Perifornical Area Urocortin-3 Neurons Promote Infant-directed Neglect and Aggression

Mammals invest considerable resources in protecting and nurturing young offspring. However, under certain physiological and environmental conditions, animals neglect or attack young conspecifics. Males in some species attack unfamiliar infants to gain reproductive advantage1–3 and females kill or neglect their young during stressful circumstances such as food shortage or threat of predation4–8. In humans, stress is a risk factor in both sexes for peripartum disorders and associated impairments in parent-infant interactions9. While recent studies have uncovered dedicated neural pathways mediating the positive control of parenting10–13, the regulation of infant-directed neglect and aggression and the relationship between these behaviours and stress are poorly understood. Here we show that urocortin-3 (Ucn3)-expressing neurons in the perifornical area (PeFAUcn3) of the hypothalamus are activated during infant-directed attacks in males and females, but not other forms of aggression. Opto- and chemogenetic manipulations of PeFAUcn3 neurons demonstrate the role of this neuronal population in the negative control of parenting in both males and females. PeFAUcn3 neurons receive input from areas associated with vomeronasal sensing, stress, and parenting, and send major projections to the ventromedial hypothalamus (VMH), ventral lateral septum (LSv) and amygdalohippocampal area (AHi). Optogenetic activation of PeFAUcn3 axon terminals in these regions triggers different aspects of infant-directed agonistic responses, such as neglect and aggression. Thus, PeFAUcn3 neurons emerge as a critical hub for the expression of infant-directed neglect and aggression, providing a new framework to examine the positive and negative regulation of parenting.

Few PeFA Ucn3 neurons were activated in virgin females engaged in pup retrieval to the nest, while this population was entirely silent in lactating females (Fig. 1h). No difference was observed in the number of PeFA Ucn3 neurons in virgin versus mated males or females (Supp. Fig. 1c). Most PeFA Ucn3 neurons expressed the vesicular glutamate transporter 2 (vglut2) suggesting an excitatory function (Supp. Fig. 1d). Thus, Ucn3-positive neurons of the rostral PeFA are specifically activated by pup-directed aggression, but no other forms of adult aggression or other behaviours.

Inhibition of PeFA Ucn3 neuronal activity in virgin males blocks infanticide
To test if PeFA Ucn3 neuronal activity is required for the display of male infanticide, we used a conditional viral strategy to express the inhibitory opsin NpHR3.0, or yellow fluorescent protein (YFP) as a control, in PeFA Ucn3 neurons and implanted bilateral optical fibres above the PeFA of virgin males (Fig. 2a, Supp. Fig. 2a). Before laser stimulation, control and NpHR3.0-injected males showed similarly high levels of pup-directed aggression with short latencies to attack (Fig. 2b-d). By contrast, optogenetic inhibition of PeFA Ucn3 neurons led to a reduction in the number of males displaying pup-directed attacks, an increase in attack latency, and a decrease in the duration of aggressive displays toward pups ( Fig. 2 b-d). Strikingly, the suppression of pupdirected aggression persisted -and even further increased -within the next 24h, suggesting a long-term effect of acute PeFA Ucn3 neuronal silencing ( Fig. 2b-d, 2 nd 'laser-off' session). Males spent equivalent time investigating pups in all conditions and did not increase their parental behaviours within the test period (Fig. 2e, f). Likewise, PeFA Ucn3 silencing did not affect motor activity or adult intruder-directed aggression (Fig. 2g). These data suggest that PeFA Ucn3 neuron activity is required for the regulation of pup-directed aggression.

Activation of PeFA Ucn3 neurons elicit infant-directed neglect in virgin females
certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
Parenting experience strongly impacts subsequent interactions of virgin females with pups 16 and indeed, sessions in which the laser off condition was tested first led to irreversible parental behaviour in mice injected with ChR2 (Supp. Fig. 2c, d). Hence, all sessions were started with the laser ON condition in both experimental and control groups, followed by laser OFF in the second session, and back to laser ON in third session (see methods). As expected, control females became increasingly more parental from session to session, with more pups retrieved and with shorter latencies in each subsequent session (Fig. 2i, j, left panels). By contrast, optogenetic stimulation of PeFA Ucn3 neurons reduced female parenting with fewer pups collected and with longer latencies in the third session compared to first (laser ON) and second (laser OFF) sessions (Fig. 2i, j right panels). Further, the duration of parental behaviours, investigation bout length, and time spent in the nest with pups were significantly reduced in the third compared to earlier sessions, and parental behaviour and time spent in the nest with pups was significantly reduced between control and ChR2 females in the third trial ( Fig. 2 k, l), and fewer females in the ChR2 group retrieved pups to the nest (Fig 2m). While the overall display of pup-directed aggression (duration of aggression and number of females displaying aggression) was not significantly increased (Fig. 2n), a subset of females (n = 3 /15) showed tail-rattling locked to laser stimulation, an overt sign of aggression never observed in control females (Supp. Video 1).
Further, activation of PeFA Ucn3 neurons did not affect motor activity or intruder-directed aggression ( Fig. 2o). Consumption of appetitive food was slightly decreased with activation of PeFA Ucn3 neurons in ChR2 females (Supp. Fig. 2e), in agreement with the previously reported anorexigenic effect of this neural population 17,18 .
To confirm the effects of PeFA Ucn3 neuron activation and assess the effects of prolonged activation in virgin females, we injected a conditional virus expressing the excitatory chemogenetic effector certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
Chemogenetic activation of PeFA Ucn3 neurons led to a decrease in pup retrieval compared to control (Cre negative) females (18% pups retrieved versus 70%, Fisher's exact test p<0.0015) ( Fig. 2q; Kolmogorov-Smirnoff test p<0.001). Further, fewer experimental females displayed parental behaviours compared to controls, and we observed a decrease in overall parental behaviour, reduced time in the nest with pups, and reduced nest building (Fig. 2r, s). Pup-directed aggression was observed in some females after activation of PeFA Ucn3 neurons, and despite diminished parenting, females with activated PeFA Ucn3 neurons investigated pups as much as control females, suggesting they were not avoidant (Fig. 2t). Together, these data suggest that activation of PeFA Ucn3 neurons reduced parenting in virgin females, and induced signs of aggression in a subset of females.

