Central amygdala Isl1 neurons control biting by integrating sensory and motivational signals

29 Feeding beh avior is controll ed by many brain circuits and n euron types 1-8 . This is because th e brain has 30 to integ rat e the many senso ry and physical attri but es of the food with th e int ero ceptive inputs from t he 31 body. The cent ral amygdala (CeA) has emerged as a brain hub regu la ting feeding behavior 8-13 . The CeA is 32 composed of distinct microcircui ts tha t e valuate app eti tive stimuli to d rive rewar d and consummatory 33 behaviors 8,9 , but also sepa ra te microcirc uits tha t suppress app eti te when r eachin g satiety and in th e 34 presence of naus ea 10,11 . The CeA has also been shown to drive hun ting-like behavi or, including pursuit , 35 grabbing and biting 14 . However , it is unclear if in additio n to th e known appe titive microcircuits, th e CeA 36 contains distinc t biting neurons. M appin g the CeA cell types and microci rcu its th a t respond to th e 37 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 5, 2026. ; https://doi.org/10.64898/2026.02.03.703447doi: bioRxiv preprint 2 variety of sensory at tributes (tast e, smell ) and physical properti es (tex tu r e, densi t y, viscosity) of food 38 and process this informa tion to gene rat e the appr opria te mot or ou tput is impor ta nt for unders tanding 39 how consumption of solid food is regulat ed. 40 Feeding req uires pr ecise coord inati on of jaw movements. These movemen ts must adapt t o different 41 food tex tur es and densi ties. Distu rbance s in their cont rol ar e linked to eati ng disorders. U nders tanding 42 how the brain regula tes jaw movements is therefo re critic al for both basic n euros cience and clinical 43 applications . The tongue/jaw-relate d pri mary motor cor tex (tjM1) plays a centr al role in initi ating and 44 controlling jaw movements and bi ting force 15-19 . Distinct excita tory neuron popu l ations within tjM1 45 contribute t o the gen era tion of specific jaw motor pat te rns 20,21 . Subcortical struc t ures also shap e 46 orofacial behavio rs. The CeA has long b e en known to evoke jaw opening, biti ng, a nd chewing across 47 species 22-24 . Yet, the iden tity of th e speci fic cell types involved remained uncle ar. 48 Previous studies demons tra ted t hat CeA biting neuro ns project t o the p arvocellul ar re ticular forma tion 49 (PCRt). This region contains mandibular a nd cervical premot or neu rons th at media te biting a ttacks 14 . 50 More r ecent wo rk showed tha t CeA proj ections to the pa rabr achial and supr atrig eminal nuclei elici t 51 rapid orofacial b ehaviors t hat dr ive inges tion of both food and n on-food objects 25 . Together, thes e 52 findings point to th e CeA as a key hub for orchestr ating or ofacial actions . 53 We hypoth esized t hat the CeA no t only initiat es jaw movements but also modul a tes th eir int ensity in 54 response to sensory inpu t. To investiga te this possibility, we focused on iden tifying the neuro nal 55 populati on involved in jaw movements. The CeA is subdivided into thre e spati al subdivisions, the 56 capsular (CeC), lateral (CeL) and medial subdivisions (CeM). The CeM is the major output r egion and it 57 has been implicat ed in at tack-rela ted bi ti ng through its projec tions to the PCRt 14 . The so far best 58 characte rized appe titive CeM populatio n marked by expr ession of the se roto nin r ecepto r Htr2 a (and 59 partially overl apping with th e Pnoc popul ation) promo tes pala tabl e food consumption and posi tive 60 reinforceme nt 8,9,12,1 3 . But unlike the bi tin g neurons, CeM Htr2a neurons media te the ir functions via the 61 parabr achial nucleus (PBN) and were t he refore an unl ikely candidat e for th e bitin g neurons. 62 Recent single-cell R NA sequ encing studi e s have highlighted a CeM popul ation ma r ked by the LIM-63 homeodomain transcrip tion facto r Isl1, h aving little overl ap with CeM Htr2a or Pnoc neurons 26 . Isl1 64 neurons wer e previously describ ed as a major CeM populati on in developm ental tracing studi es 27 and 65 were shown to pr oject to mul tiple hindb rain ta rget r egions, including PCRt 26,28 . H owever, th eir function 66 in the regul ation of feedi ng or biting beh avior remain ed unknown. 67 Here, we us ed Isl1-CreER mice to sel ectiv ely targe t this gene tically defined CeM p opulatio n and 68 examine d its rol e in modulating bi ting be havior. W e found tha t CeA Isl1 neurons a r e robustly activa ted a t 69 the onse t of biting and enc ode th e physical proper ties of food, with ha rder objects eliciting str onger 70 populati on respons es. Single-n euron an a lysis revealed functionally disti nct subpo pulations t uned t o 71 materi al-specific sensorimoto r features. Activation of CeA Isl1 neurons incre ased th e vigor of orofacial 72 movements without alt ering food intak e, whereas th eir inhibi tion impair ed efficie nt biting of solid food. 73 Projection-specific manipulatio ns demon strat ed th at activa tion of CeA → PCRt an d CeA → 74 pedunculopon tine tegmen tal nucleus (PPTg) pathways increased th e dura tion an d frequency of biting 75 behavior. Not ably, CeA → PPTg stimulation enhanced bi ting with positive mo tivat ional valence, 76 identifying a parall el pathway th at r einfo rces orofacial acti ons thr ough emoti onal drive. 77

78 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 5, 2026. ; https://doi.org/10.64898/2026.02.03.703447doi: bioRxiv preprint 3 CeA neurons show rob ust activation at bite onset across dive rse obj ects 79 Using the known localiza tion of CeA Pkcδ n eurons t o the CeL 8 , we confirmed th at Isl 1 immunoreactive 80 neurons wer e largely confined to th e CeM (Fig. 1A). We then us ed Isl1-CreER mice in combination with 81 Isl1 immunostaining to show high fidelity of Cre expression , with stro ng overlap w ith Isl1 82 immunoreactivity (fig. S1, A to C). A comparison with o ther transcrip tiona lly defined cell clusters 83 revealed tha t the I s l 1 + popula tion was di stinct from oth er CeA neu ronal sub types, including the Sst and 84 Pnoc cluste rs previously associat ed with appeti tive beh avior 9,12,2 9 ( fig. S1D). 85 To assess whether the ac tivities of individual CeA Isl1 neurons wer e associa ted wit h biting, we perform ed 86 in vivo calcium imaging in freely behavin g mice. We deliver ed Cre-dep enden t GC aMP8m-expressing 87 virus unilater ally into th e CeA of Isl1-Cre ER mice, followed by implantatio n of a gradient ind ex (GR IN) 88 lens in the same loca tion . Calcium signals in the infecte d neuro ns were r ecorde d using a head-mounted 89 miniscope (Fig. 1, B and C). We test ed an imals that wer e fasted for four h ours, un der five separa te 90 conditions: exposure t o non-food items s u ch as softwood, Styrofoam, spong e, as well as live crickets, 91 and normal chow. Rec ordings were do ne for 5 min each. To analyze how CeA Isl1 ne urons respon ded a t 92 the onse t of biting, we aligned calcium a ctivity traces t o the “first bit e”, defined a s the initial bi te of each 93 discrete b out of biting be havior, acr oss t he five conditions (Fig. 1, D and E). Notab ly, although th e 94 populati on mean activi ty revealed a pro minent peak a t bit e onse t, th e respons e profiles differed 95 betwee n mate rials, with ha rder subs tra t es (softwood, Styrofoam, chow) evoking sharper , higher 96 amplitude ac tivation compa red to softer substra tes (sponge, cricket), r egardless o f whether th ey were 97 edible or in edible (Fig. 1E). In all cases, a large prop orti on of neurons showed cl ea r modulation around 98 bite onse t, with consist ent recrui tment p att erns across bo th edibl e and inedi ble s ubstra tes. Up-99 modulated n euro ns comprised th e major ity of the popula tion in each cond ition (r anging from 58.0 % in 100 chow to 67.6 % in Styrofoam), while smaller subsets we re down-modula ted or n o n-responsive (Fig. 1F). 101 While th e freque ncy of bite events did n ot differ across mat erials (fig. S1E), the duration of individual 102 bites varied subs tan tially, with softe r objects eliciting longe r bit e dura tions (fig. S1F). This dissociation 103 raised th e possibility th at t he obse rved n eural r esponses wer e not a reflecti on of how much the animal 104 preferr ed or enjoyed biting a par ticular o bject. Ins tead , the ac tivity more likely ref lected biom echanical 105 demands or mot or out put fea tur es such as bite force o r muscle engagemen t asso ciated with differ ent 106 materi al prop erti es. For exampl e, soft wood was bitte n for less time due to high r esistance (fig. S1F), bu t 107 the popul ation me an activity was higher than for th e oth ers (Fig. S1G). The t empo ral dynamics of CeA Isl1 108 activation (peaking within seco nds after the bit e) were pr eserved acr oss conditio ns. These findings 109 indicate that CeA Isl1 ne uron ac tivity was closely tied t o the execu tion of biting be h avior and may encode 110 materi al resist ance during bi ting, as activ ity was consistently eleva ted du ring bitin g compared to non-111 biting periods (Fig. 1, G to K). 112 To examine how ne uronal activity rela te d to feeding phases , we analyzed the st r ucture of chow 113 consumption. Mic e typically began with gnawing, forceful oral inte ractio ns near t he mouth, an d in ~1-114 11 % of the time this was followed by rhythmic chewing after moving the pell et outward (fig. S1H). 115 Population mean activity was significantly higher during gnawing than during che wing or non-gnawing 116 periods (fig. S1, I and J), with no differe n ce betwe en chewing and non-chewing e pochs (fig. S1K ). 117 Correspondingly, 47.4 % of neurons wer e modulat ed during gnawing, but only 2. 9 % during chewing ( fig. 118 S1, L and M). Thus, CeA Isl1 activity was preferen tially engaged during the ini tial int eracti on with food, 119 particula rly first bite a nd gnawing, and la rgely unresponsive t o subseque nt chewi ng, suggesting a role in 120 encoding the se nsory–mo tor dema nds of bite initi ation r ath er th an gene ral food p rocessing. 121 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 5, 2026. ; https://doi.org/10.64898/2026.02.03.703447doi: bioRxiv preprint 4 To assess CeA Isl1 neuron activity during e t hologically relevan t orofacial be haviors, we performed calcium 122 imaging in freely moving mice during cric ket hunting . We defin ed pursui t as the p eriod during which th e 123 mouse actively followed the cricke t prio r to captur e, and hol ding as the e poch immediately following 124 capture , when th e mouse res train ed th e cricket in its mouth an d often r eloca ted t o a secure ar ea befo re 125 consumption. W e found th at th e averag e activity across th e popula tion did no t d iffer significantly 126 betwee n thes e behavior al epochs (fig. S1, N and O), likely because all thr ee involved orofacial 127 engagement to some e xt ent . Pursuit ofte n included at tempt ed strik es, and holdin g required con tinued 128 mouth contac t with th e prey. A significan tly higher propo rtio n of neurons was mo dulated du ring 129 gnawing (fi g. S1P to R), suggesting that CeA neur ons were mor e consist ently enga ged during forceful 130 orofacial actio ns than du ring pursuit o r prey holding. 131 Distinct populations of CeA neuro ns enc ode material-spe cific biting behavio r 132 Nex t, we asked whe ther dis tinct subpo p ulations of CeA Isl1 neur ons exhibi ted ma t erial-specific activity 133 patt erns r eflecting th e physical prope rtie s of the bitt en objects . We pe rformed lo ngitudinal cell 134 registra tion acr oss conditions for the pr e viously shown data. Hea tmaps of single-neuron calcium activity 135 revealed dive rse r esponse pr ofiles depe n ding on the object the mice int erac ted wi th. N eurons wer e 136 grouped int o five functional cluste rs bas ed on thei r dominan t activati on nea r the onset of the firs t bite . 137 Strikingly, thes e clusters e xhibi ted disti nct mate rial pref erenc es: Cluster 1 r espond ed most stro ngly to 138 silicone, Cluster 2 was prefe ren tially acti vated by softwood, Cluste r 3 respond ed to both spo nge and 139 Styrofoam, but show ed grea ter ac tivity for sponge, and Clust er 4 was selectively activated by Styr ofoam. 140 Cluster 5 showed non-resp onsive cells. T hese pat te rns were consist ent across ani mals and stable acr oss 141 trials (Fig. 2A). 142 Cluster-average d activity tr aces confirme d that r esponses wer e time-locked to bit e onset an d differed in 143 both amplitu de and dynamics dep ending on the mat erial (Fig. 2B, fig. S2, A and B) . Functional clus tering 144 of CeA Isl1 neurons reveal ed five distinct r esponse profil es across bite con ditions (F ig. 2C), with neurons 145 distribut ed across clust ers in a larg ely balanced manne r, exc ept for a small er non- responsive cluste r (Fig. 146 2D). The spatial arra ngement of t hese cl usters within the CeA Isl1 popula tion ap pe ared ra ndom (fig. S2C). 147 Together , thes e resul ts suggest tha t CeA Is l 1 neurons conta in functionally distinc t subpopulat ions tha t 148 encode mat erial-sp ecific sensory-motor features, su ppo rting a rol e for th e cent ral amygdala in 149 evaluating th e physical prope rti es of objects during orofacial b ehavior . 150 To examine how CeA Isl1 popul ation dyna mics encoded the physical pro per ties of different mat erials , we 151 performed p rincipal componen t analysis (PCA) on the pooled , trial-ave raged calci um activities across all 152 record ed neur ons and mice. The r esultin g low-dimensional traject ories r eveale d distinct popul ation 153 activity patt erns associa ted with bi ting di fferent mat erials (Fig. 2E). When pr oject ed onto the first thre e 154 principal componen ts (PC1–PC3 ), traject ories corr esponding t o silicone, sponge , softwood, and 155 Styrofoam diverged in st ate sp ace aroun d the time of bit ing, indicating mat eria l-specific neu ral 156 repr esent ations . Following th e bite , traje ctories gradually converged, su ggesting t hat CeA Isl1 popula tion 157 activity transi ently differen tia ted object i denti ty during the bi ting action b efore r e turning towa rd a 158 common post-event st ate (Fig. 2F). 159 To test whe ther n eurons differ entia ted b etween flavors du ring food consumption , we analyzed calcium 160 activity of CeA Isl1 neurons aligned to bit es of food pellets immersed in st rawbe rry, vanilla, or chocola te 161 aroma, wat er, o r quinine . The same neu r onal popula tion was tr acked across all co nditions, and n euro ns 162 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 5, 2026. ; https://doi.org/10.64898/2026.02.03.703447doi: bioRxiv preprint 5 were cluste red bas ed on th eir resp onse dynamics relative to bi te ons et (fig. S2D). Most responsive 163 neurons e xhibi ted highly similar activati o n patt erns rega rdless of flavor, with lit tle evidence for selec tive 164 responses to specific sensory iden tities (fig. S2, E and F). Similar results we re ob tai ned with non-edibl e 165 sponges infused with the sam e flavors (fi g. S2, G to J). This r esponse pa tt ern was consisten t even for 166 quinine, suggesting th at CeA Isl1 neu rons were not tuned to flavor iden tity but ins t ead encod ed gene ral 167 featur es of food-contact beh avior, such as sensory det ection o r motor engageme nt indep enden t of the 168 specific taste or a roma . Frequency and d u ration of biting behavior did no t differ s ignificantly across 169 flavors for either food o r sponges (fig. S2, J to M), suggesting th at th e observe d n eural r esponses wer e 170 not rel ate d to changes in mo tor outpu t o r behavioral p refer ence . 171 Licking and biting are both fundam ental motor acti ons requi red for consumma tor y behavior and 172 previous studies h ave shown that n euro ns in the CeA resp ond to licking 12,30,3 1 . To dete rmine whet her 173 CeA Isl1 neurons resp onded to differen t liq uids, we performed calcium imaging while mice were 174 presen ted with flavor ed or unflavore d liq uids (water, quinine , chocola te milk, vanilla milk, strawber ry 175 milk). We registe red individual n eurons a cross all conditions including differen t ha rd objects (sponge, 176 softwood, Styrofoam) to e nable withi n-neuron compar isons and cluste red ac tivity traces by respons e 177 dynamics (Fig. 2G). We identified five dis tinct neuronal clu st ers with even t-aligne d activity profiles 178 across the differ ent subst rat es. A small cl uster 1 e xhibit ed an int ermix ed resp onse patt ern, with most 179 neurons r esponding robus tly to wat er an d some to softwood. Sinc e the mice we r e water-d eprived for 180 four hours, wat er likely acted as a rewar ding stimulus in this conte xt, though clus ter 1 was distinct from 181 the flavored milks. Cluste r 2 was strongly activated by softwood, while Clust er 3 r esponded 182 prefer entially t o sponge and S tyrofoam. Cluster 4 neuro ns were sel ectively engag ed during licking of all 183 palatab le liquids r egardless of flavor (Fig. 2, H and I; fig. S3, A and B). No sp ecific cluster was 184 prefer entially t uned t o quinine , and all li quids evoked significantly higher activiti e s during licking 185 compared t o non-licking periods (fig. S3, C to G). This suggests that t he act of licking itself contribu ted t o 186 CeA Isl1 activation, even for less pala table or aversive liquids. Licking metrics reve al ed no significant 187 differences in lick frequency or dura tion across liquid types (fig. S3, H and J). Toge ther , thes e findings 188 suggest that Ce A Isl1 neurons includ e most ly material-specific and some rewa rd-sen sitive subpopula tions, 189 while also integr ating moto r-rela ted sign als. Furth ermo re, t he par tial ove rlap be t ween softwood-190 responsive and wat er-r esponsive neuron s su ggests that some CeA Isl1 neurons may integra te senso ry and 191 hedonic cues during consummat ory beh avior. 