and post-robot trials, and failing during robot trials due to robot surge. (B) Tetrode implantation in dPAG with photomicrographs of the tip (arrowhead) and unit clusters/waveforms. (C) Outbound foraging time increased during the robot session (Χ2 = 64.00, P < 0.0001, Friedman test; Ps < 0.05 for all comparisons, Dunn’s test). ***P < 0.001 compared to pre-robot and post-robot sessions. #P < 0.05 compared to the pre-robot session. (D) Pellet success rate decreased during robot session (Χ2 = 84.00, P < 0.0001, Friedman test; Ps < 0.0001 for all comparisons, Dunn’s test). ***P < 0.001 compared to pre-robot and post-robot sessions. (E) Cell type proportions showed 23.4% cells responding to robot activation (robot cells). (F) Representative dPAG robot cell raster/event histogram aligned with robot activations, and population activity of robot cells around robot activation time (t = 0) with 0.1 s and 1 s bins. Firing rates of the robot cells were higher during robot session (0-3 s blocks; Friedman test, all Χ2s > 6.952, all Ps < 0.05; Dunn’s test, all Ps < 0.05). Shaded areas indicate SEM. **P < 0.01 compared to pre-robot session. #, ##, and ### denote P < 0.05, P < 0.01, and P < 0.001 respectively, compared to post-robot session.

dPAG optical stimulation evokes fear.

(A) Virus injection, optrode implantation in dPAG, and light stimulation during single unit recordings in anesthetized rats. (B) Raster plots and peri-event time histograms for 20-Hz light stimulations (10-ms width, left; 2-s duration, center). 48% of 25 units had increased firing during 2-s light stimulation (right). (C) Virus injection, expression, and optic fiber placement in dPAG. (D) Stimulation testing: baseline trials at 75-cm distance (Long) without light; stimulation trials with 2-s light as the rat approached (∼25 cm) Long pellet; light applied as the rat approached 25-cm (Short) pellet if Long pellet unsuccessful. (E) Representative trajectories for EYFP- and ChR2-expressing rats during stimulation testing. (F) Rat behaviors during light stimulation trials. (G) ChR2 rats showed increased latency to procure pellet upon opto-stimulation (OnL) compared to EYFP rats (OffL, Z = 1.013, P = 0.311; OnL, U = 0.0, P < 0.001; Mann-Whitney U test). (H-J) ChR2 group exhibited increased latency to procure pellet compared to EYFP group based on stimulation intensity (H; Us < for all intensities < 3.5, Ps for all intensities < 0.025; Mann-Whitney U test), frequency (I; Us for 10 Hz and 20 Hz < 2.5, Ps for 10 and 20 Hz < 0.014; Mann-Whitney U test), and duration (J; Us for all durations < 4.5, Ps for all durations < 0.032; Mann-Whitney U test). *, **, and *** denote P < 0.05, P < 0.01, and P < 0.001, respectively.

dPAG optical stimulation and amygdala single-unit recordings.

(A) Virus injection in dPAG and tetrode array implantation targeting BLA. Light stimulation during single-unit recordings in freely-moving rats. (B) Stimulation testing sessions: in pre- and post-stim trials, rats freely procured pellets; in stim trials, optical stimulation prevented procurement of pellets. (C,D) During dPAG stimulation, animals showed increased outbound foraging time (C; Χ2 = 117.8, P < 0.0001, Friedman test; Ps < 0.0001, Dunn’s test) and decreased success rate (D; Χ2 = 154.0, P < 0.0001, Friedman test; Ps < 0.0001, Dunn’s test). ****P < 0.0001 compared to pre-robot and post-robot sessions. ####P < 0.0001 compared pre-robot session. (E) Subset of BLA units (10.0%) responsive to optical stimulation (Stim cells; left), and a representative (center) and all stimulation-responsive (nonStim cells; right) raster plots with PETHs. (F) Subset of animals underwent additional robot trials following the post-stim session. (G) Increased outbound foraging time during robot session compared to post-stimulation session (t(15) = 6.655, P < 0.0001; paired t-test). ****P < 0.0001. (H) Twenty-two BLA units were dPAG stimulation-responsive. (I) Representative raster plots of dPAG stimulation-responsive and -nonresponsive units. (J) Proportions of robot vs. non-robot cells differed between stimulation-responsive and -nonresponsive units (Χ2 = 11.134, P < 0.001; Chi-squared test). ***P < 0.001. (K) PETHs of stim and non-stim cells during stimulation and robot sessions. (L) Relationship between maximal firing rates during first 500 ms subsequent to robot activation and maximal firing rates during first 500 ms after stimulation onset (r(85) = 0.405, P < 0.001; Pearson correlation). (M) Population CCs with significant synchrony during robot session were higher than other sessions. Dotted vertical lines indicate 0-100 ms window for testing significance. Grey, blue, and dark yellow ***P < 0.001 compared to pre-stimulation, stimulation, and post-stimulation sessions, respectively. (N) Among synchronized BLA cell pairs, those including dPAG stimulation-responsive cell(s) (stim pairs; 61.5%) showed increased correlated firings (area under the curve, AUC, during 0-100 ms window) during the robot session compared to other sessions. In contrast, synchronized BLA cell pairs that consisted of stimulation non-responsive cells only showed no AUC differences across sessions. Grey *, blue **, and dark yellow **P < 0.05 compared to pre-stimulation, P < 0.01 compared to stimulation, and P < 0.001 compared to post-stimulation sessions, respectively. (O) Comparing CCs during testing windows (0-50 ms and 50-100 ms) between stim and non-stim pairs, stim pairs exhibited higher correlated firing than non-stim pairs during the 0-50 ms block (t(21.99) = 2.342, P = 0.0286; t-test), while displaying decreased correlated firings in the second block (50-100 ms; U = 42, P = 0.045; Mann-Whitney U test). *P < 0.05 compared to the non-stim pairs.