PeFA Ucn3 neurons receive input from parenting and stress-associated cell populations
To identify monosynaptic inputs to PeFA Ucn3 neurons, we used conditional rabies virus mediated retrograde trans-synaptic tracing in adult virgin males and females 19 (Fig. 3a). Fifteen brain areas were found to provide monosynaptic inputs, all located exclusively within the hypothalamus, lateral septum, and medial amygdala, with the highest fraction of inputs originating from PVH, MPOA, and anterior hypothalamus (AH) (Fig. 3b-d). The PVH contained the highest fractional inputs to PeFA Ucn3 neurons in both sexes, with a higher fraction in females (44.92% ± 8.1%) as compared to males (22.95% ± 4.2%) (Fig. 3d). PVH inputs were comprised of AVP (31.9% ± 11.1%), CRH (17.8% ± 7.1%) and OXT (9.1% ± 2.1%) expressing cells, i.e. Thus, PeFA Ucn3 neurons are poised to receive major inputs from vomeronasal and hypothalamic pathways associated with stress, social information, and parenting. certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not this version posted July 9, 2019. ; https://doi.org/10.1101/697334 doi: bioRxiv preprint These data prompted us to test the effect of stress on pup interactions and PeFA Ucn3 neuronal activity. Confirming previous reports 24,25 , chronic restraint stress in virgin females led to weight loss, reduced pup retrieval, and a decreased fraction of parental females (Fig. 3g-i). No difference was observed in the number of c-fos positive PeFA Ucn3 neurons in stressed females interacting with pups as compared to unstressed controls females. However, strikingly, while an inverse correlation was observed between PeFA Ucn3 neuronal activation and parental behaviour in control females (R 2 = 0.52), chronic restraint stress entirely abolished this relationship (R 2 = 0.14) ( Fig. 3k; slope difference p<0.0055). Together these data show that PeFA Ucn3 neurons function as a critical node between vomeronasal, stress, and parenting-related neural pathways, and suggest that chronic stress disrupts their normal function.
To understand the functional organization of PeFA Ucn3 neuronal projections, we performed pairwise injections of fluorophore conjugated cholera toxin B subunit (CTB-488 and CTB-647) retrograde tracers (Fig 3p, q). We focused on PeFA Ucn3 projections to LS, AHi, and VMH, based on their dense labelling ( Fig. 3m-o) and the known functions of these target areas in stress and aggression [26][27][28][29][30][31][32][33] . For each paired target injection, around 50-60% of PeFA Ucn3 neurons were labelled by both tracers (Fig. 3r), suggesting that individual PeFA Ucn3 neurons project to multiple targets. To further determine the involvement of these projection areas in pup-directed aggression, we performed CTB-mediated retrograde tracing from LS, VMH, and AHi individually and quantified the overlap between CTB labelling and C-Fos expression in the PeFA of virgin certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not this version posted July 9, 2019. ; https://doi.org/10.1101/697334 doi: bioRxiv preprint males after pup interactions (Fig. 3s, t). C-Fos labeling was found in each of the three major projections, with AHi-projecting C-Fos+ cells preferentially located in the rostral and medial subdivision of PeFA ( Fig. 3u; Supp. Fig. 3a). Together, these results suggest that PeFA Ucn3 neurons send out branched projections, and that all major projections are equivalently active during pup-directed attacks with preferential involvement of rostro-medial PeFA projections to AHi.

