Introduction

Threat-response neural circuits are conserved across species and have roles in normal behaviors and psychiatric diseases(1-4). Identifying and responding to threatening stressor is critical for survival but maladaptive changes in underlying neural circuits contribute to stress, mood, and anxiety disorders (5, 6). Active coping in response to stressors is psychosocial factor associated with resilience against stress induced mood and anxiety disorders(7). Available evidence indicates that active (e.g., escape, fighting) and passive (e.g., freezing, immobility) coping responses to stressors are governed by separable neural circuits (1, 8, 9). The functioning of circuits underling active coping could be key for overcoming anxiety and related disorders(10, 11).The neural circuits, cells, and mechanisms underlying active coping strategies remain unclear(12, 13).

The supramammillary nucleus (SuM) has been suggested to be engaged by threatening stressors and has efferent connections to stress-sensitive brain regions so may be an important regulator of responses to stressors (14-16). Research on SuM has focused on connections to the hippocampus and septum while SuM projections to other brain areas remain less understood(15, 17, 18). SuM contains distinct populations with functionally diverse roles including regulation of hippocampal activity during REM, spatial memory, arousal, and environmental interactions(19-23). In addition to functional diversity, anatomical and molecular-cellular diversity is present in SuM(24-26). Divergent populations in SuM have been defined by differential to the dentate gyrus and CA2 regions of the hippocampus, and neurochemically based on neurotransmitter expression (27, 28). Major projections from the SuM, which have yet to be examined, include the preoptic hypothalamus area (POA). We examined if SuM neurons projected to the POA and if this projection could be used to aid in separation of populations in SuM.

We used retrograde adeno associated viral (AAV) and combinatorial genetic tools to identify and characterize a population of glutamatergic SuM neurons with projections to the POA (SuMVGLUT2+::POA). We found that SuMVGLUT2+::POA neurons represented an anatomical subset of SuM neurons with extensive arborizations to brain regions including that mediate stress and threat responses including the paraventricular nucleus (PVT), periaqueductal gray (PAG), and the habenula (Hb). Thus, SuMVGLUT2+::POA neurons are positioned as hubs with spokes to many areas regulating responses to threatening stressors. We hypothesized this population could respond to and regulate responses to stressors. We found SuMVGLUT2+::POA neurons are recruited by multiple types of acute threatening stressors, encode a negative valance, do not promote anxiety-like behaviors, and evoke active coping behaviors. Further, activation of these SuMVGLUT2+::POA neurons was sufficient to convert passive coping strategies to active behaviors. These findings indicate SuMVGLUT2+::POA neurons are a central hub linked to multiple stress and threat responsive areas and can drive state transitions between passive and active responses to stress.

Results

VGLUT2+ SuM neurons projecting to the POA (SuMVGLUT2+::POA) arborize to multiple stress-engaged brain regions

The SuM contains functionally diverse and anatomically distinct populations with efferent projections to many brain regions(16, 24, 38, 39). The diversity includes glutamatergic, GABAergic, and co-expressing GABAergic/Glutamatergic populations(21, 40). Using retrograde adeno associated virus (Retro-AAV) (Retro-AAV2-DIO-tdTomato) and anterograde (AAV5-DIO-ChR2eYFP) tracing we identified a population of VGLUT2+ expressing neurons in SuM that project to the POA (SuMVGLUT2+::POA) with dense projections in the lateral preoptic area (LPO) within the POA (Supplemental Figure 1A-F) in VGLUT2-Cre mice (n=4). Using a viral construct encoding for a nuclear restricted fluorophore that switches from mCherry (red) to eGFP (green) in a Cre- dependent manner (Retro-AAV2-Nuc-flox(mCherry)-eGFP) we found SuMVGLUT2+::POA neurons were positive for VGLUT2+ in VGLUT2-Cre mice (n=4) and negative for VGAT expression in VGAT-Cre mice (n=3) (Supplemental Figure 1G-L) indicating that SuMVGLUT2::POA neurons do not belong to a GABAergic/Glutamatergic populations.

To examine arborization of SuMVGLUT2+::POA neurons we utilized a combinatorial genetic approach. Specifically, we used mice (VGLUT2-Flp) that express Flp recombinase in VGLUT2 expressing cells in a combination with Flp dependent expression (fDIO) of Cre and Cre (DIO) dependent fluorophore expression. We injected Retro-AAV2-DIO-eYFP into POA and AAV-fDIO-Cre in to SuM (Figure 1A). We thus expressed Cre in SuMVGLUT2+ neurons and, of those, only neurons projecting to the POA (SuMVGLUT2::POA neurons) were labeled with eYFP. We found SuMVGLUT2::POA neurons arborize widely projection to multiple brain regions including: the nucleus accumbens (Acb), Septum, lateral hypothalamus (LH), ventral medial hypothalamus (VMH), paraventricular nucleus (PVT), lateral habenula (lHb), the ventral lateral periaqueductal gray (VLPAG), dorsal raphe (DRD), lateral parabrachial nucleus (LPBN), laterodorsal tegmental nucleus (LDTg), and medial vestibular nucleus (MVe) (Figure 1 D-M). We visualized arborization of SuMVGLUT2::POA neurons in cleared tissue by using optical light sheet imaging. For these experiments we injected VGLUT2-Flp mice (n=3) we injected as in Figure 1A and brains were actively cleared using SHIELD for optical light sheet imaging(42). Compiled three-dimensional images of a brain hemisphere, viewed from the medial to lateral perspective (Figure 1M) or viewed from the ventral to dorsal perspective (Supplemental Figure 2M), showed projections labeled by eYFP from SuMVGLUT2+::POA to POA, areas of hippocampus, septum, Acb, and regions in the pons and midbrain. The results further established projections from SuMVGLUT2+::POA neurons to multiple brains areas and illustrate the broad arborization.

SuMVGLUT2+::POA arborize widely in the brain.

(A) Schematic of injections of Retro-AAV-DIO-eYFP in to the POA and AAV-fDIO-Cre into the SuM of VGLUT2-Flp mice to label only VGLUT+ SuM neurons that project to the POA. (B) Projections of VGLUT2+ SuM neurons labeled with eYFP seen in the POA. (C) Cell bodies labeled with eYFP in the SuM. (D-L) Arborizing processes from the SuMVGLUT2+::POA neurons are seen in multiple brain regions including the (D) AcbSh and AcbC; (E) lateral septum; (F) multiple hypothalamic areas; (G) LHb and PVT; (H) the PAG and the DRD; (I) VLPAG and LPB; (J) DRI and LDTg; (K) and DRI, and MVePC. (M) Light sheet microscopy image of a cleared mouse brain hemisphere, viewed from medial to lateral, showing eYFP labeled neurons in SuMVGLUT2+::POA neurons with cell bodies in SuM and processes in areas corresponding to septum, hippocampus, Acb, and POA. (500 μm scale bars)

Abbreviations: MPA- medial preotic, VLPO- ventral lateral preoptic, LPO- lateral preotic, HDB- nucleus of the horizontal limb of the diagonal band AcbSh- Accumbens shell, AcbC- Accumbens core, Shi- septohippocampal nucleus, LSI- lateral septal nucleus, intermediate part, LSV- lateral septal nucleus, ventral part, LV- lateral ventricle, MS- medial septal nucleus, LHb- lateral habenula, MHb- medial habenula, PVT- paraventricular thalamus, DM- dorsomedial hypothalamic nucleus, LH- lateral hypothalamic area, VMH- ventromedial hypothalamic nucleus, PAG- periaqueductal gray, DMPAG- dorsomedial periaqueductal gray, DLPAG- dorsolateral periaqueductal gray, LDTg- laterodorsal tegmental nucleus, LPAG- lateral periaqueductal gray, DRD- dorsal raphe nucleus, dorsal part, LPB- lateral parabrachial nucleus, MPB- medial parabrachial, scp- superior cerebellar peduncle, Bar- Barrington’s nucleus, DTg- dorsal tegmental nucleus, DRI- dorsal raphe, interfascicular part, LC- locus coeruleus, MVe- medial vestibular nucleus, 4V- fourth ventricle

We used tracing with Retro-AAV’s to confirm our findings from anterograde tracing by injecting unilaterally into the Acb, septum, PVT, or PAG in VGLUT2-Cre mice. Cell bodies in SuM were labeled eYFP or tdTomato (Supplemental Figure 2A-L). Each area was injected in three or more mice yielding similar results. We examine anatomic distribution of SuM::POA neuron in SuM by injected Retro-AAV2-Cre into the POA and Retro-AAV2-Nuc-flox(mCherry)-eGFP into the SuM (Supplemental Figure 3 A-B) of wildtype (WT) Cre- mice (n=3) injected with this combination of viruses we observed both neurons labeled by mCherry (Cre negative) or eGFP (Cre expressing) interspersed in the SuM. We verified the combinatorial viral selectively labeled SuMVGLUT2+::POA neurons with minimal background(43). Positive controls we injected Retro-AVV-Flpo into the POA and AAV-fDIO-Cre into the SuM labeling of cells in SuM (Supplement Figure 3 G and H). As negative controls, we injected only AVV-fDIO-Cre into the SuM in Ai14 mice, which did not yield tdTomato expression (Supplemental Figure 3C-D). To confirm that the combination of Retro-AAV2-fDIO-Cres and AAV5-DIO-ChR2eYFP did not lead to labeling of cells with eYFP in the absence of Flp we injected Retro-AAV2-fDIO-Cre into the POA and AAV-DIOChR2eYFP into SuM of WT mice. In these mice we did not observe any expression of eYFP (Supplemental Figure 3E-F). The results demonstrate the specificity of our combinatorial viral strategy. All studies replicated in a minimum of three mice.

