Central role of the habenulo-interpeduncular system in 
the neurodevelopmental basis of susceptibility and resilience to anxiety


 Experiencing stress during sensitive periods of brain development has a major impact on how individuals cope with later stress. Although many become more prone to develop anxiety or depression, some appear resilient. The mechanisms underlying these differences are unknown. Key answers may lie in how genetic and environmental stressors interact to shape the circuits controlling emotions. Here we studied the role of the habenulo-interpeducuncular system (HIPS), a critical node of reward circuits, in early stress-induced anxiety. We found that a subcircuit of this system, characterized by Otx2 expression, is particularly responsive to chronic stress during puberty, which induces HIPS hypersensitivity to later stress and susceptibility to develop anxiety. We further show that Otx2 deletion restricted to the HIPS counteracts these effects of stress.
Together, these results demonstrate that Otx2 and stress interact, around puberty, to shape the HIPS stress-response, revealed here as a key modulator of susceptibility/resilience to develop anxiety.


Introduction
Most psychiatric illnesses occur in adolescents and adults. They are often triggered by physical or psychological, trauma in individuals presenting a susceptibility that was latent (Bale, Baram et al. 2010). On the contrary, a fraction of the population appears resistance (Hegde and Mitra 2020). A major challenge is to understand how susceptibility or resilience to psychiatric disorders is established.
Since many psychiatric diseases are now considered to have a neurodevelopmental origin, caused by an interplay between genes and environment during specific sensitive periods of development (Assary, Vincent et al. 2018, Lopatina, Panina et al. 2021), susceptibility or resilience should have their roots at these periods. While this hypothesis is widely accepted, the neural circuits and the modifications involved remain to be discovered. Major contributing factors include chronic stresses endured during early life (Ross, Foster et al. 2017), especially during periods termed "sensitive" or "critical", when environmental inputs combined with genetic or epigenetic factors shape neural circuits and influence their long-term functioning. It is therefore presumed that a sensitization or steeling effect of the stress would then set vulnerability or resilience to diseases, respectively (Rutter 2012). The nature of these opposite effects, which should be observed prior to detectable disease, remains mysterious.
Considering that chronic stress is a predisposing factor for depression and anxiety (Ross, Foster et al. 2017), investigators have applied it to animal models to identify the circuits and mechanisms involved. Among the validated form of chronic stress that can induce long-term anxiety-like and depressionrelated behaviors in rodents, restraint has been shown to be very efficient (Chiba, Numakawa et al. 2012) (Seewoo, Hennessy et al. 2020), and has many advantages. The psychological and physiological changes induced by the restraint result from the distress of the deprivation of movements. This procedure is thus painless and does not have the confounding effects of pain or injury associated with other physical stresses. It can be applied at any stage of postnatal development including pre-and postweaning unlike maternal separation or social defeat, and its intensity can be modulated over a wide range. Depending on the developmental stages when the stress is applied, different important processes can be affected ranging from neurogenesis to synaptic plasticity. Among critical periods, stress endured during pubertal period increases risk for anxiety and depression throughout adolescence and into adulthood (Horovitz, Tsoory et al. 2012, Holder and Blaustein 2014, Winer, Parent et al. 2016. Regarding the brain circuits involved, the reward system is considered a prime candidate to be the substrate for stress induced anxiety-like and depression-like behaviors. Indeed, there is a strong connectivity between stress response and reward circuits and, in fact, a considerable overlap between the two (Bouarab, Thompson et al. 2019, Metzger, Souza et al. 2021. This highly interconnected system is considered important not only for reward value, but also for overall adaptive behavior in the face of stressful events. Moreover, a central component of reward circuits, the habenulo-interpeduncular system (HIPS), is also involved in the regulation of mood, novelty attraction, motivation, and memory (Viswanath, Carter et al. 2013, McLaughlin, Dani et al. 2017). This network is composed of the medial habenula (MHb) and its unique output the interpeduncular nucleus (IPN). Clinical studies are linking HIPS dysfunction to major depressive and bipolar disorders (Boulos, Darcq et al. 2017) and animal model studies have shown that the HIPS contributes to the pathological mechanism of notably anxietyand depressive-like disorders (McLaughlin, Dani et al. 2017). Thus, following the neurodevelopmental hypothesis, HIPS developmental abnormalities may be at the origin of these disorders.
While the plasticity of other components of reward circuits, such as the amygdala and the ventral tegmental area (VTA), in response to early adversity has started to be studied (Pena, Kronman et al. 2017, Birnie, Kooiker et al. 2020), this has never been investigated for the HIPS so far: studies of this circuit are mainly focusing on adult physiology (Lee, Yang et al. 2019, Vickstrom, Liu et al. 2020, Yoo, Yang et al. 2021, and the few developmental studies of the HIPS have not addressed the direct relation between HIPS developmental abnormalities and stress-induced pathologies (Aizawa, Amo et al. 2011, Hsu, Morton et al. 2016. Importantly, although the HIPS is implicated in many aspects of behavioral control and is known to contain heterogenous neuronal populations, it is unclear which function is associated with which population. This complexity may be at the origin of the discrepancies found in the literature. For example, one study in mice concluded that the dorsal part of the MHb (dMHb) plays a role in controlling fear responses and learning (Yamaguchi, Danjo et al. 2013). Another study found no effect on these parameters (Hsu, Wang et al. 2014), but showed that this circuit is involved in motivation and hedonic state. In any cases, what is clear is that the HIPS represents a critical and integrative node of the feedback from cortical and subcortical regions to the brainstem in order to elicit adapted emotional responses. Regarding clinical research, the precise defects underlying psychiatric status has been difficult to address because of the genetic heterogeneity and the unique and complex patient's trauma history. It is therefore difficult to isolate the parameters that condition vulnerability or resilience. Lifting these scientific barriers requires a model designed to identify the type of alteration leading to vulnerability or resilience.
We have recently characterized a genetically-recognizable circuit within the HIPS (Ruiz-Reig, Rakotobe et al. 2018). The habenula-and interpeduncular-parts of this subcircuit both critically require the expression of Otx2 to develop and are marked by it, throughout development and into adulthood. The part where Otx2 is highly expressed in the medial habenula (Otx2 High ) corresponds to the centralmost part of both the dMHb and ventral MHb (vMHb). It also corresponds to the latest born habenular neurons and preferentially targets IPN regions that contain high density of Otx2+ IPN neurons with which it establishes synaptic contacts (Ruiz-Reig, Rakotobe et al. 2018). Accordingly, we named this subcircuit the HIPOPS (Habenulo-InterPeduncular-Otx2-Positive-System). Otx2 is a developmental gene that encodes a homeoprotein with many important brain functions. Its expression is highly conserved throughout evolution (Ranade, Yang-Zhou et al. 2008, Sen, Reichert et al. 2013, making it a good marker of conserved brain circuits. Based on all these evidences regarding Otx2 and HIPS functions, we hypothesize that the HIPOPS plays important and conserved functions inside the reward system. Beyond the HIPOPS, Otx2 marks several other interconnected brain regions that also belong to the reward circuits. For example, Otx2 has recently been shown to play major roles during specific critical periods of postnatal development of the ventral tegmental area (VTA) (Pena, Kronman et al. 2017).