PeFA Ucn3 neuronal projections govern discrete aspects of infant-directed neglect and aggression
To test the respective functions of PeFA Ucn3 projections, we performed optogenetic activation of PeFA Ucn3 terminals in virgin females injected in the PeFA with a conditional ChR2-expressing virus and with optical fibres implanted bilaterally above each projection target. We initially focused on projections to VMH, based on the documented role of this area in the control of aggression 30,32 (Fig 4a). Stimulation of PeFA Ucn3 to VMH projections reduced pup retrieval across sessions and shortened individual bouts of pup investigation ( Fig. 4b; Supp. Video 2).
However, it did not affect the overall number of parental females nor the overall time engaged in parental interactions, latency to retrieve, time spent in the nest with pups, and pup grooming ( Fig. 4c-e). Further, we did not observe any pup-directed aggression (Fig. 4f). Thus, activating PeFA Ucn3 to VMH projections inhibits sustained interactions with pups and pup retrieval, but does not elicit aggression toward pups.
We next stimulated PeFA Ucn3 to LS projections (Fig. 4g) and found that this manipulation inhibited parental behaviour in a significant fraction of females, as well as reduced cumulative pup retrieval and pup grooming across the three sessions ( Fig. 4 h, i). Moreover, stimulation of PeFA Ucn3 to LS projections triggered instances of pup-directed aggression (such as aggressive handling and biting) in 30% of females, although the overall time displaying aggression remained negligible (Fig. 4j). PeFA Ucn3 to LS stimulation had no significant effect on time spent certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not this version posted July 9, 2019. ; https://doi.org/10.1101/697334 doi: bioRxiv preprint in the nest with pups, pup investigation, overall parenting time, or latency to retrieve (Fig. 4 k, l).
In a subset of females (n = 2 / 10), PeFA Ucn3 to LS stimulation led to an escape response, which was never observed in other manipulations (Supp. Video 3). Together, these results suggest that PeFA Ucn3 to LS projections inhibit parenting and mediate signs of aggression.
Finally, we examined the function of PeFA Ucn3 to AHi projections (Fig. 4m) and found that 55% of females displayed infant-directed aggression such as biting and aggressive carrying of pups without completing retrieval in a manner similar to the stereotyped infanticidal behaviours observed in males with increased time displaying aggression (Fig. 4n, Supp. Video 4).
Additionally, PeFA Ucn3 to AHi stimulation dramatically reduced cumulative retrieval over the course of three sessions, time spent in the nest with pups, number of parental females, and time spent pup-grooming (Figure 4o, p), with no effect on retrieval latency, pup investigation bout length, or time parenting (Fig. 4q, r). Thus, PeFA Ucn3 to AHi projections strongly affect the expression of infant-directed aggression and the suppression of parental behaviours.
Stimulation of PeFA Ucn3 projections to VMH, LS, and AHi did not impact locomotion, appetitive feeding, or conspecific aggression (Supp. Fig. 3b-h). Altogether, these data suggest that various aspects of infant-directed neglect and aggression are mediated across PeFA Ucn3 projections, with PeFA Ucn3 to VMH projections suppressing pup investigation, PeFA Ucn3 to LS projections mediating reduced pup handling and low-level aggression, and PeFA Ucn3 to AHi projections inducing stereotyped displays of pup-directed aggression (Fig. 4s).

