Activation of Parabrachial Tachykinin 1 Neurons Counteracts Some Behaviors Mediated by Parabrachial Calcitonin Gene-related Peptide Neurons

—Many threats activate parabrachial neurons expressing calcitonin gene-related peptide (CGRP PBN ) which transmit alarm signals to forebrain regions. Most CGRP PBN neurons also express tachykinin 1 (Tac1), but there are also Tac1-expressing neurons in the PBN that do not express CGRP (Tac1+;CGRP (cid:1) neurons). Chemogenetic or optogenetic activation of all Tac1 PBN neurons in mice elicited many physiological/behavioral responses resembling the activation of CGRP PBN neurons, e.g., anorexia, jumping on a hot plate, avoidance of photostimulation; however, two key responses opposed activation of CGRP PBN neurons. Activating Tac1 PBN neurons did not produce conditioned taste aversion and it elicited dynamic escape behaviors rather than freezing. Activating Tac1+;CGRP (cid:1) neurons, using an intersectional genetic targeting approach, resembles activating all Tac1 PBN neurons. These results reveal that activation of Tac1+;CGRP (cid:1) neurons can suppress some functions attributed to the CGRP PBN neurons, which provides a mechanism to bias behavioral responses to threats. (cid:1)


INTRODUCTION
The ability to identify and appropriately respond to threats is critical for survival. Threats to homeostasis may be subtle (e.g., the first hints of nausea signaling an impending bout of food poisoning) or startling (e.g., a pouncing predator) and require a range of behavioral responses for an animal to survive. Upon encountering a novel threat, animals will engage in innate defensive reactions. Learning associations between threats and the cues that predict them is essential for increasing long-term survival. The psychological literature has documented that animals form strong associations between biologically relevant predictive cues and the consequences they signal, a phenomenon called selective associations (e.g., Garcia & Koelling, 1966;Lin et al., 2014). However, the neural underpinnings of this phenomenon remain unclear. Candidate neural circuits need to involve the sensory processing of cues and threats and interface with circuits capable of generating behavioral responses. The parabrachial nucleus (PBN) is a likely node in these neural circuits because it is activated by a wide variety of threats and it has reciprocal connections with relevant forebrain regions, e.g., hypothalamus, extended amygdala, insular cortex, and brainstem.
Within the PBN, an astonishing range of threatening stimuli (e.g., somatic pain, visceral malaise, itch, and most strong sensory stimuli) activate calcitonin generelated peptide (CGRP; encoded by the Calca gene) expressing neurons in the external lateral parabrachial nucleus (CGRP PBN ) neurons (Carter et al., 2013;Han et al., 2015;Campos et al., 2018;Kang et al., 2022). Photoactivation of these CGRP neurons induces freezing behavior, anxiety, autonomic arousal, and anorexia https://doi.org/10.1016/j.neuroscience.2023.03.003 0306-4522/Ó 2023 The Author(s). Published by Elsevier Ltd on behalf of IBRO. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). (Carter et al., 2013;Campos et al., 2016;Bowen et al., 2020). Furthermore, they are both necessary and sufficient for establishing conditioned taste aversion (CTA) resulting from the pairing of a novel tastant with visceral malaise as well as forming fear-learning associations between external stimuli (e.g., tones) and somatic pain (e.g., foot shocks) Chen et al., 2018). This breadth of tuning and critical role in learning led to the characterization of CGRP PBN neurons as a general alarm system (Palmiter, 2018). Interestingly, fear and taste learning have long been viewed as distinct based on experiments showing that animals are adept at forming biologically relevant cue:threat associations such as taste:visceral malaise (interoceptive stimuli) or tone:shock pairing (exteroceptive stimuli) (Freeman & Riley, 2009;Garcia & Koelling, 1966;Gemberling et al., 1980;Gemberling & Domjan, 1982;Holland & Wheeler, 2009;Rusiniak et al., 1982) as compared with biologically mismatched associations (Krane and Wagner, 1975;Delamater and Treit, 1988). Thus, the fact that CGRP PBN neurons are necessary for both forms of learning is surprising, offering an opportunity to probe for other neuronal populations that interface with the general alarm signal of CGRP PBN neurons to generate appropriate behavioral responses to a particular threat.
Neurons within the PBN express many neuropeptides including CGRP, cholecystokinin, neurotensin, prodynorphin, pituitary adenylate cyclase-activating peptide, proenkephalin and tachykinin 1 (Pauli et al., 2022). Notably, most CGRP PBN neurons co-express tachykinin 1. Tachykinin 1 (Tac1 gene), which encodes the precursor of both neurokinin A and substance P. Tac1 is widely expressed, often along with CGRP, in pain circuitry including dorsal root ganglia, spinal cord, trigeminal neurons and PBN (Gutierrez-Mecinas et al., 2017;Edvinsson et al., 2021). Recently Tac1 PBN neurons were shown to be necessary and sufficient for generating escape responses, particularly jumping on a hot plate (Barik et al., 2018). Thus, we hypothesized that they might be involved in overriding the general alarm from CGRP PBN neurons from freezing into active avoidance (or escape) responses to threats involving somatic pain. To address this possibility, we infused adeno-associated viruses carrying recombinase-dependent genes in the PBN of transgenic mice to visualize the cell bodies and axonal projections of either Tac1, CGRP, or Tac1+; CGRPÀ as well as manipulate their activity. Using these methods, we provide evidence that activating Tac1 neurons biases defensive reactions towards escape-like behaviors and suppresses the formation of conditioned taste aversions generated either by activation of CGRP PBN neurons or malaise-inducing agents. This mechanism may provide a check on the general alarm and facilitate appropriate, defensive responses.