4
Early life stress (ELS) reduced Otx2 expression in the VTA only during the stress period and this reduction was the main factor that induced long-term susceptibility to depression-like behavior. Thus, Otx2 is an important player of critical periods of neuronal development and disturbing its expression may lead to long-term effects on brain functions related to the reward system. In these particular periods, an association between Otx2 expression and the maturation of one type of cells, the parvalbumin cells (PV cells), appears to be one of the driving forces that makes these periods, critical windows of development (Lee, Bernard et al. 2017). During these developmental phases, circuits are more plastic and are subjected to long term changes, under the influence of environmental cues (Hensch 2005). One such change is the appearance of perineuronal nets (PNNs), notably, around PV cells. PNNs may lock and stabilize existing synaptic contacts. Coincidentally, Otx2 is linked to human pathologies involving the reward system. It was found to be a susceptibility gene for bipolar diseases and stress-related disorders (Sabunciyan, Yolken et al. 2007, Maheu andRessler 2017). Because HIPS dysfunctions may be involved in these diseases, it is possible that the role of Otx2 in critical periods of HIPOPS development is important in the process.
In this study, we have developed a new paradigm to investigate the neurodevelopmental basis of risk/resistance to specific psychiatric diseases and focused on the HIPOPS. We used restraint, as a form of psychological and physical stress. We employed a two-hit restraint stress protocol in which the first stress applied during development is meant to generate susceptibility or resilience that translate into a pathological state only following a second period or "hit" of stress in later life (Daskalakis, Bagot et al. 2013). That way, we uncovered the existence of a critical period of HIPOPS development under the influence of environmental stressors. Individuals undergoing restraint stress during this period exhibited a specific hyperactivity of the HIPOPS compared to other circuits of the HIPS. These changes did not by themselves provoke any state of anxiety, but they induced long-term modifications in the circuit that made animals more prone to develop anxiety-like behaviors when facing new periods of stress, later. Furthermore, Otx2 is a contributing factor since reducing its expression specifically in the HIPOPS altered the stress response and abolished this long-term susceptibility.