SuMVGLUT2+::POA neurons are an anatomical distinct subset of all SuMVGLUT2+ neurons

Recent studies have highlighted functionally divergent roles of SuM neurons and suggested differential projections, particularly to regions of hippocampus and PVT, may identify functionally distinct populations (19, 20). To examine if SuMVGLUT2+::POA neurons are a subset of SuMVGLUT2+ neurons we qualitatively compared the projections of total SuMVGLUT2+ neurons to SuMVGLUT2+::POA neurons. Similar to projections for SuMVGLUT2+::POA neurons (Figure 1) we observed labeled projections from the total SuMVGLUT2+ neurons in the POA, PVT, LHB, and the CA2 field of the hippocampus (Figure 2 A-E). Importantly, we found areas that received projections from the total SuMVGLUT2+ and not from SuMVGLUT2+::POA neurons. As schematized (Figure 2K), projections from the total SuMVGLUT2+ but not SuMVGLUT2+::POA populations were present in dentate gyrus (DG) and medial habenula (mHb) (Figure 1G and Figure 2 F-J). Thus, SuMVGLUT2+::POA population can be further defined as a subset of the total SuMVGLUT2+ neuronal population with distinct projection targets. Some areas, such as septum, may receive projections from multiple populations.

SuMVGLUT2+::POA neurons project to a subset of brain regions compared to all SuMVGLUT2+ neurons.

(A) Schematic of injections in VGLUT2-Cre mice of AAV-DIO-ChR2eYFP into SuM. (B) eYFP (green) labeled neurons present in SuM and (C) processes were observed in the POA. (D) Process from SuM VGLUT2+ neurons were also present in MHb, LHb and PVT. (E) In hippocampus process were observed in DG and CA2. (F) Schematic of injections in VGLUT2-Flp mice of Retro-AAV-fDIO-Cre and AAV-DIO-ChR2eYFP into SuM. (G-H) Cells in SuM and process in POA were labeled by ChR2eYFP. (I) Labeled processes in LHb were evident but none seen in MHb. (J) In hippocampus, labeled processes were present in CA2 but not observed in DG. (500 μm scale bars) (K) Schematic summary showing the projection to DG and MHb present in the total SuM VGLUT2+ population (outlined arrows) but absent in the SuMVGLUT2+::POA population (filled arrows).

Threatening stressors not spontaneous higher velocity movement recruit SuMVGLUT2+::POA neurons

Neurons in SuM can be activated by acute stressors, and we sought examine the recruitment of SuMVGLUT2+::POA neurons by threatening stressors (14, 44, 45). We found that forced swimming induced cFos expression in SuM. Labeling SuMVGLUT2+::POA neurons using Retro-AAV-DIO-mCherry revealed an increase in the number of mCherry and cFos labeled cells following forced swim (Supplemental Figure 4A-G). To further examine recruitment of SuMVGLUT2+::POA neurons by acute stressors we tested a diverse set of acute threatening stressors using in vivo Ca2+ detection via fiber photometry. We expressed GCaMP7s in SuMVGLUT2+::POA neurons (Figure 3A-B) using the combinatorial viral genetic approach detailed for anatomic studies. We developed a forced swim assay that utilized a moveable platform allowing mice to be placed into and removed from water while obtaining fiber photometry recordings. Mice were dunked for a 30-second swim trial every 2 minutes for a total of 10 trials (n=8 mice). During the assay mice exhibited almost exclusively active swimming and climbing behaviors reflected in the quantification of mean time mobile (Supplemental Figure 4I). Analysis of Ca2+dependent (470 nm excitation) and isosbestic control (415 nm excitation) signals from fiber photometry recordings reveal rapid rise of approximately 5 standard deviations in GCaMP Ca2+ dependent signal with start of the swim trial that was sustained through the 30-sec swim period. The Ca2+dependent signal returned to baseline following the platform raising removing the animal from the water (Figure 3F and G). These data show the dunk assay evoked active coping behavior and recruited SuMVGLUT2+::POA neurons.

Acute stressors recruit SuMVGLUT2+::POA neurons.

(A) Schematic of VGLUT2-Flp mice injected with AAV-DIO-GCaMP7s in SuM and Retro-AAV-fDIO-Cre in the POA with a (B) fiber ced over SuM for fiber photometry recordings from SuMVGLUT2+::POA neurons. (C) While mice were connected for fiber photometry they were (C) subjected to ten 30-second trials of forced swimming, (D) to a 2-second foot shock following a 30 second tone for five trials, or (E) to mock predator ambushed by remote-controlled spider. (F) Heat map and mean ± 95% CI Z-score for recordings obtained from single animal during the repeated forced swim showing increase Ca2+ signal during the swim session. (G) The mean ±95% CI Z-score of 10 trials for all animals (n=8) in the dunk assay. (H) Heat map and an ± 95% CI Z-score for recordings from a single animal during the five shock trials showing increase Ca2+ signal. (I) Mean ± 95% CI Z-score for recordings of five trials for all animals (n=6) in the foot shock assay. (J) Heat map, mean ± 95% CI Z-score (blue), mean ± 95% CI velocity (gold), for recordings obtained from animals (n=9) during the ambush showing an increase in Ca2+ signal as the animals flee from the remote-controlled spider with the mean ±95% confidence interval Z-score of 1 trial bush) for all 9 animals. (K) Heat map, mean (± 95%CI) Z-score (blue), mean (± 95% CI) velocity (gold), for the same mice but ambush in the predator assay. Time frame gated for spontaneous locomotion. Ca2+ signal does not increase during spontaneous locomotion. Mean peak velocity was not significantly different.

We also tested a foot shock paradigm (Figure 3D) and a predator ambush assay (Figure 3E). In foot shock assays VGLUT2-Flp mice (n=6), prepared as above (Figure 3A and B), were subjected to five pseudo randomly spaced trials with 30 sec of tone proceeding a two-second shock. We observed a rapid increase in the Ca2+ dependent GCaMP7s signal following the foot shock (Figure 3H and I). For the predator ambush we adapted a previously demonstrated paradigm using a mock mechanical spider attack (46). In this assay a remote-controlled mechanical spider was hidden in a box with swing door. The box was inside a larger walled arena. Mice were able to freely explore the arena and at a moment when the mouse was in proximity to the box opening the spider was moved out towards the mouse. The mice fled, often to a corner, stopped, turned to face the spider (Supplemental video 1). In mice (n=9) subjected to the ambush paradigm we observed an increase in Ca2+ dependent GCaMP signal at the time of the ambush followed by suppression below the initial baseline (Figure 3J). Populations of neurons in SuM have previously been found to correlate with future movement velocity (41), so we examined if recruitment of SuMVGLUT2+::POA neurons was correlated to periods of spontaneous higher velocity movement. Analyzing data collected prior to the ambush event we examined if increased velocity, at speeds similar to the fleeing induced by the ambush, was correlated with increased Ca2+ dependent GCaMP signal in SuMVGLUT2+::POA neurons (Figure 3L). We found no correlation (Figure 3K). We further examined the data for a correlation of increased Ca2+ dependent signal in SuMVGLUT2+::POA neurons to movement velocity during open field exploration in a sperate cohort of mice (n=13). Here we found no evidence for correlation of velocity with Z-score of Ca2+ dependent signal including a cross correlation analysis to account for a potential temporal offset (Supplemental Figure 5). In aggregate, the data support recruitment of SuMVGLUT2+::POA neurons by diverse threatening stressors but not during times of higher velocity spontaneous movement.

SuMVGLUT2+::POA neurons evoke active coping behaviors

To examine how activation of SuMVGLUT2+::POA neurons contributes to responding to threatening stressors we assessed behavioral changes evoked by photostimulation of SuMVGLUT2+::POA neurons. We injected VGLUT2-Cre or WT (Cre-) mice with Retro-AAV5-DIO-ChR2eYFP in the POA, and an optic fiber placed over SuM (Figure 4A). We employed a paradigm with a 15-minute trial divided into three 5-minute periods, pre-stimulation, stimulation at 10 Hz, and post-stimulation. Based on review of videos obtained during the 15-min trials, we observed behaviors that could be classified into nine distinct categories: grooming, immobile, walking, chewing of bedding, rearing, rapid locomotion (movement was limited by the size of the arena), digging (moving bedding towards the tail), treading (moving bedding forward with front paws), and jumping (Figure 4B). A blinded observer scored the behaviors in 10-second intervals for the predominant behavior observed during each interval and each behavior was assigned a color for visualization. Color coded representation of trials from 16 Cre+ (n=16) and WT (n=16) mice is illustrated (Figure 4B). The behavioral pattern in the pre-stimulation period is similar between the Cre+ and WT mice. During the stimulation period, a clear shift in behavior was evident in Cre+ mice. Photostimulation induced rearing, treading, digging, rapid locomotion, and jumping. During the post-stimulation period we observed a new pattern with Cre+ mice spending time immobile and grooming. An example of a Cre+ animal is shown in Supplemental video 2.

Photostimulation of SuMVGLUT2+::POA neurons evokes active coping behaviors.