192 The activity of CeA neu rons is positively correlated with the hard ness of the bitt en obj ect 193 To examine if CeA Isl1 neur al respons es were cor rela ted with bi ting force, we com pared t he calcium 194 activities of CeA Isl1 neuro ns when the mic e were biti ng the same objec t (silicone), but of different 195 firmness. Population h eatm aps reveal ed that a frac tion of neur ons was activate d more stro ngly when 196 biting hard compar ed to soft silicon e (Fig. 2, J and K), consisten t with more r obust engagement o r 197 sensory feedback. Ha rd silicone elici ted h igher mean activity acr oss the en tir e pop ulation and gr ea ter 198 peak respons es of up-modulated n euro n s than soft silicone (Fig. 2, L and M). Conversely, hard silicone 199 elicited mor e pron ounced suppr ession of down-modulated ne urons r elative to ba seline tha n soft 200 silicone (Fig. 2N). Non-modula ted ne uro ns remained flat across conditio ns (Fig. 2O). These resul ts 201 suggest that h arde r mate rials drive st ron ger and more dive rse neu ral resp onses, poten tially reflec ting 202 great er senso ry input, incr eased bi ting force, or longe r engageme nt dur ations . This finding supports the 203 interp re tati on tha t CeA Isl1 neuro ns encod e the mecha nosenso ry featur es of biting in a materi al-specific 204 manner, in tegra ting both s ensory feedba ck and stimulus identity . 205 CeA Isl1 ne urons are sufficient to inc rease bite force and frequen cy 206 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 5, 2026. ; https://doi.org/10.64898/2026.02.03.703447doi: bioRxiv preprint 6 Given tha t CeA Isl1 neur onal activity cor rel ated with the ha rdness of the bi tt en obje ct, we hypothesiz ed 207 that these n eurons may also scale with b ite force . If so, artificial ac tivation may e nhance bit e str ength 208 compared t o no or weak stimula tion . To test this, we optog ene tically activat ed th ese neur ons in Isl1-209 CreER mice bilate rally injected with a Cr e -dependen t AAV e xpr essing ChR2 in the CeA and induced Cre 210 expr ession by Tamoxifen (Fig. 3A). Mice were 4-hour food dep rived and t hen pr e sented with thr ee 2-cm 211 linguine pieces (a ribbon-shap ed thick an d relatively ha rd pasta tha t requi res subs tanti al bite forc e to 212 break) per session, wit h or without 473 nm light stimu lation acr oss test ing days (Fig. 3B). Optogen etic 213 stimulation of CeA Isl1 neurons significantl y increased th e frequ ency of biting bouts compared to no n-214 stimulate d sessions, but o nly at 20 Hz (Fig. 3C), suggesting a frequency-dependen t effect on moto r 215 behavior. Con trol mice e xpr essing a Cre-depend ent EGFP virus did not show th is effect. Despit e the 216 increased bo ut numbe r, th e dura tion of i ndividual feeding bouts was significantly shorte r in ChR2-217 expr essing mice compared t o EGFP contr ols during 20 Hz light stimulatio n (Fig. 3D), suggesting less 218 controll ed or mor e forceful biting. Consis tent wit h this, th e to tal amoun t of linguine consumption was 219 similar betwe en groups (Fig. 3E), but ChR2 mice left more fragment ed pieces sca tt ered on the floor , 220 suggesting that bi tes may have bee n ove rly forceful or imprecise. Thes e resul ts suggest that Ce A Isl1 221 activation e nhances t he vigor of orofacia l movements but may disrupt fine con tro l of bite force, lea ding 222 to inefficient food eng agement wit hout i ncreased consump tion. 223 To further de te rmine whet her CeA Isl1 ne urons were sufficient t o drive biting b eh avior, we prese nte d fed 224 mice with food or Styrofoam blocks (Fig. 3F). Blue light (473 nm) was delivered in an alte rnat ing epoch 225 design consisting of thre e cycles of 1-min ute OFF a nd 1-minute O N peri ods. Photo activation of CeA Isl1 226 neurons significantly incre ased th e time and frequency tha t mice spen t biting foo d (Fig. 3, G to I) or 227 Styrofoam (Fig. 3, L to N), compared t o b oth OFF e pochs and cont rol mice. The sa me effect was 228 observed in mice afte r 4 hours of food d eprivatio n, suggesting tha t th e biting beh avior was not much 229 influenced by inter nal sta te (fig. S4, A to C). In addition to o bject-direc ted bi ting, photoac tivation of 230 CeA Isl1 neurons also induc ed robus t fictive feeding behavior in the abs ence of edi ble (food) (Fig. 3, J and 231 K, fig. S4, D and E) or inedible (Styrofoam ) stimuli (Fig. 3, O and P). Fictive feeding behaviors wer e 232 characte rized by fore paw-to-mouth mov ements r esembling consumption . These behaviors wer e 233 spatially biased : fictive feeding was significantly more likely to occur a t grea te r distances from th e object 234 compared t o object-dir ected bi ting (Fig. 3, Q and R). These observa tions suggest t hat activa tion of CeA Isl1 235 neurons is sufficient to engage th e core motor prog rams underlying feeding be ha vior, even in the 236 absence of ex te rnal senso ry cues or inge stible mat erial . 237 To test whe ther CeA Isl1 ne urons influence food consumption, we measur ed food in take in fed mice using 238 both opt ogene tic and chemogen etic ac ti vation. Fo r optog enetics, mice r eceived b lue light stimula tion in 239 alter nating 20-minut e ON/OFF epochs, a nd food intake was averag ed across cond itions (fig. S4, F and G). 240 For chemogene tic activa tion, CNO was a dminister ed 30 minutes befor e t esting, a nd food intake was 241 measured a t 0.5, 1, 2, an d 3 hours (fig. S4, H and I). In bot h cases, activa tion of CeA Isl1 neurons did no t 242 significantly alter t otal food in take in sa t ed animals. 243 Our calcium imaging expe riments r eveal ed tha t CeA Isl1 neurons we re str ongly acti vated during gnawing 244 on crickets. To t est whet her this activity patt ern r eflecte d a causal rol e in preda to ry behavior, we 245 performed a cricke t hunti ng assay in mice food-deprived for 4 hours to incre ase motivation . Mice wer e 246 presen ted with thre e live crickets, and bl ue light (473 nm) was delivered in an alt erna ting 1-minute OFF 247 / 1-minute ON design for thr ee cycles. C onsisten t with our imaging dat a, opt oge netic activa tion did no t 248 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 5, 2026. ; https://doi.org/10.64898/2026.02.03.703447doi: bioRxiv preprint 7 alter the time spen t investigating o r purs uing crickets (fig. S4, J and K), but signific antly increas ed the 249 time spent ca pturing pr ey and engaging i n fictive feeding in the absenc e of cricket s (fig. S4, L and M). 250 Similarly, in a novel object e xplor ation ta sk, CeA Isl1 neurons stimula tion did no t increase the tim e spent 251 near or fr equency of visits to the n ovel o bject (fig. S4, P to R), indicating that activ ation do es not 252 enhance mo tivation for e xplo rati on or n ovelty-seeking behavior . No tably, th e ti me spent ea ting and th e 253 total numb er of crickets killed did no t dif fer betwe en ChR2 and contr ol (EGFP) mi ce (fig. S4, N and O), 254 suggesting that ph otoac tivation of CeA Isl1 neurons promo tes bi te-rel at ed behavior s without enha ncing 255 overall pre dat ory efficiency. 256 CeA Isl1 ne urons are nec essary for efficie nt biting of solid food 257 Having established that activatio n of CeA Isl1 neurons enhanc ed vigor of orofacial movements during 258 linguine consumption, we n ex t asked whethe r thes e neur ons were n ecessary for execu ting efficient 259 biting. To tes t this, we op togen etically in hibited CeA Isl1 neurons u sing archa erhod opsin (Arch) while mice 260 engaged in a solid food consumptio n assay (Fig. 3S). We first challenged th e mice with uncooked 261 linguine. During light-induced inhibi tion ( 532RiUDnm), mice showed significantly prolonged feeding bou ts 262 compared t o Arch-OFF an d EGFP control s (Fig. 3, T and U), while the frequency of feeding bouts 263 remained unch anged (fig. S5A). This sugg ested that Ce A Isl1 neuronal activity was not requi red to initia te 264 feeding per se, bu t was critical for sust ai ning efficient oromot or ou tput . No tably, total consump tion 265 remained unch anged (fig. S5, B and C), indicating a disrupti on in ex ecutio n rat her than app eti te. To 266 dete rmine whet her this disrupti on dep e nded on food hard ness, we ne xt used a more fragile subst rat e of 267 similar tast e—uncook ed spaghe tti with a smaller diamet er (1 mm) ( fig. S5D). In this condition, CeA Isl1 268 neuron inhibi tion in fast ed mice led to a partial dis rupti on in feeding efficiency, with significantly longer 269 consumption times obs erved for t he sec ond pasta pi ece in Arch-O N mice (fig. S5, E to I). These resul ts 270 suggested tha t CeA Isl1 neur ons were n ece ssary for efficient biting behavior which became most obvious 271 when mice processed r esistan t food item s or when they re ached sati ety. W hen m ice were t ested with 272 softer food, such as stan dard chow pell et s, no impairment was obse rved (fig. S5, J to O). 273 To examine how CeA Isl1 neu ron activity in fluences orofacial mot or out put, we r eco rded 274 electr omyographic (EMG) activity from two jaw closing mus cles, masseter an d te mporalis muscles using 275 flexible Myoma trix a rrays 32 implanted in awake, freely moving mice (Fig. 3V). In Isl1-CreER mice 276 expr essing the inhibi tory DREADD hM4Di in CeA Isl1 neurons, food-biting be havior genera ted r obust EM G 277 signals under saline condi tions (Fig. 3W). Chemogenetic inhibi tion of CeA Isl1 neuro ns with CNO markedly 278 reduced the ampli tude of EMG bu rsts in both muscles (Fig. 3, W to Y), indicating t hat CeA Isl1 neur ons 279 activity is requir ed for normal r ecruitm e nt of jaw-closing motor units during feed ing. 280 We ne xt t este d wheth er th e contrib ution of CeA Isl1 neuron activity to jaw-muscle recruitmen t dep ended 281 on the mechanic al prop erti es of the obje ct being bitt en. Chemogen etic inhibi tion of CeA Isl1 neurons 282 markedly reduced masse te r and t empor alis EMG amplitud es during biting of har der substr at es, 283 including softwood and Styrofoam (fig. S5, P to S). By contrast, the effect was at te nuated for soft er 284 materi als: soft silicone pr oduced only pa rtial r eductions (fig. S5, T and U), and spo nge biting showed 285 minimal or no change (fig. S5, V and W). These resul ts indicat e tha t CeA Isl1 neuron activity is 286 prefer entially r equir ed for gener ating hig h-force jaw closure, with weaker eng age ment during low-effort 287 biting of compliant mat erials . 288 Activation of CeA Isl1 neur ons supports re inforcem ent and r eward-associated lear ning 289 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 5, 2026. ; https://doi.org/10.64898/2026.02.03.703447doi: bioRxiv preprint 8 Neuro ns in the CeA h ave previously been shown to be activat ed by both ave rsive and appe titive 290 stimuli 11,33 , raising th e questi on of whet her CeA Isl1 neur ons contribu te to th e posi tive or nega tive 291 motivational value of exp erienc e. W e u sed real-t ime place pr efer ence (RTPP) to test wheth er CeA Isl1 292 neurons suppo rt ed positive r einforceme nt. During tes ting, ChR2-expr essing mice showed a strong 293 prefer ence for th e light-pair ed compar t ment, which revers ed when stimula tion contingencies wer e 294 switched. Contr ol mice showed no such prefer ence, indica ting tha t CeA Isl1 neuro n activation is 295 intrinsically reinfo rcing (Fig. 4, A and B). To further t est t he re inforcing prope rti es of CeA Isl1 neurons, we 296 conducted an in tracr anial self-stimula tio n (ICSS) assay, in which mice nose-poked for bilater al CeA Isl1 297 neuron pho tos timulatio n. ChR2 mice rap idly developed rob ust ICSS behavio r, whereas EG FP controls did 298 not (Fig. 4C), confirming that CeA Isl1 neur onal activati on is sufficiently rewarding t o support op eran t 299 reinforceme nt. 300 Natural rewa rds not only r einforce inst rumental beh avior bu t als o su ppor t Pavlovian learning, whe re 301 predictive cues acqui re incen tive value a nd promot e appro ach even in th e absenc e of the rewa rd 34 . We 302 next asked wheth er CeA Isl1 neur on activa tion could drive condi tione d place pr efer ence (CPP ). In this 303 assay, ChR2-expressing mice develop ed a significant prefer ence for th e light-pair ed conte xt , and 304 similarly, chemogene tic activatio n of CeA Isl1 neurons led to a significant p refer ence for the CNO-pair ed 305 contex t, while con trol mice did no t show a prefere nce in eit her condi tion (Fig. 4, D and F), suggesting 306 that CeA Isl1 neuron act ivation is su fficient for forming reward–con tex t associa tions . Finally, to assess 307 whether Ce A Isl1 neurons enc oded t he pr e dictive value of cues during reward l earni ng, we performed in 308 vivo calcium imaging during a Pavlovian c onditioning t ask. Mice wer e tr ained t o as sociate on e cue (CS⁺) 309 with food delivery and ano the r (CS⁻) with no outcome (Fig. 4G). Be haviorally, ani mals learned the 310 association, sh owing a progressive red uction in lat ency to re tri eve the r eward acr oss training days (Fig. 311 4H). However, CeA Isl1 neuro nal activity di d not differ betw een CS⁺ and CS⁻ trials, e ven after le arning (Fig. 312 4I), suggesting that while these n eurons can drive reinforcem ent an d suppor t associative lea rning, th ey 313 do not encod e cue-specific predict ive value during Pavlovian conditioning. 314 Distinct CeA–brainstem pathways drive biting behavior 315 To begin addressing th e circuit mechanis m by which CeA Isl1 neurons contr ol biting behavior, we 316 anatomically mappe d the lo ng-range out puts of CeA Isl1 neurons. We found th at th ey projected to 317 hindbrain t arge ts including the PCRt, PB N, PPTg, nucleus of the solitary tr act (NT S), microcellular 318 tegment al nucleus (MiTg), and supratrig eminal nucleus (Su5) (Fig. 5, A and B). These ta rget r egions have 319 all be shown to be di rectly or indi rectly i nvolved in motor cont rol. The PCRt is involved in coordina tion of 320 orofacial movements a nd auto nomic functions 14,35 . The PBN is a critical meal-ter mination cen ter that 321 integra tes visceral an d nociceptive signal s to suppress feeding and p romot e aversi on 36,37 . The P PTg 322 regulat es locomotio n, playing a key role i n movement initia tion an d mainte nance 38 , 3 9 . The NTS receives 323 visceral afferents from th e vagus nerve a nd drives reflexive o r pat te rned mot or r e sponses, such as 324 swallowing, gagging, or vomiting 39, 40 . The MiTg is likely a premotor brains tem nucl eus involved in 325 coordina ting specific motor pat te rns 41 and the Su5 provides di rect pr emot or cont r ol of orofacial motor 326 neurons 42,43 . 327 In additi on, we found tha t CeA Isl1 neur on s projected t o oth er foreb rain and midb r ain targe ts including 328 the subth alamic nucleus (STh), substanti a nigra pars lat eral is (SNL) and pars compacta (SNC), the 329 periped uncular nucle us (PP ), the bed nuc leus of the stri a te rminalis (BNST), the r e ticular nucleus of th e 330 thalamus (Rt), the p oste rior triangula r th alamic nucleus (PoT), the ventral post ero later al (VPL), posterior 331 intralamin ar (PIL), and ventral post erom edial (VPM), thalamic nuclei (Fig. 5, A and B). These regions ar e 332 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 5, 2026. ; https://doi.org/10.64898/2026.02.03.703447doi: bioRxiv preprint 9 known to be involved in motor r egulat io n (STh, SNL, SNC) 44 ,45 , reward and arousal (BNST, SNC) 46 , 333 sensorimot or int egrati on (PP, PoT, V PL, VPM, PIL) 47 , and sensory gating or att enti onal filtering (Rt) 48,49 . 334 Together , thes e projec tion pa tte rns are c onsisten t with a model th at CeA Isl1 neuro ns coordinat e orofacial 335 motor prog rams and visceral refl exes wit h broade r behavior al sta te and se nsorim otor con te xt. 336 To dissect the downs tream circui ts by which CeA Isl1 neurons regula te bi ting beha vior, we examin ed two 337 major brainst em projectio ns: th e CeA Isl1 → PCRt and the CeA Isl1 → PPTg pathways . To investigate the 338 CeA Isl1 → PCRt projection, we injecte d A AV-DIO-ChR2-EYFP into the CeA of Isl1-CreER mice and 339 implanted op tic fibers above the PCRt (Fig. 5C). Mice were presen ted with five no n-edible Styrofo am 340 objects tha t eleva ted calcium signals in t he majority of CeA Isl1 neurons (Fig. 1). Op togene tic stimula tion 341 of CeA Isl1 axon termin als in the PCRt signi ficantly increased bo th th e dura tion and frequency of biting 342 behavior dir ected toward the Styr ofoam (Fig. 5, D and E, fig. S6A). Similar increase s in duration an d 343 frequency of biting behavior we re obse r ved towards pieces of norm al chow (Fig. 5, D and F, fig. S6B). 