PVT as a potential mediator of dPAG signals to amygdala and a hypothesized model of the dPAG-amygdala circuit in anti-predatory behaviors.

(A) Retrograde tracer CTB injected into the BLA and AAV-CaMKII-EYFP injected into the dPAG to explore candidate areas conveying dPAG signals to the BLA. (B) Representative images of AAV and CTB expressions in dPAG and BLA, respectively. (C) Rats encountering the robot predator failed to procure pellets, while foraging-only animals succeeded (Base, U = 26, P = 0.6926; Test, U = 0, P = 0.0001; Mann-Whitney U test). ***P < 0.001 compared to the foraging-only group. (D) Terminal expressions of AAV injected into the dPAG cell bodies were predominantly observed in the midline nuclei of the thalamus. (E) PVT showed higher c-Fos-positive cells in robot-experienced rats compared to foraging-only control rats (U = 3.0, P = 0.0016; Mann-Whitney U test), while other midline thalamic areas showed no differences (Us > 12.0, Ps > 0.129; Mann-Whitney U test). **P < 0.01 compared to the foraging-only group. (F) Representative photomicrographs of PVT c-Fos staining from foraging-only (upper) and robot-experienced (bottom) rats. (G) Schematic of triple staining in the PVT. (H) Triple staining of AAV, CTB, and c-Fos in the PVT suggests it might mediate predator information from the dPAG to the BLA. (I) A putative predatory fear circuit model: Predator surge detected by visual pathways, such as the superior colliculus (Furigo et al. 2010, Rhoades et al. 1989), is transmitted to the dPAG, then to the PVT, which excites the BLA. The BLA projects to regions involved in executing escape responses, such as the ventral striatum (Li et al. 2021, Menegas et al. 2018) and ventromedial hypothalamus (Silva et al. 2013, Kunwar et al. 2015, Wang, Chen, and Lin 2015).

Cell types of the dPAG units.

(A) A subset (35.1%) of dPAG neurons increased firing rates in response to the robot, food pellet, or both. Among them, 66.7% of dPAG neurons responded exclusively to the robot. Units were classified as “robot cells” if their z-score was greater than 3 in one or more bins during the robot phase and less than 3 during the pre-robot phase. Conversely, units were classified as “pellet cells” if their z-score was greater than 3 during the pre-robot phase and less than 3 during the robot phase. If the z-score was greater than 3 in both the pre-robot and robot phases, those units were classified as “BOTH cells.” (B) Raster plots showing all robot, pellet, and BOTH cells during pre-robot, robot, and post-robot sessions. (C-E) The top row in each group of cells shows the mean firing rate and movement speed (± SEM; shaded areas) during the pre-robot, robot, and post-robot sessions. The bottom rows show the correlation coefficients between the firing rate and movement speed during each session in the robot (C), pellet (D), and BOTH (E) cell groups.

Optrode recording of the dPAG.

(A) Virus was injected in the dPAG and an optrode was implanted targeting the dPAG (left). Photomicrographs show virus expression and optic fiber/electrode tips, as well as a representative dPAG neuron (right). Light stimulation was delivered while single units were recorded in anesthetized rats. (B) Response latencies of the stimulation-responsive dPAG cells. (C) There was one dPAG cell showing inhibited responses to the optical stimulation of the dPAG.