EXPERIMENTAL PROCEDURES Subjects
All subjects were adult male and female mice (P56-P140). Tac1 Cre/+ mice were obtained from Jackson Laboratory (stock #021877), and Calca Cre mice were generated and maintained as described by Carter et al. (2013) and are available at Jackson Laboratory (stock #033168). Calca Flpo (Arthurs et al., 2023) mice were generated by our lab. All mice were backcrossed onto a C57BL/6J background for more than six generations. Genotypes of experimental mice were established using polymerase chain reaction. All studies involved mice of both sexes; no sex differences were noted but sample size was too small for statistical analysis. Mice were group-housed before surgery and singly housed after recovering from surgery. They were housed on a 12-h light/dark cycle at 22°C with food and water available ad libitum unless otherwise noted in the experimental procedure. Experiments were conducted following guidelines from the NIH Guide for the Care and Use of Laboratory Animals, and all experimental procedures were approved by the University of Washington Institutional Animal Care and Use Committee.

Stereotaxic surgery
Mice were anesthetized with isoflurane (1-4%), treated with ketoprofen (5 mg/kg, IP), and secured in a stereotaxic frame (Kopf Model 1900). An incision site on the scalp was prepared by removing hair and cleaning the skin with alternating washes of betadine and ethanol. The cranial sutures were exposed via a midline incision allowing visualization of landmarks on the skull. A drill mounted to the stereotaxic frame was used to drill holes in the skull overlying the PBN. Targeting coordinates were referenced from bregma for AP and ML coordinates and dura for the DV coordinate and were as follows: PBN fibers and virus (AP: À4.8, ML: ±1.4, and DV: À3.3 for virus and À3.0 for fiber optic cannulae). Virus (200-300 nL) was injected bilaterally targeting the PBN using a glass capillary micropipette attached to a Nanoject II (Drummond Scientific). An additional hole was drilled in the skull to allow the anchoring of a bone screw (FST Item No. 19101-00). The fiber optic cannula and the bone screw were cemented to the skull with cyanoacrylate hardened with dental acrylic monomer. A headcap reinforcing all implants and filling the incision was formed from dental acrylic and the ends of the wound were secured with suture. Isoflurane anesthesia was discontinued, and mice were allowed to recover for at least 3 weeks before undergoing experiments.

Histology and microscopy
Mice from behavioral experiments received a fatal overdose of pentobarbital (EuthasolÒ, 270 mg/kg, IP, Merck) and were perfused transcardially with ice-cold PBS followed by 4% paraformaldehyde in phosphatebuffered saline. Brains were post-fixed for 2-24 hours at 4°C depending on the quality of the perfusion and then cryoprotected in 30% sucrose, frozen in TissuePlus OCT (ThermoFisher), and stored at À80°C before sectioning (15-35 mm) on a cryostat (Leica Microsystems). For in-situ hybridization, fresh frozen brains were sectioned at 15-20 mm and directly mounted to glass slides whereas for all other purposes tissue was sectioned at 30 mm and collected into cold PBS for immunohistochemical processing and wet mounting on glass slides. All images were collected using either a Keyence BZ-X710 microscope or Olympus Fluoview FV-1200 confocal microscope. Images were minimally processed for brightness and contrast.

