The lateral periaqueductal gray and its role in controlling predatory hunting and social defense.

Evasion from imminent threats and prey attack are opposite behavioral choices critical to 2 survival. The lateral periaqueductal gray (LPAG) is a key player in these behaviors, it 3 responds to social threats and prey hunting while also driving predatory attacks and active 4 defense. Our results revealed that distinct neuronal populations in the LPAG drive prey 5 hunting and evasion from social threats. We show that the LPAG provides a putative 6 glutamatergic projection to the lateral hypothalamic area (LHA). LPAG > LHA pathway 7 optogenetic inhibition impaired insect predation but did not alter escape/attack ratio during 8 social defeat. The results suggest that the LPAG control over evasion to a social attack may 9 be regarded as a stereotyped response depending probably on descending projections. 10 Conversely, the LPAG control over predatory behavior involves an ascending pathway to 11 the LHA that likely influences LHA GABA neurons driving predatory hunting and may 12 provide an emotional drive for appetitive rewards. 26 for Fos protein and fluorescent in situ hybridization for c- fos mRNA in animals exposed 27 sequentially to IP and SD. We next examined the GABAergic and glutamatergic nature of the 28 LHA neurons activated during IP and SD. The results showed that the number of LHA GABA - 29 activated neurons was significantly higher after exposure to IP compared to SD whereas the 30 number of LHA Glut -activated neurons was comparable in both conditions. To address the behavioral role of LHA GABA and LHA Glut neurons on predatory hunting and social defensive