Concluding Remarks
The last few decades have seen considerable progress in the use of experimental model systems to identify the neural basis of parental behaviour [11][12][13][14]34 . These studies have uncovered critical sensory cues underlying parent-infant interactions, as well as the neural circuit basis of specific repertoires of nurturing displays in males and females [11][12][13][14]34 . By contrast, the neural control of infant-directed aggression remains largely unaddressed. Initially considered a certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not this version posted July 9, 2019. ; https://doi.org/10.1101/697334 doi: bioRxiv preprint pathological behaviour, pioneering work in langurs and other mammals instead suggested that infanticide is an adaptive behaviour that enables males to gain advantage over rivals and females to adapt reproduction to adverse circumstances 1,2 . Although a number of brain areas have been described as negatively affecting parental behaviour [35][36][37][38][39][40] , the existence of a dedicated control hub for infant-directed agonistic interactions, including infanticide, had not yet been explored.
In the present study, we demonstrate that PeFA Ucn3 neurons are specifically activated by pup-directed aggression in males and females, but not by feeding, conspecific aggression, or acute stress, and that PeFA Ucn3 neurons are required for infant-directed aggression in males.
While activation of PeFA Ucn3 neurons did not elicit widespread aggression toward pups in females except tail rattling in a subset of animals, we showed that inducing PeFA Ucn3 neuron activity strongly reduced maternal behaviour including pup retrieval, time in the nest, and overall parenting time. We identified significant direct inputs to PeFA Ucn3 neurons from largely inhibitory MPOA Gal neurons, shown to be critical for the positive control of parenting 11,13 . Because MPOA Gal neurons are active in females exposed to pups, MPOA Gal -mediated inhibition may repress PeFA Ucn3 neurons in females and thus occlude the full expression of pup-directed agonistic behaviour during artificial activation of PeFA Ucn3 neurons.
Previous studies have shown that electrical stimulation of the PeFA leads to a so-called 'hypothalamic rage response' comprising typical aggressive displays including piloerection, teeth baring, and bite attacks 29,30,33 . Indeed, we observe c-fos positive neurons in the PVH and PeFA of males and females displaying conspecific aggression, but these neurons are Ucn3negative, suggesting that distinct PeFA populations mediate aggression toward either adults or infants. In addition, Ucn3 in the PeFA has been implicated in the regulation of social recognition and stress responsivity 18,[41][42][43] . However, our experiments with targeted inhibition or stimulation of PeFA Ucn3 neurons specifically affected parental behaviour with no effect on adult social certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not this version posted July 9, 2019. ; https://doi.org/10.1101/697334 doi: bioRxiv preprint behaviours, indicating that PeFA Ucn3 neurons may be specifically tuned to social cues from infants and/or juveniles rather than adult conspecifics.
Specific tuning of these neurons to social cues is supported by our finding that input areas to PeFA Ucn3 neurons include the MPOA, MeA, and PVH, known to be essential for integration of social cues sensed by the vomeronasal system which has been well-established as a crucial pathway for infanticide 13,44 . MPOA Gal neurons directly send input to PeFA Ucn3 neurons, suggesting a direct relationship between this well-established hub for pro-parental behaviour and the putative infant-directed aggression neuron population described here. Input neurons from PVH mostly express AVP and CRF, and AVP-expressing PVH neurons have previously been implicated in male-typical social behaviour 20,21 , while CRF-expressing PVH neurons are essential mediators of the physiological stress response 45,46 . Our results show that PeFA Ucn3 neurons do not respond to acute restraint stress, but given their involvement in pupdirected aggression and the role of stress in poor parenting, we reasoned that chronic stress may activate the PeFA Ucn3 neurons, resulting in decreased maternal behaviour. We found a strong correlation of PeFA Ucn3 activation and lower levels of parental behaviour in animals under non-stressed conditions suggesting a strong contribution of PeFA Ucn3 neuron activity to baseline variability in maternal care. Surprisingly, rather than enhancing this relationship, chronic stress disrupted the effect of PeFA Ucn3 neuron activity on parenting revealing that PeFA Ucn3 neurons are sensitive to stress in a manner that decouples their activity, as measured by c-fos expression, from strength of parental behaviour. Alternatively, PeFA Ucn3 neurons may be chronically activated during prolonged stress, rendering differences in c-fos expression inaccurate to measure changes in neural activity in this cell population during chronic stress.
Previously identified hypothalamic neuronal populations critical for feeding (Agrp neurons in the arcuate nucleus) and parenting (MPOA Gal neurons) have been shown to be organized in distinct pools, each largely projecting to a distinct downstream area 11,47 . In contrast, individual PeFA Ucn3 neurons predominantly project to multiple targets, in a similar manner as certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not this version posted July 9, 2019. ; https://doi.org/10.1101/697334 doi: bioRxiv preprint VMH neurons extending collaterals to downstream defensive circuits 48 . Interestingly the three major PeFA Ucn3 target areas VMH, LS and AHi convey overlapping but distinct aspects of infantdirected neglect and aggression (Fig. 4s). Specifically, PeFA Ucn3 to AHi projections appear to be directly responsible for triggering pup-directed attacks. The function of the AHi is not well understood, although recent data suggest it may send sexually-dimorphic projections to MPOA Gal neurons essential for parenting 11 . LS and VMH projections [49][50][51] appear to play similar roles in the regulation of agonistic interactions with infants. Based on its role in conspecific aggression 32,52 , we hypothesized a role for PeFA Ucn3 to VMH projections in pup-directed attack.
However, we found instead that this projection regulates pup investigation. Activation of the PeFA Ucn3 to LS projections also resulted in sustained reductions in pup interactions, a finding that is consistent with a recent study showing that LS neurons expressing corticotrophinreleasing factor receptor 2 (CRFR2), the high-affinity receptor for Ucn3, may control persistent anxiety behaviour 27 . In addition, PeFA Ucn3 to LS projection activation led to incidences of pupdirected aggression. We noticed that the AHi-projecting neurons that are active during infanticide were located predominantly in the rostral PeFA, as were the majority of PeFA Ucn3 neurons associated with infanticide. Together with our functional data showing that activation of this projection leads to more pup-directed aggression compared to other projections, we propose that the pattern of activation within subsets of PeFA Ucn3 neurons defined by their projection targets correlates with the degree of anti-parental behaviour displayed by the individual, from neglect or avoidance, to genuine pup directed attack (Fig. 4s).
Parental behaviour, and by extension infant-directed aggression, is highly regulated by context and previous social experience. Previous studies have shown that females with repeated exposure to pups will become more parental over time 16 . Our experiments with virgin females also showed a similar strong effect of pup exposure on subsequent infant-directed behaviour (Supp. Fig. 2). We observed that males allowed to interact with pups during suppression of PeFA Ucn3 neurons displayed long lasting reduced aggression toward pups, and certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not this version posted July 9, 2019. ; https://doi.org/10.1101/697334 doi: bioRxiv preprint that activation of PeFA Ucn3 neurons in females induced prolonged reduction in parenting. These results stand in contrast to the effects of stimulating MPOA Gal neurons, which confers rapid, reversible facilitation of parental behaviour 11,13 . Together, these data suggest that PeFA Ucn3 neurons may undergo significant functional plasticity or induce long lasting changes in downstream parental circuitry.
Our functional characterization of the PeFA Ucn3 neurons and their major projection targets uncover this neural population as essential for pup-directed aggression, with specific projections controlling aspects of anti-parental behaviour such as pup investigation and handling as well as stereotyped displays of pup-directed agonistic behaviour. The discovery of a dedicated circuit for pup-directed aggression reinforces the notion that this behaviour might have adaptive advantages in the wild. This study reveals a novel framework for understanding how parental behaviour is modulated at a circuit level and opens new avenues for studying the context-and physiological-state dependent expression of parenting.