(A) Illustration of injections and fiber implant in VGLUT2+ Cre mice and schematic of photostimulation paradigm of 5 min pre, 5 min stimulation, and 5 mins post. Behaviors was evaluated during each epoch. (B) For Cre+ (n=16) and Cre- (n=16) mice the behavior in 10 seconds bins was scored based on the predominate behavior displayed during each 10 period for the 15-minute trial into grooming, stationary, walking, chewing, rearing, rapid locomotion, digging (moving bedding towards the tail), treading (move beading forward), and jumping. The graphic show a by animal scoring of the 15-minute trial color coded for each of the behavior categories. (C) During the stimulation period Cre+ (n=15) mice show significantly (p=0.0012) greater jumps than Cre- (n=16) mice and jumping behaviors was not significantly different during pre and post stim periods. (D) Behavior was also scored for time spent moving beading (digging or treading). Cre+ mice showed significantly (p=0.004) greater time engaging in digging/ treading behaviors during the stimulation period and significantly (p=0.002) less time during the post stimulation period compared to Cre- mice. (E) Behavior was also scored for time spent engaging in grooming behaviors and Cre+ mice spent significantly (p<0.001) less time grooming during the stimulation period and significantly (p<0.001) more during the post stimulation period compared to Cre- mice. All data plotted as mean ±SEM

We analyzed behavior quantifying specific behaviors during each period. Quantification of jumping revealed that Cre+ mice (n=16) engaged in significantly (p=0.0012) more jumps during the stimulation period compared to WT (n=16) mice (Figure 5C). During the pre- and post-stimulation periods there was not a significant difference in the number of jumps. We also observed that during the stimulation period the Cre+ mice engaged in bouts treading/digging, vigorously moving bedding forwards and backwards, as previously described for defensive burying (47). Defensive burying, characterized by moving bedding forward or backwards often in alternating pattern, is evoked by threatening and noxious stimuli, and is a described active stress coping strategy in rodents(47-50). We conservatively quantified together the movement of bedding as treading/digging that may include spontaneous digging (Figure 4D) and found no significant (p=0.7) difference during the pre-stimulation period. In the stimulation period there was a significant (p=0.004) increase in the time spent treading/digging, and surprisingly, during the post-stimulation period time treading/digging was significantly (p=0.0028) decreased in Cre+ mice. The variability in behaviors can be attributed to exclusion of one behavior by the other, with individual mice having variability in the predominant behavior displayed during the photostimulation period but all mice shifted to increased activity coping behaviors. The behaviors evoked by photostimulation fit with active coping behaviors. Specifically, escape (jumping, rapid locomotion), defensive burying/treading (digging, pushing beading forward).

Activation of SuMVGLUT2+::POA neurons drives aversion but does not promote anxiety-like behavior

(A) Schematic of injection of Retro-AAV-ChR2eYFP into the POA and placement of a midline optic fiber over SuM. (B) Representative heat maps for Cre+ and Cre- mice at baseline (no stim) and 10Hz photostimulation in two-sided arena with photostimulation associated with one side. (C) Quantification of time spent of stimulation side for Cre+ (n =19) and control Cre- (n=21) mice showing frequency dependent increase in avoidance of the stimulation side of the area (***p<0.001, ****p<0.0001). (D) Mean ±SEM entrances per min (in 1-min bins) to the during 10 Hz stimulation trials were not significantly (p=0.87) different between Cre+ (n=14) and Cre- (n=13) mice. (E) Mean time spent on the stimulation side for Cre- and Cre + in 1-minute bins during 10 Hz stimulation trials was significantly (p<0.0001) lower in Cre+ mice. (F) (Inset) Diagram of light/dark arena with photo stimulation provided on the dark side of the arena and quantification of time spent on dark (stimulation) side demonstrating that Cre- (n=12) animals show a baseline preference for the dark side of arena that is overcome by photostimulation with Cre+ (n=13) mice spending significantly (**p=0.002, ***p<0.001, ****p<0.0001) less time on the stimulation side. (G) Representative heat map of Cre- and Cre+ during 10Hz photostimulation in open field test. (H) In open field test time spent in center and perimeter were not significantly different (p=0.19) between Cre+ (n=18) and Cre- (n=17) during 10 Hz photostimulation. (I) Distance traveled during open field testing was significantly (p<0.001) increased in Cre+ compared to Cre- mice. (J) Schematic of real time light/dark choice testing with stimulation provided at 10Hz throughout the arena. (K) Both Cre+ (n=8) and Cre- (n=9) mice showed a preference for the dark portion of the arena but were not significantly (p=0.9) different. (L) The total distance traveled by Cre+ mice were significantly (p=0.003) greater than Cre- mice. All data plotted as mean ±SEM.

We also quantified grooming behaviors in each of the three periods (Figure 4E). During the pre-stimulation period there was not a significant difference. During the stimulation period the Cre+ mice did not engage in grooming leading to a significant (p<0.0001) decrease in time spent grooming compared to Cre- mice. In the post-stimulation-period Cre+ mice showed a significant (p<0.0001) increase in time spent grooming compared to Cre- mice. A reasonable interpretation of rise in grooming post stimulation is that photostimulation evoked a stressed like state and cessation of photostimulation led to selfcare grooming as seen following acute stressors (51, 52). In summary, the analysis of behaviors elicited by photostimulation of SuMVGLUT2+::POA neurons without conditioned cues or concomitant stressors demonstrates a dramatic shift in behavior to escape oriented (jumping, rapid locomotion) and threat response behavior, including rearing and defensive burying, during the photostimulation period. In contrast to freezing, the behaviors elicited by activation of SuMVGLUT2+::POA neurons indicate that they alter active coping strategies.

Photostimulation of SuMVGLUT2+::POA neurons drives real-time avoidance

To examine whether activation of SuMVGLUT2+::POA neurons may contribute to the aversive aspects of threatening stress we carried out real-time place aversion testing (RTPA) by pairing one side of the chamber with photostimulation of SuMVGLUT2+::POA (Figure 5A) neurons at multiple frequencies (1,5,10, and 20 Hz). Photostimulation of SuMVGLUT2+::POA neurons produced significant (p<0.0001) aversion at all frequencies in Cre+ (n=19) mice compared to Cre- (n=21) control and higher stimulation frequencies evoked greater aversion (Figure 5B and C). Example of Cre+ mice with stimulation at 1, 5, and 10Hz as well as Cre- mice is in Supplemental Video 3. WT and Cre+ mice explored equivalently, with similar (p=0.87 mixed-effects model) number of entries to the stim side (Figure 5D). Cre+ mice quickly left the simulation side and the mean time spent on the stimulation side of the area was significantly (p<0.0001) lower (Figure 5E). These data indicate surprisingly that photostimulation, although aversive, did not generate aversive pairing, as found for other brain areas (53). To examine the relative aversiveness of SuMVGLUT2+::POA photostimulation we carried out RTPA experiments using an arena with a dark side and a bright side (Figure 5F inset). We paired photostimulation with the dark side. As expected, Cre- mice show a preference for the dark side of the arena (Figure 5F), however photostimulation of the SuMVGLUT2+::POA neurons yielded significant (1 Hz p=0.002, 5 Hz p=0.0004, and 10, 20 Hz p<0.0001) aversion of Cre+ (n=13) compared to Cre- (n=12). Cre+ mice spent nearly the entire trial in the brightly light side demonstrating photostimulation of SuMVGLUT2+::POA neurons drove avoidance sufficient overcome a mildly aversive stimulus (bright light).

Reports have described multiple populations of neurons in supramammillary including neuronal populations that project to the dentate gyrus of the hippocampus that release both GABA and glutamate. Our anatomic studies indicated that SuMVGLUT2+::POA neurons do not project to the dentate gyrus (Figure 2) and are not GABAergic (Supplemental Figure 1G-L). Thus, SuMVGLUT2+::POA neurons are not part of either a GABAergic or dual transmitter population. To examine if GABAergic neurons in SuM (SuMVGAT+) can mediate real-time place aversion or preference, we injected VGAT-Cre mice with AAV to express ChR2 (n=26) or control (n=26) in SuM and carried out real-time place aversion testing. Photostimulation at 10 Hz yielded no significant aversion or preference compared to baseline (Supplemental Figure 6A-D). This is in contrast to the robust aversion caused by photostimulation of SuMVGLUT2+::POA neurons.

Photoactivation of SuMVGLUT2+::POA neurons does not cause anxiogenic-like behavior

Threat can induce anxiety-like, risk aversion behavioral states (54-56). To test whether activity of SuMVGLUT2+::POA neurons contributes to anxiety-like behaviors we used two established assays for anxiety-like behavior, open field, and light-dark exploration (57-60). During open field testing 10 Hz photostimulation was applied to SuMVGLUT2+::POA neurons. We quantified the time spent in the perimeter (outer 50%) vs. the center (inner 50%) of the arena and found VGLUT2-Cre+ (n=18) and WT (n=17) mice did not significantly (p=0.17, and 0.19) differ (Figure 5G and H). The total distance traveled was significantly (p<0.001) increased in in Cre+ mice compared to WT mice (Figure 5I). Increased distance traveled in some mice was not surprising, because prior testing (Figure 4) revealed the mice engaged in escape behaviors during photostimulation. We observe behaviors including jumping during open field testing as well. We also carried out open field testing using VGAT-Cre mice to selectively activate SuMVGAT+ neurons and observed no significant (p=.99) effect on time in center or perimeter (Supplemental Figure 6F), but mice expressing ChR2eYFP displayed a small but significant (p=0.035) decrease in distance traveled (Supplemental Figure 6G). We tested if photostimulation of SuMVGLUT2+::POA neurons would increase preference for the dark area, a potential sign of elevated anxiety (Figures 5J). We found photostimulation of SuMVGLUT2+::POA neurons did not significantly(p=0.9) alter preference (Figure 5K) in VGLUT2-Cre+ mice (n=8) compared to WT (n=9). The activation of SuMVGLUT2+::POA neurons evoked escape behaviors in open field testing, as seen in Figure 4, leading to a significant (p=0.003) increase in total distance moved (Figure 6L). The findings we obtained in open field and light-dark exploration testing do not support a role for SuMVGLUT2+::POA neurons in driving anxiety-like behaviors.