Fluorescent in-situ hybridization
Brains were collected from mice anesthetized with isoflurane and euthanized via decapitation. After extraction, brains were frozen for approximately 60 sec in 2-methylbutane cooled with Dry Ice to À30°C. Samples were stored at À80°C until cut into 20-mm sections on a cryostat. Sections were processed using RNAscope Multiplex Fluorescent Assay (version 1) to label Tac1 (Atto 647) and Calca (AF 488) RNA according to manufacturer instructions (Advanced Cell Diagnostics).

BioDAQ consummatory behavior
Mice were individually housed in BioDAQ recording chambers (Research Diets) and acclimated to consuming food or water from the monitored hopper apparatus. Animals were habituated for at least 7 days before experimental manipulation or either food or water intake. Cumulative food and water intake during the dark cycle, was analyzed via BioDAQ Viewer. Clozapine-N-oxide (CNO) was injected (1 mg/kg, IP) 30 min before the start of the dark cycle to activate neurons expressing hM3D(Gq).

Hot plate
Mice were injected with CNO (1 mg/kg, IP) 30 min before being placed on a 55°C hot plate (Bioseb) for 30 sec. Behavior on the hotplate was recorded and video was scored offline for latency to lick a hind paw or jump, as well as the number of times animals, jumped up off the hotplate.

Small arena
Mice were prepared for photostimulation by connecting the implanted fiber-optic cannulae to a blue-light laser via a fiber-optic cable before being placed in a welllighted arena (10 Â 10 cm). After habituating to the experimental setup for 1 min, mice were photostimulated (30 Hz, 10 sec) twice with a 1-min interstimulus interval. Video of the behavioral responses to photostimulation was scored offline. Freezing was defined as rigid uninterrupted immobility that was usually terminated by the initiation of head movement. Locomotion was defined as a continuous period of forward movement.

Big arena
Mice were prepared for photostimulation by connecting the implanted fiber-optic cannulae to a blue-light laser via a fiber optic cable before being placed in a welllighted arena (10x50 cm). Mice were habituated to the apparatus for 5 min before undergoing two bouts of photostimulation (30 Hz, 10 sec) separated by a 1-min interstimulus interval. On another day, a partial low wall (12 cm) extending 7.5 cm into the arena was added creating a small hide area that was 10 cm square with an entrance of about 2.5 cm and a small, shaded corner. The habituation period was extended to 10 min to encourage animals to explore outside the hide area. The laser was activated for 10 sec (30 Hz) when mice were at the end of the arena opposite the hide area. Behavioral responses were recorded and scored as for the small arena test with the addition that time spent within the hide box was also quantified when applicable.

Real-time place conditioning
Mice were prepared for photostimulation as above and placed into a two-chamber real-time place conditioning apparatus. Each chamber measured 28 cm by 28 cm with distinct visual cues (i.e., vertical versus horizontal stripes). Photostimulation (15 Hz) on one side of the apparatus began after an initial 5-min baseline period to assess initial preferences. 15-Hz photostimulation was selected so that movement phenotypes (i.e., freezing, jumping) would not interfere with their ability to move freely between chambers.

Conditioned taste aversion
Our standard conditioned-taste-aversion paradigm has been described (Chen et al., 2018). Mice were acclimated to individual housing in custom cages with two ports for attaching drinking tubes and then adapted to scheduled water access 30 min in the morning and 1 hour in the afternoon for three days. Intake was recorded in grams and converted to milliliters. The three conditioning trials consisted of pairing 30 min of 5% sucrose access with an injection of CNO (1 mg/kg, IP) as the unconditioned stimulus followed by scheduled water access that afternoon and the following day. Finally, animals were given a 2-bottle test consisting of 30-min access to both sucrose and water. For optogenetic CTA experiments, the unconditioned stimulus was 30 min of optical stimulation via a blue laser (30 Hz, 3-sec on 2-sec off). To examine the capacity of Tac1 PBN neuronal stimulation to attenuate a LiCl-induced CTA, mice were administered CNO (1 mg/ kg, IP) immediately after 30-min access to 0.2% saccharin and 30 min before being administered LiCl (180 mg/kg, IP) on each of two conditioning trials. Finally, animals were given a 2-bottle test consisting of 30-min access to both saccharin and water. For this experiment we made several changes to our standard CTA paradigm. We extended the interstimulus interval between the CS and US for CNO to be maximally effective when the US was administered. We also found that sucrose was ineffective as a CS with this longer interstimulus interval and switched to saccharin, as it evokes a stronger neophobic reaction (e.g., Kronenberger and Me´dioni, 1985).