344 We also found rob ust fictive feeding beh avior induced by photo activati on of CeA Isl1 → PCRt projection in 345 both cases (Fig. 5G). These r esults d emo nstra te th at CeA Isl1 → PCRt project ions ar e sufficient to drive 346 object-direct ed or ofacial actions, sugges t ing a functional role for this pathway in p romoting biting-347 relat ed moto r progr ams. 348 To assess whether CeA Isl1 → PCRt stimul ation car ries motiva tional valenc e, we pe rformed RTPP testing 349 (Fig. 5, H to K). Mice did not develop p ref erence for the pho tos timulatio n-paired c hamber (Fig. 5, H and 350 I), indicating tha t CeA Isl1 → PCRt activati on lacks positive reinforc ement . Locomo tor activity an alysis 351 revealed n o significant changes in velocit y or total dis tance t ravele d (Fig. 5, J and K), demonstrating that 352 the beh avioral effects wer e not due to g eneral arou sal o r hyperac tivity. Togeth er , these d ata id entify 353 the CeA Isl1 → PCRt pathway as a moto r-promoting circuit that ex ecutes t he act of biting without 354 encoding motivatio nal value. 355 Stimulatio n of CeA Isl1→ PPTg projections (Fig. 5L) also produced robus t increas es i n biting toward bo th 356 inedible and edible objec ts, as well as fictive feeding, similar t o the Ce A Isl1 → PCRt pathway (Fig. 5, M to 357 P, fig. S6, C and D ). However, in the RTPP assay, the CeA Isl1 → PPTg pathway was p ositively reinforcing, as 358 mice spent significantly more tim e in the light-paired compar tmen t (Fig. 5, Q and R). These mice also 359 exhibi ted dec reas ed velocity but unchan ged tot al movement dist ance (Fig. 5, S a nd T), suggesting 360 enhanced a pproach mo tivation . Togeth e r, thes e findings reveal th at CeA Isl1 → PCRt and CeA Isl1 → PPTg 361 projections cons titute distinc t componen ts of an emotion al–mo tor ne twork: t he CeA Isl1 → PCRt 362 projection p rovides a descend ing command for bite e xecut ion, while th e CeA Isl1 → PPTg projection 363 energiz es and reinfo rces biting beh avior through posi tive motivati onal drive . 364 Nex t, we aimed a t iden tifying the monos ynaptic inputs to Ce A Isl1 neurons with a f ocus on the CeA Isl1 365 subpopulati on tha t projec ts to th e PCRt. A Flp-depend ent r abies glycoprot ein (RVG) and TVA recept or 366 were co-injected in to th e CeA, while Cre- depend ent Flp r ecombinase was inject ed into the PCRt and 367 delivered r et rograd ely via rAAV2-ret ro in to the Ce A and ot her incoming brai n are as (fig. S6E). This 368 strat egy ensured that o nly CeA Isl1 neuron s projecting to th e PCRt would expr ess th e necessary 369 components for r abies virus infection a n d trans-synaptic spr ead. Foll owing rabies virus injection into th e 370 CeA, we found the st art c ells mainly located in th e ant erio r par t of CeM (fig. S6F). Input ne urons wer e 371 identified throughou t th e brain . Quan tita tive analysis reveal ed substan tial inpu t fr om several foreb rain 372 and brainst em regions . No tably, th e stro ngest sources of input o riginat ed from th e late ral par t of the 373 central amygdala , the b asolat eral amygdala (BLA), latera l hypothal amus (LH), insu lar cort ex, a nd several 374 regions of the mot or and se nsory cort ex (fig. S6G). These results suggest t hat CeA Is l 1 → PCRt neurons 375 integra te divers e motivati onal, sens ory, and motor-r elat ed signals to r egulat e bra instem moto r outpu t. 376 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 5, 2026. ; https://doi.org/10.64898/2026.02.03.703447doi: bioRxiv preprint 10 Building on thes e circuit- and beh avior-le vel findings, we next e xamined whe the r CeA Isl1 neurons wer e 377 conserved across speci es. Transcrip tional analyses reveal ed tha t Isl1 + ne urons wer e presen t in th e CeA 378 or CeA-like regions of human 50,51 , mouse 52 , chicken 51 , lizard 53 , and even salamand er 54 (fig. S7, A to E). We 379 compared Isl1 with Prkcd , a can onical marker of CeA neurons tha t media te ave rsi ve and defensive 380 behaviors, bec ause this con tras t highlights functional diversi ty within the CeA . In both mouse and 381 human, Pkcδ + and Isl1 + ne urons wer e lar gely non-overlapping (fig. S7, C to E, H to J), indicating tha t they 382 repr esent dis tinct subtyp es. In liz ard, onl y minimal overlap was observed (fig. S7, A and F), whereas in 383 chicken approxima tely 50% of PRKCD + cel ls also expr essed ISL1 (fig. S7, B and G), suggesting a less 384 specialized CeA o rganiza tion. In salaman der, Prkcd e xpressi on was absent , and o nly Isl1 + cells were 385 detec ted in th e CeA-like regi on (fig. S7, K to M). Thus, Isl1 emerges as a highly con served molecula r 386 signature of CeA-like ne urons across ver t ebra tes (fig. S7N), underscoring its funda mental rol e in th e 387 evolutiona ry blueprin t of this struc ture . 388

389 Our resul ts uncover a pr eviously unrecog nized function of CeA neu rons in encodin g the physical 390 proper ties of a bit ten obj ect, with ne uro nal activity scaling propor tionally t o the object’s hardn ess. 391 Specifically, we show that stimula tion of CeA Isl1 neurons no t only triggers biti ng be havior but also 392 disrupts th e fine contr ol of bite force , whereas inhibi tion of th ese neu rons impairs bite efficiency. These 393 findings indicate tha t CeA Isl1 neur ons hel p calibrat e jaw motor out put t o match th e physical demands of 394 different objec ts, ensuring t hat forc eful, adaptive bi ting is appropri at ely scaled to substra te hard ness. 395 Our calcium imaging data r evealed that e dible and non-edi ble food it ems with differen t physical 396 proper ties (densi ty, viscosity) activated t he majority of CeA Isl1 neurons t o varying degrees . Depending on 397 the it em, CeA Isl1 neuro ns were r ecruit ed i nto differen t ensembl es and th e hard er t he object, t he 398 stronger the ac tivation , suggesting tha t t hese neu rons encod e the range of physical prope rties of food 399 items. In con tras t, th e sensory a ttrib utes (taste, smell) of food had minor effects o n CeA Isl1 neurons, with 400 most flavors eliciting similar activa tion p att erns. Thes e charac teris tics differenti at e CeA Isl1 neurons well 401 from other CeA appe titive ne urons. Ce M Htr2a neurons were previously shown to d rive the consumpti on 402 of food and water and CeM Ss t neuro ns to drive water consump tion only 12 . Both po pulations incr ease 403 their ac tivities duri ng consumption of re warding substances. When switching be t ween rewards of 404 different physical at tribu tes (e.g. soli d food and liquid Fresub in), the ac tivity of CeM Htr2a neurons often 405 remained s table , suggesting tha t th ese n eurons pay lit tle r egard t o the physical a t tribut es of food. 406 Mechanistical ly, these two popula tions p rolong food or wat er consumptio n by conveying stimulus-407 specific signals to downstream r eward ce nters, rath er th an promo ting biting or lic king directly 12,33 . 408 Calcium imaging further reveal ed tha t Ce A Isl1 neurons ar e prefe ren tially engaged d uring gnawing but 409 show minimal responses during subsequ ent chewing, highlighting a function al dis tinction in how th e 410 CeA modulat es orofacial beh aviors. This specificity suggests that CeA circui ts facilitat e motor p rograms 411 that d emand forceful, pr ecise int erac tion s with objects, such as biting and gnawin g, rathe r than the 412 automa tic, rhythmic mastica tion govern ed by local brainst em centr al pat ter n ge nera tors. Inde ed, 413 previous work has shown that dis tinct pr emotor circui ts within th e supra trigemin al nucleus and 414 trigeminal mot or system differen tially co ordinat e gnawing and chewing movements, with gnawing 415 requiring highe r jaw-closing forces and incisors, whereas chewing prima rily involving rhythmic molar 416 action 55 42,56 . This i s in line with older wor k demonstra ting tha t as the h ardness of food increas es, both 417 the numbe r of chewing cycles and the le vel of jaw-closing muscle EMG activity also increase, bu t th at 418 EMG activity progr essively decreas es dur ing the sequenc e as the food is br oken d own 57,58 . This suggests 419 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 5, 2026. ; https://doi.org/10.64898/2026.02.03.703447doi: bioRxiv preprint 11 that forc eful, adap tive bite con trol is esp ecially critical during th e initi al inte ractio n with hard food items, 420 the phas e during which CeA Isl1 neurons a re most active in ou r study. O nce th e bol us is broken down, the 421 demands for precise forc e modulati on di minish and chewing proceeds th rough well-defined rhythmic 422 circuits in the br ainst em 56 . Together , th e se observati ons suppor t a model in which CeA Isl1 neurons 423 couple sensory evalua tion of substa nce p roper ties t o the ini tia tion and calib rati on of bite force, linking 424 motivational s tat e to skilled o rofacial mo tor outpu t . 425 Our pho toactiva tion e xpe riments showe d that CeA Isl1 neurons ar e su fficient to inc rease bi te force an d 426 frequency of biting, withou t increasing f ood consumption . The activity of CeA Isl1 neurons is not much 427 influenced by the hunge r sta te of the ani mals, in sharp contr ast t o CeM Htr2a neuro ns which are activate d 428 by the hunger ho rmone ghre lin 13 . Biting activity towards edi ble and non- edible i t ems was instead 429 spatially biased : object-dir ected bi ting oc curred a t a distanc e of appro x. 2 cm from the mouth, whe reas 430 fictive feeding behavior was favored a t la rger distanc es. The circuit p rope rti es und erlying this spatial 431 bias are curr ently uncle ar and will be an i nter esting futur e res earch focus. Photoi n hibition e xpe riments 432 showed tha t the ac tivity of CeA Isl1 neuro ns is necessary for efficient biting of solid food that r equir es 433 strong bit e force, consist ent with a role i n scaling motor outpu t to ma tch physical demands. 434 Our circuit an alysis reveale d tha t CeA Isl1 neurons r egulat e biting beh avior thr oug h two paralle l outpu t 435 pathways tha t differenti ally contribute t o motor e xecution and mot ivational rein forcement . Activati on 436 of the CeA Isl1 → PCRt projection robustly drove biting, indica ting direc t contr ol of orofacial moto r circuits. 437 In contr ast, stimul ation of th e CeA Isl1 → PPTg pathway not only elicited biting but also produ ced a 438 reinforcing effect, suggesting r ecruitm en t of reward-rel at ed mechanisms. The PPTg is a key 439 mesencephalic hub that modu lat es dopa minergic neurons in t he substa ntia nigr a and ventral tegmen tal 440 area . Previous work has shown that PPTg stimulation evokes s tria tal dopamin e rel ease via cholinergic 441 and glutamat ergic tr ansmission 59-61 , supporting th e idea that the CeA Isl1 → PPTg projection may link 442 emotion al or senso rimoto r signals from the amygdala to midbrai n reward syst ems. CeA Isl1 → PPTg 443 stimulation a lso reduc ed locomot or velo city despite its reinforcing prop er ties. Thi s reduction likely 444 reflects a beh avioral shift from e xplor ati on to focused consummato ry engageme nt—a hallma rk of 445 reward an ticipati on and goal-dir ected ap proach sta tes— rath er t han moto r suppr ession. Conversely, t he 446 CeA Isl1 → PCRt pathway likely mediates t h e direct mo tor compon ent of biting thro ugh reticula r premo tor 447 circuits controll ing jaw and tongue move ments. Toge ther , thes e resul ts uncover a dual outpu t 448 architec tur e in which the CeA orch estr at es both th e kinematics and mo tivatio n of orofacial actions . This 449 organiza tion provides a n eural subs tra te through which emoti onal sta tes can dyn amically transform 450 simple motor pa tt erns—like bi ting—in to goal-direct ed and affectively charged b e haviors. 451 Underst anding th e afferen t inputs t o dist inct CeA cell popula tions is essen tial for deciphering how 452 different ne uron typ es integr ate s ensory, emotional , and motiva tional informa tio n to orches tra te 453 behavior. F or e xample, Ce A Htr2a neurons, which are linked to h edonic feeding, r ece ive strong input from 454 homeosta tic and r eward-rel ate d regions such as the arcua te nucle us (Arc), parasu bthalamic nucleus 455 (PSTN), substantia nigra pars la ter alis (SN L), and dorsal raphe nucl eus (DR) 8 . In co ntras t, CeA Isl1 neuro ns 456 show minimal input from these classic fe eding and neur omodula tory cent ers but i nstead r eceive rob ust 457 inputs from limbic and associative ar eas i ncluding the basola te ral amygdala (BLA), medial amygdala 458 (MeA), ventral pall idum (VP), piriform co rte x (Pir), and caudate pu tamen (CPu). T his patt ern suggests 459 that CeA Isl1 ne urons may play a broad er r ole in integr ating valenc e, social, and mo tivational signals . 460 Notab ly, the st rong input from t he Me A, which is absent for CeA Htr2a neurons, posi tions CeA Isl1 neuro ns 461 to process social o r pher omonal cues th a t could shape be haviors such as differen t iating aggressive bit es 462 from parent al pup re tri eval. Similarly, su bstanti al input from th e int ersti tial nucle us of the poste rior limb 463 of the ant erio r commissure (IPAC) implies that CeA Isl1 neurons may integra te in ter nal stat e signals like 464 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 5, 2026. ; https://doi.org/10.64898/2026.02.03.703447doi: bioRxiv preprint 12 stress and r eward 62-64 to flexibly adjust b ehavioral ou tputs , for ex ample, modula ti ng feeding or hunting 465 in response to thr ea t. Fut ure work combi ning precise circuit mapping with b ehavi oral tes ting will be 466 crucial to reve al how thes e inputs shap e specific motivated beh aviors. 467 The ext ensive projec tions of CeA Isl1 neur ons to multiple b rainst em and foreb rain regions unde rscore t he 468 CeA’s role as a hub tha t dynamically inte grates motiva tional a nd sensory signals with premot or circuits . 469 For ex ample, pr ojections t o brains tem ar eas such as the PBN, NTS, and Su5 likely facilitat e th e 470 coordina tion of orofacial mo tor prog ram s with visceral feedback 36,42, 65 . Inputs to t he PPTg and substantia 471 nigra may further modula te a rousal, l oco motor drive , and reinfo rcemen t lear ning, there by linking 472 emotion al or motiva tional st ates to acti o n selection an d behavior al vigor 66,67 . Proj ections to the B NST 473 and PP could contribute to sustain ed affective, social, an d defensive sta tes by int egrating multimod al 474 sensory and motivati onal informa tion t o modulate a nxie ty-like behaviors, ma te rn al responses , social 475 inter action, and cont ex t-depend ent thre at or ar ousal re actions 68,69 . Meanwhile, c onnections to multipl e 476 thalamic rel ay nuclei, including somatos e nsory (VP L, VPM) and polymodal (PIL, Po T) regions, raise th e 477 possibility tha t the CeA ex erts top-down modulation ove r sensory gain, biasing pe rception in a s tat e-478 depend ent mann er to p romot e or suppr ess specific behavioral re per toir es 70,71 . Futu re work will be 479 neede d to dissect h ow specific CeA Isl1 subpopu lati ons differen tially engage th ese downstream targe ts to 480 coordina te discre te asp ects of orofacial a nd locomotor b ehaviors. More over, func tional e xpe riments 481 targe ting the in ter actions b etwe en CeA outputs to moto r nuclei and asce nding sensory relays could 482 elucidat e how motivation al circuits over r ide or resh ape senso ry processing to sup port goal-dir ecte d 483 actions. This distri but ed outpu t archi tect ure may be especially r elevant for un der standing how 484 emotion al stat es contr ibut e to malad apti ve conditions such as compulsive gnawing, pica, or str ess-485 induced sleep b ruxism. 486

487 We thank Sofia delgado for help with managemen t of the animal colony; Helena Weltzien for help with 488 behavior exp erimen t; Yu e Zhang (Depart ment of Synaps es – Circui ts – Plas ticity, Max-Planck Insti tut e fo r 489 Biological Intellig ence) for help with data analysis; Karl-Klaus Conzelmann (Gene Center Munich, LMU) 490 for providing EnvA G-del et ed rabi es virus. The Myomat rix Arr ay u sage r epor ted in the publicatio n was 491 support ed by th e NIH B RAI N I nitia tive u nder awa rd numbe rs N IH U24 NS126936 and N IH R01NS109237 . 492 A.P.A . was support ed by FAPESP fellowship no. 2023/02896-3. This study was support ed by the M ax-493 Planck Society and the Europ ean Res ear ch Council under the Europ ean Unio n’s Horizon 2020 rese arc h 494 and innovation p rogramme (no. 885192, BrainR edesign). 495 496 Author co ntributions 497 WD and RK conceptualiz ed and designe d the study. WD conduct ed th e exp erim ents and an alyzed dat a . 498 AP assisted with surgery and be havior e xperim ents. WK and CQ assisted with da ta analysis. WD and RK 499 wrote th e pap er with inpu t from all auth ors. R.K. sup ervised and p rovided fundin g. 500 501 Declaration of interests 502 The autho rs declar e no competi ng inter e sts. Supplem enta ry Informati on is availa ble for this pape r. 503 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 5, 2026. ; https://doi.org/10.64898/2026.02.03.703447doi: bioRxiv preprint 13 Data Availability 504 The data supporting the curr ent study are available from the correspo nding author upon reaso nabl e 505 reques t. 506 507