SuMVGLUT2+::POA neurons drive can drive instrumental actionoutcome operant behavior

(A) Schematic of injection and implant in VGLUT2+ Cre mice and paradigm of testing in 10-minute trials with 4 days before a progressive ratio (PR) trial on day 5. (B) Illustration of the testing paradigm and set up. 10 Hz photostimulation was applied during the trials. Activation of the active port trigger the house light and paused stimulation for 10 seconds. Also shown is the progressive ratio used with number of required port activations per reward on vertical axis and reward number on the horizontal. (C) Cre+ mice (n=11) activated the active port triggering significantly (****p<0.0001) more pauses in stimulation than Cremice (n=11) mice on all 4 days of testing. (D) Cre+ mice activated the port significantly (****p<0.0001, ***p<0.001) than the Cre- mice on all 4 trials. Cre+ activated the active port significantly (p<0.001) on during the fourth trial than the first. (E) The time to first activation of the active port was not significantly (p=0.32) different during the first trial but Cre+ mice became significantly (**p<0.01, ***p<0.001) on subsequent trials. (F) On the fifth day after four 10-minute trials mice were tested for 30 minutes using a progressive ratio. Cre+ performed significantly (p<0.0001) more active port activations that Cre- mice and activated the inactive port significantly more (p<0.005) times. Cre+ mice also triggered significantly (p<0.001) more pauses in photostimulation that Cre- mice. (G) Individual data for a representative cohort of Cre+ mice showing cumulative active port activations by Cre+ mice (n=7) as function of time during the progressive ratio test are shown illustrated ongoing engagement of the active port throughout the 30-minute trial. (H) The cumulative pauses in photostimulation (rewards) earned as a function of time during the progressive ratio trial are show for individual Cre+ mice. Mice earned between 7 and 11 pause reward during the trial. All data plotted as means ±SEM.

SuMVGLUT2+::POA neurons can drive instrumental action-outcome behavior

Current theoretical frameworks for examining threat responses divide behaviors into two broad categories, innate (fixed) and instrumental (action-outcome)(1, 61). Additionally, stress responses can be divided into active vs passive actions. Behaviors in these separable categories are mediated by distinct neural circuits(1, 62). For example, areas involved in responding to specific threatening stimuli (looming threat) drive an innate fixed behavior repertoire(10). Photostimulation of SuMVGLUT2+::POA neurons evoked multiple active behaviors displayed when responded to threatening or noxious stimuli (Figure 4). To test if SuMVGLUT2+::POA neurons can promote flexible repertoires of behavior including goal-directed, as opposed to only innate behaviors (ex. fleeing), we used an operant negative reinforcement paradigm (Figure 6A-B). Because our results show photostimulation of SuMVGLUT2+::POA neurons is aversive photostimulation of SuMVGLUT2+::POA neurons could be used as a negative reinforcer. We hypothesized if stimulation of SuMVGLUT2+::POA evokes fixed innate defensive behaviors, or simply promotes high amounts of locomotion, then performance of an operant task (an action-outcome behavior) would be impaired because these behaviors would conflict with performance of the operant task. Importantly, results obtained can help differentiate general elevation in locomotor activity from directed coping behavior. We employed a paradigm using two nose poke ports, one active and one inactive. SuMVGLUT2+::POA neurons were photostimulated at 10 Hz during a 10-minute trial. Activation of the active port by a nose poke triggered a 10-second pause in photostimulation and turned on a house light for 10 sec (Figure 6A and B). Following this 10-sec pause the house light switched off and photostimulation resumed until the subsequent activation of the active port. We tested animals for 4 consecutive days with the 10-min trials without prior training. Importantly, performance of the task had to be completed during photostimulation.

Analyses of port activations during the 10-min trials revealed that Cre+ mice (n =11) paused the photostimulation (reward) significantly (p<0.0001) more on all four days of testing compared to Cre- (n=11) mice (Figure 6C). The number of 10-second pauses approached the maximum, 60 possible in a 10-min period. Cre+ mice engaged the active port significantly (p<0.001) more than Cre- mice (Figure 6D). Indicating no difference in baseline performance of the task, the time to the first engagement of the active port was not significantly (p=0.32) different for Cre+ compared to Cre- mice during the first test trial. During the three subsequent tests Cre+ mice were significantly (p<0.01) faster to engage the active port after the start of the trial compared to Cre- mice. Cre+ mice also were significantly (p<0.01 paired t-test) faster in trials two through four at engaging the active port compared the first trial. Interestingly, Cre+ mice engaged the active port significantly (p=0.009 paired t-test) fewer times on the fourth day than the first day but achieved more pauses (Figure 6E).

To examine effort-related motivation we employed a 30-min test using a progressive ratio requiring exponentially increasing number of active port activations to trigger a pause in the photostimulation on the fifth day of operant behavior testing (Figure 6B) (63). During trials using a progressive ratio, Cre+ mice activated the active port and paused stimulation significantly (p<0.001) more times than Cre- mice (Figure 6F). Examination of individual cumulative active port activation data showed Cre+ mice continue to engage the active port throughout progressive ratio trials despite the increasing work required to generate each pause in photostimulation. No breakpoint was observed during the trial (Figure 6G and H). This finding indicates that activation of SuMVGLUT2+::POA neurons remains salient and motivating through the tested period and that photostimulation of does not evoke behaviors precluding completion of the task. Examination of reward behavioral epochs for representative animals shows that for lowest total number of rewards earned (green dots) the next pause required 20 activations of the active port and for the highest (red dots) the next pause required 50 activations of the active port. Taken together, results from instrumental reinforcement tasks demonstrate that activation of SuMVGLUT2+::POA neurons does not solely evoke innate stereotyped behaviors and can drive active coping in the form of instrumental behaviors.

SuMVGLUT2+::POA neurons are recruited during active coping behaviors in forced swim testing

The forced swim test is increasingly understood as an assay of coping strategy with passive and active components (64, 65). During a first trial of forced swimming mice shifted from active escape behaviors (wall climbing, robust swimming) to in passive (immobile floating) behaviors. We found that SuMVGLUT2+::POA neurons were recruited in a short dunk assay (Figure 3F and G), which elicited active escape behaviors (Supplemental Figure 4A). To test if recruitment of SuMVGLUT2+::POA neurons changes with the shift from active to passive behaviors during a longer swim assay we carried out experiments using fiber photometry paired with automated computer-based behavior analysis(66). The automated behavior scoring enabled high temporal resolution analysis of behaviors (Supplemental Video 4) and each frame of the video was coded for the classified behavior and plotted as color coded point. We observed 4 types of behavior during forced swimming: climbing, swimming, hindpaw swimming, and immobile floating and trained a model to identify each. Analysis of behavior showed that after the first minutes of the trial mice shifted from swimming and climbing to immobile floating with intermittent swinging or hindpaw swimming. We examined if recruitment of SuMVGLUT2+::POA neurons changed in correlation with this shift to more passive behavior. As shown in representative trace (Figure 7A) the Ca2+-dependent signal decreased the shift in behavior was occurring. To combine behavioral data we assigned numerical values, one through four, to each behavior: climbing (4), swimming (3), hindpaw swimming (2), and immobile floating (1). The mean was plotted as a heat map (red higher values to blue lower values) to be compared to time locked fiber photometry data (Figure 7B). Examining the timeframe around the behavioral shift in each animal we found that as behavior shifted from active coping (climbing, swimming) to more passive strategies and greater immobility the Ca2+ signal substantially declined (Figure 7B). Examining data after this initial transition we observed that many elevations in the Z-score were accompanied by changes in behavior to swimming or hindpaw swimming (Figure 7A). To further examine this possible correlation, we examined the time frame around transition to hindpaw swimming for changes in Ca2+ signal. In a 20-sec window centered on the onset of hindpaw swimming we hind swimming was accompanied by a rise in the Ca2+ dependent signal across multiple events within a trial (Figure 7C). A similar analysis across events in multiple animals demonstrated a similar rise in the Ca2+ dependent signal from SuMVGLUT2+::POA neurons during the shift from immobility to hindpaw swimming or swimming behaviors (Figure 7D). Time series analysis of 20-sec epochs centered on random time intervals did not reveal any rise in the Ca2+ dependent signal (Figure 7E). These data indicate that recruitment of SuMVGLUT2+::POA neurons fluctuates with changes in coping strategy during a forced swim assay with decreased engagement of this circuit during times of immobility.

SuMVGLUT2+::POA neurons are recruited during active coping and activation is sufficient drive a switch to active coping.