Data analysis
All data were analyzed using OriginPro (OriginLab) software to conduct two-way repeated measures ANOVAs, one-way ANOVAs, or t-tests as appropriate with Fisher post hoc tests. Notably, due to software limitations, only the first ten hours (out of 12 total) of cumulative intake were analyzed in two-way repeated measures ANOVAs for experiments examining food and water intake during the dark cycle.

RESULTS
Chemogenetic activation of either Tac1 PBN or CGRP PBN neurons causes anorexia, adipsia, and jumping on a hotplate Barik et al. (2018) described a role for parabrachial Tac1expressing neurons in mediating escape behaviors to noxious stimuli, including a hot plate and inflammatory pain. Their experiments using chemogenetic activation (Armbruster et al., 2007) indicated that there are at least two distinct populations of Tac1 neurons in the PBN; those that project to forebrain regions such as the central nucleus of the amygdala (CeA) and those that project to the reticular formation in the medulla, referred to as MdD. Non-selective inactivation of Tac1 neurons in the PBN suppressed escape behaviors, while selective activation of the projection to MdD (but not the CeA projection) promoted jumping behavior on the hot plate. They compared the results of activating hM3Dq in Tac1 PBN neurons with activating neurotensin (Nts) expressing PBN neurons (NTS PBN ) or CGRP PBN neurons and concluded that only activation of the Tac1 neurons promoted escape behaviors. They reported minimal Calca and Tac1 mRNA overlap in the PBN, with Tac1 being expressed in the lateral PBN and Calca in the external lateral PBN. However, the Allen Mouse Brain Atlas reveals robust Tac1 expression in the external lateral region like Calca and Nts, and we have shown by scRNA-Seq and RiboTag approaches that Calca and Tac1 are extensively coexpressed in the external lateral PBN (Pauli et al., 2022).
To compare the responses to painful stimuli, a novel cohort of mice was administered CNO and then 30 min later placed on a 55°C hot plate for 30 sec. The latency to respond (10-15 s) after being placed on the hot plate was the same for Tac1-hM3 PBN (t(10) = À1.340, p > .05) and CGRP-hM3 PBN (t(12) = À0.804, p > .05) neurons compared to YFP controls (Fig. 1H, J).
However, the nature of the response was different. The initial response of YFP controls was always licking the hind paw, whereas activation of either Tac1 or CGRP mice resulted in jumping. Specifically, the majority of Tac1-hM3 PBN (t(5) = 4.757, p < .01) or CGRP-hM3 PBN (t(12) = À4.547, p < .001) jumped (i.e., escape-like behavior) as the initial response (5 of 6 for Tac1-hM3 and 5 of 7 for CGRP-hM3), whereas only two of the YFP mice jumped during the entire test and never as the initial response (Fig. 1I, K). The number of jumps by the Tac1-hM3 PBN mice reported here is like that reported by Barik et al. (2018), but unlike their results, the CGRP-hM3 PBN mice jumped just as much. Jumping on the hot plate by the CGRP-hM3 PBN mice is consistent with our prior results showing that inactivation of CGRP PBN neurons with tetanus toxin blocked jumping  and photoactivation of axon terminals of CGRP-ChR2 PBN neurons in the CeA promoted jumping (Bowen et al., 2020). We conclude that jumping in response to thermal pain is not an exclusive phenotype of Tac1 PBN projections to the MdD.