Data and code availability statement
The data and code that support the findings of this study are available upon request from the corresponding author.

Ethical statement
All animal experiments were approved by the Harvard University Institutional Animal Care and Use Committee. All experiments were performed in compliance with our Harvard University IACUC approved protocols.

Acknowledgements
We thank S. Sullivan for help with behaviour scoring and mouse husbandry, and R. Hellmiss at MCB Graphics for work on illustrations. We thank members of the Dulac lab for comments on certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not this version posted July 9, 2019. ; https://doi.org/10.1101/697334 doi: bioRxiv preprint the manuscript, A. Leonard, E. Owolo, and L. Moussalime for assistance with data collection. A.E.A. and C.D. wrote the manuscript with input from all authors.

Competing Interests
The authors declare no competing financial interests. certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not this version posted July 9, 2019. ; https://doi.org/10.1101/697334 doi: bioRxiv preprint certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not this version posted July 9, 2019. ; https://doi.org/10.1101/697334 doi: bioRxiv preprint certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not this version posted July 9, 2019. ; https://doi.org/10.1101/697334 doi: bioRxiv preprint certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not this version posted July 9, 2019. ; https://doi.org/10.1101/697334 doi: bioRxiv preprint The copyright holder for this preprint (which was not this version posted July 9, 2019. ; https://doi.org/10.1101/697334 doi: bioRxiv preprint the vomeronasal pathway, as well as brain areas involved in the regulation of stress, physiological state, and parenting. PeFA Ucn3 neurons send bifurcating processes to major projection targets including the AHi, LS, and VMH. Stimulation of these projections specifically impact pup-directed behaviour including aggression (AHi), and neglect (LS and VMH). Inhibitory MPOA Gal neurons send input to PeFA Ucn3 neurons and to subsets of PeFA Ucn3 projection targets such as VMH and LS, likely ensuring the repression of agonistic interactions with infants when animals are in a parenting state. certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not this version posted July 9, 2019.

Behaviour assays
Mice were individually housed for at least 1 week prior to testing. Experiments were conducted during the dark phase under dim red light. Tests were recorded by Geovision surveillance system or by Microsoft LifeCam HD-5000 and behaviours were scored by an observed blind to experimental condition using Observer XT11 Software or Ethovision XT 8.0 (Noldus Information Technology). Animals were tested for a single behaviour per session with at least 24 hours between sessions.

Parental behaviour
Parental behaviour tests were conducted in the mouse's home cage as previously described 11,13 .
Mice were habituated to the testing environment for 10 minutes. One to two C57BL6/J pups 1-4 days old were presented in the cage in the opposite corner to the nest. Test sessions began when the mouse first closely investigated a pup (touched the pup with its snout) and lasted for 5-15 minutes. In the event that the mouse became aggressive by biting and wounding the pup, the certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not this version posted July 9, 2019. ; https://doi.org/10.1101/697334 doi: bioRxiv preprint session was immediately halted and the pup was euthanized. The following behaviours were quantified: latency to retrieve, sniffing (close contact), grooming (handling with forepaws and licking), nest building, time spent in the nest, crouching, latency to attack (latency to bite and wound), aggression (roughly handling, aggressively grooming, aggressive carrying with no retrieval), and tail rattling. A 'parenting behaviour' index was calculated as the sum of duration of sniffing, grooming, nest building, time spent in the nest, and crouching. Mice were categorized as 'parental' if they showed pup-retrieval behaviour and 'non-parental' if they did not retrieve pups.
Mice were categorized as 'aggressive' if they showed bite attacks, tail rattling, aggressive pup carrying, or aggressive pup-grooming and 'non-aggressive' if they did not display any of these behaviours. Optogenetic experiments were limited to 5 minutes as previously performed for infanticidal testing, thus limiting the ability to score male retrieving behaviour which is typically displayed between 4 and 7 minutes 11,13 . Chemogenetic experiments lasted 15 minutes.

Conspecific aggression
Conspecific aggression assays were performed as previously described 30 . Briefly, castrated males were swabbed with ~40 µL of urine from an intact male and introduced into the home cage of the test mouse for 5 minutes (session started at first close contact).

Locomotor behaviour
Locomotor behaviour assays were performed in a 36 x 25 cm empty cage. Velocity and distance moved were tracked for 5 minutes.

Feeding behaviour
Feeding behaviour was assessed by introducing a small piece of palatable food (chocolate chip cookie) into the far corner of a mouse's home cage. Session started when mouse approached the food and behaviour was recorded for 5 minutes. Parameters rated were duration sniffing and eating the cookie.

Fluorescence in situ hybridization
certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not this version posted July 9, 2019. ; https://doi.org/10.1101/697334 doi: bioRxiv preprint Fluorescence in situ hybridization (FISH) was performed as previously described 13,53 . Briefly, fresh brain tissue was collected from animals housed in their home cage or 35 min after the start of the behaviour tests for immediate early gene (c-fos) studies. Only animals that engaged in a particular behaviour were used. Brains were embedded in OCT (Tissue-Tek) and frozen with dry ice. 20µm cryosections were used for mRNA in situ. Adjacent sections from each brain were collected over replicate slides to stain with multiple probes. The staining procedure is as previously described 13,53  Slides were mouted with Vectashield with 4',6-diamidino-2-phenylindole (DAPI, Vector Labs).