(A) Color coded per video frame (dot color) behavior combined with normalized Ca2+ dependent and isosbestic fiber photometry signals. Scoring and analysis of climbing, swimming, hindpaw swimming, and immobility was completed using deep learning based classification and quantification (LabGym) of recorded behavior. Inset shows injection and implants and behavioral assay. Box highlights the time frame shown in B. (B) Heat maps for recordings obtained mice (n=9) during forced swim, averaged color-coded behaviors, and mean ±SEM Z-score of Ca2+ dependent signal for 2-min period around behavioral shift from climbing and swimming to hindpaw swimming and immobility. (C) Representative heat map, averaged behavior, and mean ±SEM Z-score of Ca2+ dependent signal for 20-second window around onset of hindpaw swim from a representative animal. (D) Representative heat map and mean ±SEM Z-score of Ca2+ dependent signal for 20-second window around onset of hindpaw swim from for events from nine mice. Engaging in hindpaw swim correlates with brief increase in Ca2+ signal. (E) Representative heat map and mean ±SEM Z-score of Ca2+ dependent signal for 20-second window around random time points. (F) Schematic of injection and implant in VGLUT2-Cre or VGAT-Cre mice. (G) Illustration of testing paradigm on second of forced swim test following 15-minute swim on first day. (H) The average time spent immobile during the pre-stimulation period was not significantly (p=0.48) different. During 10 Hz photostimulation VGLUT2-Cre+ mice engaged in vigorous swimming and the time spent immobile was significantly (***=p<0.001) less than Cre- mice. In the post-stimulation period, the time spent immobile remained significantly (*=0.016) lower in Cre+ mice compared to Cre-. (I) In VGAT-Cre+ mice expressing ChR2 (n=10), the average time spent immobile during the pre-stimulation period was not significantly (p=0.08) different compered to eYFP controls. Photostimulation did not significantly increase time immobile (p=0.64). However, there was a significant (*** p=0.0004) decrease in the time spent immobile compared in the post-stimulation period.

SuMVGLUT2+::POA neurons promote active coping during forced swim test

We found that SuMVGLUT2+::POA neurons are activated by acute stressors (Figure 3), are recruited during times of greater active coping behaviors (Figure 7). To test if SuMVGLUT2+::POA neurons can drive a switch in coping strategy in the context of ongoing stressor we used a two-day forced swim stress test (65, 67). We tested animals during the second day, when immobile floating is the predominant behavior. As in previous experiments, we used VGLUT2-Cre (Cre+) or WT (Cre-) mice injected in the POA bilaterally with Retro-AVV-DIO-ChR2eYFP. On day one we subjected mice to 15 mins of forced swim. On the subsequent day we repeated the forced swim for six mins divided into three periods, pre-stimulation, stimulation at 10Hz, and post-stimulation (Figure 7F). Trials on the second day were recorded and scored by a blinded observer for time spent immobile in each 2-min period. Example of Cre+ mouse with stimulation 10Hz is shown in Supplemental Video 3. During the pre-stimulation period Cre+ (n=9) and Cre- (n=11) mice there was not a significantly (p=0.45) different amount of time immobile. During the stimulation period Cre+ mice began swimming and attempting to climb the wall of the circular swim arena, reflected by significantly (p<0.0001) less time immobile (Figure 7H). Interestingly, in the post stimulation phase the difference in time spent immobile between Cre+ can WT mice decreased but remained significantly (p=0.016) different. These results indicate that in the context of an ongoing stressor activation of SuMVGLUT2+::POA neurons is sufficient to trigger a change in coping strategy from a largely passive (floating) to and active (swimming, climbing). The persistent effect of the acute activation into the poststimulation phase suggests that activation of SuMVGLUT2+::POA neurons may shift how the stressor is processed or approached.

We next examined the effect of 10 Hz photostimulation of SuMVGAT neurons on coping strategy using the same two-day swim paradigm. VGAT-Cre mice were injected in the SuM to express ChR2eYFP (n=10), or control (eYFP) (n=11) and a fiber optic was placed over SuM (Figure 7F). Quantification of behavior revealed no significant difference in time spent immobile during the pre-stimulation or the stimulation periods. Surprisingly, in the post-stimulation period we observed a dramatic shift in behavior marked by significant (p=0.0004) decrease in time spent immobile (Figure 7I). The amplitude of change in behavior was similar as we observed during the stimulation phase of the experiments on SuMVGLUT2+::POA neurons (Figure 7H). One interpretation of these data is that release of sustained local inhibition leads to rebound activity of output SuMVGLUT2+::POA neurons.

Suppression of SuMVGLUT2+::POA neurons is required for feeding

Feeding and responding to threats are conflicting actions and SuMVGLUT2+::POA neurons may play a role in switching between behavioral paradigms (e.g., feeding vs escape). We have found threatening stressors recruit SuMVGLUT2+::POA neurons. Responding to an acute threatening stressor would prevent feeding. We sought to examine SuMVGLUT2+::POA neuron activity in relation to feeding using fiber photometry. To promote varying drive for food mice were given ad lib access to food (fed condition) or food deprivation for 24 hours and on testing day presented with a chow pellet while Ca2+ dependent and isosbestic signals were recorded. In the fed state mice spent significantly (p<0.001) less time interacting with the food (Figure 8C) and ate significantly (p<0.001) less (Figure 8D) compared to food deprived state. We analyzed the Ca2+ dependent and isosbestic signals around presentation of food to fed and food deprived mice. In the food deprived state mice spent more time with the food and ate more food (Figure 8E and F). At the time of food presentation, the Ca2+ dependent signal decreased but not the isosbestic signal. The change in the Ca2+ dependent was larger in the food deprived state (Figure 8E and F). Together this data supports suppression of SuMVGLUT2+::POA neural activity during consummatory behavior with greater consumption associated with greater suppression.

Consummatory behavior suppresses SuMVGLUT2+::POA neuron activity and SuMVGLUT2+::POA neuron blocks food consumption.

(A) Schematic of injections in VGLUT2-Flp mice with AAV-DIO-GCaMP7s in SuM and Retro-AAV-fDIO-Cre in the POA and a fiber placed over SuM. (B) Mice were given ad lib food access or food deprived for 24 hours prior to testing. After being placed in arena and allowed to habituate mice were given a 20-minute trial divided into 5 mins baseline and 15 mins after a chow pellet was added to the arena. (C) Food deprived mice (n=9) spent significantly (p< 0.001) more time interacting with the food pellet and (D) ate significantly (p<0.001) more than in the fed state. (E) Heat map and mean ±SEM z-scores for Ca2+ dependent and isosbestic signals for recordings obtained from 9 animals show a small drop in (E) fed mice state compered to a sharp decrease in Ca2+ activity after the introduction of the food pellet following addition in (F) food deprivation. (G) Mice were fasted for 24 hours prior to testing and the schematize 20 min paradigm was used. Following food deprivation animals were given access a chow pellet. The trial was divided in to a 5-minute pre- and post-stimulation periods with 10-minute period of stimulation. (H) The time spent interacting with the food was quantified for each period. Cre+ (n=16) and Cre- (n=17) mice both rapidly engaged with the food pellet and the average time spent interacting was not significantly (p=0.37) different. During the stimulation period and the post stimulation period the average time interacting with the food was significantly (p=0.0002, p=0.0001) lower in Cre+ mice compared to Cre- mice. (I) The food eaten during the total trial was calculated based on pellet weights and on average Cre+ mice ate significantly (p=0.001) less food during the trial than Cre- mice. (**p=0.001, ***p<0.001, ****p<0.0001)

To examine whether activation of SuMVGLUT2+::POA neurons disrupted consummatory behavior we tested the impact of photostimulation on feeding behavior in food deprived mice. Using VGLUT2-Cre (Cre+) or WT (Cre-) mice injected with Retro-AAV5-DIO-ChR2eYFP into the POA and implanted with a fiber optic midline over SuM we examined how feeding behaviors in food deprived mice were altered by 10 Hz photostimulation of SuMVGLUT2+::POA neurons (Figure 8G- and D). We used trials lasting 20 mins with unrestricted access to food added at the start of the trial. Each trial was divided into three periods, a 5-min pre-simulation period, 10-min stimulation period, and a 5-min post-stimulation period. Trials were recorded and scored for time spent interacting with the food pellet. During the pre-stimulation period Cre+ (n=17) and Cre- (n=16) mice spent interacting with the food was not significantly (p=0.37) different. With the start of photostimulation the Cre+ mice stopped eating and interacting with the food. During the stimulation period Cre+ mice spent significantly (p=0.0002) less time interacting with the food compared to Cre- mice. Surprisingly, the decrease in time spent interacting with food continued into the post-stimulation phase when Cre+ mice interacted with the food significantly (p<0.0001) less (Figure 8E). Reflecting the decrease in time spent with the food, Cre+ mice consumed significantly (p=0.001) less food during the 20-min trial than Cre-mice. These results indicate that SuMVGLUT2+::POA neurons can redirect behavior away from consumption and suppression of SuMVGLUT2+::POA neurons is required for feeding even in a food deprived state. Ethologically, animals must choose rapidly between responding to threats and feeding when foraging. The findings here implicate this pathway in control of switching between these behaviors.