508 Animals 509 Male and femal e mice tha t were a t leas t 2 months old were used in al l exp erimen ts, following 510 regulati ons from the governm ent of Upp er Bavaria . The mice were h oused in th ei r home cages with 2-5 511 mice per cage, und er a 12-hour light/12- hour dark cycle, with food and wat er fre ely available. The Isl1-512 CreER transgenic lin e (Isl1 tm1 (cre/Esr1*) Krc /S evJ) was purchased from Jackson Labora t ories. Prkcd-Cre (Tg 513 (Prkcd-gl c-1/ CFP, - Cre) EH124Gsat) BAC mice were import ed from the Mutan t M ouse Regional R esourc e 514 Center (ht tps://www.mmrrc.org/ ). Ai9 ( B6. Cg- Gt (ROSA) 26Sor tm9 (CAG- tdTomato ) Hze /J) mouse line was as 515 described pr eviously, Transgenic mice were br ed ont o a C57BL/6NRj background (Janvier Labs - 516 http://www.janvier-labs.com ). For op tog enetic an d chemogene tic manipula tions, animals were handl ed 517 and housed singly on a 12-hour invert ed light cycle for at least 3 days prio r to t he expe riments . Excep t 518 during food deprivati on for feeding e xpe riments, mice wer e given ad libitum food access. All behavior 519 tests wer e carri ed out a t th e same time each day during the d ark peri od (1 p.m.– 6 p.m.). 520 521 Viral vectors 522 The following AAV viruses were obta ined from Addgene: A AV9-pAAV-hSyn-DIO-hM3D (Gq)-m Cherry 523 (Addgene, 44361), pAAV-hSyn-DIO-hM4 D (Gi)-m Cherry (Addgene, 44362), AAV2-pAAV-hSyn-DIO-524 mCherry (Addgene, 50459), AAV5-Ef1a-D IO-ChR2-EYFP (Addgene, 35509), AAV5-E f1a-DIO-EGFP 525 (Addgene, 27056). AAV5-EF1a-DIO-eA rch3.0-EYFP were produced at the Gene Th e rapy Center Vec tor 526 Core at th e Universi ty of North Car olina Chapel Hill. pssAAV-2-hSyn1-dlox-jGCaM P8m(rev) -dlox-WPRE-527 SV40p(A) ( v628-1 ) was purchased from V iral Vector Facili ty (VVF) in Neuroscience Center Zurich (ZNZ) at 528 the Universi ty of Zurich and ETH Zurich. rAAV2/8-nEf1α-F DIO-RVG-WPRE-hGH polyA, rAAV2/8-nEF1α-529 fDIO-EGFP-T2A-TVA-WPRE -hGH polyA, a nd rAAV2 (Retr o)-EF1α-DIO-Flp-WPRE-h GH polyA were 530 produced by BrainVTA (Wuha n). EnvA G-delet ed rabi es for long-range monosyna ptic tracing was a gift 531 from Karl-Klaus Conzelmann (Gene Cen t er Munich, LMU). 532 533 Ster eot actic surgery 534 Mice were an esth etiz ed with 1.5-2% isoflurane and received o xygen at 1 .0 lite r pe r minute befo re being 535 placed in a ste reo taxic fram e (Kopf Instr uments). A hea ting pad was used t o keep the body temp era tur e 536 stable . Carprofen (5 mg/kg bodyweight) was used subcutaneo usly as an analgesic . 537 Once th e mouse skull was expose d, we d rilled a crani al window (1–2 mm 2 ) unilate rally (calcium imaging 538 expe riments , monosynaptic r abies and a nterog rade tracing e xperim ents) or bila t erally (optogen etic and 539 chemogene tic exp erimen ts). Ne xt, a glas s pipett e (#708707, BLAUBR AND intr aM ARK) was lowered into 540 the window to delive r 300 nl of viral vector to th e ar ea of inte rest (coordi nat es: CeA: −1 .22 mm anteri or 541 to bregma, ± 2.6 mm late ral from midline and, −4 .85 mm vertical from the br ain surface; PCRt: − 5.6 mm 542 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 5, 2026. ; https://doi.org/10.64898/2026.02.03.703447doi: bioRxiv preprint 14 ante rior t o bregma, ± 1.27 mm later al from midline and, − 4.7 mm vertical from t he brain surface ; PPTg: 543 − 4.7 mm anteri or to b regma, ± 1 .2 mm later al from midline and, − 3.5 mm vertic al from the brain 544 surface). In th e same surgery, mice used in optogen etic e xperim ents wer e bilat er ally implanted with 545 optic fibers (200 μm core, 0.39 NA, 1.25- mm ferrule (Thorlabs)) above the CeA (− 4.35 mm ventral), the 546 PCRt (− 4.2 mm ventral), or the PPTg (- 3.0 mm ventral). Implants wer e secur ed with cyanoacrylic glue, 547 and the e xpos ed skull was covered with dental ac rylic (Paladur). For all oth er mice, the incision was 548 closed with sutur es. 549 Isl1-CreER mice used for rabies tracing were first unil ate rally injected in the PCRt with rAAV2 (Retr o)-550 EF1α-DIO-Flp and in the CeA with th e sta rter rAAV2/8-nEf1α-FDIO-RVG and rAAV 2/8-nEF1α- fDIO-EGFP-551 T2A-TVA. After 4 weeks, th e same mice were injecte d with the rabies virus. S eve n days later, mice wer e 552 killed, and th eir brai n tissue was collect e d and processed for immunohis tochemis try. 553 554 GRI N lens implant ation and basepl at e fixation 555 Four weeks afte r GCaMP8m viral injectio n in the CeA, mice wer e implant ed with a gradient ind ex (GR IN) 556 lens. At the same co ordina tes of th e inje ction, a small cranio tomy was made and a 23G needl e was 557 slowly lowered into th e brain to clea r the path for th e lens to a d epth of -4.5 mm from bregma. Aft er 558 retr action of th e ne edle, a GRI N lens (ProView lens; diamet er, 0.6 mm; leng th, ~7 . 3 mm, Inscopix) was 559 slowly implanted above th e CeA and the n fixed to th e skull using UV light-curable glue (GRADIATM 560 DIRECT FLO) and iBOND® Universal (Kulzer). 4 weeks after GRI N lens implan tatio n , mice were 561 “baseplate d” under an esthesi a. Bri efly, in the ste reo ta xic setup, a basepla te (BPL-2; Inscopix) was 562 positione d above the GRI N lens , adjustin g the dist ance and the focal plane until t he neur ons wer e visible. 563 The basepla te was fixed using UV light-curable glue (GR ADIATM DIRECT FLO) C& B Meta bond (Parkell). A 564 basepla te cap (BCP-2, Inscopix) was left in place to pro tec t the l ens. 565 566 Tamoxifen pro tocol 567 Tamoxifen solutio n was prepa red in 90% corn oil and 10% ethan ol. The injecti on d ose was dete rmined 568 by weight (using appro ximat ely 200 mg t amoxifen/kg body weight) and was given for 2 consecu tive days. 569 For all AAV expr ession exp erimen ts, tam oxifen was administe red 1 day aft er s ter eotac tical A AV injectio n. 570 571 Pharmacological tr eatme nts 572 For chemogene tic ex perime nts, mice we re given an intr aperi ton eal (IP) injection of CNO (2 mg /kg for 573 hM3D (Gq) and 0.4 mg/k g for hM4D (Gi) diluted in salin e) or the equivalen t volume of saline before the 574 expe riment and were all owed to recover in their home cages for 30 minut es. 575 576 Optoge netic manipul atio ns 577 Mice were bila ter ally te ther ed to o ptic fiber patch co rds (Prizmatix) connect ed to a multi wavelength 578 LED (Prizmatix) via a rotary joint (Prizmat ix) and mating sleeve (Thorlabs). For ph o toactiva tion 579 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 5, 2026. ; https://doi.org/10.64898/2026.02.03.703447doi: bioRxiv preprint 15 expe riments , 5 ms, 473-nm light pulses at 20 Hz and 10-15 mW were used. Const ant 532-nm light at 18 580 mW was used for photoinhibi tion e xpe ri ments. The LED were trigge red, an d puls es were cont rolled with 581 EthoVision XT16 software (Noldus Infor mation Technologi es). 582 583 Feeding beh avior 584 Mice were habi tua ted t o the b ehavior al contex t for daily 10-min sessions, for 2 days before the 585 expe riment . For 4h-fast ed feeding t est, mice were food res trict ed for 4 hours. Mi ce were pres ent ed with 586 a regular food p ellet , and allowe d to fee d. The weight of the food p elle t, includin g the food debris lef t in 587 the cage floor aft er t est, was measured t o calcu late the food int ake. Fo r ‘fed’ fee d ing test, mice wer e 588 not food dep rived befor e the test . In th e optogen etic e xper iment , the ligh t was starte d just after the 589 mice were put in to th e tes ting cage for 20 minutes, th en the ligh t was off for 20 minutes. The food 590 intake was measur ed for both p eriods . F or the ph armacological e xpe riment , CNO were injected 20-30 591 minutes befor e the tes t. The food int ake was measured for 0.5 h, 1 h, 2 h and 3 h. All of the feeding t ests 592 were perfo rmed be tween 1 pm to 6 pm. 593 594 Optical in tracr anial self-stimula tion 595 Behaviour al trai ning and tes ting occurre d in mouse operan t chambe rs (21.6 cm W x 18.6 cm D x 12.7 cm 596 H, Med Associa tes Inc .) interfaced with o ptogene tic stimula tion equi pment . Prior to the firs t session, 597 mice were familiarize d with purified foo d pellets (20 mg, TestDiet), which is slightly sweet, in th eir home 598 cage for 2 cons ecutive days. Mic es wer e food res trictio n by tim e-limiting th e avail ability of food (1 to 2 h) 599 in the home cages (85 % to 90 % of their free-feeding bodyweight) to pr omot e ap parat us explo rati on. 600 Mice expl ored op era nt bo xes cont aining an active and inac tive nosepok e. On the pre-tr aining day, both 601 active and inactive p orts wer e bait ed wit h food trea ts to e ncourage explo rat ion. F or this study, 602 appeti tive stimuli would have impact ed l atency measur emen ts on the p re-tr ainin g day and were not 603 used. Training Sessio n length was 60 min, during which time mice were fre e to r espond at any nosep oke 604 port, no food tre ats wer e deliver ed. The i nactive nosepok e produc ed no resul t. Th e active nosep oke 605 delivered a 20-Hz t rain of 5 ms pulses of bluelight for 3 s, th e active nosep oke wa s accompanied by a cue 606 light above th e nosepoke and a ton e (3k Hz, 5 s), which were insufficient to prom ote nos epoking in the 607 absence of optical s timulati on. Testi ng occurred once p er day for 4 days and port /frequency assignment 608 was counterbal anced . Nosep oke activity was recorded with Noldus EthoVision XT software and visually 609 monitor ed via infrared came ra. 610 611 Real-time plac e prefe rence test 612 Mice with optic fiber p atch cables tet her ed were allow ed to e xplor e a two-compa rtmen t aren a (50 cm × 613 25 cm × 25 cm ). Mice were tes ted acr oss two sessions. In session on e, on e side was assigned as the 614 photost imulatio n chamber . Every time t he mouse en ter ed this chambe r, 5-ms, 473-nm light pulses at 20 615 Hz and 10-15 mW (measured at th e tip o f optic fibers) were delivere d intr acranial ly for activation 616 expe riment . Photos timulati on ceased wh en the mous e left th e phot ostimula tion s ide. In the second 617 session, we assigned th e oth er chambe r as the pho tostimula tion sid e and rep ea t ed tes ting. The 618 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 5, 2026. ; https://doi.org/10.64898/2026.02.03.703447doi: bioRxiv preprint 16 behavior of the mice was r ecorde d using a camera. E thovision XT16 software (Nol dus Information 619 Technologies) was used to deliver ligh t p ulses and analyze b ehavioral p arame te rs. 620 621 Conditioned pl ace prefe renc e tes t 622 A two-compartme nt CPP apparatus (35 cm L x 18cm W x 29 cm H, connected by a closable door 5 cm x 5 623 cm) was used for the CP P test. The two c ompartmen ts have distinc t wall pat terns and floor grids. Animal 624 location was tr acked, and the tim e spent in each box was assessed with Eth ovision XT 16 software 625 (Noldus). For opt ogene tic exp erimen ts: T he CPP test consisted of 4 days. On Day 1, each mouse was 626 allowed to e xplo re th e en tire ap para tus freely for 15 min (Pre_Test). Afte r th e pre -test th e initial 627 prefer ence of each mouse for a given sid e compartme nt was reco rded . With ou r appara tus design, most 628 of the mice showed an initi al prefe rence for one of the compa rtmen ts. On d ays 2 and 3, we placed th e 629 mice into the ir initi ally non-prefer red co mpartmen t and delive red 20-Hz light sti mulation for 20 min. 630 Appro ximat ely 4 h later, we pl aced th e mice into the othe r side of the comp art ment without light 631 stimulation . On day 4, we place d the mic e back into th e appar atus with bo th com partmen ts accessible 632 for 15 min (without light stimulation) an d the time the mouse sp ent in t he comp artmen t is calculat ed. 633 For chemogene tic ex perime nts: The CPP test consist ed of 6 days. On Day 1, each mouse was allowed to 634 explo re th e ent ire app ara tus freely for 1 5 min (Pre_Test). After the pr e-tes t the i nitial pr eferenc e of 635 each mouse for a given side compar tme nt was record ed. Wi th our a ppara tus des ign, most of the mice 636 showed an initial p refer ence for on e of t he compartm ents. Condi tioning was initi ated on d ay 2 and 637 encompassed four sessions p erformed o n four consecutive days. In the first condi tioning session, mice 638 were injecte d i.p. with CN O (2 mg/kg ) an d placed for 1 h in the init ially non-prefe r red (I.N .P.) 639 compartmen t. O n day 3, during the s eco nd conditioning session , mice were inject ed with vehicle (2 % 640 DMSO) and confined for 1 h in the oppos ite (tha t is, initially pr eferr ed, I .P.) compartmen t. On d ay 4 and 641 day 5 the first and second cond itioning s essions were r epea ted , respec tively. On day 6, the mice were 642 test ed for th eir side compar tmen t prefe r ence by placing them in th e left compar t ment and allowing 643 them to e xplo re th e en tire ap para tus fre ely for 30 min (post-test). Animal loca tio n was tracked, and the 644 time spent in each box was assess ed wit h Ethovision XT 16 software (Noldus). Preference in dices (P.I.) 645 were calculat ed as [(Time in Initially Preferre d (I.N .P.) chamber − Time in Init ially Preferred (I .P.) 646 chamber)/Total time]. 647 648 Pavlovian conditioning 649 Training was conducted in mouse op eran t chambers (21.6 cm W x 18.6 cm D x 12.7 cm H, Med 650 Associates I nc.). The chambe r had a pell e t dispense r tha t deliver ed one 20-mg pell et into the po rt when 651 triggere d. This pelle t dispens er can moni tor th e head access by infrared de tect ors . The chamber 652 containe d a multipurp ose sound gen era t or (ENV-230, Med Associates Inc), which delivered eith er a 653 single clear ton e (‘‘ton e’’) or click noise (“click” ) when activated. Bo th stimuli wer e presen ted a t 75 dB. 654 All condition ed stimuli wer e 5 s in durati on, separ ate d by a variable int er trial in te rval (ITI) with a mean 655 of 2 min (range = 2 to 6 min). The two auditory stimuli describ ed wer e used in th e exper iments ; whethe r 656 the ‘‘click’’ o r ‘‘t one’ ’ stimulus was the C S+ was counter balance d across mice. Mi ce were first food-657 restric ted t o 90 % of their baselin e body weight and habit uat ed to sucr ose pell ets (TestDiet 5TUT) in 658 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 5, 2026. ; https://doi.org/10.64898/2026.02.03.703447doi: bioRxiv preprint 17 their hom e cage for 3 days. On th e first d ay of behavioral tr aining, mice lea rned t o retri eve pelle ts from 659 the food dispens er. During t his session, mice received fifteen 20-mg sucrose pell ets across a 30-min 660 period . Afte r food port traini ng, mice rec eived conditioni ng sessions, which lasted 7 days before the 661 imaging test session. Each sessio n laste d 30 min and included 10 presen tati ons of each CS presen ted in a 662 pseudoran domized o rder and sepa rat ed by a variable ITI. During th ese sessions, t erminati on of one of 663 the cues was followed 1 s late r by deliver y of one sucrose pelle t; this cue was desi gnated th e CS+. The 664 other s timulus was presen ted al one with out food and was designa ted th e CS-. During the Pavlovian 665 appeti tive e xtincti on tes t, ide ntical pr oce dures were followe d exce pt th e rewa rd was not deliver ed. 666 667 Cricket hunting 668 To habitua te mice to cricke t hunti ng, eac h mouse was individually placed in an e mpty cage with thre e 669 live crickets daily for thre e consecutive d ays. Mice were all owed to fre ely inte ract with and consume the 670 crickets; th e session end ed once all thre e crickets were ea ten . On t he t est day, mi ce were t ested un der 671 mild food restricti on (2.5RiUDg standar d cho w per day) to increase mo tivation . At the beginning of each 672 trial, thre e crickets wer e intr oduced in to the cage ne ar one co rner , while th e mouse was positioned in 673 the diagon ally opposite co rner . For op to genetic stimula tion e xpe riments , light was delivered in 674 alter nating cycles of 1-minute stimula tio n followed by 1-minute no stimulatio n, for a to tal dura tion of 6 675 minutes. Fo r calcium imaging experim en ts, recor ding continu ed until t he mouse c onsumed the final 676 cricket, allowing captu re of complet e ne ural dynamics throughou t th e hunting bo ut. 677 678 Single-Unit EMG r ecording 679 For implanta tion of th e EMG el ectr odes, the fur over t he he ad and one ch eek is shaved. The scalp is 680 disinfected, an d the skin is re trac ted t o e xpose th e skull. Small incisions ar e made above the masse te r 681 and tempo ralis muscles. As much fascia as possible is removed from th e muscle, and the targe t site for 682 implanta tion is iden tified. Using blun t for ceps, a subcutan eous tunn el is crea ted fr om the face to the 683 head/neck are a of the animal . The el ectr ode leads of myomatri x arr ays (RF-4x8-BVS-5) are carefully 684 thre aded t hrough th e tunn el until they e merge above th e ta rget muscle . They are then ancho red a t 685 several loca tions within th e muscle—two leads in the masse ter and two in th e te mporalis. The skin is 686 closed around the a rray, and the conn ect or (Omnetics connec tor, 13 mm, 1.36 g) is secured to the skull. 687 During behavior tes t session, conn ect th e Inta n heads tage t o the Omnetics conn ector . The oth er end of 688 the wire from th e heads tage is plugged i nto th e recor ding board (Op en Ephys). 