Laser-capture microscopy and microarray analysis
Males were sacrificed 35 minutes after onset of attacking behaviour for peak c-fos expression and fresh brain tissue was frozen in OCT. 20µm sections with alignment holes punched into the tissue by a glass pipet were prepared in a series with every other section collected on superfrost plus slides for in situ hybridization and the other for laser capture microdissection and collected on PEN coated membrane slides (Leica). Double fluorescence in situ hybridization (FISH) for cfos and oxytocin was performed as previously described 13,53 . Sections prepared for laser-capture certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not this version posted July 9, 2019. ; https://doi.org/10.1101/697334 doi: bioRxiv preprint were dehydrated and stained in a 50% ethanol solution containing cressyl violet. Using a Zeiss PALM microscope, slides from adjacent FISH and laser-capture sections are aligned via tissue pinches using PALM Software (Zeiss). After alignment, tissue from the area containing c-fos+, oxytocin+ cells, and an outside region were identified from FISH sections and collected from fresh laser sections by laser capture. For each region, tissue from 6 brains were pooled to prepare total RNA (RNEasy micro kit, Qiagen) for reverse transcription and amplification to cDNA (Ovation Pico WTA kit, Nugen). Three biological replicates of cDNA (n=18 total brain samples; 6 pooled per replicate) were prepared for hybridization to the Affymetrix Exon ST 1.0 array as described by the manufacturer. Differential expression analysis was performed using the Affymetrix Expression Console software to analyze gene expression levels. We then performed a paired ttest to determine genes with significantly altered gene expression in the c-fos+ area compared to bordering cell regions. Finally, in order to rank genes taking into account the fold change and expression level, we performed a residual analysis on the regression (Fig. 1d) and found that Avp and Ucn3 had the highest residual values (Avp 5.66, Ucn3 4.80).

pS6 pulldowns and RNAseq analysis
Adult male mice were habituated in their home cages to a testing room for 4 hours. Males were then either presented with a 1-4 day old C57BL/6J pup or no stimulus and allowed to interact for 10 minutes, after which time behaviour was recorded and the pup was removed. Sixty minutes after stimulus presentation, males were sacrificied and a punch (1 mm diameter) was collected from a 2mm thick brain slice containing the perifornical area, the paraventricular nucleus, and part of the anterior hypothalamus. The tissue was processed for pS6 immunoprecipitation as described previously 15 . Briefly, punches were held in (ice-cold homogenization buffer (10mM HEPES, 150mM KCl, 5mM MgCl2 and containing 1X phosphatase inhibitor cocktail, 1X Calyculin. Protease inhibitor cocktail (EDTA free), 250ug/ml Cycloxeximide and 2.5ul/ml Rnasin) and 10 punches were pooled per sample, with 3 biological replicates taken from infanticidal and naïve groups. Tissue was homogenized, centrifuged for 10 minutes at 2000 g, subjected to treatment certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not this version posted July 9, 2019. ; https://doi.org/10.1101/697334 doi: bioRxiv preprint with 0.7% NP40 and 20mM DHPC and centrifuged for 10 minutes at (20,000rcf) at 4° C to obtain the supernatant containing the input fraction. A (10 µL sample) of the input was kept for sequencing and the remaining input fraction was incubated with Protein A Dynabeads (Invitrogen) conjugated to pS6 antibody (Life Technologies 44923G) at a concentration of around 4 µg/IP for 10 minutes at 4° C. Beads were collected on a magnet and washed in 0.15M KCl wash buffer, then associated RNAs were lysed into RLT buffer and RNA was isolated using the RNAeasy Micro kit (Qiagen) according to manufacturer's instructions. RNA was checked for quality and quantity on an Agilent tape station. High quality samples from matched input and IP fractions were prepared using the (Low-input Library Prep kit (Clontech).
Sequenced reads were mapped with STAR aligner 54 , version 2.5.3a, to the mm10 reference mouse genome and the Gencode vM12 primary assembly annotation 55 , to which non-redundant UCSC transcripts were added, using the two-round mapping approach. This means that following a first mapping round of each library, the splice-junction coordinates reported by STAR, across all libraries, are fed as input to the second round of mapping. The parameters used in both mapping rounds are: outSAMprimaryFlag ="AllBestScore", outFilterMultimapNmax="10", outFilterMismatchNoverLmax="0.05", outFilterIntronMotifs="RemoveNoncanonical". Following read mapping, transcript and gene expression levels were estimated using MMSEQ 56 . Following that transcripts and genes which cannot be distinguished according to the read data were collapsed using the mmcollapse utility of MMDIFF 57 , and the Fragments Per Killobase Million (FPKM) expression units were converted to natural logarithm of Transcript Per Million (TPM) units since the latter were shown to be less biased and more interpretable 58 . In order to test for differential gene activation between the infanticidal males and naïve males, we used a Bayesian regression model 59 . We defined the response (for each transcript or gene) as the difference between means of the posterior TPMs of the IP and input samples (i.e., paired according to the tissue punches from which they were obtained). The uncertainty in the response was computed as square root of the sum of the variances of the posterior TPMs of the IP and input samples certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not this version posted July 9, 2019. ; https://doi.org/10.1101/697334 doi: bioRxiv preprint divided by the number of posterior samples (1,000). The categorical factor for which we estimated the effect was defined as the behaviour of the animal, meaning infanticidal and naïve, where naïve was defined as baseline. Hence, our Bayesian regression model estimates the posterior probability of the natural logarithm of the ratio between infanticidal IP TPMs divided by input TPMs and naive IP TPMs divided by input TPMs being different from zero. In order to define a cutoff of this posterior probability, above which we consider the activation (i.e. effect size) significant, we searched for a right mode in the distributions of posterior probabilities across all transcripts and genes (separately for each), and placed the cutoff at the local minima that separates the right mode from the left mode, which represents transcripts and genes for which the likelihood of differential gene activation is very weak and hence their posterior probability is dominated by the prior (Supp. Fig. S1a).