Discussion

The presented data show SuMVGLUT2+::POA population of neurons have broad arborizations to brain regions involved in responding to threat and stress (Figures 1, 2, and Supplemental Figure 2), are recruited by threatening stressors (Figures 3 and 7), can contribute to multiple aspects of behavioral responses to threatening stressors, and promote active coping behaviors (Figures 6 and 7). We also found that SuMVGLUT2+::POA can drive aversion and although aversive stimuli can promote anxiety-like behaviors, photostimulation of SuMVGLUT2+::POA stimulation did not (Figure 5). This finding suggests SuMVGLUT2+::POA neurons may play a role in suppressing anxiety and passive fear behaviors to promote active engagement. Finally, examining the conflicting behaviors required for threat response and feeding we SuMVGLUT2+::POA were suppressed during feeding and that activation of SuMVGLUT2+::POA neurons blocked feeding activity both during stimulation in the period after stimulation (Figure 8). SuMVGLUT2+::POA neurons are distinct subset neurons projecting to many but not all areas receiving projections from the total SuMVGLUT2+ population. Notably, mHb and DG projections were present in the SuMVGLUT2+ population but absent in the SuMVGLUT2+::POA projections (Figure 3K), consistent with a separation of DG and CA2 projecting SuM neurons (68). The extensive arborization of SuMVGLUT2+::POA neurons challenges the often-employed neural circuit concept that projection from a single area to a second to mediate or modify a behavior. The projections to PVT, Hb, and PAG are of particular interest because these areas regulate stress coping (69-72). Dopamine receptor 2 expressing neurons in the PVT have been implicated in promoting active coping behaviors and may be a potential target of SuMVGLUT2+::POA neurons (8, 9). The habenula also regulates active vs passive coping behaviors, and the lHb functions in aversive processing making the selective projections of SuMVGLUT2+::POA neurons to the lHb of similar interest (69, 73-76). In addition, the PAG, which receives projections from SuMVGLUT2+::POA neurons, has been implicated in regulating defensive behaviors and coping strategies (70, 71, 77). The arborization to these and other areas position SuMVGLUT2+::POA neurons to be a central hub for regulating behavioral responses to threatening stressors by concurrent recruitment of multiple areas.

Superficially, the aversiveness of SuMVGLUT2+::POA neuron stimulation we report contrasts with a recent study that found photostimulation of SuM neurons projections could be reinforcing. Photostimulation of projections from SuM to the septum drove place preference (septum) but, the same report found photostimulation of SuM projections in the PVT was aversive(19). We report here that SuMVGLUT2+::POA neurons also project to the PVT and the septum (Figure 2). A plausible interpretation is that SuM contains separable molecularly defined populations of neurons which gate a host of distinct behavioral strategies through their connectivity and or co-transmitter content. Delineating the overarching circuitry will require further studies, but our results together with recent studies establish SuM is an important and poorly understood node for regulating appetitive and aversive motivated behaviors.

We addressed the challenge of potential confounds from locomotor activity inherent in active coping behaviors, which is of special concern here because a population of SuM neurons have been found to correlate with velocity of movement, by using multiple behavioral assays with fiber photometry and photostimulation. We found multiple threatening stressors (dunk, shock, ambush) with differing evoked behaviors all recruit SuMVGLUT2+::POA neurons (Figure 3). We also found increased recruitment of SuMVGLUT2+::POA neurons during active coping behaviors (Figures 3 and 7). We did not find evidence for recruitment of SuMVGLUT2+::POA neurons during elevated locomotion in two cohorts in assays of free movement (Figure 3K-M and Supplemental Figure 5). These data are in contrast to recent studies that found activity of some neurons in SuM correlated future velocity (41). We carefully examined the behaviors elicited by activation of SuMVGLUT2+::POA neurons both in absence of stressors (Figure 4) and during ongoing stressors (Figure 7) and found stimulation of SuMVGLUT2+::POA neurons promoted active coping behaviors. Because escape behaviors could be innate fixed behaviors we examined if SuMVGLUT2+::POA neurons could also drive performance of an operant task, which would be impaired if activation of SuMVGLUT2+::POA neurons lead to fixed escape behaviors. We found mice effectively performed the operant task. In trials using a progressive ratio mice performed the task repetitively during sustained photostimulation because the number of port activations required to terminate the stimulation was high. Taken together the results from the behavioral tests indicate that SuMVGLUT2+::POA promote flexible stressor appropriate coping behaviors and not generalized locomotion.

Although neurons that release both GABA and glutamate projecting to dentate gyrus are present in SuM, our anatomic studies indicate and SuMVGLUT2+::POA neurons were discrete from GABAergic neurons do not project the dentate gyrus (Figure 2 and Supplemental Figure2)(21, 40). SuMVGAT could be a functionally distinct population. We tested the effects of photostimulation of SuMVGAT neuron and in contrast to SuMVGLUT2+::POA neurons, found no effect on place aversion, minor decreases in locomotion, and no significant change in active coping behaviors during stress (FST) during stimulation. Surprisingly, immediately after photostimulation stopped active coping behavior increased dramatically (Figure 7I). These data demonstrate functional separation of SuMVGAT and suggest connection between the SuMVGLUT2+::POA and SUMVGAT+ populations.

The aversiveness of SuMVGLUT2+::POA neuron stimulation we report contrasts with a recent study that found photostimulation of SuM neurons projections could be reinforcing. However, mor closer examination suggests the potential underlying diversity of SuM neurons may explain the results. Photostimulation of projections from SuM to the septum drove place preference but photostimulation of SuM projections in the PVT was aversive(19). We report here that SuMVGLUT2+::POA neurons also project to the PVT and the septum (Figure 2). A plausible interpretation is that SuM contains separable populations. One set (PVT projecting) could drive aversion and the other (non-PVT projecting) may regulate reinforcement. Delineating the overarching circuitry will require further studies. Our results together with recent studies establish SuM is an important and poorly understood node for regulating appetitive and aversive motivated behaviors.

In conclusion, SuMVGLUT2+::POA neurons arborize to multiple areas involved in stress and threat response that promote active coping behaviors. Passive coping strategies have been associated with unescapable stress, anhedonia, and depression(7, 78-80). Targeting neuromodulation of a circuit able to act across many brain areas may represent a therapeutic avenue for common psychiatric conditions and SuMVGLUT2+::POA neurons are a newly identified node in critical approach-avoidance circuitry.

Acknowledgements

This work was supported by the NIH through 5R01MH112355 to MRB and P30DA048736 (MRB) and 5K08MH119538 to AJN. Support was also provided by the Hope Center Viral Vectors Core and the Genome Technology Resource Center at Washington University in St. Louis (NIH P30CA91842 and UL1TR000448). This work was supported by a Pilot Project Award from the Hope Center for Neurological Disorders and by the Hope Center Viral Vectors Core at Washington University School of Medicine.

Materials and Methods

Methods

Key Resources Table

For further information regarding reagents and resources, contact Aaron Norris, norrisa@wustl.edu.

Experimental model and subject details

Adult (25-35 g, at least 8 weeks of age upon experimental use) male and female VGLUT-Cre (RRID: IMSR_JAX:016963), VGLUT2-Flp (RRID: IMSR_JAX: 030212), Ai14 (RRID: IMSR_JAX: 007908), VGAT-Cre (RRID: IMSR_ JAX: 028862) and C57BL/6J (RRID: IMSR_JAX: 000664) mice (species Mus musculus) were group housed (no more than five littermates per cage) in a 12 hr:12 hr light:dark cycle room with food and water ad libitum(81). The Washington University Animal Care and Use Committee approved all procedures which adhered to NIH guidelines.

Stereotaxic surgery

Injections and implantations were done as described previously(34, 82). Briefly, in an induction chamber, mice were anesthetized (4% isoflurane) before being placed in a stereotaxic frame (Kopf Instruments). Anesthesia was maintained with 2% isoflurane. Mice were then injected unilaterally or bilaterally, depending on the combination of virus(es) used and brain regions injected. A blunt needle Neuros Syringe (65457-01, Hamilton Con.) and syringe pump (World Precision Instruments) were used to perform the injection schemes below. After surgery, a warmed recovery chamber housed the animal while it recovered from anesthesia before being returned to its home cage.

Injections were made at a rate of 50 nl/min, with the injection needle being withdrawn 5 min after the end of the infusion. Fiber optics for photostimulation or optical fibers for fiber photometry were implanted after injections for all behavioral experiments. Fiber optics for photostimulation or optical fibers for fiber photometry were implanted after injections for all behavioral experiments.

Fiber optic implants for photostimulation were fabricated as previously described using 200μm glass fibers and implanted midline over SuM (34, 83). For implantation, the skull was cleaned and etched with OptiBond® (Kerr) and the fiber was cemented to the skull with Tetric N-Flow® (Ivoclar Vivadent). Blue light was used to cure and harden cement. Mice were allowed to recover for at least seven days before the start of behavioral experiments. The same process was used for implantation for fiber photometry fibers, which were trimmed to length (NeuroPhotometrics).

Anatomical tracing

For anterograde viral tracing experiments, virus (was injected at least six weeks prior to transcardial perfusions with 4% paraformaldehyde to allow for anterograde transport of the fluorophore. AAV5-EF1a-DIO-hChR2eYFP or AAV5-EF1a-DIO-eYFP were used. Alternatively, to label only SuMVGLUT2+::POA neurons for anterograde tracing Retro-AAV2-DIO-eYFP or Retro-DIO-ChR21eYFP was injected into the POA with AAV-fDIO-Cre injected in the SuM. A minimum of 6 weeks was allowed prior to sacrifice, harvesting or brains, and sectioning (30μM). Serial 30 μM sections approximately 60μM apart were examined. For retrograde studies viruses were injected (see figure legends and text for specific viruses) at the targeted site and table for specific viruses(84-86). Three weeks were allowed to elapse prior to harvesting brains following injections. Images were obtained on a Leica DM6 B upright microscope and processed using Thunder imaging station (Leica).