689 690 In vivo calcium imaging of freely moving mice 691 All in vivo imaging experiments we re con ducted on fre ely moving mice. GCaMP8m fluorescence signals 692 were acquir ed using a miniatur e int egrat ed fluorescence microsco pe system (nVista – Inscopi x) secured 693 in the basep lat e holde r befor e each imag ing session. Mice were h abituate d to th e miniscope procedu r e 694 for 3 days before behavior al exp erime nts for 30 min per day. Set tings were kep t constan t within 695 subjects and across imaging sessions. Im age acquisition an d behavior we re synchronized using th e dat a 696 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 5, 2026. ; https://doi.org/10.64898/2026.02.03.703447doi: bioRxiv preprint 18 acquisition bo x of the nVista Imaging System (Inscopix) thro ugh a TTL box (Noldus) connected to the 697 USB-IO bo x from the Eth ovision system (Noldus). 698 Behavioral recordi ngs were conduct ed in a transpa ren t acrylic chamber equi pped with a 45RiUD°C angled 699 mirror posit ioned a t th e bot tom. This set up allowed unobst ruct ed, bo ttom-up video captu re of the 700 mouse’s behavior via a came ra positi one d below the chambe r. 701 Object biting : 702 Mice were pr esen ted with five distinct o bjects designed to vary in mechanical p r opert ies: softwood , 703 Styrofoam, sponge ma teri al, hard silicon e, soft silicone, an d chow pellet . All mat e rials were cut in to 704 standar dized cubes (0.5RiUDcm × 0.5RiUDcm × 0.5RiUDcm). Each imaging session lasted 5 minutes, du ring which 705 the mouse could fre ely expl ore an d inte r act with the o bjects while calcium signals were recor ded . 706 Flavored milk drinking: 707 For liquid consumption trials , mice were water-dep rived for 4 hours pri or to testi ng. Animals remain ed 708 in their hom e cage and wer e sequen tiall y given access to different liquids: st rawb erry-flavored milk, 709 chocolate-flavor ed milk, vanilla-flavored milk, water, and 100RiUDmM quinin e soluti on. Each liquid was 710 presen ted for 5 minut es, during which calcium activity and behavioral r esponses were reco rded . For 711 imaging data processing and an alysis, we used IDEAS (Inscopix) version 25.1.7 . 712 Flavored food or spong e biting: 713 Flavored food or spong e were pr epa red by soaking standard food pell ets or spo n ge cubes in 10 % 714 flavoring solutions (strawber ry, vanilla, c hocolat e), water, or 100RiUDmM quinine sol ution. After d rying 715 overnight, the flavored i tems were p rese nted in ran domized o rder . Mice wer e all owed to freely bi te th e 716 flavored food or sponge pi eces during 5-minute sessions while calcium activity was continuously 717 record ed. 718 719 Linguine feeding assay with optogen etic manipulatio n 720 Mice were fast ed for 4 hours prio r to t es ting to enh ance motiva tion. During each session, animals were 721 individually placed in a clean beh avioral arena a nd pres ent ed with thr ee pr e-weighed pieces of raw 722 linguine (approxima tely 2RiUDcm each). Mic e were allowe d to fre ely inter act with an d consume the pas ta 723 for 10 minutes. Each mouse und erwen t t wo test sessions on sep ara te days in a co unterb alanced orde r: 724 one session with light stimul ation and on e without . For th e opt ogene tic activati on exper iment , 5 Hz, 10 725 H and 20 Hz blue light (473RiUDnm) was deli vered thr oughout the 10-minut e session. For the inhibi tion 726 expe riment , yellow light (532RiUDnm) was d elivered con tinuously th roughout the 10- minute t est. R esidual 727 linguine was weighed after each session to det ermin e food consumption . 728 729 Biting behavior assay with food or Styrof oam 730 Mice were pr esen ted with small objects made of eithe r regula r chow or styrofoa m. Prior to th e tes t, five 731 uniform cubes (approxima tely 0.5RiUDcm × 0.5RiUDcm × 0.5RiUDcm) of food or styrofoam were placed in a cle an 732 transpa ren t chamber . Mice wer e allowe d to freely int erac t with the o bjects for t he dura tion of the 733 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 5, 2026. ; https://doi.org/10.64898/2026.02.03.703447doi: bioRxiv preprint 19 session. Op togen etic stimula tion was del ivered in a struc tur ed alt erna ting pat ter n of 1RiUDminute OFF 734 followed by 1 minute ON, r epe at ed thr e e times over a total of 6 minutes. Blu e light (473RiUDnm) was 735 delivered via a fiber-op tic patch cabl e co nnected to a lase r thr ough a rot ary joint, with light power 736 calibrat ed at the fiber tip (10–15RiUDmW). Light delivery was controll ed using a pulse genera tor t o achieve 737 consistent timing across animals. B ott om -up video capture of th e mouse’s b ehavi or via a camera 738 positione d below th e chamber an d biting events were scor ed manually. 739 Cell registra tion 740 To identify the same individual cells ac ro ss diverse imaging settings, we used cen t roid ex trac tion, stiff 741 alignment, and optimal on e-to-on e matc hing. For each imaging conditi on, we first extr acte d spatial 742 footprin ts from the bina rized cell maps a nd computed th e cent roid of each segme nted foot print . On e 743 condition was tr eat ed as th e refe rence map, and cent roids from the o the r condi tion were co-r egiste red 744 to this r eferenc e using a rigid transform a tion (transla tion an d rot ation) es timat ed from the top n ear est-745 neighbor cen troid pai rs via singular value decomposition . Following alignment , we computed th e whole 746 pairwise cent roid-dista nce matri x be twe en conditions an d used th e Hungaria n ap proach to achi eve th e 747 best one- to-one assignmen t; cell pai rings with a post-regist rati on cent roid dista nce of less than 35 pixels 748 were consider ed match ed. 749 750 PCA 751 We perfo rmed principal compo nen t anal ysis ( PCA) on event-aligned populatio n a ctivity to charact eriz e 752 the dominan t low-dimensional st ru cture. For each neuron, z-score d calciu m trace s were ex trac ted from 753 –4 to +6 s rel ative t o bite o nset . Ne urons with sparse or unr eliabl e respo nses wer e exclud ed by requiring 754 that > 5 % of samples exceed ed zer o acr oss trials. Ac tivity in each conditi on was baseline-align ed by 755 subtracti ng the mea n signal from a pre-e vent window (– 0.5 to 0 s), and peak-nor malized across 756 conditions so th at no single uni t domina t ed the po pulati on embeddi ng. To estima te th e share d low-757 dimensional a xes, we const ructe d a neur on × time mat rix by concat ena ting post- bite activity (0 to + 5 s) 758 across conditions an d cent ered each tim e bin before d ecompositio n. Conditi on-specific trajecto ries 759 were th en obt ained by projec ting the ev ent-aligned ac tivity back onto t he PCA loadings. Impor tan tly, no 760 tempor al smoothing was applie d prior to PCA; traces were only spline-in ter pola te d for visualization, an d 761 all loadings and variance estima tes were computed from unsmoot hed signals. 762 763 Pairwise Pearson corr elati on analysis. 764 To quantify repres ent ation al similarity ac ross conditions, we compu ted Pears on correl ation co efficients 765 based on conditi on-averag ed neur al acti vity. For each neu ron, we e xt ract ed z-scored calcium signals 766 within a – 6 s to + 8 s event window. M e an z-scored activity was calculat ed for ea ch neuron × condi tion, 767 yielding a neuron-by-condition respons e matrix . Ne ural r esponses wer e averag ed across neurons for 768 each mate rial condi tion, and p airwise Pe arson corr elati ons were computed be tw een mat erial-level 769 response vect ors to quantify similarity ac ross conditions . 770 771 Statis tical analysis 772 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 5, 2026. ; https://doi.org/10.64898/2026.02.03.703447doi: bioRxiv preprint 20 All statis tics are d escribed whe re used . S tatis tical analyses wer e conduct ed using GraphPad Prism 9 773 software (Grap hPad). No sta tistical me th ods were used t o pred et ermine sampl e sizes, but t he numbe r 774 of samples in each group were similar to those r epor ted in pr evious publicati ons. Data collectio n and 775 analysis were not p erformed bli nd to th e conditions of the e xpe rimen ts. Mice t ha t, afte r histological 776 inspection, h ad the l ocation of the viral i njection (repor te r pro tein) or of the o pti c fiber (s) outside the 777 area of int eres t were exclud ed. All da ta were rep rese nted as the me an ± SEM. Si gnificance levels are 778 indicated as follows: *P < 0.05 ; **P < 0.0 1; ***P < 0.001 ; ** **P < 0.0001 . 779 780 781

782 1 Apont e, Y., A tasoy, D. & Ster nson, S. M . AGRP neur ons are sufficient to orch estra te feeding 783 behavior ra pidly and without traini ng. N ature Neur oscien ce 14, 351-355 (2011). 784 https://doi. org:10.1038/nn.2739 785 2 Atasoy, D., B etl ey, J. N ., Su, H . H. & Ster n son, S. M. Decons tructi on of a neural circ uit for hunger . 786 Nat ure 488 , 172-+ (2012). https://doi.org :10.1038/natur e11270 787 3 Betl ey, J. N ., Cao, Z. F. H ., Rit ola, K. D. & Sternson , S. M . Parallel, R edund ant Circuit Organiza tion 788 for Homeosta tic Control of Fe eding Beh a vior. Cell 155 , 1337-1350 (2013). 789 https://doi. org:10.1016/j.cell .2013.11.00 2 790 4 Betl ey, J. N . et al. N eurons for hung er an d thirst t ransmit a n egative-valenc e te aching signal. 791 Nat ure 521 , 180-185 (2015). https://doi.org:10.1038/nature14416 792 5 Chen, Y., Lin, Y. C., Kuo, T. W. & Knight, Z. A. Sensory de tec tion of food rapidly mo dulates 793 arcuat e feeding circuits . Cell 160 , 829-841 (2015). https://doi.org:10.1016/j.cell .2015.01.033 794 6 Jennings, J . H. e t al. Visualizing Hypot hal amic Network Dynamics for Appeti tive a nd 795 Consummatory Behaviors . Cell 160 , 516-527 (2015). https://doi.org:10.1016/j.cell .2014.12.026 796 7 Nectow, A . R. e t al. Iden tification of a Br a instem Circuit Contr olling Feeding . Cell 170 , 429-+ 797 (2017). https://doi.org:10 .1016/j.cell.201 7.06.045 798 8 Douglass, A. M. e t al. Cen tral amygdala ci rcuits modulat e food consumptio n thro u gh a positive-799 valence mechanism. Na t Ne u r o s ci 20, 1384-1394 (2017 ). https://doi.org:10 .1038/nn.4623 800 9 Hardaway, J . A. e t al. Cen tral Amygdala P repron ociceptin-Ex pressing N eurons M e diate Pala table 801 Food Consumption and R eward . Neur on 102 , 1088 (2019). 802 https://doi. org:10.1016/j.neuron .2019.0 4.036 803 10 Cai, H., Haube nsak, W. , Anth ony, T. E. & Anderson , D. J. Cen tral amygdala PKC-delta(+) neurons 804 mediate the influenc e of multiple an ore x igenic signals. Nat Neur osci 17, 1240-1248 (2014). 805 https://doi. org:10.1038/nn.3767 806 11 Ding, W. Y., Wel tzien , H., Pet ers, C. & Klein, R. Naus ea-induced supp ression of fee ding is 807 mediate d by central amygdala Dlk1-expr essing neurons. Cell R ep 43 (2024). 808 https://doi. org:A RTN 11399010.1016/j.celrep .2024.113990 809 12 Fermani, F . et al . Food and wa ter in take are regul at ed by distinct cen tral amygdal a circuits 810 revealed using in ters ection al genet ics. N at Com mun 16, 3072 (2025). 811 https://doi. org:10.1038/s41467-025-58144-3 812 13 Peters, C. e t al. Transc riptomics reve als a mygdala neuron regul ation by fasting an d ghrelin 813 ther eby promoting fe eding. Sci Ad v 9 (2023). https://doi.o rg:ARTN 814 eadf652110.1126/sciadv.adf6521 815 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 5, 2026. ; https://doi.org/10.64898/2026.02.03.703447doi: bioRxiv preprint 21 14 Han, W. e t al. Int egrat ed Contr ol of Predatory Hunti ng by the Centr al Nucleus of t he Amygdala. 816 Cell 168 , 311-324 e318 (2017 ). https://doi.org:10 .1016/j.cell.2016.12 .027 817 15 Hoffman, D. S. & Luschei, E. S. Responses of monkey precentral co rtical cells du rin g a controlled 818 jaw bite task. J N e u r op hy s i o l 44, 333-348 (1980). https://doi.org:10.1152/jn.1980. 44.2.333 819 16 Dubner, R. Oral-facial senso ry and motor mechanisms. Science 170 , 1130-1132 (1970). 820 https://doi. org:10.1126/science .170.396 2.1130 821 17 Lund, J. P. & Lamarre, Y. Activity of neuro ns in the lower pr ecen tral cor te x during voluntary and 822 rhythmical jaw movements in the monke y. Exp Brain R es 19, 282-299 (1974 ). 823 https://doi. org:10.1007/BF00233235 824 18 Luschei, E. S., G arthwai te , C. R. & Armstr ong, M. E. Rel ationshi p of firing patte rns of units in face 825 area of monkey prec entr al cort ex to con ditioned jaw movements . J Ne uro physiol 34, 252-261 826 (1971). https://doi.org:10 .1152/jn.1971.34.4.552 827 19 Yoshino, K., Mikami, A . & Kubota, K. N eu ronal activiti es in the ven tral p remo tor c orte x during a 828 visually guided jaw movement in monkeys. Neurosci Res 30, 321-332 (1998). 829 https://doi. org:10.1016/s0168-0102(98) 00014-5 830 20 Tamura, K. et al . Cell-class-specific orofacial motor maps in mouse n eocor tex . Curr Biol 35, 1382-831 1390 e1385 (2025). https://doi.org:10.10 16/j.cub.2025.01.056 832 21 Mercer Lindsay, N. et al. O rofacial Move ments Involve Paralle l Corticobulb ar Projections from 833 Motor Cor te x to Trigemina l Premoto r Nuclei. Neur on 104 , 765-780 e763 (2019). 834 https://doi. org:10.1016/j.neuron .2019.0 8.032 835 22 Baldwin, M., F rost, L. L. & Wood , C. D. In vestigation of th e Primate Amygdala - Movements of 836 the Face a nd Jaws. N eurol ogy 4 , 586-598 (1954). https://doi.org:Doi 10.1212/Wnl .4.8.586 837 23 Kaada, B. R ., And erse n, P. & Jansen, J. Sti mulation of the Amygdaloid Nuclea r Complex in 838 Unanest heti zed Cats . Neur olog y 4 , 48-64 (1954). https://doi.org:Doi 10.1212/Wnl .4.1.48 839 24 Sasamoto, K. & Oh ta, M . Amygdaloid-ind uced jaw opening and facilitati on or inhi bition of the 840 trigeminal mot oneurons in th e rat . Com p Bioche m Physiol A Co mp Physiol 73, 349 -354 (1982 ). 841 https://doi. org:10.1016/0300-9629(82)9 0166-9 842 25 Lafferty, D. S. et al . An amygdalopon tine pathway promot es moto r programs of in gestion. 843 bioRxi v (2025). https://doi.o rg:10.1101/2025.06.05.657686 844 26 O'Lea ry, T. P. et al. Neuronal cell types, p rojections, a nd spatial organiza tion of th e centr al 845 amygdala. iScienc e 25, 105497 (2022). https://doi.org:10 .1016/j.isci.2022.105497 846 27 Waclaw, R. R ., Ehrman, L. A ., Pierani , A. & Campbell, K. Developmental o rigin of the neuro nal 847 subtypes th at compr ise t he amygdala r fe ar circuit i n th e mouse . J Neur osci 30, 69 44-6953 (2010). 848 https://doi. org:10.1523/J NEUROSCI .577 2-09.2010 849 28 Wang, Y. et al . Mul timodal mapping of cell types and pr ojections in th e cent ral n ucleus of the 850 amygdala. Elife 12 (2023). https://doi.org:10.7554/eLife.84262 851 29 Kim, J., Zhang, X., Mur alidhar , S., LeBl anc , S. A. & Tonegawa, S. Basola ter al to Cen t ral Amygdala 852 Neur al Circuits for Appe titive B ehaviors . Neuro n 93, 1464-1479 e1465 (2017). 853 https://doi. org:10.1016/j.neuron .2017.0 2.034 854 30 Lafferty, D. S. et al . An amygdalopon tine pathway promot es moto r programs of in gestion. 855 bioRxi v (2025). https://doi.o rg:10.1101/2025.06.05.657686 856 31 Yang, T. et al. Plastic and s timulus-specific coding of salient events in th e cent ral a mygdala. 857 Nat ure 616 , 510-519 (2023). https://doi.org:10.1038/s41586-023-05910-2 858 32 Chung, B. et al . Myomatri x arr ays for high-definition muscle reco rding. Elife 12 (2023). 859 https://doi. org:A RTN RP8855110.7554/eLife.88551 860 33 Yang, T. et al. Plastic and s timulus-specific coding of salient events in th e cent ral a mygdala. 861 Nat ure 616 , 510-+ (2023). https://doi.org :10.1038/s41586-023-05910-2 862 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 5, 2026. ; https://doi.org/10.64898/2026.02.03.703447doi: bioRxiv preprint 22 34 Everitt, B . J ., Mor ris, K. A. , Obri en, A . & Robbins, T. W. The B asola ter al Amygdala Ventral S tria tal 863 System and Condition ed Place Prefere nce - Further Evidenc e of Limbic Striatal I nt eracti ons 864 Underlying Reward-R elat ed Processes . N euroscie nce 42, 1-18 (1991). https://doi. org:Doi 865 10.1016/0306-4522(91 )90145-E 866 35 Yasui, Y., Tsumori, T., Oka , T. & Yokota, S. Amygdaloid axon t erminals ar e in cont a ct with 867 trigeminal pr emot or neu rons in the p arvi cellular re ticula r formation of th e ra t me dulla 868 oblongata . Brai n Res 1016 , 129-134 (200 4). https://doi.o rg:10.1016/j.brain res.20 04.04.080 869 36 Campos, C. A., Bowen, A. J ., Schwar tz, M. W. & Palmiter, R. D. Para brachial CG RP Neuro ns 870 Control Me al Termina tion. Cell M eta b 23 , 811-820 (2016 ). 871 https://doi. org:10.1016/j.cmet .2016.04. 006 872 37 Carter , M. E., So den, M . E., Zweifel, L. S. & Palmiter, R. D. G ene tic identifica tion o f a neural 873 circuit tha t suppress es appe tit e. Na t u re 503 , 111-114 (2013 ). 874 https://doi. org:10.1038/na ture12596 875 38 Rosebe rry, T. K. et al . Cell-Type-Specific C ontrol of Br ainst em Locomotor Circuits b y Basal 876 Ganglia. Cell 164 , 526-537 (2016). https:/ /doi.org:10.1016/j.cell .2015.12.037 877 39 Altschule r, S. M ., Bao , X. M., Bi eger, D. , Hopkins, D. A. & Miselis, R. R . Viscerot opic 878 repr esent ation of t he uppe r alimen tary t ract in th e rat : sensory ganglia and nucl ei of the solitary 879 and spinal trigemin al tr acts. J C omp N eur ol 283 , 248-268 (1989 ). 