Introduction
The periaqueductal gray (PAG) commonly has been recognized as a downstream site in neural 2 networks for the expression of a variety of behaviors, for example, sexual, maternal, and 3 defensive behaviors with accompanying modulation of nociceptive transmission, autonomic 4 changes, and vocalization [1][2][3][4][5][6][7][8][9]. By and large, PAG-related responses have been regarded as 5 being mostly stereotyped and dependent on descending projections to the lower brainstem and 6 spinal cord. However, the PAG is also known to influence complex events like approach and 7 avoidance responses to perform risk assessment of potential threats [10-13], fear memory 8 processing [14][15][16][17][18][19], and reward-seeking [20-23]. 9 It is particularly puzzling that the PAG controls both aversive and rewarding behaviors. In this 10 regard, we call special attention to the lateral part of the PAG (LPAG) that has been implicated 11 in mediating the opposite behavioral choices of predatory attack and evasion from a conspecific 12 aggressor. Previous studies from our laboratory have associated the LPAG with the organization 13 of prey hunting. Insect predation was associated with increased Fos expression in the LPAG [20].
14 Behavioral observations have shown that LPAG NMDA lesions interfere with prey hunting; the 15 animals lost their motivation to pursue and attack prey without affecting general levels of arousal, 16 locomotor activity, and regular feeding [22]. In line with these observations, the LPAG contains 17 reward-excited neurons, which exhibited an increased firing rate in response to appetitive food 18 [23]. Conversely, the LPAG has also been shown to respond to a conspecific aggressor. Thus, a 19 significant Fos upregulation in the LPAG was found in animals exposed to dominant conspecifics 20 [24]. In this regard, chemical and optogenetic stimulation experiments suggest that the LPAG 21 mediates active defensive responses such as circa-strike defense, and that it supports a role in 22 active social defensive behaviors to escape from a dominant's attacks [25 -27]. 23 Here, we investigated how the LPAG controls the opposite behavioral choices of predatory 24 hunting and evasion from a social attack. We started by asking whether there is a differential 25 activation of the neural populations in the LPAG activated during insect predation (IP) and social 26 defeat (SD). To this end, we used an approach based on the temporal translational dynamics of 27 the c-fos gene by combining immunofluorescence for Fos protein with fluorescent in 28 situ hybridization for c-fos mRNA (Fos protein/c-fos mRNA IF-FISH) in animals exposed 29 sequentially to IP and SD. We next examined the functional role of neural populations in the 30 LPAG responsive to IP or SD by using pharmacogenetic silencing in Fos DD-Cre mice [28], 31 which allowed targeting neural populations activated by a specific stimulus. Further, because the 32 lateral hypothalamic area (LHA) is one of the main targets of the LPAG [29], we used 1 optogenetic inhibition and examined how LPAG projections to the LHA influence predatory and 2 social defensive behaviors. 3 The LHA, in its turn, is involved in controlling predatory hunting and evasion from imminent 4 threats [30]. In view of the fact that the LHA is heavily targeted by the LPAG, we also 5 investigated how the LHA mediates IP and SD. First, we differentiated neuronal populations in 6 the LHA responding to IP and SD using Fos protein/c-fos mRNA IF-FISH. Next, we determined 7 the GABAergic and glutamatergic nature of LHA neurons responding to IP and SD using Fos 8 protein/VGAT mRNA and Fos protein/VGLUT2 mRNA IF-FISH. Finally, using VGAT-IRES- 9 Cre and VGlut2-IRES-Cre mice, we applied pharmacogenetic tools to silence the activity of the 10 LHA GABA and LHA Glut neurons selectively and then explored the functional role of LHA GABA 11 and LHA Glut neurons on predatory hunting and social defensive behavior. 12 We here show that the LPAG contains two segregated neural populations that separately control 13 opposite behavioral choices of predatory hunting and evasion from a social attack. Our findings 14 are in line with the idea that the PAG works as a unique hub driving stereotyped responses and 15 supplying primal emotional tone to influence complex aversive and appetitive responses [see 31]. 16 Our results show that LPAG control over evasion in response to a social attack may be regarded 17 as a stereotyped response depending most likely on descending projections to lower brainstem 18 sites organizing the behavioral output. In contrast, the LPAG control over predatory behavior 19 involves an ascending glutamatergic pathway to the LHA that likely influences LHA GABA 20 neurons to drive predatory hunting. This path putatively exerts a role in providing emotional 21 drive for prey hunting and may conceivably have more widespread control on the motivational 22 drive to seek other appetitive rewards. 25 To differentiate neuronal populations in LPAG responding to insect predation (IP) and social 26 defeat (SD), we combined immunofluorescence for Fos protein (IF) and fluorescent in 27 situ hybridization for c-fos mRNA (FISH) following the protocol described by Marin-Blasco et 28 al. [32]. Exposure to a first treatment increases c-fos mRNA levels, which reach a maximum level 29 at about 20 min and decline to basal levels two hours later when the Fos protein is close to its 30 maximum [33][34][35]. The animals are then exposed to a second stimulus, and a second peak of c-31 fos mRNA will be observed about 20 min later. Thus, neurons responding to the first stimulus 1 will show Fos protein (IF+/FISH-) mainly, neurons responding to the second stimulus will show 2 mostly c-fos mRNA (IF-/FISH+), and those neurons activated by both stimuli will appear as 3 double-labeled (IF+/FISH+). To determine the optimum exposure and sacrificing time points for 4 our specific experimental conditions, we quantified the number of Fos protein-and c-fos mRNA 5 positive-neurons in LPAG at different time points after IP and SD (see Figure S1). As shown in 6 Figure S1, two hours after the onset of IP, we found maximum levels of Fos protein levels and 7 imperceptible c-fos mRNA signal; we obtained maximum c-fos mRNA levels 20 minutes after 8 exposure to SD. However, low but perceptible newly synthesized Fos protein emerged at this 9 time point in response to SD. Thus, we decided to shorten this time point to 15 minutes when c-10 fos mRNA levels are close to the maximum and less significant Fos protein levels were detected 11 ( Figure S1). Because exposure to a stressful stimulus such as SD affects motivation for hunting, 12 we exposed animals first to IP and subsequently to SD. We also included two control groups of 13 animals exposed twice to the same stimulus (groups IP+IP and SD+SD) to discard neuronal 14 activation due to the mere manipulation of animals. Animals repeatedly exposed to the same 15 stimulus showed reactivation of about 50% of neurons (IF+/FISH+ neurons) and a low number 16 of IF-/FISH+ neurons ( Figure 1A, C). Animals exposed sequentially to IP and SD showed 17 reactivation of about 40% of neurons (IF+/FISH+). GzLM analysis of the number of newly 18 activated neurons (IF-/FISH+) in animals exposed sequentially to IP and SD showed a significant 19 effect of group factor (groups IP+IP, SD+SD, and IP+SD) [χ 2 (2) = 165.57, p < 0.001]. Pairwise 20 comparisons indicate that the Group IP+SD presented a higher number of newly activated 21 neurons (IF-/FISH+) when compared to IP+IP and SD+SD groups (p < 0.001) ( Figure 1C).