Optogenetic activation of either Tac1 PBN or CGRP PBN neurons elicits opposing defensive behaviors
The previous results based on chemogenetic technology reveal that activation of either Tac1 PBN or CGRP PBN neurons resulted in similar behaviors; however, the degree of activation of hM3Dq with CNO is less than what can be achieved by photoactivation of channelrhodopsin.
Robust (30 Hz, 15 mW) photoactivation of CGRP-ChR2 PBN neurons results in profound freezing behavior in which mice remain in awkward positions and resist moving even when prodded (Bowen et al., 2020). When we compared the photoactivation of Tac1-ChR2 PBN and CGRP-ChR2 PBN neurons, the responses of the mice were distinctly different. AAV1-DIO-ChR2:YFP (or YFP only as control) was Mice typically consume food throughout the dark cycle (Group YFP; n = 6); however, activating Tac1-hM3 (n = 7) neurons with CNO 30 min before the dark cycle results in robust anorexia. E. As with food, mice consume water steadily throughout the dark cycle (Group YFP) and activation of Tac1 PBN neurons via CNO suppresses this intake (Group Tac1-hM3). F. As previously demonstrated, activation of CGRP PBN neurons (CGRP-hM3; n = 7) causes anorexia relative to YFP controls (n = 6). G. CGRP-hM3 activation via CNO also causes adipsia. H. When placed on a 55°C hotplate Tac1-hM3 activation does not cause a significant change to the latency to respond. (n = 6 per group) I. Tac1-hM3 activation causes mice to jump off the hot plate repeatedly. J. CGRP-hM3 activation does not alter the latency to respond to placement on a hot plate (n = 7 for both groups) K. CGRP-hM3 activation causes mice to jump off the hotplate repeatedly. (n = 7 for both groups).
injected into the PBN of Tac1 Cre or Calca Cre mice, and fiber-optic cannulas were implanted above the PBN ( Fig. 2A-C). After several weeks for viral expression, mice were placed in a small (10 Â 10 cm) arena and photoactivated at 30 Hz (10 mW) for three 10-sec intervals. The percent time freezing (F(2,11) = 507.27, p < .0001) or locomoting (F(2,11) = 210.48, p < .0001) during stimulation was assessed from video recordings. As predicted, photoactivation of CGRP-ChR2 PBN neurons resulted in freezing (i.e., immobility of the head and body with no locomotion) with negligible locomotion; in contrast, photoactivation of Tac1-ChR2 PBN neurons resulted in bouts of rapid locomotion indicative of escape behavior with no freezing compared to the YFP control mice (p < .05) (Fig. 2D, E). When this experiment was repeated using a larger arena (10 Â 50 cm) that would allow more room for locomotion, the results were similar but less dramatic; photoactivation of CGRP-ChR2 PBN neurons promoted freezing (F (2,11) = 13.599, p < .01), although only 70% of the time (Fig. 2F), while again photoactivation of Tac1-ChR2 PBN neurons only promoted locomotion (F(2,11) = 15.755, p < .001); YFP controls also moved more in the larger arena the (Fig. 2G). In the third repeat of this experiment, a partial wall was installed 10 cm from one end of the larger arena to create an accessible ''hide" area.  Box and whisker plots of hide time (% of stimulation time) for YFP (n = 4), CGRP (n = 5), or Tac1 (n = 5) mice placed in a big (10 Â 50 cm) arena with a partial hide box at one end (10 Â 10 cm). Individual data points are represented as hollow diamonds, solid diamond indicates group means, the box represents the median, 25th, and 75th percentiles, and whiskers represent the 5th and 95th percentiles. Between-group differences indicated by significant between-group difference *P < .05.
Photoactivation of the CGRP-ChR2 PBN neurons at the end opposite to the hide area still resulted in prolonged freezing (F(2,11) = 273.68, p < .0001) (Fig. 2H), but after activation of all Tac1-ChR2 PBN neurons the mice spent more time locomoting compared to YFP or CGRP-ChR2 PBN (F(2,11) = 7.379, p < .01) (Fig. 2I). A one-way ANOVA revealed only a trending effect for time spent in the ''hide" area (F(2,11) = 3.244, p = 0.078) (Fig. 2J). However, it is interesting that CGRP PBN stimulated mice spent no time in the ''hide" area. These results reveal that robust activation of Tac1-ChR2 PBN neurons promotes escape like behaviors -running if there is no choice but hiding some of the time if that is possiblewhereas the CGRP-ChR2 PBN mice freeze regardless of the options available.

Distinct populations of neurons co-express Tac1 and Calca versus neurons that express Tac1 without Calca
The results in Fig. 1 are compatible with Tac1 and Calca being co-expressed; however, with more intense stimulation, some phenotypes are distinct (Fig. 2), which is difficult to reconcile with strict co-expression. We used RNAscope in-situ hybridization to locate the coexpressing neurons and ascertain whether there are neurons that express Tac1 without Calca (Tac1+; CGRPÀ) and, if so, where they reside in the PBN. The RNAScope experiments revealed extensive coexpression of Tac1 and Calca in the external lateral region of the PBN as expected; however, many Tac1+; CGRPÀ neurons were present in other subdivisions of the PBN, including dorsal lateral, superior lateral, central lateral and even medial regions (Fig. 3). We have reported elsewhere that the majority of Tac1 neurons ($55%) do not co-express Calca (Arthurs et al., 2023; see also Pauli et al., 2022).