880 https://doi. org:10.1002/cne .902830207 881 40 Shapiro, R . E. & Miselis, R. R . The cent ral organiza tion of the vagus nerve inn ervat i ng the 882 stomach of the ra t. J Comp Neur ol 238 , 473-488 (1985 ). https://doi.org :10.1002/cne.902380411 883 41 Takatoh, J . et al. Construc ting an adul t or ofacial premot or atl as in Allen mous e CCF. Elife 10 884 (2021). https://doi.org:10 .7554/eLife.67 291 885 42 Stanek, E. t., R odriguez , E., Zhao, S ., Han, B. X. & Wang, F. Supra trigemin al Bilat er ally Projecting 886 Neuro ns Maint ain Basal Ton e and Enable Bilate ral Phasic Activation of Jaw-Closing Muscles. J 887 Neurosci 36, 7663-7675 (2016). https://doi.org:10 .1523/JNEUR OSCI.0839-16.2016 888 43 Ohta , M. & Moriyama, Y. Sup rat rigemina l neurons media te the shor tes t, disynapti c pathway 889 from the cent ral amygdaloid nucleus to t he contr alat eral trigeminal mo toneu rons in the ra t. 890 Comp Bio che m Physiol A Co mp Physiol 83 , 633-641 (1986 ). https://doi.org :10.10 16/0300-891 9629(86)90702 -4 892 44 Alex ander , G. E . & Crutcher, M . D. Functi onal archit ectu re of basal ganglia circuits : neural 893 substra tes of parall el processing. Tre nds Neurosci 13, 266-271 (1990). 894 https://doi. org:10.1016/0166-2236(90)9 0107-l 895 45 Smith, Y. & Parent, A. N eurons of th e sub thalamic nucleus in prima tes display glut amate bu t not 896 GABA immuno reac tivity. Brai n Res 453 , 353-356 (1988 ). https://doi.org :10.1016/0006-897 8993(88)90177 -1 898 46 Jennings, J . H. e t al. Distinct ex tend ed a mygdala circuits for divergent motiva tion al stat es. 899 Nat ure 496 , 224-228 (2013). https://doi.org:10.1038/nature12041 900 47 MacLaren, D. A ., Ma rkovic, T. & Clark, S. D. Assessment of sensorimo tor ga ting following 901 selective lesio ns of cholinergic peduncul opontin e neur ons. Eur J Neuros ci 40, 3526-3537 (2014 ). 902 https://doi. org:10.1111/ejn.12716 903 48 Crabtre e, J . W. Intr ath alamic sensory con nections media ted by th e thal amic reticu lar nucleus. 904 Cell Mol Life Sci 56, 683-700 (1999). https://doi.org:10.1007/s000180050462 905 49 McAlonan, K. & Brown, V. J. The t halamic reticula r nucleus: mor e than a sensory n ucleus? 906 Neurosci en tist 8 , 302-305 (2002). https://doi.org:10.1177/107385840200800405 907 50 Siletti , K. et al . Transcrip tomic diversity o f cell types across the adul t human br ain. Science 382 , 908 eadd7046 (2023). https://doi.o rg:10.112 6/science.add7046 909 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 5, 2026. ; https://doi.org/10.64898/2026.02.03.703447doi: bioRxiv preprint 23 51 Yu, B. et al . Molecul ar and cellul ar evolu t ion of the amygdala across species a naly zed by single-910 nucleus transcri ptome p rofiling. Cell Disc ov 9 , 19 (2023). https://doi.org:10 .1038/s41421-022-911 00506-y 912 52 Peters, C. e t al. Transc riptomics reve als a mygdala neuron regul ation by fasting an d ghrelin 913 ther eby promoting fe eding. Sci Ad v 9 , ea df6521 (2023 ). https://doi.org :10.1126/sciadv.adf6521 914 53 Hain, D. et al . Molecul ar diversi ty and evolution of neu ron types in t he amnio te b rain. Scie nce 915 377 , eabp8202 (2022). https://doi.o rg:10 .1126/science.abp8202 916 54 Woych, J. e t al. Cell-type p rofiling in sala manders iden tifies innovati ons in vert ebr ate for ebrain 917 evolution . Scienc e 377 , eabp9186 (2022). https://doi.o rg:10.1126/science .abp918 6 918 55 Nakamura, Y. , Yanagawa, Y., M orriso n, S. F. & Nakamura, K. M edullary Re ticular N eurons 919 Media te N europ eptid e Y-Induced M etab olic Inhibition and Mastic ation . Cell Met a b 25, 322-334 920 (2017). https://doi.org:10 .1016/j.cmet.2 016.12.002 921 56 Moore , J. D., Klei nfeld, D. & Wang, F. Ho w the brainst em contr ols orofacial be haviors comprised 922 of rhythmic actions. Tren ds in Neur oscie nces 37, 370-380 (2014). 923 https://doi. org:10.1016/j.tins .2014.05.0 01 924 57 Peyron, M. A ., Blanc, O., Lund, J. P. & Wo da, A. I nfluence of age on ad aptab ility of human 925 mastication . J Ne uro physi ol 92, 773-779 ( 2004). https://doi.o rg:10.1152/jn.01122. 2003 926 58 Peyron, M. A ., Lassauzay, C. & Woda, A. Effects of increased hardn ess on jaw movement and 927 muscle activity during chewing of visco-e lastic model foods. Ex p Brai n Res 142 , 41-51 (2002 ). 928 https://doi. org:10.1007/s00221-001-0916-5 929 59 Forste r, G . L. & Blaha, C. D. Pedunculopo ntine t egment al stimulati on evokes stri at al dopamine 930 efflux by activation of acetylcholine and glutamat e recep tors in the midbra in and pons of the rat. 931 Europe an J ourn al of Neur oscie nce 17, 75 1-762 (2003 ). https://doi.org :10.1046/j.1460-932 9568.2003.02511.x 933 60 Nort on, A . B., Jo, Y. S. , Clark, E. W. , Taylor, C. A. & Mizumori, S . J. Indep enden t ne ural coding of 934 reward and movemen t by pedunculop on tine t egmenta l nucleus neu rons in freely navigating rats . 935 Eur J Neur osci 33, 1885-1896 (2011). https://doi.org:10.1111/j.1460-9568.2011.0 7649.x 936 61 Yoo, J. H. e t al . Activati on of Pedunculop ontine Glut amate Neu rons Is Reinfo rcing. Jour nal of 937 Neurosci en ce 37, 38-46 (2017). https://doi.org:10 .1523/Jneurosci.3082-16.2016 938 62 Chang, S. et al . Tripar tite ex tend ed amygdala-basal ganglia CRH circuit drives loco motor 939 activation an d avoidance b ehavior. Sci A dv 8 , eabo1023 (2022). 940 https://doi. org:10.1126/sciadv.ab o1023 941 63 Furlan, A . et al . N euro tensin ne urons in t he ex tend ed amygdala contr ol die tary choice and 942 energy homeos tasis. Na t Ne u r o s c i 25, 14 70-1480 (2022 ). https://doi.org:10 .1038/s41593-022-943 01178-3 944 64 Liu, H. M. et al. IPAC integ rat es rewardi n g and environment al memory during th e acquisition of 945 morphine CPP. Sci Adv 9 (2023). https://doi.org:10 .1126/sciadv.adg5849 946 65 Carter , M. E., So den, M . E., Zweifel, L. S. & Palmiter, R. D. G ene tic identifica tion o f a neural 947 circuit tha t suppress es appe tit e. Na t u re 503 , 111-+ (2013). https://doi.org:10.103 8/nature12596 948 66 Ferre ira-Pinto, M . J . et al . Function al dive rsity for body actions in th e mesenceph al ic locomotor 949 region. Cell 184 , 4564-4578 e4518 (2021) . https ://doi.org:10 .1016/j.cell.2021.07.0 02 950 67 Steinb erg, E. E. et al. Amygdala-Midb rain Connections Modul ate Appe titive and A versive 951 Learning. Ne uro n 106 , 1026-1043 e1029 (2020). https://doi.org:10 .1016/j.neuron .2020.03.016 952 68 Kim, S. Y. et al. Diverging neur al pathway s assemble a behaviou ral sta te from sep arable fea tur es 953 in anxie ty. Na ture 496 , 219-223 (2013). https://doi.o rg:10.1038/na ture12018 954 69 Cai, H. R. et al. Comparison of th e conne ctivity of the poste rior in tral aminar th ala mic nucleus 955 and perip eduncular nucl eus in ra ts and mice. Fron t Neur al Circui ts 18, 1384621 ( 2024). 956 https://doi. org:10.3389/fncir.2024.1384 621 957 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 5, 2026. ; https://doi.org/10.64898/2026.02.03.703447doi: bioRxiv preprint 24 70 Lanuza, E., Mo ncho-Bogani, J. & Ledoux, J. E. Uncondi tion ed stimulus pat hways to the amygdala: 958 effects of lesions of the post erio r intr ala minar thalamus on foo t-shock-induced c-Fos expr ession 959 in the subdivisions of the la ter al amygdala. Neur oscie nce 155 , 959-968 (2008). 960 https://doi. org:10.1016/j.neuroscience .2 008.06.028 961 71 Ledoux, J. E., Ruggie ro, D. A. , For est, R ., S torne tt a, R. & Reis, D. J. Topogra phic org anizati on of 962 convergent pr ojections t o the thalamus f rom the inferi or colliculus and spinal co r d in the ra t. J 963 Comp Ne urol 264 , 123-146 (1987). https: //doi.org:10.1002/cne.902640110 964 965 966 Figure lege nds 967 Figure 1. CeA Isl1 neu rons respo nses duri ng obje ct-biting behaviors. 968 A. Repr esent ative image showing Isl1 im munostaining in th e CeA of Pkcδ-CreER : Ai9 mouse. Scale b ar, 969 30 μm. 970 B. Schematic of a fre ely moving mouse expressing GCaMP8m in CeA Isl1 neuro ns and carrying a head-971 mounted miniscope . Left pan el: raw calci um imaging frame; right panel : manually segmented r egions of 972 inter est (ROIs) corr esponding t o individual neurons . 973 C. Post hoc histological valida tion showin g GCaMP8m expression (green) in th e medial cent ral amygdala 974 (CeM), with the lens track visible above t he CeA. Scale b ar, 100 μm. 975 D. Hea tmaps of individual neu rons (rows) showing z-scored calcium activity (color scale) aligned to th e 976 first bite (vertical red line) of each subst r ate: sof twood, Styr ofoam, sponge, cricke t, and chow. N eurons 977 are sor ted by respo nse type . N= 12 mice per group . 978 E. Average r esponses of th e z-scored calc ium activity for each condition . Shad ed a reas rep rese nt s. e.m. 979 F. Proporti ons of neurons classified as up -modulated (red), down-modula ted (blu e), or non-modulat ed 980 (white) for each substrat e. 981 G–K. Me an neuron al activity during object-biting versus non-biting epochs: soft wood ( G ), Styrofoam ( H ), 982 sponge ( I), cricket ( J ), and chow (K ) (n= 12 mice per group. Wilco xon signed-r ank t est; * *PRiUD<RiUD0.01, 983 ***PRiUD<RiUD0 .001, ** **PRiUD<RiUD0 .0001). 984 985 Figure 2. The mate rial-specific enc oding properties of CeA Isl1 n eurons. 986 A . Heatm aps of the r esponse t o differen t materials . Each row repr esen ts the activities of one neuron . 987 Neuro ns were group ed by clustering bas ed on thei r mean r esponse p rofiles acros s conditions (n= 6 988 mice). 989 B . Averag e activity tr aces for each clust e r. Distinct t emporal r esponse p at terns an d stimulus selectivity 990 emerged acr oss clusters . Shaded areas r epres ent s. e.m. 991 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 5, 2026. ; https://doi.org/10.64898/2026.02.03.703447doi: bioRxiv preprint 25 C. Neu ronal ac tivity across five functiona l clusters during differen t bit e conditio ns: silicone, softwood, 992 sponge, and Styr ofoam. (n= 6 mice per g roup; * ** *p < 0.0001; two-way ANOVA with Tukey's post hoc 993 test). 994 D. Propor tion of neurons in each clu st er. 995 E . Principal component a nalysis (PCA) of trial-aver aged popul ation ac tivity across mice reveals distinc t 996 traject ories for bi tes on silicone (re d), sponge (green), softwood (blue), and Styro foam (purple). Each 997 traject ory repr esen ts the tempo ral evolution of popu la tion activity pr ojected on t o the first t hre e 998 principal componen ts (PC1–PC3 ), with time progressing along th e tr ajectory (arr o w). 999 F. Time courses of the first two principal components (PC1 and PC2 ) aligned to bit e onset . Distinct 1000 materi al-specific dynamics emerge aroun d the time of bit ing, indicating t hat CeA Isl 1 ensemble activity 1001 transie ntly encod es the physical iden tity of the bit ten mat erial . 1002 G . Heatma ps of the resp onse t o different stimuli. Each row repr esen ts the ac tiviti es of one neuro n. 1003 Sponge, softwood, and Styrofoam align e d to bite o nset . Quinine , wate r, vanilla m ilk, chocolate milk and 1004 strawber ry milk aligned to lick onset . Ne urons were grou ped by clusteri ng based on their me an resp onse 1005 profiles across conditi ons (n= 6 mice). 1006 H. N euron al activity across five functiona l clusters during differen t conditi ons: spo nge, softwood, 1007 Styrofoam, quinin e, wate r, vanilla milk, chocolat e milk and strawbe rry milk. (n= 6 mice per group, * ** *p 1008 < 0.0001, two-way ANOVA with Tukey's post hoc tes t). 1009 I. Propor tion of neur ons in each cluste r. 1010 J . Hea tmaps of respons es to har d silicone and soft silicone. Each r ow repr esen ts t he activities of on e 1011 neuron (n = 6 mice). 1012 K. Fract ion of neuro ns classified as up-modulat ed, down-modula ted, o r non-mod ulated . Blue is ha rd 1013 silicone, orang e is soft silicone (n= 6 mice; N.S .; two-way ANOVA wit h Tukey's post hoc test). 1014 L-O . Average r esponses of all neurons ( L ) and neurons classified as up-modula ted ( M ), down-modulated 1015 (N ), or non-modulated ( O ). Shad ed ar eas repres ent s. e.m. 1016 1017 Figure 3. CeA Isl1 neu ron activation enha nces biting efficie ncy, freque ncy, a nd fic tive feeding b ehaviors. 1018 A. Schematic of optic-fiber pl acemen t ab ove CeA Isl1 neurons e xpressing ei the r ChR2 or EGFP. 1019 B. Repr esen tative fram e from video sho wing a mouse grasping and biting linguine. 1020 C. Mean n umber of biting bou ts across st imulation condi tions (n= 12 mice per gro up; **p < 0.01, two-1021 way ANOVA with Tukey's post hoc t est). 1022 D. Biting bou t dur ation du ring 20 Hz stimulation (*p < 0 .05, unpair ed t-t est). 1023 E. Weight of linguine consumed by 4 h fasted animals (N.S .; unp aired t-tes t). 1024 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 5, 2026. ; https://doi.org/10.64898/2026.02.03.703447doi: bioRxiv preprint 26 F. Schematic of a mouse fre ely biting pie ces of chow/Styrofoam. 1025 G. Raste r plot showing biting bou ts (blue ) and fictive feeding (orange) across individual fed ChR2 mice 1026 during alte rna ting light OFF and light ON epochs in biting food assay. 1027 H-K. Quan tification of be haviors during ON and O FF perio ds for both ChR2 mice and EGFP mice in biting 1028 food assay. H. Perce ntage of tim e spent biting food. I. M ean numbe r of food-biti ng bouts. J . Perc entag e 1029 of time engaged in fictive feeding. K . Me an number of fictive feeding bouts . N= 1 3 mice per group. Two-1030 way ANOVA with Tukey's post hoc t est , * PRiUD<RiUD0.05, **PRiUD<RiUD0 .01, ** *PRiUD<RiUD0.001, * ***PRiUD<RiUD0 .0001. 1031 L. Same as in panel G but for S tyrofoam pieces. 1032 M-P. Quan tification of be haviors during ON and O FF perio ds for both ChR2 mice and EGFP mice in biting 1033 Styrofoam assay. M. Pe rcent age of time spent biting S tyrofoam. N. Mean numb er of Styrofoam-biting 1034 bouts. O. Percen tage of time e ngaged in fictive feeding. P. M ean numbe r of fictive feeding bouts . N= 13 1035 mice per group. Two-way ANOVA with T ukey's post hoc tes t , *PRiUD<RiUD0.05, **PRiUD<RiUD 0.01, ** *PRiUD<RiUD0.001, 1036 *** *PRiUD<RiUD0.0001. 1037 Q. Probability densi ty distribu tions of distances of the animal to the food sourc e d uring biting (blue) and 1038 fictive feeding (orange) events. 1039 R. Quan tification of mean dis tances for e ach behavior across anim als (n = 7 mice, *** *P < 0.0001, two-1040 tailed unp aired t-tes t). 1041 S. Schematic of optic-fiber plac ement a b ove CeA Isl1 neurons e xpressing ei the r Arc h3.0 or EGFP. 1042 T. Rast er plo ts showing eating bou ts (grey for eating bouts and gr een for e ating b outs longer than 130 s) 1043 across light OFF and ON ep ochs in Arch3. 0and EGFP groups during linguine consu mption assays. 1044 U . Duration of individual feedi ng bouts a cross light OFF and O N epochs in A rch3.0 and EGFP groups 1045 during linguine consumption assays (n = 7 mice per group; * p < 0.05, **p < 0 .01; t wo-way ANOVA with 1046 Tukey's post hoc tes t). 1047 V. Schema tic of Myomatrix micro arr ays (bottom), e ach containing eigh t elec trod e contacts on a flexi ble 1048 substra te, implan ted in to th e masset er a nd tempor alis muscles for elect romyogra phic (EMG) recordings. 1049 W. Repr esen tat ive EMG t races from th e right masset er and t empor alis muscles of a freely behaving Isl1-1050 CreER mouse exp ressing the in hibito ry DREADD hM4Di in the central amygdala . Traces show activity 1051 during food biting afte r intr aperi ton eal i njection of saline or CN O (2 mg kg⁻¹ ). 1052 X. EMG amplitud e measur ements from t he massete r muscle during food biting af ter salin e or CNO (0.4 1053 mg kg⁻¹ ) injection in Isl1-CreER; hM4Di mice. n=4; *** *P < 0.0001. 1054 Y. EMG amplitud e measur ements from t he tempo ralis muscle during food biting after salin e or CNO (0.4 1055 mg kg⁻¹ ) injection in Isl1-CreER; hM4Di mice. n=4; *** *P < 0.0001. 1056 1057 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 5, 2026. ; https://doi.org/10.64898/2026.02.03.703447doi: bioRxiv preprint 27 Figure 4. Activation of CeA Isl1 n euro ns supports positive rei nforce ment and condi tioned reward 1058 learning. 1059 A. Heatma ps showing repres enta tive occ upancy plots of EYFP (top) and ChR2 (bottom) mice. 1060 B . Quan tification of time sp ent in th e light-paired (O N) and non-pair ed (OFF) compartmen ts (n = 13 mice 1061 per group; ** **p < 0 .0001, two-way AN OVA with Sidak’s multipl e comparisons). 1062 C. Int racra nial self-stimulati on assay (ICSS). Nose poke be havior across four days of training in ChR2 1063 (blue) and EGFP (black) mice (n = 11 mi c e per group ; ** **p < 0.0001, two-way A NOVA with Sidak’s 1064 multiple compariso ns). 1065 D. Schema tic of CPP chamber. Ini tially pr eferre d (I.P.) chamber and ini tially non-p referr ed (I.N .P.) 1066 chamber. D1. Op togen etic assay: mice underwen t a pre-t est (Day 1), two days of conditioning (light-1067 paired I .N .P. or unpai red I .P. compar tme nts); and a tes t session withou t stimula ti on (Day 4 ). D2. 1068 Chemogenetic act ivation assay: mice un derwent a pre-t est (Day 1). On altern atin g days ( Day 2 and 4), 1069 mice received CNO pai red I .N .P. compar t ment), while on Day3 and 5, they rec eive d vehicle paired with 1070 I.P. compar tment , and a t est session (Da y 4). 1071 E. Preference ind ex t oward I .N .P. chamber on t est versus pr e-test d ays in ChR2 (b lue) and control (grey) 1072 mice. N = 9 mice per group; *p < 0.05; tw o-way ANOVA with Sidak's mu ltip le com parisons. 1073 F. Preferenc e index toward I.N .P. chamb er on t est versus pr e-test d ays in hM3Dq (green) and control 1074 (grey) mice. (n = 7 mi ce per group; * *p < 0.01, ** *p < 0.001, *** *p < 0.0001 ; two- way ANOVA with 1075 Sidak's multiple compa risons). 