22
It is important to note that the maximum levels of c-fos mRNA occur 20 minutes after the second 23 stimulus (SD). However, we decided to sacrifice the animals 5 minutes earlier to minimize the suggest that exposure to SD induced activation of neurons already recruited by IP and also led to 27 the recruitment of new neurons. Therefore, LPAG neuronal populations responding to predatory 28 hunting and social defensive responses appear to be partially differentiated. 30 In Fos DD-Cre mice, Cre recombinase is fused to an E. coli dihydrofolate reductase (ecDHFR)- Exclusively Activated by Designer Drug) fused to an mCherry reporter (AAV5-hSyn-6 hM4D(Gi)-DIO-mCherry) to silence selectively the activity of LPAG neurons responsive to IP 7 or SD (Figure 2A, B). Previous tract-tracing studies showed that LPAG neurons send substantial 8 ascending projections to the LHA [22]. Therefore, we also aimed to detect possible differences 9 in LHA projection patterns from LPAG-activated neurons in response to IP or SD. To this end, 10 animals also received a paired injection of Cre-dependent AAV expressing an EYFP reporter 11 (AAV5-hSyn-DIO-EYFP) for tracing ascending projections to the LHA ( Figure 2B) and the 12 descending projections to the brainstem ( Figure S2). The EYFP reporter was used to trace LPAG 13 projections because it yields a much stronger fluorescent signal than the reporter with mCherry 14 fused to hM4D(Gi). The administration of TMP in animals previously exposed to IP or SD 15 stabilizes DD-Cre expressed in active neurons and leads to Cre-dependent expression of 16 hM4D(Gi) and EYFP (Figure 2A). Animals were subsequently treated with saline or clozepine-17 N-oxide (CNO) [38] and re-exposed to IP and SD for behavioral testing.

18
Previous studies showed that cytotoxic lesions in the LPAG increased the latency to start hunting 19 and decreased the number of captured prey [22]. Therefore, during predatory hunting, we 20 quantified the latency to catch the first cricket and the total number of crickets captured. 21 Conversely, chemical and optogenetic stimulation experiments suggest that the LPAG mediates 22 predominantly active defensive responses such as the circa-strike defense. This supports a role 23 in active social defensive behaviors aimed at escaping from resident's attacks [25][26][27]. Thus, 24 during the social agonistic interaction, we analyzed the number of intruder escapes in response 25 to resident attacks (escape/attack ratio) as a measure of active defensive behavior. presented a significant increase in the latency to start hunting (p<0.001) ( Figure 2D). IP Cre 31 stabilized treated with CNO also presented a reduced number of captured crickets ( Figure 2D).

32
In contrast, CNO-treated animals that received TMP Cre stabilization for SD did not differ from 1 the saline-treated animals in the latency to start hunting (Tukey's HSD test, p>0.26) and the 2 number of captured prey ( Figure 2D). presented a significant decrease in the escape/attack ratio observed during the social agonistic 8 interaction (p<0.001) ( Figure 2D). In contrast, CNO-treated animals that received TMP Cre 9 stabilization for IP did not differ from the saline-treated animals in the escape/attack ratio 10 observed during SD (p>0.46) ( Figure 2D).