Brain clearing and light sheet microscopy

Tissue clearing and imaging was carried out on brains collected and fixed in 4% PFA as described above by LifeCanvas Technologies. Briefly, brains were fixed using SHIELD post ix and cleared for 7 days in SmartClear II Pro. Index matched with EASYIndex at room temperature. Samples were mounted ventral side up and imaged at 3.6x with pixel size 1.8 x 1.8 mm, axial resolution < 4 μm, z step 4 μm in 488nm channel. Fos was labeled by Alexa Flour 488. SuM boundaries were defined by -2.6 to -2.95 rostral to Bregma. The medial mammillary nucleus and the mammillary recess of the 3rd ventricle marked the medial and ventral boundaries while fornix marked the lateral, and fasciculus retroflexus the dorsal boundaries. Images were quantified by a trained laboratory member who was blind to the experimental conditions.

Immunohistochemistry

Mice were intracardially perfused with 4% PFA and then brains were sectioned (30 microns) and placed in 1x PB until immunostaining. Free-floating sections were washed in 3 times in 1x PBS for 10 min intervals. Sections were then placed in blocking buffer (0.5% Triton X-100% and 5% natural goat serum in 1z PBS) for 1 hr at room temperature. After blocking buffer, sections were placed in primary antibody rabbit Phospho-c-Fos (Ser32) antibody (1:500, Cell Signaling Technology) overnight 4° C temperature. After 3 x 10 min 1x PBS washes, sections were incubated in secondary antibody goat anti-rabbit Alexa Fluor 488 (1:2500, Invitrogen) for 2 hr at room temperature. Sections were washed in 1x PBS (3 x 10 min) followed by 2 x 10 min 1x PB washes. After immunostaining, sections were mounted on Super Frost Plus slides (Fisher) and covered with Vectashield Hard set mounting medium with DAPI (RRID:AB_2336788, Vector Laboratories) and cover glass prior to being imaged.

Imaging and cell quantification

‘The Mouse Brain in Stereotaxic Coordinates’ provided the framework to label brain sections relative to bregma(87). A Leica DM6 B epifluorescent microscope was used to image all sections. For eYFP visualization, a YFP filter cube (Excitation: 490-510, Dichroic: 515, Emission: 520-550) was used and for tdTomato visualization, a Texas Red Filter Cube (Excitation: BP 560/40, Dichroic: LP 585, Emission: BP 630/75) was used. Fos was labeled by Alexa Flour 488. SuM boundaries were defined by -2.6 to -2.95 rostral to Bregma. The medial mammillary nucleus and the mammillary recess of the 3rd ventricle marked the medial and ventral boundaries while fornix marked the lateral, and fasciculus retroflexus the dorsal boundaries. Images were quantified by a trained laboratory member who was blind to the experimental conditions.

Forced-Swim Test

For stress induction via forced swim (Figure 4)(65, 88), mice were individually placed in a cylindrical container (18 cm in diameter) filled with water at 25 +/- 1 °C for 15 min. Prior to stress force swimming, mice were habituated to the arenas for three days prior to the beginning of FST to minimize stress. Control mice were brought to the behavioral testing area but remained in the home cage until sacrifice. Water was replaced after every animal.

Fiber Photometry

For all fiber photometry experiments the same strategy to selectively label SuMVGLUT2+::POA neurons was used. VGLUT2-Flp mice were injected bilaterally in POA with AAV-Retro-EF1a-fDIO-Cre and with AAV-syn-FLEX-GCaMP7s-WPRE in SuM. After two weeks, mice were implanted with fiber-optic cannulas (200μm) in SuM (D/V -4.3-5)(89). Mice recovered a minimum of one week prior to behavioral testing. Recording of Ca2+ dependent and isosbestic signals were obtained using previous described methods with Bonsai software and FP3002 (Neurophotometrics)(63, 90). 470 and 415 nm LEDs were used to record interleaved isosbestic and Ca2+ dependent signals following the manufacturer directions.

For repeated forced swim experiments (dunk tank) mice were placed inside a custom-made plastic rectangular enclosure (20 cm x 20cm x 23 cm) filled with water 30 +/- 1 °C. The enclosure has a square wire mesh platform that could be raised and lowered without touching the animal. The platform was lowered for the swim session and raised at the conclusion of each session. During testing mice were tethered to a fiberoptic patch cabled that was attached to a counter balanced arm that prevented downward force on the animal. Time mobile was quantified by a trained observer for the ten 30s trials.

For foot shock stress testing, mice were individually placed inside a custom-made clear plastic box (15.24 L x 13.34 W x 14 H cm) inside a sound-attenuated cabinet. A speaker was placed 4 cm above the chamber for the delivery of auditory cues (75dB). A constant current aversive stimulator (ENV-414S) delivered foot shocks through a grid floor (0.7mA). Five shocks of 2 sec were delivered after a 30 s tone. Intertrial interval ranged from 90 to 180 sec. After the session, the animal was removed from the chamber and placed back in its home cage. The chamber and grid were wiped down with 70% ethanol between animals.

For open field testing for velocity measurements, mice were tethered to a fiberoptic patch cable and placed inside a custom-built square arena (50.8 cm x 50.8 cm x 50.8 cm). Mice were allowed to explore the arena for a 20 min session. Velocity was quantified using scripts in Bonsai 2.4. Bedding in the arena was replaced between animals, and the floors and walls of the arena were wiped down with 70% ethanol.

Similar to the repeated forced swim experiments, for forced swim mice were placed on top of a square wire mesh platform that could be raised and lowered inside a custom-made plastic rectangular enclosure (20 cm x 20 cm x 23 cm) filled with water at 25 +/- 1 °C. The platform was lowered at the beginning of the test for 15 min and raised at the conclusion of the test. During testing, mice were tethered to a fiberoptic patch cable that was attached to a counter balanced arm that prevented downward force on the animal. Water was replaced between animals and the enclosure was wiped down with 70% ethanol and rinsed with water. Time immobile, hindpaw swim, swimming, and climbing were quantified using LabGym (Hu et al., 2022).

To test the response to a natural stressor that mice can potentially encounter in their natural habitat(91). in SuMVGLUT2+::POA neurons, a remote-controlled spider was used to simulate an ambush of a potential predator. The remote-controlled spider (17 cm x 16 cm; Amazon) was placed inside a polylactic acid (PLA) enclosure (19 cm x 20 cm x 23 cm) that was then placed inside a custom-built square arena (50.8 cm x 50.8 cm x 50.8 cm). Mice were tethered to a fiberoptic patch cable and placed inside the arena for 10 min. Baseline activity was recorded for 5 min before the ambush. After 5 min, the animal was ambushed once it moved in close proximity to the spider’s enclosure (Supplemental video 4). Velocity was measured and analyzed using DeepLabCut (Mathis et al., 2019) within Bonsai. The arena and spider’s enclosure were wiped down with 70% ethanol between animals.

For experiments that examined recruitment of SuMVGLUT2+::POA neurons during consummatory behavior, mice were food deprived or given adlib access to food for 24 hours prior to testing. On test day, mice were placed inside an 18 cm diameter round arena and allowed to habituate for 30 min before they were tethered to the fiberoptic path cable. Once tethered, mice were placed inside the arena for 20 min. Baseline activity was recorded for 5 min before the introduction of the chow pellet and for 15 min after the chow pellet was introduced. Chow pellets were weighed before and after the 20-min trial. The difference is reported as food eaten. The same procedure was followed for control mice with the only difference being that these animals were not food deprived for 24 hours.

For fiber photometry data analysis, the interleaved isosbestic and Ca2+ dependent signals were recorded at 60 fps (30 fps each). Deinterleaved signals were analyzed using methods as previously reported (90). Briefly, raw signals were smoothed using a moving average, fitted with an exponential curve using a non-linear least squares function for baseline correction, signals were standardized using the mean value and standard deviation (Z-Score), the standardized isosbestic and Ca2+ signals were scaled a non-negative robust linear regression, and normalized dF/F was calculated. In experiments shown in figures 7A and 8 E-F there was sustained step drop evident in the Ca2+ dependent signal reflecting change in population activity because it was not seen in the isosbestic signal. The nature and duration of the change of the signal precluded fitting a curve to the Ca2+ dependent signal. In these cases, we show both the Ca2+ dependent and isosbestic signals. Z-scores were calculated without baseline correction for both Ca2+ dependent and isosbestic signals based on the variability in the baseline state.

Real-time place aversion testing

For real-time place preference testing with optogenetic photostimulation we used custom-made, unbiased, balanced two-compartment conditioning apparatus (52.5 x 25.5 x 25.5 cm) as described previously(34, 82, 92). Mice were tethered to a patch cable that allowed free access to the entire arena for 30 min. Entry into one compartment triggered photostimulation, 1Hz, 5Hz, 10Hz, or 20 Hz (473 nm laser, 10 ms pulse width) that persisted while the mouse remained in the light paired side. The side paired with photostimulation was counterbalanced across mice. Ordering was counterbalanced with respect to stimulation frequency and placement. Bedding in the behavior apparatus was replaced between every trial, and the floors and walls of the apparatus were wiped down with 70% ethanol. Time spent in each chamber and total distance traveled for the entire 30-min trial was measured using Ethovision 10 (Noldus Information Technologies).