1076 G. Schematic of Pavlovian training pro to col. From Days 1–7, a condition ed stimul us (CS⁺) was paired 1077 with food delivery, while a second s timulus (CS⁻ ) was never paired. Day 8 served a s a probe t est. 1078 H . Lat ency to re trieve food (n = 7 mice). 1079 I. Mean Ce A Isl1 neuron ac tivity in respons e to CS⁺ (green) and CS⁻ (black) cues across training days. 1080 1081 Figure 5. CeA Isl1 → PCRt terminal stimulation promot es obje ct-dir ected biting be havior. 1082 A . Repr esen tative images of brai n are as innervat ed by CeA Isl1 neurons. Scal e bar , 100 μm. 1083 B. Major br ain regions inn ervat ed by CeA Isl1 neurons depict ed as perc ent age (mean ± s.e.m .) of total 1084 outputs (n = 3 mice). 1085 C. Schematic showing viral ta rgeti ng of C eA Isl1 neurons a nd te rminal stimula tion i n the PCRt using 473 1086 nm light in Isl1-CreER mice. 1087 D. Rast er plo t showing biting bouts (pink ) and fictive feeding bouts (orange) acros s individual ChR2 mice 1088 during alte rna ting light OFF and light ON epochs in biting Styrofoam an d food assa y. 1089 E-G. Quantification of beh aviors during O N and OFF pe riods for bo th ChR2 (pink) mice and EGFP (grey) 1090 mice in biting assay. E. Duration of biti ng Styrofoam. F. Dura tion of biting food. G. Duration of fictive 1091 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 5, 2026. ; https://doi.org/10.64898/2026.02.03.703447doi: bioRxiv preprint 28 feeding. N = 5 mice pe r group. Two-way ANOVA with Tukey's pos t hoc tes t , *PRiUD< RiUD0.05, ** *PRiUD<RiUD0.001, 1092 *** *PRiUD<RiUD0.0001. 1093 H. Rep resen ta tive hea tmap showing occupancy plots of CeA Isl1 → PCRt ChR2 mic e. 1094 I. Quantifica tion of time spe nt in th e ligh t-paired (O N) and non-paire d (OFF) compartmen ts (n = 5 mice 1095 per group; N.S. , two-tail ed unpai red t- te st). 1096 J. Average vel ocity in the light-pai red (O N) and non-paired (OFF) compar tments ( n = 5 mice per group; 1097 N.S. , two-tail ed unpair ed t-t est). 1098 K. Total dis tance moved in th e light-pair e d (ON) and non-paired (OFF) compar tme nts (n = 5 mice per 1099 group; N .S., two-tailed un pair ed t-t est). 1100 L. Schematic showing viral targe ting of CeA Isl1 neurons a nd te rminal stimula tion i n the PPTg using 473 1101 nm light in Isl1-CreER mice. 1102 M. Rast er plo t showing biting bouts (gre en) and fictive feeding bouts (orange) acr oss individual ChR2 1103 mice during alte rnati ng light OFF and ligh t ON ep ochs in biting Styrofoam an d food assay. 1104 N-P. Quan tification of beh aviors during ON and O FF perio ds for both ChR2 (gree n) mice and EGFP (grey) 1105 mice in biting assay. N. Durati on of biting Styrofoam. O. Dura tion of biting food . P. Duration of fictive 1106 feeding. N = 5 mice pe r group. Two-way ANOVA with Tukey's pos t hoc tes t , *PRiUD< RiUD0.05, **PRiUD<RiUD0 .01. 1107 Q. Repres ent ative he atmap showing occ upancy plots of CeA Isl1 → PPtg ChR2 mic e. 1108 R. Quan tification of time sp ent in th e lig ht-paired (O N) and non-pair ed (OFF) compartmen ts (n = 5 mice 1109 per group; *p < 0.05, two-tail ed unpair e d t-test). 1110 S. Average velocity in th e light-pair ed (O N) and non-paired (OFF) compar tments ( n = 5 mice per group; 1111 **p < 0.01 , two-tail ed unpair ed t-t est). 1112 T. Total dist ance moved in th e light-pair e d (ON) and non-paired (OFF) compar tme nts (n = 5 mice per 1113 group; N .S., two-tailed un pair ed t-t est). 1114 Abbrevia tions: B NST, th e bed nucleus of the stri a te rminalis; MiTg, microcellul ar t egmental nucl eus; NTS, 1115 nucleus of the solit ary trac t; PAG , peri aq ueductal gray; PB N, par abrachial nucl eus ; PCRt; parvicellular 1116 reticula r nucleus; PP, perip eduncula r nucleus; PPTg, pedunculopontin e tegmen tal nucleus; PIL, poste rior 1117 intralamin ar th alamic nucleus; PoT, post erior thalamic nuclea r group, triangula r Part; R t, r eticula r 1118 thalamic nucleus; S NL, substan tia nigra , later al par t; SNC, subst anti a nigra, compa ct part ; STh, 1119 subthalamic nucleus ; Su5, supra trigemin al nucleus; scp superio r cere bella r pedun cle (brachium 1120 conjunctivum); VP L, ventral post erola te r al thalamic nucleus ; VPM, ventral pos ter omedial tha lamic 1121 nucleus. 1122 1123 1124 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 5, 2026. ; https://doi.org/10.64898/2026.02.03.703447doi: bioRxiv preprint Biting Non-biting -20 0 20 40 60 80 Mean activity ✱✱✱✱ Softwood Time from event (s) 0 2 4 6 8-2-4-6 Down modulated Up modulated Non modulated 65.6 % 23.8% 10.6 % Styrofoam Time from event (s) 0 2 4 6 8-2-4-6 67.6 % 17.6 % 14.8 % Sponge Time from event (s) 0 2 4 6 8-2-4-6 62.5 % 26.6 % 10.9 % Cricket Time from event (s) 0 2 4 6 8-2-4-6 58.8 % 34.1 % 7.1 % Chow Time from event (s) 0 2 4 6 8-2-4-6 58.0 % 31.3 % 10.7 % 200 150 100 50 0 Neurons D First bite First bite First bite First bite First bite 2z-0.5z AAV-DIO-Gcamp8m B C Isl1-CreER AAV-DIO-Gcamp8m Softwood Styrofoam Sponge Cricket ChowH I J K Figure 1 E F G ROIs Biting Non-biting -20 0 20 40 60 80 Mean activity ✱✱ Biting Non-biting -20 0 20 40 60 80 Mean activity ✱✱✱✱ Biting Non-biting -20 0 20 40 60 80 Mean activity ✱✱✱ 0.4z Biting Non-biting -20 0 20 40 60 80 Mean activity ✱✱ Ai9/Isl1 CeL CeM A Pkcδ-Cre:Ai9 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 5, 2026. ; https://doi.org/10.64898/2026.02.03.703447doi: bioRxiv preprint 1 2 3 4 5 -1 0 1 2 3 4 Mean activity (z score) ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ 1 2 3 4 5 -0.5 0.0 0.5 1.0 Mean activity (z score) ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ Sponge Figure 2 Hard 347 1 Soft 0 8-6 0 8-6 Time from event (s) J Softwood Sponge Styrofoam 0 8-6 0 3Time from event (s) 147 1 A Silicone 0 8-6 0 8-6 0 8-6 Cluster1 Cluster2 Cluster3 Cluster4 Cluster5 Softwood Sponge StyrofoamSilicone 2 0 Time from event (s) B Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 80 8-6 0 8-6 0-6 Time from event (s) 80 8-6 0 8-6 0-6 0 8-6 0 8-6 Sponge Softwood Styrofoam Quinine Water Vanilla Chocolate StrawberryG 0 4 1 196 26.5 % 24.3 % 22.1 % 17.6 % 9.6 % Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 C D 29.8 % 20.2 % 16.7 % 30.8 % 2.5 % Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 H I Sponge Styrofoam SoftwoodSilicone Cluster Strawberry Styrofoam QuinineChocolate 0 0 0 0 Time from event (s) Vanilla Water Softwood Cluster Up Down Non 0 50 100 Fraction of neurons z score Mean activity 0-6 L 8 Non-modulated 0-6 8 -0.1 0.0 0.1 0.2 0.3 0.4 M Down-modulated 0-6 8 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 0-6 8 Up-modulated -0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 N O HardSoft K PC1 PC3 PC2 TimeSilicone Styrofoam Softwood Sponge E F PC10 1 2 -1 -2 0 1 -1 -2 0 1 2-1-2-3-4 0 1 2-1-2-3-4 Time (s) PC2 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 5, 2026. ; https://doi.org/10.64898/2026.02.03.703447doi: bioRxiv preprint ChR2 GFP 0 50 100 150 200 250 Bout duration (s) ✱ 0 Hz 5 Hz 20 Hz 0 5 10 15 20 Mean number of biting bouts ✱ ✱ ChR2 EGFP 0.00 0.05 0.10 0.15 0.20 0.25 Linguine consumed (g) Fictive feeding ON OFF ON OFF 0 5 10 15 20 25 Time of fictive feeding (%) ✱ ✱ ✱ ✱ ✱ ON OFF ON OFF 0 20 40 60 80 Time of biting food (%) ✱ ✱ ✱ ✱ ✱ ON OFF ON OFF 0 20 40 60 80 100 Time of biting styrofoam (%) ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ON OFF ON OFF 0 10 20 30 40 Mean number of styrofoam-biting bouts ✱ ✱ ✱ ✱ ✱ ✱ ✱ ON OFF ON OFF 0 5 10 15 20 Mean number of food-biting bouts ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ON OFF ON OFF 0 5 10 15 20 25 Mean number of fictive feeding bouts ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ON OFF ON OFF 0 2 4 6 8 Mean number of fictive feeding bouts ✱ ✱ ChR2-1 ChR2-2 ChR2-3 ChR2-4 ChR2-5 ChR2-6 ChR2-7 ChR2-8 ChR2-9 ChR2-10 ChR2-11 ChR2-12 ChR2-13 50 100 150 200 250 300 350 Time (s) 0 ChR2-1 ChR2-2 ChR2-3 ChR2-4 ChR2-5 ChR2-6 ChR2-7 ChR2-8 ChR2-9 ChR2-10 ChR2-11 ChR2-12 ChR2-13 50 100 150 200 250 300 350 Time (s) 0 Biting bouts ONOFF ONOFF ONOFF ONOFF ONOFF ONOFF Biting food Biting styrofoam ON OFF ON OFF 0 10 20 30 The time of fictive feeding Eating linguine B C D Fictive Feeding Biting 0 1 2 3 4 Distance (cm) ✱ ✱ ✱ ✱ Distance (cm) 0 1 2 3 4 5 6 7 8 Probability density0.0 0.2 0.4 0.6 0.8 Fictive feeding Biting Isl1-CreER mouse A E G H I J K L M N O P Q R Figure 3 AAV-DIO-EGFP AAV-DIO-ChR2-EYFP 473 nm F ON OFF ON OFF 0 200 400 600 Bout duration (s) (during linguine assay) ✱ ✱ ✱ ✱ AAV-DIO-EGFP AAV-DIO-Arch3.0-EYFP 532 nm Isl1-CreER mouse S UT Second since first bite 0 200 400 600 Arch-OFF Arch-ON EGFP-ON EGFP-OFF Eating bouts Long bouts EMG TemporalisMasseter Myomatrix arrays Masseter Temporalis hM4Di-Saline hM4Di-CNO 10 s5 mV Saline CNO 0 500 1000 1500 Amplitude (μV) ✱ ✱ ✱ ✱ Saline CNO 0 500 1000 1500 2000 Amplitude (μV) ✱ ✱ ✱ ✱ V W X YMasseter Temporalis (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 5, 2026. ; https://doi.org/10.64898/2026.02.03.703447doi: bioRxiv preprint Light No light EGFPChR2 Min Max RTPP 1 2 3 4 0 500 1000 Training day Nose pokes ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ChR2-active ChR2-inactive EGFP-active EGFP-inactive Pre-test Test Pre-test Test -1.0 -0.5 0.0 0.5 Preference index for I.N.P chamber ✱ Day 1 Day 2 Day 3 Day 4 Pre- test I.N.P~Light; after 4 h; I.P~No light I.N.P~Light; after 4 h; I.P~No light Test Pre-test Test Pre-test Test -1.0 -0.5 0.0 0.5 1.0 Preference index for I.N.P chamber ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ Days 1-7 CS+ Day 8 Food CS+ CS- CS- 0 2 4 6 8 0 10 20 30 40 Day Latency to food (s) 1 2 3 4 5 6 7 8 0 2 4 6 8 Day Mean activity Pavlovian Training CPP ON OFF ON OFF 0 20 40 60 80 100 Time spent in compartments (%) ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ICSS Figure 4 A B C D E F G H I Day 1 Day 2 Day 3 Day 4 Day 5 Day 6 Pre- test I.N.P~ CNO I.P~ Veh I.N.P~ CNO I.P~ veh Test D2 D1 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 5, 2026. ; https://doi.org/10.64898/2026.02.03.703447doi: bioRxiv preprint ON OFF 0 10 20 30 40 50 Time of fictive feeding (s) ✱ ✱ ✱ PCRt Bregma:0.14 mm BNST -3.08 mm -4.36 mm -5.34 mm -5.80 mm -7.20 mm ic 4V PCRt NTS SNLSNC PPTg PBN 12N Sp5 scp A Figure 5 % total output Su5 PPTg MiTg PBN PCRt SNL SNC BNST Rt PIL VPL VPM 0 5 10 15 20 PP PoT STh NTS B Midbrain Palidum Subthalamus Thalamus Hindbrain OFF ON OFF 0 1 2 3 4 5Velocity cm/s ✱✱ ON OFF 0 500 1000 1500 2000 2500 Distance moved (cm) ON OFF 0 500 1000 1500 Duration in the compartment (s) ✱ CeAIsl1→ PPTg ON OFF 0 1000 2000 3000 Distance moved (cm) ON OFF 500 600 700 800 900 Duration in the compartment (s) ON OFF 0 1 2 3 4 5Velocity cm/s CeA Isl1 → PCRt I J 473 nm CeA AAV-DIO-ChR2-EYFP C RTPP ON OFF CeA 473 nm AAV-DIO-ChR2-EYFP PPTg RTPP ON D ON OFF 0 50 100 150 Time of biitng styrofoam (s) ✱ ✱ ✱ ON OFF 0 10 20 30 40 50 Time of biitng food (s) ✱✱✱✱ ✱✱✱ ✱✱✱✱ E Food Styrofoam ONONON 50 100 150 200 250 300 Time (s) 0 OFF OFF OFF ChR2-1 ChR2-2 ChR2-3 ChR2-4 ChR2-5 ChR2-1 ChR2-2 ChR2-3 ChR2-4 350 F Food Styrofoam ONONON 50 100 150 200 250 300 Time (s) 0 OFF OFF OFF ChR2-1 ChR2-2 ChR2-3 ChR2-4 ChR2-5 ChR2-1 ChR2-2 ChR2-3 ChR2-4 350 ChR2-5 ON OFF 0 50 100 150 200 Time of biitng styrofoam (s) ✱✱ ✱✱ ✱✱ ON OFF 0 50 100 150 200 Time of biting food (s) ✱ ✱ ✱Fictive feeding Biting bouts Fictive feeding Biting bouts G ON OFF 0 5 10 15 20 Time of fictive feeding (s) ✱✱✱✱ ✱✱✱✱ ✱✱✱✱ H K L M N O P Q R S TP Q R S (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 5, 2026. ; https://doi.org/10.64898/2026.02.03.703447doi: bioRxiv preprint Gnawing down modulated Chewing Non chewing -20 0 20 40 60 80 Mean activity Gnawing Non gnawing -20 0 20 40 60 80 Mean activity ✱✱ Gnawing Chewing -20 0 20 40 60 80 Mean activity ✱✱ Non modulated Isl1-CreER:Ai9 Isl1/Ai9/DAPI B C D Isl1+ Isl1- 0 50 100 Percentage of Ai9+ cells (%) Ai9+ Ai9- 0 50 100 Percentage of Isl1+ cells (%) AAV-DIO-EGFP Isl1-CreER A Softwood Styrofoam Sponge Cricket Chow 0 10 20 30 40 Mean number of biting bouts E Softwood Styrofoam Sponge Cricket Chow 0 50 100 150 200 250 Bite duration (s) ✱✱ ✱✱✱ ✱✱ F Chewing modulated 97.1 % 2.9 %Gnawing up modulated I J K L M 32.3 % 52.6 % 15.1 % Gnawing Holding -10 0 10 20 30 40 Mean activity Gnawing Pursuit -10 0 10 20 30 40 Mean activity 30.3 % 38.5 % 31.2 % 6.6 % 93.4 % 4.3 % 95.7 % Non modulated Holding modulated Non modulated Pursuit modulated Cricket hunting Feeding behavior N O P Q R H First bite Gnawing Chewing 0 20 40 60 80 100 120 140 180 0 5z score Isl1 Prkcd Sst Drd2 Penk Calcrl Dlk1 Nr2f2 Pnoc Drd1 Tac1 Tac2 Htr2a -1 2 Time from event (s) 0 2 4 6-2-4-6 8 0.4z Softwood Styrofoam Sponge Cricket ChowG (s) Non modulated Gnawing down modulated Gnawing up modulated Non modulated (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 5, 2026. ; https://doi.org/10.64898/2026.02.03.703447doi: bioRxiv preprint Figure. S1. CeAIsl1 neurons were activated at the bite onset. A. Representative histological image of EGFP expression in the CeA. Scale bar, 100 μm. B. Representative image showing Isl1 immunostaining (green) and Ai9 reporter expression (red) with DAPI (blue) in an Isl1-CreER: Ai9 mouse. The CeA subregions are outlined. Scale bar, 30 μm. C. Quantification of the proportion of Ai9⁺ neurons expressing Isl1 and the proportion of Isl1⁺ neurons expressing Ai9 (n = 3 mice). D, Heatmap of marker gene expression for CeA neuronal subpopulations. Cells are grouped and color-coded by transcriptionally defined clusters (top), as described in a previous study1. Red outlines and red text highlight the Isl1- positive population. E. Quantification of mean number of biting bouts for objects with varying physical properties (n=10 mice; N.S.; one-way ANOVA test) F. Quantification of bite duration for objects with varying physical properties (n=10 mice; ***p <0.001, **p <0.01; one - way ANOVA test). G. Average responses of the z-scored calcium activity for each condition: softwood, Styrofoam, sponge, cricket, and chow. Shaded areas represent s.e.m. H. Representative z-scored calcium trace from a single neuron aligned to behavioral events: first bite (red), eating (blue), and chewing (green). I-M, Feeding behavior. I. Mean neuronal activity during gnawing and chewing epochs (n=10 mice; **P < 0.01; Wilcoxon signed-rank test). J. Mean neuronal activity during gnawing versus non-gnawing epochs (n=10 mice; **P < 0.01; Wilcoxon signed-rank test). K. Mean neuronal activity during chewing versus non-chewing epochs (n=10 mice; N.S.; Wilcoxon signed-rank test ). L. Proportion of neurons up-modulated, down-modulated, and non-modulated during gnawing. M. Proportion of neurons modulated during chewing. N-R. Cricket hunting behavior. N. Mean neuronal activity during gnawing and pursuit epochs (n=10 mice; N.S.; Wilcoxon signed-rank test ). O. Mean neuronal activity during gnawing and holding epochs (n=10 mice; N.S.; Wilcoxon signed-rank test ). P. Proportion of neurons up-modulated, down-modulated, and non-modulated during gnawing. Q. Proportion of neurons modulated during pursuit. R. Proportion of neurons modulated during holding. (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 5, 2026. ; https://doi.org/10.64898/2026.02.03.703447doi: bioRxiv preprint Strawberry Chocolate Vanilla Water Quinine 0 5 10 15 Flavor food bite frequency Mean number of biting bouts Cluster 5Cluster 4Cluster 3Cluster 2 Strawberry Chocolate Vanilla Water Quinine 0 5 10 15 Flavor sponge bite frequency Mean number of biting bouts Strawberry Chocolate Vanilla Water Quinine 0 50 100 150 Flavor food bite duration Bite duration (s) Chocolate Quinine Strawberry Vanilla Water 200 40 80 120 160 2 1 0 2 1 0 40 80 120 160 Chocolate Quinine Strawberry Vanilla Water Flavored food Flavored sponge Mean activity 0 -0.1 -0.2 0.1 0.2 0.3 0.4 0-2-4-6 2 4 6 Mean activity 0 -1 -2 1 2 3 0-2-4-6 2 4 6 0 -0.2 0.1 0.2 0.3 0.4 -0.1 -0.3 Mean activity Mean activity 0 -1 -2 1 2 3 0-6 8 0-6 8 0-6 8 0-6 8 0-6 8 0-6 8 0-6 8 0-6 8 0-6 8 0-6 8 Time from event (s) Time from event (s) Time from event (s) Time from event (s) D E F G H I J K Strawberry Chocolate Vanilla Water Quinine 0 50 100 150 200 250 Flavor sponge bite duration Bite duration (s) L M Number of neurons Neural response delay (s) after biting onset Silicone 0 5 10 15 20 0 2 4 6 8 Softwood 0 5 10 15 20 0 2 4 6 8 Number of neurons Sponge 0 5 10 15 20 0 2 4 6 8 Number of neurons Styrofoam 0 5 10 15 20 0 2 4 6 8 Number of neurons A C Mouse 1 Mouse 2 Mouse 3 Mouse 4 Mouse 5 Mouse 6 Cluster 1 Silicone Softwood Sponge Styrofoam Pearson correlation 0 1 B (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 5, 2026. ; https://doi.org/10.64898/2026.02.03.703447doi: bioRxiv preprint Figure. S2. Neuronal responses to flavored food and flavored sponge. A. Histograms showing the distribution of response delays (s) after biting onset for neurons recorded during interactions with silicone (blue), softwood (orange), sponge (green), and Styrofoam (purple). Each bar indicates the number of neurons exhibiting peak activity within the corresponding time bin. B. Correlation coefficient matrixes of the responses of all neurons for each condition. C. The spatial locations of individual extracted neurons in the field of view (FOV) in the CeA of six mice. clusters labeled by different color code. D. Heatmaps showing normalized activity of individual neurons aligned to bite onset for different flavored food. Each row represents one neuron, and color scale indicates z-scored activity (n= 4 mice). E. Mean population activity traces for each flavored food, showing the average z-scored activity aligned to bite events. Shaded areas represent ± s.e.m. F. Box plots summarizing