Activation of Tac1 PBN neurons suppresses conditioned taste aversion
Pairing access to a novel food with activation of CGRP PBN neurons, using either hM3Dq and CNO or photoactivation of ChR2, suppresses the consumption of that food on subsequent exposure -a phenomenon called conditioned taste aversion (CTA) (Carter et al., 2015;Chen et al., 2018). We compared CTA responses of mice expressing hM3Dq, or YFP controls, in Tac1-hM3 PBN or CGRP-hM3 PBN neurons (Fig. 4A). We paired 30-min access to 5% sucrose in water-deprived mice with CNO on two separate days, followed two days later by giving the mice a choice of sucrose or water. Tac1-hM3 PBN mice failed to learn a CTA despite the overlap between neurons expressing Tac1 and Calca (t(9) = 1.116, p = .293) (Fig. 4B). As expected, mice with activation of CGRP-hM3 PBN neurons had a low preference for sucrose, indicative of a CTA (t(10) = 4.738, p < .001) (Chen et al., 2018) (Fig. 4C). Due to the overlap between Tac1 PBN and CGRP PBN neurons, these results suggest that activating Tac1 PBN neurons overrides the effect of activating CGRP PBN neurons and prevents the establishment of a taste aversion. Alternatively, activating the subset neu-rons that express both Tac1 and CGRP ($45%) may be insufficient to establish a CTA. In a second cohort of mice, we tested whether activating Tac1 PBN neurons could attenuate the formation of a CTA based on visceral malaise produced by LiCl. Tac1 Cre mice were injected with AAV-DIO-hM3Dq-YFP or AAV-DIO-YFP as control. In third cohort of mice, we used a similar CTA paradigm except for access to the saccharine (the conditioned stimulus; CS) was followed immediately with an injection of CNO (1 mg/kg, IP) and then 30 min later by injection of LiCl (the unconditioned stimulus, US; 180 mg/kg, IP) a classic CTA-inducing agent. Activation of Tac1-hM3 PBN neurons significantly attenuated LiCl-mediated CTA compared to YFP controls (t(13) = À3.638, p < .01) (Fig. 4D). Next, in a fourth cohort of mice, we compared the CTA responses of mice expressing ChR2 in four different groups of mice (Fig. 4E): Tac1 PBN neurons (Tac1 Cre ), Tac1+;CGRPÀ neurons (Tac1 Cre ::Calca FLPo , Fig. 4F), Tac1 and CGRP neurons (Tac1 Cre ::Calca Cre mice, Fig. 4G) and CGRP PBN (Calca Cre ) neurons. We paired 30-min access to sucrose in water-deprived mice with photoactivation of ChR2 (30 Hz, 3 sec on, 2 sec off) on two separate days, followed two days later by giving the mice a choice of sucrose or water. As expected, the CGRP-ChR2 PBN mice developed a CTA and avoided the sucrose, whereas none of the other groups developed a CTA and continued to prefer sucrose (F (3,16) = 15.885, p < .0001) (Fig. 4H). Thus, like the suppression of freezing behavior, activation of Tac1 PBN neurons blocks or attenuates the CTA generated either by activating CGRP PBN neurons or classical LiCl-induced visceral malaise. Activation of CGRP-ChR2 PBN neurons is aversive and results in a place aversion in a real-time place preference paradigm (RTPP) (Bowen et al., 2020). We asked whether activation of Tac1+;CGRPÀ neurons would also generate a place aversion in an RTPP paradigm. Photoactivation at 15-Hz of mice from the Tac1 +;CGRPÀ mice from the CTA experiment produced a profound place aversion in an RTPP paradigm and the effect remained after stimulation was discontinued (F (2,15) = 19.287, p < .001) (Fig. 4I).