11
Thus, our functional analysis of Fos DD-Cre mice revealed that CNO inhibition of LPAG 12 neurons activated in response to IP led to an increase in the latency to start hunting and a decrease 13 in the total number of crickets captured but did not alter escape/attack ratio during SD.
14 Conversely, CNO inhibition of LPAG SD-responding neurons did not change predatory behavior 15 parameters and led to a decrease in the escape/attack ratio during SD.

16
The results of our pathway tracing analysis in Fos DD-Cre mice revealed that LPAG IP-and SD-17 responding neurons yielded a similar EYFP anterograde labeling in the LHA ( Figure 2E) and 18 brainstem ( Figure S2). 20 The LPAG projects densely to the LHA, which is one of the main targets of the LPAG [22]. We to discard light stimulation behavioral effects. The 561-nm laser light was continually delivered 1 to the LHA during the 5 min of IP or SD through surgically implanted dual-fiber optic elements. 2 For the latency to catch the first prey, univariate ANOVA revealed a significant effect for the 3 factor virus (eArch+ and control, F1,8=21.28; p<0.002) and the factor treatment (laser on / off, 4 F1,8=16.87; p<0.004), and a significant interaction between the factors (virus x treatment, 5 F1,8=32.21; p<0.001). Post hoc pairwise comparisons (Tukey's HSD test) revealed that eArch+ 6 animals during laser on presented a significant increase in the latency to catch the first prey 7 ( Figure 3D, p<0.001). Photoinhibition of the LPAG > LHA projection during IP also reduced the 8 number of captured crickets ( Figure 3C). For the escape/attack ratio, univariate ANOVA  Thus, our functional analysis revealed that photoinhibition of LPAG > LHA pathway impaired 12 IP and increased the latency to start hunting while decreasing the total number of captured 13 crickets; it did not alter escape/attack ratio during SD ( Figure 3C). 15 To differentiate neuronal populations in LHA responding to IP and SD, we next applied IF-FISH 16 for Fos protein and c-fos mRNA [32]. Animals repeatedly exposed to the same stimulus (groups     that CNO treated HM4D+ animals presented a significant decrease in the escape/attack ratio 10 (#p<0.01) ( Figure 5D). For latency to start hunting, univariate ANOVA revealed no significant 11 effect for the factors virus (F1,8= 0.98; p= 0.35) and treatment (F1,8= 0.04; p= 0.84). CNO 12 treated HM4D+ animals also did not change the number of captured crickets ( Figure 5D).

13
Our functional analysis revealed that pharmacogenetic inhibition of LHA GABA neurons impaired 14 IP by increasing the latency to start hunting and decreasing the total number of captured crickets, 15 but it did not alter the escape/attack ratio during SD ( Figure 5C). Conversely, pharmacogenetic 16 inhibition of LHA Glut neurons decreased the escape/attack ratio during SD but did not influence 17 IP ( Figure 5D).

19
The experiments reported here address how the LPAG mediates the seemingly opposite 20 behavioral choices of fleeing from a conspecific aggressor and prey hunting. First, we found that 21 LPAG neuronal populations responding to predatory hunting and social defensive responses 22 appear to be partially differentiated. Our functional analysis using Fos DD-Cre mice revealed 23 that CNO-induced inhibition of LPAG neurons activated in response to IP disrupted prey hunting 24 but did not alter escape/attack ratio during SD. Conversely, CNO-induced inhibition of LPAG 25 SD-responding neurons did not change predatory behavior parameters and led to a decrease in 26 the escape/attack ratio during SD. Thus, our functional analysis supports the idea of diverse 27 neuronal populations in the LPAG influencing the predatory and social-defensive responses. In 28 addition, we observed that photoinhibition of the LPAG > LHA pathway impaired IP but did not 29 alter escape/attack ratio during SD. This suggests distinct axonal pathways from the LPAG play 30 a role in organizing IP and SD.