Optogenetics
Ucn3::Cre mice 8-12 weeks old were used for these experiments. Pilot experiments and our previous observations indicated that mated parents were insensitive to manipulations aimed at reducing parental behaviours 13 , so we used virgin males and females in order to maximize our opportunity to see decrements in parenting. Virgin males were prescreened for infanticidal behaviour. We injected 200 nL of AAV1/DIO-hChR2-eYFP for activation or AAV1/FLEx-NpHR3.0-eYFP for inhibition (or AAV1/FLEx-YFP or AAV1/FLEx-tdTomato as control virus) bilaterally into the perifornical area (AP -0.6, ML ±0.3, DV -4.2). Mice recovered for 1-2 weeks and in a second surgery a dual fibre optic cannula (300 µm, 0.22 NA, Doric Lenses) was implanted 0.5 mm above the target area and affixed to the skull using dental cement. Cannula positions were as follows: PeFA = AP -0.6mm, ML ± 0.5mm, DV -3.7mm; LS = AP 0.3mm ML ± 0.5mm DV -3.2mm; VMH= AP ML ± 0.5mm DV; AHi = AP -2.3mm, ML ± 2.5mm, DV -4.7mm. Mice recovered for an additional week prior to the start of behavioural testing. In pilot studies we found that order of testing impacted parental behaviour results, so rather than randomizing test order, we certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not this version posted July 9, 2019. ; https://doi.org/10.1101/697334 doi: bioRxiv preprint alternated laser on and off conditions in the same order for both control and manipulated mice. For locomotion, feeding, and conspecific aggression tests, we randomized the order of laser on/off presentation. On test days, the fibre implant was connected to an optical fibre with commutator connected to either a 473 nm laser (150 mW, Laserglow Technologies), or a 460 nm LED (50W, Prizmatix) for ChR2 stimulation, or to a 589 nm laser (300mW, OptoEngine) for NpHR3.0 inhibition. Behaviour tests were 5 minutes in duration for each condition and behaviour category.
For parental behaviour assays with ChR2, the 473 nm laser or 460 nm LED was triggered in 20ms pulses at 20 Hz for 2 seconds when the mouse contacted the pup with its snout, and the order of laser sessions was on first, off second, on third. For locomotor experiments with ChR2 activation, the laser was triggered (20ms, 20 Hz, 2 seconds) every 10 seconds to approximately equivalent duration as the time laser was on during the parental sessions. For cookie experiments with activation, the laser was triggered when the mouse contacted the cookie. For conspecific aggression experiments with activation, the laser was triggered when the test mouse made close contact with the intruder mouse. The power exiting the fibre tip was 5mW, which translates to ~2 mW/mm 2 at the target region (calculator provided by Deissertoth lab at http://www.stanford.edu/group/dlab/cgi-bin/graph/chart.php). For all behaviour assays with NpHR3.0, the 589 nm laser was illuminated (20 ms, 50 Hz) throughout the duration of the 5 minute session at a power of 15 mW, or ~8.4 mW/mm 2 . For parenting sessions with inhibition, the order of laser sessions was off first, on second, off third.
Cre positive and cre negative females were administered 0.3mg/kg clozapine-n-oxide dissolved in saline intraperitoneally and habituated to the testing environment for 30 minutes. Females certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not this version posted July 9, 2019. ; https://doi.org/10.1101/697334 doi: bioRxiv preprint were presented two C57BL6/J pups in the corner of their homecage opposite the nest and parental behaviours were recorded for 15 minutes.