Light /Dark choice

For light/dark choice the same areas as used for real time preference testing were modified and used as previously described (93). On the light side a small sport light was placed overhead, and the walls were covered with white laminated paper. Light levels on this side measure 580-590 lux. For the dark side, an infrared spotlight was placed over head, to allow for video tracking of the mice. The walls were covered with matte black plastic. Light levels in the center of the dark side measure 100-110 lux. Animals were recorded using a USB web cam without an infrared filter. For real-time aversion testing photostimulation was paired with the dark side of arena as described above. For anxiety-like behavior testing stimulation was provided uniformly during the trials. Time spent in each chamber and total distance traveled for the entire 30-min trial was measured using Ethovision 10.

Observational behavioral assay

To observe behaviors evoked by photostimulation of SuMVGLUT2+::POA neurons we mice were habituated for at least 3 days prior to testing to round (18 cm diameter) clear arena with counterbalanced optical commutators to minimize impact of the head tether on movement. Testing occurred after habitation and approximately 2 cm of bedding material was placed in the arena. Behavior was recorded from the side and scored by a blinded observer.

Negative reinforcer two nose port operant behavior testing

For operant behavior testing we used Med Associates mouse operant conditioning chamber with dual nose ports and house light as previously described(63). Briefly, mice were tethered via cantilevered counterweighted optical commutator to a laser light source. A five-day protocol call was used. Day one through four were 10-min trials. Photostimulation was provided at 10 Hz and activation of the active port resulted in a 10-second pause in the photostimulation and activation of the house light inside the arena. On the first four days each activation of the active port outside of a 10-second pause resulted in a new pause. On day five was a 30-min trial using a progressive ratio protocol the number activation of the active port to generate a pause increased with each activation based on the number of activations (j) = [5e(0.2j) − 5] round to the nearest integer generating the schedule 1, 2, 4, 6, 9, 12, 15, 20, 25, 32, 40, 50….(94). The trial was limited to 30-mins due to concern for animal welfare due to the head tether and confined space. Photostimulation terminated at the conclusion of the trial.

Forced Swim with photostimulation

Forced swim trials with optogenetic stimulation were done as a two-day test. On the first day all mice were subjected to a 15-min forced swim, dried and returned to the home cage. On the second day they were tethered and subjected to a 6-min forced swim divided into three periods, each two mins in length, pre stimulation, stimulation, and post stimulation. 10 Hz photostimulation was provided during the trial. Mice were closely monitored during each trial of swimming. Trials were recorded and scored by a blinded observer for time spent immobile on the second day of the test.

Feeding after food deprivation

For tests involving brief access to food after deprivation, the same 18 cm diameter round arenas with counterbalanced optical commutators were used. Mice were habituated to the arenas for a minimum 3 days prior to testing. Mice were food deprived by removal of food from the home cage 24 hours prior to testing. Mice were placed in the arena and allowed to habituate prior to introduction of a chow pellet. The 20-minutue trial with the foot pellet was recorded and scored by a blinded observer. Chow pellets were weighed before and after the 20-min trial. The difference is reported as food eaten.

Open Field Test

For Open Field testing we used a purpose-built 20in square behavior arena. Mice were tethered to a patch cable and placed into the behavioral arena. The laser frequency was set to 10 Hz and was left on for 20 mins. Distance moved for the 20-min trial was quantified using Ethovision 10. Bedding in the arena was replaced between every trial, and the floors and walls of the arena were wiped down with 70% ethanol.

Statistical analyses

Statistical analyses were conducted using GraphPad Prism software. Data are shown as mean ±SEM except for z-scores which as shown as ± 95% confidence interval, as noted in text. Values for individual p values are given in the text and figure legends. Significance was held at α less than 0.05. Unless noted in the test, nonparametric Mann-Whitney tests were used for statistical comparisons. In cases of multiple comparisons, Holm-Šídák method was used to correct for multiple comparisons. Paired testing was used with comparing within cohorts with repeated measurements and unpaired between cohorts. All “n” values represent the number of animals in a particular group for an experiment. For fiber photometry statistical analysis, the mean signal of the baseline and stimulus windows was used, and comparisons were made using the Wilcoxon ranked-sum test, with α=0.95.

SuMVGLUT2+ neurons project to the preoptic area of the hypothalamus and are not VGAT+.

(A) Schematic of injections into VGLUT2-Cre mice of Retro-AAV-DIO-tdTomato into the POA and AAV-DIO-ChR2eYFP into the SuM. (B and C) Projections of VGLUT2+ SuM neurons and evident in the POA with dense labeling in LPO and the site of viral injection in SuM (500um scale bar). (D) The brain regions, POA and SuM, shown in (B,C, E and F). (E and F) Cell bodies of VGLUT2+ SuM neurons are evident in Sum with injection of Retro-AAVs in POA. (G) Schematic of injection in POA of fluorophore switching construct in Retro-AAAV in VGLUT2-Cre or VGAT-Cre mice. (H and I) Labeling of Cre- (red) and Cre+ neurons in POA but only Cre+ seen in SuM of VLGUT2-Cre mice showing all SuM::POA neurons are VGLUT2+. (J) Quantification of labeling of Cre+ and Cre- cells in SuM following injections in POA in VGLUT2-Cre and VGAT-Cre mice. (K and L) Labeling of Cre- (red) and Cre+ neurons in POA but only Cre- seen in SuM of VGAT-Cre mice. (100um scale bar)

Retrograde verification of projection targets of SuMVGLUT2+::POA.

(A, D, G, and J) Schematic showing the injection sites of Retro-AAV’s in VGLUT2-Cre mice. (B-C) In VGLUT2-Cre mice injection of Retro-AAV-DIO-eYFP unilaterally in the (B) accumbens nucleus resulted in labeling of cell bodies in (C) SuM. (E-F) Similarly, injection of Retro-AVV-DIO-tdTomato into (E) septum resulted in labeling of cell bodies in (F) SuM. (H-I) injection of Retro-AAV-DIO-eYFP targeting the (H) PV and LHb resulted in cell bodies in (I) SuM being labeled with eYFP. (K-L) Retro-AVV-DIO-tdTomato injected into (K) PAG labeled cells in (L) SuM. (M) The cleared brain hemisphere from a VGLUT2-Flp mouse injected in the POA with injected Retro-AAV-DIO-eYFP show in Figure 2M displayed from a perspective of ventral to dorsal showing cells labeled in SuM and processes extending widely including in the hippocampus and POA. (100 and 500 μm scale bars)

Combinatorial viral and genetic approach is effective with minimal background.

(A) Schematic of injection of Retro-AAV-Cre into the POA and AAV-Nuc-flox(mCherry)-eGFP into the SuM of WT mice. (B) In SuM eGFP (green) Cre positive and mCherry (red) Cre negative labeled neurons are seen based on expression of Cre mediated by POA injected retrograde AAV constructs showing SuM::POA neurons are a subset of all cells in SuM. (C and D) In wildtype mice injected in POA with Retro-AAV-fDIO-Cre and in SuM with AAV-DIO-ChR2eYFP expression of ChR2eYFP (green) is not seen. (E-H) In Ai14 reporter mice injected with AAV-fDIO-Cre into SuM tdTomato expression is not observed unless Retro-AAV2-Flp is injected. (100 μm scale bars)

SuMVGLUT2+::POA neurons show Fos induction after FST and dunk test evokes active coping.

Sections of brains from mice (A) left in the home cage or (B) subjected to 15 minutes of forced swimming 90 min prior to sacrifice were probed with anti-Fos antibodies (cyan) showed (scale bars 500μm) (C) a significant (p= 0.029) increase in them mean ±SEM number of anti-Fos labeled neurons in brains from mice subjected to force swim (n=5) compared to control mice (n=4). (D) In VLGUT2-Cre mice Retro-AAV-DIO-mCherry was injected in the POA of mice (n=5) subjected to 15 minutes forced swim. (E) Region of SuM shown with higher magnification in (F) where SuMVGLUT2+::POA neurons labeled by mCherry show overlap with cells (arrows) labeled by anti-Fos. (Scale bar 100μm) (G) A significant increase (p= 0.016) in the number of VGLUT2+ cells labeled with mCherry and anti-Fos in mice subjected to forced swim. (H) Mice subjected to 10 30-second trials of “dunk” forced swimming. (I) Quantitation of average time mobile for every 30-second trial during “dunk” forced swim.

SuMVGLUT2+::POA neurons are not recruited during increased locomotor activity in the absence of acute stressor.

(A) Heat maps showing velocity for 12 episodes of rapid locomotion from a representative animal during free movement in open field with mean ±95% CI for the trials. (B) For that same animal heat map of Ca2+ dependent signal was analyzed during average peak velocity showing no change in activity during peak velocity. Representative heat map for a single animal with the mean ±95% confidence interval of the 12 trials for same animal. (C) Mean ±95% confidence interval for Z-score of 12 trials for all 13 animals and Boxplot for average Z-score. (D) Cross-correlation analysis shows no correlation between Ca2+ signal from SuMVGLUT2+::POA neurons spontaneous velocity increases.

GABAergic neurons in SuM do not drive place preference.

(A) Schematic of viral injection into VGAT-Cre mice in SuM. (B) VGAT+ neurons labeled and outline of implanted fiber. (C) Real-time arena showing the pair of photostimulation on one side of the arena. (D) Photostimulation at 10Hz did not drive a significant place preference. (eYFP (control) n=26 and ChR2=26). All data plotted as mean ±SEM. (E) Representative path traces of VGAT-Cre mice with ChR2 or eYFP (control) in open field testing. (F) No significant differences in center or perimeter time when comparing the ChR2eYFP expressing to controls VGAT-Cre mice. (G) Chr2eYFP expressing mice traveled slightly but significantly (p=0.034) less distance compered to control.