mean activity across different flavored food (n=4 mice; N.S.; Wilcoxon signed-rank test). G. Heatmaps showing normalized activity of individual neurons aligned to bite onset for different flavored sponge. Each row represents one neuron, and color scale indicates z-scored activity (n= 3 mice). H. Mean population activity traces for each flavored sponge, showing the average z-scored activity aligned to bite events. Shaded areas represent ± s.e.m. I. Box plots summarizing mean activity across different flavored sponge (n=3 mice; N.S.; Wilcoxon signed-rank test). J. Quantification of mean number of biting bouts for different flavored food (n=4 mice; N.S.; Wilcoxon signed-rank test). K. Quantification of bite duration for different flavored food (n=4 mice; N.S.; Wilcoxon signed-rank test). L. Quantification of mean number of biting bouts for different flavored sponge (n=3 mice; N.S.; Wilcoxon signed-rank test). M. Quantification of bite duration for different flavored sponge (n=3 mice; N.S.; Wilcoxon signed-rank test). (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 5, 2026. ; https://doi.org/10.64898/2026.02.03.703447doi: bioRxiv preprint Licking Non-licking -20 0 20 40 60 80 Mean activity ✱ Water Licking Non-licking -20 0 20 40 60 80 Mean activity ✱ Quinine Licking Non-licking -20 0 20 40 60 80 Mean activity ✱✱✱✱ Chocolate milk Licking Non-licking -20 0 20 40 60 80 Mean activity ✱✱✱✱ Strawberry milk Licking Non-licking -20 0 20 40 60 80 Mean activity ✱ Vanilla milk Strawberry Chocolate Vanilla Water Quinine 0 5 10 15 20 Mean number of licking bouts Strawberry Chocolate Vanilla Water Quinine 0 50 100 150 200 Lick duration (s) Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 Quinine Water Vanilla Chocolate Strawberry 0.0 1.5 A C D E F G H I Pearson correlation Chocolate milk 0 1 Quinine Softwood Sponge Strawberry milk Styrofoam Vanilla milk Water Chocolate milk Quinine Softwood Sponge Strawberry milk Styrofoam Vanilla milk Water B Figure. S3. The encoding properties of CeAIsl1 neurons. A. Summary matrix of mean peak response across stimuli in each cluster. Dot size and color represent average response magnitude, and density plots indicate the distribution of neuron responses to each stimulus. B. Correlation coefficient matrixes of the responses of all neurons for each condition. C-G. Mean neuronal activity during licking versus non-licking epochs: water (C), quinine (D), chocolate milk (E), strawberry milk (F), and vanilla milk (G) (n= 6 mice; *P < 0.05, ****P < 0.0001, Wilcoxon signed-rank test). H. Quantification of mean number of licking bouts or different flavored milk (n=6 mice; N.S.; one-way ANOVA test). I. Quantification of lick duration for different flavored milk (n=6 mice; N.S.; one-way ANOVA test). (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 5, 2026. ; https://doi.org/10.64898/2026.02.03.703447doi: bioRxiv preprint ON OFF ON OFF 0 10 20 30 Mean number of fictive feeding bouts ✱ ✱ ON OFF ON OFF 0 20 40 60 80 100 Time of biting food (%) ✱ ✱ ON OFF ON OFF 0 5 10 15 20 Mean number of food biting bouts ✱ ✱ ✱ ✱ Biting food (Food deprived for 4 h) Fictive feeding ChR2-1 ChR2-2 ChR2-3 ChR2-4 ChR2-5 ChR2-6 ChR2-7 ChR2-8 ChR2-9 ChR2-10 ChR2-11 ChR2-12 ChR2-13 50 100 150 200 250 300 350 Time (s) 0 Biting bouts ONOFF ONOFF ONOFF A ON OFF ON OFF 0.0 0.2 0.4 0.6 0.8 Food intake (g) hM3Dq-saline hM3Dq-CNO mCherry-saline mCherry-CNO 20 min off20 min on 30 60 120 180 0.0 0.1 0.2 0.3 0.4 0.5 Food intake (g) Time (minutes) B C D E F G I J K L ON OFF ON OFF 0 10 20 30 40 50 Time of fictive feeding (%) ✱ ✱ ON OFF ON OFF 0 5 10 15 20 Cricket - Capture Time (s) ✱ ON OFF ON OFF 0 20 40 60 80 100 Cricket - Eat Time (s) ON OFF ON OFF 0.0 0.5 1.0 1.5 2.0 2.5 Cricket - Pursuit Time (s) ON OFF ON OFF 0 5 10 15 20 Cricket - Investigation Time (s) ChR2 EGFP 0 1 2 3 Crickets killed Crickets # ON OFF ON OFF 0 20 40 60 80 Cricket - Fictive feeding Time (s) ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ M N O H AAV-DIO-hM3Dq Isl1-CreER mouse OFF ON OFF 0 2 4 6 8Time in object area (s) OFF ON OFF 0 2 4 6 8 Mean number of entering object area P Q R Figure. S4. Activation of CeAIsl1 neurons in feeding, hunting, and exploratory behaviors. A. Raster plot showing biting bouts (blue) and fictive feeding (orange) across individual 4-h food-deprived ChR2 mice during alternating light OFF and light ON epochs in biting food assay. B-E. Quantification of behaviors during light ON and OFF periods for both ChR2 mice (n = 14) and EGFP mice (n = 4) in food biting assay. B. Percentage of time spent biting food. C. Mean number of f food-biting bouts. D. Percentage of time engaged in fictive feeding. E. Mean number of fictive feeding bouts. Two-way ANOVA with Tukey's post hoc test, **P < 0.01, ****P < 0.0001. Blue is ChR2 group, and grey is EGFP. F. Schematic of the paradigm for testing the effects of photoactivation on feeding behavior. G. Food intake by fed ChR2 mice (n = 16) and EGFP mice (n = 9). N.S.; two-way ANOVA with Tukey's post hoc test. H. Schematic of CeAIsl1 neurons expressing either the excitatory DREADD hM3Dq or control mCherry. I. Cumulative food intake of fed Isl1-CreER animals that expressing the excitatory DREADD hM3Dq or control mCherry in the CeA, after i.p. injections of saline and CNO (2 mg/Kg). N = 7 mice per group; N.S.; two-way ANOVA with Tukey's post hoc test. J-O. Quantification of different behavioral epochs during light ON and OFF period for both ChR2 mice (n = 13) and EGFP mice (n = 7) in cricket hunting tasks. J. Time spent investigating crickets. K. Time spent pursuing crickets. L. Time spent capturing crickets. M. Time spent on fictive feeding. N. Time spent eating crickets. O. Number of crickets killed. Two-way ANOVA with Tukey's post hoc test, *P < 0.05, ***P < 0.0001, ****P < 0.0001. P. Schematic of the paradigm for testing the effects of photoactivation on novel object exploration. Q. Time spent in the object area for both ChR2 mice (n = 8) and EGFP mice (n = 3); N.S.; two-way ANOVA with Tukey's post hoc test. R. Mean number f entering the object area for both ChR2 mice (n = 8) and EGFP mice (n = 3) N.S.; two -way ANOVA with Tukey's post hoc test. AAV-DIO-mCherry (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 5, 2026. ; https://doi.org/10.64898/2026.02.03.703447doi: bioRxiv preprint ONOFF ONOFF 0 100 200 300 2nd spaghetti duration (s) ✱ ✱ ✱ 1 s tONOFF ONOFF ONOFF ONOFF ONOFF ONOFF0.00 0.05 0.10 0.15 Linguine consumed (g) 2nd 3rd 1st 2nd 3rd ONOFF ONOFF 0 10 20 30 40 Mean number of feeding bouts (during linguine assay) ArchT GFP 0.00 0.05 0.10 0.15 0.20 0.25 Food consumed (g) Food deprived for 20 h ArchT GFP 0.00 0.02 0.04 0.06 0.08 Food consumed (g) Food deprived for 4 h ON OFF ON OFF 0 50 100 150 Bout duration (s) spaghetti C N O ON OFF ON OFF 0.0 0.1 0.2 0.3 Linguine consumed (g) ON OFF ON OFF 0 50 100 150 200 1st spaghetti duration (s) ON OFF ON OFF 0 50 100 150 3rd spaghetti duration (s) G H I B J Eating bouts Arch-OFF Arch-ON EGFP-ON EGFP-OFF Second since first bite 0 200 400 600 ONOFF ONOFF 0 20 40 60 80 Bout duration (s) (during chow assay) ON OFF ON OFF 0.00 0.05 0.10 0.15 0.20 0.25 Chow consumed (g) ON OFF ON OFF 0 5 10 15 20 25 Mean number of feeding bouts (during chow assay) K L M A D Eating spaghetti F E Arch-OFF Arch-ON EGFP-ON EGFP-OFF Eating bouts 2nd spaghetti Second since first bite 0 200 400 600 Saline CNO 0 500 1000 1500 2000 2500 Softwood Masseter Amplitude (μV) ✱ ✱ Saline CNO 0 500 1000 1500 2000 2500 Softwood Temporalis Amplitude (μV) ✱ ✱ ✱ Saline CNO 0 500 1000 1500 2000 2500 Styrofoam Masseter Amplitude (μV) ✱ ✱ Saline CNO 0 500 1000 1500 2000 2500 Styrofoam Temporalis Amplitude (μV) ✱ ✱ ✱ ✱ Saline CNO 0 500 1000 1500 2000 2500 Soft silicone Temporalis Amplitude (μV) ✱ ✱ ✱ Saline CNO 0 500 1000 1500 2000 2500 Soft silicone Masseter Amplitude (μV) Saline CNO 0 500 1000 1500 2000 2500 Sponge Masseter Amplitude (μV) ✱ Saline CNO 0 500 1000 1500 2000 2500 Sponge Temporalis Amplitude (μV) P Q R S T U V W (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 5, 2026. ; https://doi.org/10.64898/2026.02.03.703447doi: bioRxiv preprint Figure. S5. Inhibition of CeAIsl1 neurons does not affect food consumption. A. Mean number of feeding bouts across light OFF and ON epochs in Arch3.0 and EGFP groups during linguine consumption assays (N.S.; two-way ANOVA with Tukey's post hoc test). B. Cumulative consumption of the first, second, and third linguine pieces across light OFF and ON epochs in Arch3.0 and EGFP groups (N.S.; two-way ANOVA with Tukey's post hoc test). C. Linguine intake by 4-h food-deprived Arch3.0 mice (n = 7) and EGFP mice (n = 5) during alternating light OFF and light ON epochs. N.S.; two-way ANOVA with Tukey's post hoc test. D. Example frame of a mouse consuming spaghetti. E. Raster plots showing eating bouts (black) and consumption duration of the second spaghetti piece (green) across light OFF and ON epochs in Arch3.0 and EGFP groups during spaghetti consumption assays. F. Duration of second spaghetti consumption across light OFF and ON epochs in Arch3.0 and EGFP groups (n = 7 mice per group; *p < 0.05, **p < 0.01; Two-way ANOVA with Tukey's post hoc test). G. Duration of first spaghetti consumption across light OFF and ON epochs in Arch3.0 (n = 8) and EGFP (n = 5) groups. N.S.; two-way ANOVA with Tukey's post hoc test. H. Duration of third spaghetti consumption across light OFF and ON epochs in Arch3.0 (n = 8) and EGFP (n = 5) groups. N.S.; two-way ANOVA with Tukey's post hoc test. I. Duration of individual feeding bouts across light OFF and ON epochs in Arch3.0 (n = 8) and EGFP (n = 5) groups during spaghetti consumption assays. N.S.; two-way ANOVA with Tukey's post hoc test. J. Raster plots showing eating bouts (orange for Arch3.0 and grey for EGFP) across light OFF and ON epochs in Arch3.0 and EGFP groups during chow consumption assays. K. Duration of individual feeding bouts across light OFF and ON epochs in Arch3.0 and EGFP groups during chow consumption assays (n = 7 mice per group; N.S.; Two-way ANOVA with Tukey's post hoc test). L. Mean number of feeding bouts across light OFF and ON epochs in Arch3.0 and EGFP groups during chow consumption assays (n = 7 mice per group; N.S.; Two-way ANOVA with Tukey's post hoc test). M. Cumulative consumption of the chow across light OFF and ON epochs in Arch3.0 and EGFP groups (n = 7 mice per group; N.S.; Two-way ANOVA with Tukey's post hoc test). N. Food intake by 20-h food-deprived Arch3.0 mice (n = 7) and EGFP mice (n = 5). N.S.; two-way ANOVA with Tukey's post hoc test. O. Food intake by 4-h food-deprived Arch3.0 mice (n = 7) and EGFP mice (n = 5). N.S.; two-way ANOVA with Tukey's post hoc test. P. EMG amplitude measurements from the masseter muscle during softwood biting after saline or CNO (0.4 mg kg⁻¹) injection in Isl1-CreER; hM4Di mice. n=4; **P < 0.01; paired t-test. Q. EMG amplitude measurements from the temporalis muscle during softwood biting after saline or CNO (0.4 mg kg⁻¹) injection in Isl1-CreER; hM4Di mice. n=4; ***P < 0.001; paired t-test. R. EMG amplitude measurements from the masseter muscle during Styrofoam biting after saline or CNO (0.4 mg kg⁻¹) injection in Isl1-CreER; hM4Di mice. n=4; **P < 0.01; paired t-test. S. EMG amplitude measurements from the temporalis muscle during Styrofoam biting after saline or CNO (0.4 mg kg⁻¹) injection in Isl1-CreER; hM4Di mice. n=4; ****P < 0.0001; paired t-test. T. EMG amplitude measurements from the masseter muscle during soft silicone biting after saline or CNO (0.4 mg kg⁻¹) injection in Isl1-CreER; hM4Di mice. n=4; N.S; paired t-test. U. EMG amplitude measurements from the temporalis muscle during soft silicone biting after saline or CNO (0.4 mg kg⁻¹) injection in Isl1-CreER; hM4Di mice. n=4; ***P < 0.001; paired t-test. V. EMG amplitude measurements from the masseter muscle during sponge biting after saline or CNO (0.4 mg kg⁻¹) injection in Isl1-CreER; hM4Di mice. n=4; *P < 0.05; paired t-test. W. EMG amplitude measurements from the temporalis muscle during sponge biting after saline or CNO (0.4 mg kg⁻¹) injection in Isl1-CreER; hM4Di mice. n=4; N.S; paired t-test. (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 5, 2026. ; https://doi.org/10.64898/2026.02.03.703447doi: bioRxiv preprint Figure. S6. Brain network architecture of CeAIsl1 neurons A. Mean number of Styrofoam-biting bouts during ON and OFF periods for both CeA Isl1 → PCRt ChR2 (pink) mice and EGFP (grey) mice in biting assay. N = 5 mice per group. Two-way ANOVA with Tukey's post hoc test, ***P < 0.001, ****P < 0.0001. B. Mean number of food-biting bouts during ON and OFF periods for both CeA Isl1 → PCRt ChR2 (pink) mice and EGFP (grey) mice in biting assay. N = 5 mice per group. Two-way ANOVA with Tukey's post hoc test, ****P < 0.0001. C. Mean number of Styrofoam-biting bouts during ON and OFF periods for both CeA Isl1 → PPtg ChR2 (green) mice and EGFP (grey) mice in biting assay. N = 5 mice per group. Two-way ANOVA with Tukey's post hoc test, **P < 0.01, ***P < 0.001. D. Mean number of food-biting bouts during ON and OFF periods for both CeA Isl1 → PPtg ChR2 (green) mice and EGFP (grey) mice in biting assay. N = 5 mice per group. Two-way ANOVA with Tukey's post hoc test, N.S.. E. Schematic of the strategy for monosynaptic retrograde rabies virus (RV) tracing. F. Representative image of the injection site and input areas. High-magnification panel indicate starter cells expressing G + TVA (green) and RV (red) (arrows). Scale, 30 μm. G. Major brain regions projecting to CeAIsl1 neurons depicted as percentage of total inputs (n = 3 mice). Abbreviations: ACo, anterior cortical amygdaloid nucleus; AI, agranular insular cortex; AID, agranular insular cortex, dorsal part; AIP, agranular insular cortex, posterior part; AIV, agranular insular cortex, ventral part; APir, amygdalopiriform transition area; BNST, the bed nucleus of the stria terminalis; BLA, basolateral amygdaloid nucleus, anterior part; BLP, basolateral amygdaloid nucleus, posterior part; BMP, basomedial amygdaloid nucleus, posterior part; BMA, basomedial amygdaloid nucleus, anterior part; CeL, lateral part of central amygdala; CPu, caudate putamen (striatum); CxA, cortex-amygdala transition zone; DI, dysgranular insular cortex; DLO, dorsal lateral olfactory tract; D3V, dorsal 3rd ventricle; FrA, frontal association cortex; GP, globus pallidus; IPAC, interstitial nucleus of the posterior limb of the anterior commissure; LA, lateral amygdaloid nucleus; LH, lateral hypothalamic area; MePD, medial amygdaloid nucleus, posterodorsal part; MePV, medial amygdaloid nucleus, posteroventral part; M2, secondary motor cortex; MO, medial orbital cortex; PAG, periaqueductal gray; PBN, parabrachial nucleus; PPTg, pedunculopontine tegmental nucleus; PLCo, posterolateral cortical amygdaloid nucleus; PVA, paraventricular thalamic nucleus, anterior part; Po, posterior thalamic nuclear group; PIL, posterior intralaminar thalamic nucleus; PRh, perirhinal cortex; PSTh, parasubthalamic nucleus; Pir, piriform cortex; SNL, substantia nigra, lateral part; STh, subthalamic nucleus; scp superior cerebellar peduncle (brachium conjunctivum); Tu, olfactory tubercle; VP, ventral pallidum; VPM, ventral posteromedial thalamic nucleus; ZI, zona incerta. PCRtCeA 1. AAV-FDIO-RVG+ AAV-FDIO-TVA 2. After 4 week, rabies virus (RV)rAAV-DIO-Flp Isl1-CreER mouse E RFP(RV)/G+TVA/DAPI Bregma: -1.06 mm -1.34 mm -1.58 mm 2.58 mm 0.14 mm 0.14 mm -0.34 mm -0.58mm -1.34 mm -2.30 mm -2.30 mm -2.46 mm -5.34 mm FrA DI AIP BNST PVA D3V LH BLP BMP PLCo Po PF PSThcp VPM scp PBN CeM CeC BLA CeMAV BLA BMA CeLCeM BLA CeL CeM BMA F BLA BMP BMA BLP ACo APir PLCo LA PRh PBN PPTg LH PSTh STh ZI AID AIP AIV AI DI PAG M2 CxA VP BNST GP Pir FrA MO DLO IPAC CPuTu VPM Po PIL % total input 0 2.5 5.0 7.5 10.0 12.515.0 17.5 Association Cortex Hindbrain Hypothalamus Amygdala Motor Cortex Olfactory areas Pallidum Insular Cortex Midbrain Prefrontal Cortex Striatum Thalamus Piriform Cortex G MePV MePD CeL SNL ON OFF 0 2 4 6 8 10 Mean number of food-biting bouts ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ON OFF 0 2 4 6 8 Mean number of styrofoam-biting bouts ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ ✱ BA ONOFF 0 10 20 30 Mean number of food-biting bouts ON OFF 0 10 20 30 40 Mean number of styrofoam-biting bouts ✱✱✱ ✱✱ ✱✱✱ C D CeA Isl1 → PCRt CeAIsl1→ PPTg (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 5, 2026. ; https://doi.org/10.64898/2026.02.03.703447doi: bioRxiv preprint Chicken PRKCD PRKCDPrkcd ISL1 ISL1Isl1 Mouse Human Lizard PRKCD ISL1 Isl1 Prkcd Salamander (from stage 36, 41, 46, and 50 larvae telencephalon) PRKCD ISL1 Human Siletti et al., 2023Yu et al., 2023Peters et al., 2023Yu et al., 2023Hain et al., 2022 Woych et al., 2022 A F B G C H D I E J K L M N Lizard Salamander Mouse Human Chicken Figure. S7. Conserved segregation of Isl1 and Prkcd neuronal populations in the central amygdala across vertebrates A–J, Single-cell transcriptomic maps showing expression of Isl1 (A-E) and Prkcd (F-J) across CeA clusters in lizard2, chicken3, mouse1, and human3,4. K–M, UMAP visualization of salamander telencephalon (stages 36, 41, 46, and 50)5 showing annotated brain regions (K) and expression of Isl1 (L) and Prkcd (M). Red outlines indicate putative CeA regions. N, Schematic phylogenetic tree illustrating the species analyzed. (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 5, 2026. ; https://doi.org/10.64898/2026.02.03.703447doi: bioRxiv preprint