31
To differentiate neuronal populations in LPAG responding to insect predation (IP) and social 1 defeat (SD), we combined immunofluorescence for Fos protein (IF) and fluorescent in 2 situ hybridization for c-fos mRNA (FISH) using sequential exposure to IP and SD. In response 3 to the second stimulus (SD), we found a 35% increase in newly activated neurons. However, this 4 measure of newly activated neurons may be an underestimate because we sacrificed the animals  Conversely, animals exposed to a conspecific aggressor present strong activation in the The LHA is known to control predatory hunting, and the LPAG glutamatergic ascending 5 projection to the LHA could provide motivational drive to LHA on predatory hunting. Our 6 findings support the hypothesis that apart from organizing stereotyped responses, the PAG plays 7 more complex roles by providing primal emotional tone to forebrain sites mediating complex 8 behavioral responses [29]. In line with this view, an opioid-dependent mechanism in the LPAG 9 has been shown to switch the motivation from maternal care to prey hunting in lactating rats [21, 10 46]. In addition, LPAG cytotoxic lesions impair place preference induced by morphine [47], thus 11 suggesting an influence on the motivational drive to seek appetitive reward. Conversely, the 12 ascending projection to the LHA from LPAG glutamatergic neurons does not seem to be critical 13 for modulating SD. For this behavior, descending pathways from the LPAG to the brainstem are 14 more likely to be involved in controlling escape responses during SD. In this regard, we have 15 presently described the LPAG SD-responding neurons projections to several targets in the 16 brainstem (see Figure S2). 17 The LHA has been implicated in controlling predatory hunting and threat evasion. LHA GABA 18 neurons promote feeding [30, [40][41][42] 24 We found that LHA neuronal populations responding to predatory hunting and social defensive 25 responses appear to be partially differentiated as observed by combining immunofluorescence 26 for Fos protein and fluorescent in situ hybridization for c-fos mRNA in animals exposed 27 sequentially to IP and SD. We next examined the GABAergic and glutamatergic nature of the      animals received no treatment; IP+IP GROUP (n = 4), animals were exposed to IP (5 minutes), 32 1 hour and 40 minutes later re-exposed to IP (5 minutes), and after the second IP, were left for 1 10 minutes undisturbed before perfusion; SD+SD GROUP (n = 4), animals were exposed to SD 2 (5 minutes), 1 hour and 40 minutes later re-exposed to SD (5 minutes), and after the second SD, 3 were left for 10 minutes undisturbed before perfusion; IP+SD GROUP (n = 6), animals were 4 exposed to IP (5 minutes), 1 hour and 40 minutes later exposed to SD (5 minutes), and after the 5 SD, were left for 10 minutes undisturbed before perfusion. Groups of animals exposed repeatedly 6 to the same treatment (IP+IP and SD+SD) were included to discard neuronal activation due to 7 the mere manipulation of animals. Immediately after the treatments, animals were anesthetized to IP (5 minutes) and left for 1 hour and 55 minutes undisturbed before perfusion; and SD 14 GROUP (n = 6), animals were exposed to SD (5 minutes) and left for 1 hour and 55 minutes 15 undisturbed before perfusion. Animals were anesthetized and perfused after the treatments.  Techonologies). On experimental day 1, animals (groups CONTROL and EARCH) were 2 exposed to predatory behavior with the yellow laser turned OFF. On experimental day 2, animals 3 were exposed again to predatory behavior with the laser ON for the optogenetic inhibition of 4 LHA terminals originating from LPAG. After three days (day 6), animals of both groups were 5 exposed to social defensive behavior with the yellow laser turned OFF. On the next day (day 7), 6 animals were animals were re-exposed to social defensive behavior with the laser ON. ON/OFF 7 laser cycles were of 5 minutes for insect predation and 5 minutes for defensive behavior.       (FAPESP) Research Grants #2014/05432-9 (to NSC) and #2016/18667-0 (to SCM). FAPESP 3 fellowships to IJMB (#2017/04830-9) and to MJRJ (#2017/09753-2). We thank Prof. Larry 4 W. Swanson for insightful discussion and comments on the manuscript. The authors have no competing financial interests or potential conflicts of interest.