Parabrachial Tac1 and CGRP neuron axon projections have distinct post-synaptic targets
To accurately compare the projections of Tac1 PBN and CGRP PBN neurons, we co-injected AAV1-DIO-YFP and in Tac1 Cre ::Calca Flpo mice (n = 3). Fig. 5A-D show the expression of both fluorescent proteins at 4 coronal (rostral to caudal) levels of the PBN, revealing distinct expression patterns. There is substantial overlap of the fluorescent proteins only in the external lateral PBN. Based on labelling we could divide neurons into Tac1+; CGRPÀ (green), Tac1/CGRP (yellow), and CGRP+; Tac1À (red).
We saw projections from all three populations in each downstream target structure. However, this approach allowed us to discern subtle differences in the axon projections within target brain regions. For example, there were green Tac1+;CGRPÀ fibers in the ventral part of the lateral septum (Fig. 5E) and insular cortex (Fig. 5F), whereas red fibers from CGRP neurons were rare. Within the bed nucleus of the stria terminalis (BNST) there was a mixture of fiber patterns with most co-labeled fibers in the oval subdivision (BNSTov) with a more distinct Tac1+;CGRPÀ and CGRP+;Tac1À fibers in the ventral portion of the BNST (BNSTv) (Fig. 5G). In the thalamus there were few Tac1+;CGRPÀ fibers but CGRP-only fibers were present in several thalamic nuclei including the nucleus of reuniens (Re) and rhomboid nucleus (Rh). There were some co-labeled and Tac1+;CGRPÀ fibers in the region of the rostral border of parasubthalamic nucleus (Fig. 5H). Within the amygdala, Tac1+;CGRPÀ fibers extended more into the lateral and medial regions of the rostral CeA than the CGRP+;Tac1À fibers. However, there were also many co-labeled fibers in the lateral capsular CeA (CeC) (Fig. 5I). A less distinct partial segregation of colors was apparent in the caudal CeA, VPMpc, and PSTN (Fig. 5J, K), with green and red fibers in proximity but little co-expression. There were strong Tac1+; CGRPÀ fibers in the posterior portion of the basolateral amygdala (Fig. 5L). Within the medullary reticular nucleus dorsal part (MdD), which was a focus of the Barik et al. (2018) study, we see diffuse green and red fibers indicating widespread projections from both Tac1 +;CGRPÀ and CGRP+;Tac1À neurons with few if any yellow fibers noted (Fig. 5M, N). These results suggest that even within the same brain region distinct postsynaptic cells are likely to innervated by the Tac1+; CGRPÀ versus the Tac1/CGRP co-expressing neurons.

DISCUSSION
While Tac1 and CGRP are co-expressed in the external lateral PBN, there are many neurons in the PBN that express Tac1 without CGRP. The CGRP PBN neurons are activated by a wide variety of sensory stimuli, somatic and visceral threats, as well as cues that predict harm; consequently, they appear to serve as a general alarm. A common response to virtually all threats is transient suppression of non-essential behaviors to facilitate attention being directed towards the immediate threat. For example, animals will cease consummatory behaviors upon sensory detection of potential predators and either flee, freeze or attack depending on the proximity of the predator (Bolles, 1970;Apfelbach et al., 2005;Blanchard et al., 2005;Lenzi et al., 2022;Narushima et al., 2022). Beyond this alarm role, CGRP PBN neurons play an important role in learning to avoid future threats; they convey the unconditioned stimulus (US) in classical conditioning experiments. Pairing chemogenetic or optogenetic activation CGRP PBN neurons with innocuous cues results in condi- tioned responses, while inactivation of these neurons suppresses the ability to learn to avoid natural threats Campos et al., 2016Campos et al., , 2018Chen et al., 2018;Palmiter, 2018;Bowen et al., 2020). An effective alarm system needs a mechanism to discern the nature of the threat to allow an appropriate response. The CGRP PBN neurons project their axons to several forebrain regions, so perhaps their US signals converge with sensory conditioned stimuli (CS) in different post-synaptic neurons to elicit appropriate responses. Alternatively, while most threats activate CGRP PBN neurons, they may also activate other neurons within the PBN as well, such that the constellation of neurons that are activated dictates appropriate responses. This idea is consistent with the observation that while most threats activate Fos in CGRP PBN neurons, Fos is also expressed in many other undefined PBN neurons as well (e.g., Carter et al., 2013;Chen et al., 2022;Lee et al., 2021). We show that activation of Tac1+; CGRPÀ neurons can suppress some functions of CGRP PBN neurons and thereby preclude the freezing behavior or taste conditioning that would result from activating only the CGRP PBN neurons.