Anterograde tracing
AAV2.1-FLEX-tdTomato virus (UPenn Vector Core) and AAV2.1-FLEX-synaptophysin-eGFP virus (plasmid gift of Dr. Silvia Arber; custom virus prep, UNC Vector Core) were mixed at a ratio of 1:2 and stereotaxically injected unilaterally into the PeFA (Bregma coordinates AP: -0.6mm; ML: -0.3mm; DV: -4.2mm) using a Drummond Nanoject II (180 nL to cover entire structure) into adult virgin male (n=3) and female (n=3) Ucn3::cre mice. After 2 weeks, mice were transcardially perfused with 4% paraformaldehyde. Tissue was cryoprotected (30% sucrose) and serial sections were prepared on a freezing microtome (60 µm). Every third section was immunostained using Anti-GFP antibody (described in immunohistochemistry methods), mounted to superfrost plus slides, coverslipped with Vectashield plus DAPI and imaged at 10X magnification using the AxioScan (Zeiss). Synaptic densities were quantified by ImageJ software after background subtraction and a density ratio was prepared by dividing by the density at the injection site which was normalized to 100%. Density ratio was compared across brain regions by One-way ANOVA followed by Dunnett's test for multiple comparisons and sexual dimorphisms compared using a two-tailed t-test.
The copyright holder for this preprint (which was not this version posted July 9, 2019. ; https://doi.org/10.1101/697334 doi: bioRxiv preprint 488 and CTB-647 was quantified. In control experiments, a 1:1 mixture of CTB-488 and CTB-647 was injected into the LS and AHi. In C57BL6/J mice, males were habituated to a testing environment for two hours then presented with one foreign C57BL6/J pup in the far corner of their homecage. When the male attacked the pup, the pup was removed from the cage and euthanized. Males were perfused 80-90 minutes after pup exposure and brains were postfixed overnight, cryoprotected in 30% sucrose overnight, and 60 µm sections were prepared on a sliding microtome. Tissue was stained with C-Fos antibody (see details in immunostaining section), imaged at 10X (Axioscan, Zeiss), and the number of C-Fos positive cells labeled with CTB-488 was quantified.

Monosynaptic input tracing
A 1:1 ratio (180 nL to cover the entire region) of AAV2.1-FLEx-TVA-mCherry (avian TVA receptor) and AAV2.1-FLEx-RG (rabies glycoprotein) was stereotaxically injected unilaterally into the PeFA (Bregma coordinates AP: -0.6mm; ML: 0.3mm; DV: -4.2mm) of adult Ucn3::cre male and female mice (n=3/sex). Two weeks later, g-deleted rabies virus (deltaG-RV-eGFP) (300 nL) was injected into the same coordinate. Seven days later, mice were transcardially perfused with 4% paraformaldehyde. Tissue was postfixed in PFA overnight and cryoprotected (30% sucrose) then serial sections were prepared on a freezing microtome (60 µm). Every third section was imaged at 10X using the AxioScan (Zeiss). Cell bodies labeled by eGFP in anatomical areas were quantified and an input ratio was prepared by dividing the number of input cells in each region by the total number of input cells in each brain. Input ratio was compared across brain regions by One-way ANOVA followed by Dunnett's test for multiple comparisons and sexual dimorphisms compared using a two-tailed t-test.

C-Fos antibody staining
To visualize C-Fos protein in combination with CTB, perfused tissue was sliced on a freezing microtome at 30 µm, and every 3 rd section throughout the perifornical area was stained. Briefly, certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not this version posted July 9, 2019. ; https://doi.org/10.1101/697334 doi: bioRxiv preprint sections were rinsed in PBS with 0.1% Triton (PBST), then blocked in 2% normal horse serum and 3% fetal bovine serum diluted in PBST for 2 hours at room temperature. Primary antibody Rabbit anti-C-Fos (Synaptic Systems) was diluted 1:2000 in PBS and sections were incubated for 72 hours at 4°C. After rinsing with PBST, secondary Anti-Rabbit A568 (1:1000) was applied at 1:1000 dilution in PBS for 48 hours at 4° C. Sections were rinsed in PBS, mounted to superfrost plus slides, coverslipped with Vectashield containing DAPI, and imaged (see details above).

GFP antibody staining
To amplify synaptophysin densities in the anterograde tracing experiments, the GFP signal was enhanced by antibody staining. Perfused tissue was sliced on a freezing microtome at 60 µm, and every 3 rd section throughout the brain was stained. Briefly, sections were rinsed in PBS with 0.1% Triton (PBST), then blocked in 1.5% normal horse serum diluted in PBST for 1 hour (blocking solution) at room temperature. Primary antibody Rabbit anti-GFP (Novus Biologicals NB600-308) was diluted 1:2000 in blocking solution and sections were incubated overnight at 4° C. After rinsing with PBST, secondary anti-Rabbit-A488 was applied at 1:200 dilution in blocking buffer and incubated overnight at 4° C. Sections were rinsed in PBS, mounted to superfrost plus slides, coverslipped with Vectashield containing DAPI, and imaged (see details above).

Oxytocin/vasopressin staining
Perfused tissue was sliced on a sliding microtome at 60 µm. Every 3 rd section from the perifornical area was stained. Briefly, sections were rinsed in PBS with 0.1% Triton (PBST), then blocked in 10% fetal bovine serum diluted in PBST. Primary antibody Rabbit antivasopressin (Immunostar) or Rabbit anti-oxytocin (Millipore) were diluted 1:1000 in PBST and sections were incubated overnight at 4° C. After rinsing with PBST, secondary anti-Rabbit A647 was applied at 1:200 dilution in blocking buffer and incubated overnight at 4° C. Sections were rinsed in PBS, mounted to superfrost plus slides, coverslipped with Vectashield containing DAPI, and imaged on the Axioscan (Zeiss) at 10X magnification. certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not this version posted July 9, 2019. ; https://doi.org/10.1101/697334 doi: bioRxiv preprint certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.