Characteristics of BLA cells.

(A) Among 85 units collected from the BLA, 25.9% units immediately responded to the robot. (B) Representative waveforms of the BLA cells. (C) A representative raster plot and peri-event time histograms to the 20-Hz light stimulations (10-ms pulse width, 2-s duration). (D) Response latencies of the BLA cells that responded to the dPAG stimulation. (E) A pre-event time histogram showing the BLA activity aligned to the pellet procurement (pre-stim and post-stim sessions), dPAG stimulation (stim session), or robot activation (robot session). (F) The firing rates of the robot cells and movement speeds were plotted for the pellet procurement (post-robot session) and the robot activation (robot sessions). (G) All BLA cells, except for one, showed no significant correlation between the firing rate and the movement speed. (H, I) dPAG stimulation-responsive units showed higher levels of robot-evoked firings compared to the stimulation-non-responsive units during the short (H; 0-0.5 s after robot activation; U = 339.0, P = 0.0001; Mann-Whitney U test) and long-lasting (I; 0-5 s after robot activation; U = 384.5, P = 0.0009; Mann-Whitney U test) time windows. *** denotes P < 0.001 compared to the non-stim pairs.

Spike synchrony of the dPAG stimulation-responsive BLA neurons under predatory threats.

(A) Among the 185 CCs of the simultaneously recorded BLA neurons, 14.1% of the pairs showed significant (z > 3) peaks during the robot session. (B, C) The mean CC AUCs (B) and CC peak values (C) were higher during the robot session compared to the other sessions (Zs > 29.36, P < 0.0001, Friedman test; Ps < 0.01 for pre-stimulation vs. robot, stimulation vs. robot, and post-stimulation vs. robot sessions comparisons, Dunn’s test). **, ***, and **** denote P < 0.01, P < 0.001, and P < 0.0001 compared to the robot session, respectively. (D) Among the BLA cell pairs that significantly synchronized during the robot session, 61.5% of the pairs (stim pairs) included dPAG stimulation-responsive BLA cells while 38.5 % of the pairs (non-stim pairs) did not. (E) Spike synchrony of the stim pairs. A representative CC of a stim pair (top) showed increased correlated firing during the robot session. The mean CC AUC (bottom, left) and CC peak (bottom right) of stim pairs during the robot session were higher than those during the other sessions (Zs > 17.42, P < 0.001, Friedman test; Ps < 0.037 for pre-stimulation vs. robot, stimulation vs. robot, and post-stimulation vs. robot sessions comparisons, Dunn’s test). *, **, and *** denote P < 0.05, P < 0.01, and P < 0.001 compared to the robot session, respectively. (F) Spike synchrony of the non-stim pairs. The representative CCs show significant correlated firing during the robot session in the non-stim pairs (top). The CC AUC (bottom left) did not differ between the robot and pellet-only sessions (pre- and post-stim sessions) while peak area during the robot session was higher than that during the stim session (Z = 13.84, P < 0.01, Friedman test; P = 0.0017 for the stimulation vs. robot sessions comparison, Dunn’s test). The CC peak (bottom right) during the robot session was higher than those during the stimulation and post-stimulation sessions (Z = 15.92, P < 0.01, Friedman test; Ps < 0.0335 for stimulation vs. robot and post-stimulation vs. robot sessions comparisons, Dunn’s test), but not the pre-stimulation session. *, **, and *** denote P < 0.05, P < 0.01, and P < 0.001 compared to the robot session, respectively. (G) Stim pairs showed higher correlated firings than the non-stim pairs during the 20-40 ms time period after the paired cell fired (U = 32, P = 0.011; Mann-Whitney U test) while they showed lower correlated firings than the non-stim pairs during the 60-100 ms time window of the CCs (60-100 ms; Us < 37, Ps < 0.023; Mann-Whitney U test). * denotes P < 0.05 compared to the non-stim pairs. (H) The CC peaks of the stim pairs during the robot session tended to be higher than those of the non-stim pairs (U = 43.0, P = 0.051; Mann-Whitney U test).

Histological reconstructions of the recording sites in the dPAG and BLA, and optic fiber locations in the dPAG.

(A) Red bars indicate the trajectory of the tetrode recording sites in the dPAG. (B) Red bars indicate the trajectory of the optrode recording sites in the dPAG. (B) Red and gray circles indicate the optic fiber locations of the ChR2 and EYFP rats, respectively. (D) Orange circles and red bars indicate the optic fiber locations in the dPAG (left) and recording trajectories in the BLA (right). The numerical values represent AP coordinates relative to the Bregma.