Calca
The co-expression of Tac1 and CGRP in the external lateral PBN explains the nearly identical effects of chemogenetically activating Tac1 PBN and CGRP PBN neurons in the generation of anorexia and adipsia as well as biasing behavioral responses on a hot plate toward active escape (i.e., jumping) rather than the typical paw licking behavior seen in control animals. However, this similarity highlights the remarkable differences observed after the optogenetic activation of these populations. Optogenetic stimulation of CGRP PBN neurons generated robust freezing behavior as previously reported (Bowen et al., 2020); however, stimulation of Tac1 PBN neurons had an opposite effect causing robust locomotion behavior, while animals actively avoid stimulation of either neuronal population in the RTPP assay (Bowen et al., 2020). We assume that activation of hM3Dq with CNO amplifies ongoing signaling events whereas photoactivation may provide an extra (unnatural) boost in activity to the Tac1+;CGRPÀ neurons, allowing them to suppress the function of CGRP PBN neurons and prevent freezing behavior. We predict that when a predator is detected at a distance, the CGRP PBN neurons are initially activated to promote freezing but as a predator approaches the Tac1+;CGRPÀ neurons are activated, which suppresses the activity of the CGRP PBN neurons and promotes active escape-like behaviors.
Remarkably, chemogenetic activation of Tac1 PBN neurons failed to produce CTA in the same behavioral paradigm that resulted in robust CTA after stimulating CGRP PBN neurons. We used Tac1 Cre ::Calca Flpo mice along with dual-recombinase INTRSECT viruses to express ChR2 in Tac1+;CGRPÀ neurons and found that pairing sucrose with activation of these neurons failed to generate a CTA. Furthermore, chemogenetic activation of Tac1 PBN neurons attenuated a LiCl-induced CTA. Because CGRP PBN neurons are activated by visceral malaise induced by lithium chloride, lipopolysaccharide, or GDF15 (Carter et al., 2013;Sabatini et al., 2021) and are necessary for establishing a CTA (Chen et al., 2018), we predict that the Tac1+; CGRPÀ neurons can suppress the function of CGRP PBN neurons to prevent CTA. This inhibition could be due to preferential activation of Tac1+;CGRPÀ neurons under some conditions (e.g., somatic pain) coupled with an indirect, negative-feedback circuit onto CGRP PBN neurons or their post-synaptic targets. Several possible negativefeedback circuits can be envisioned. Because Tac1+; CGRPÀ and CGRP PBN neurons are glutamatergic (express Vglut2) the circuit needs to include intervening inhibitory neurons (Pauli et al., 2022). One possibility is a short-loop within the PBN: Tac1+;CGRPÀ ? GABA interneuron ? CGRP. Alternatively, there could be a long-loop, inhibitory circuit from post-neurons, e.g., in CeA or BNST, regions that are known to have GABAergic projections to the PBN (Jia et al., 2005;Cai et al., 2014;Douglass et al., 2017;Lundy, 2020;Raver et al., 2020;Luskin et al., 2021;Bartonjo and Lundy, 2022). We show that both CGRP PBN and Tac1 PBN neurons project axons to the CeA and BNST but their anatomical destinations within these regions differ. For example, the CGRP terminals are largely restricted to the lateral capsule of the CeA, whereas the Tac1 terminals extend into the lateral and medial region, which are presumably from the Tac1 +;CGRPÀ neurons. RNAscope in situ hybridization studies reveal Calcrl mRNA (encodes CGRP receptor) in the lateral capsule and Tacr1 mRNA (encodes substance P receptor) in lateral/medial regions of the CeA, and both populations of neurons are GABAergic (Bowen et al., 2023). These observations suggest that GABAergic, Tacr1-expressing, neurons in the CeA that are activated by Tac1 PBN neurons could either project back to the PBN to inhibit CGRP PBN neurons or inhibit Calcrl-expressing neurons in the CeA. Additional experiments will be needed to examine these hypothetical feedback circuits. Overall, these experiments reveal both the value and dangers of studying the behavioral consequences of activating specific neuronal populations. The value comes from being able to describe what activating or inhibiting specific neurons can do. However, the danger is that those specific neurons may never be selectively activated; consequently, activating the ensemble of neurons is more physiological. While there are now several ways to trap and manipulate neuron ensembles (Guenthner et al., 2013;Franceschini et al., 2020), understanding the intercellular mechanisms involved depends on knowing the individual roles of all the neurons in the ensemble.