HIPOPS circuits are interconnected and Otx2 High MHb region drives Otx2 cIPN activity
Previous results showed that MHb Otx2 HIgh neurons preferentially make synapses with IPN neurons that express Otx2, located in two subnuclei of the caudal IPN, the cIPN (centro-caudal IPN) and lat-IPN (latero-caudal IPN). From these results we hypothesized that Otx2 marks a subcircuit inside the HIPS that may be preferentially interconnected (Ruiz-Reig, Rakotobe et al. 2018). To test this hypothesis further, we asked whether chemogenetic activation of MHb Otx2 High neurons that are glutamatergic would preferentially drive the activation of their hypothesized targets in the cIPN. We performed stereotactic injection of AAVs expressing a floxed Gq-coupled hM3D DREADD fused with mCherry under the control of human synapsin promoter into the MHb of Otx2 CreERT2/+ and wild type male mice (Fig.1a,b). To monitor neuronal activation, we examined the expression of immediate-early genes that are strongly and specifically induced after neuronal activation (Zangenehpour and Chaudhuri 2002). As expected, in wild type mice, after tamoxifen induced recombination and CNO mediated activation of the DREADDs, no evidence of neuronal activation could be detected in the MHb or in the IPN (Supplemental Fig.S1). The absence of activation was also found in Otx2 CreERT2/+ mice when CNO was omitted, despite the presence of DREADDs visualized by mCherry decorated MHb Otx2 High neurons (Supplemental Fig.S1). In contrast, upon CNO injection in Otx2 CreERT2/+ mice, a strong Fosimmunoreactivity was found in dMHb neurons expressing hM3D in the Otx2 High region ( Fig.1c-f). In the IPN target, neuronal activity could be detected in neurons of the cIPN and lat-IPN (Fig.1g,h). Moreover, only cells that were Otx2 + were strongly labeled ( Fig.1h1-h2,'arrows). In the lat-IPN, only cells in the central-most part of it (C-lat-IPN) were Fos-immunoreactive ( Fig.1c; green cells). Thus, forced activation of Otx2 High neurons in the MHb triggers activity specifically in Otx2 + neurons in the IPN.
These results show that MHb Otx2 High neurons are able to modulate the activity of Otx2 + IPN neurons and also confirm that these neurons form a sub-circuit within the HIPS that can be marked by the expression of Otx2, which we therefore called the HIPOPS.
Interestingly, in the MHb, Fos-immunoreactivity was not only found in DREADD + neurons. Many neurons within the MHb and some in the lateral habenula were also labelled ( Fig.1f; arrows) suggesting that Otx2 High neurons can activate local habenular neurons in addition to the neurons in the cIPN that express Otx2.

HIPOPS critical period of response to stress
We then wanted to investigate the response of habenular neurons and their IPN targets upon stressinduced neuronal activation. As we wanted to test different developmental stages, we selected physical restraint.
We tested the effect of a chronic stress (CS) on HIPS activity by applying the same restraint stress 2 hours a day for 7 consecutive days, on male mice. After the last stress, c-Fos and Egr1 (Krox-24, Zif-268), another immediate early gene (Zangenehpour and Chaudhuri 2002), were strongly induced in the dMHb compared to control in a part that corresponded to the Otx2 High region ( Fig.2a-h). Egr1 appeared an even more robust marker of stress-induced activity (Fig2.c,d versus Fig.2c',d'); and was, thus, used thereafter. Egr1 induction was particularly evident post-weaning into the pre-adolescence and adolescence stages diagram Fig.2g and graph Fig.2h). Following the same trend, activation of the IPN was strong and more evident at these stages ( Fig.3 and graph Fig.3k). Interestingly, the neurons particularly activated belonged to Otx2 + neurons of the cIPN and of the C-lat-IPN ( Fig.3a-h and graph 3k). We then tested the effect of an acute stress (1 session of 2 hours restraint) which did not significantly activate neurons of the MHb (Supplemental Fig.S3) nor Otx2 + neurons of the cIPN. Such stress, however, activated IPN neurons that were Otx2or Otx2 + confined to the L-lat-IPN (Fig3a purple cells and Fig.3i). Those Otx2 + neurons displayed a different molecular identity than the ones activated by a chronic stress. They expressed parvalbumin (Fig.4a) and were surrounded by perineuronal nets (PNN), a particular form of the extracellular matrix (Fig.4b). PNN structures were particularly evident from the pre-adolescence stage suggesting that these neurons were immature prior to that (Fig.4c). We therefore called them, Otx2 PNN+ neurons.
To verify that this activation was due to the stress and not to any other type of event that the animals would experience, we tested the response of the HIPOPS following an exposure to a pleasurable event such as pleasant-tasting sucrose containing water. In contrast to stress, this condition elicited very little response ( Fig.3j) confirming the specificity of the IPN response towards unpleasant stressful event.
Thus, the HIPS responds to stressful events. Chronic stress activates MHb neurons in the Otx2 high region and consequently, Otx2 + neurons in the cIPN and C-lat-IPN. The peak of activation is around the preadolescent period (P30-P36) (Fig.2g). By contrast, acute stress only activates Otx2 + neurons in the L-lat-IPN characterized by PV expression and surrounded by PNNs (Otx2 PNN+ neurons) as well as some Otx2neurons. The maturation of Otx2 PNN+ neurons coincides with the time window when the HIPOPS is strongly activated by stress.

Restraint Stress during hyper-responsive period precipitates anxiogenic effects of adult stress
We then wanted to test the consequences of chronic stress applied during the pre-adolescent period on the susceptibility to develop anxiety-like behavior on the long term. In accordance with the two-hit stress model, we wanted to apply a mild stress such as the one we used above (2 hours of restraint for 7 consecutive days) during the HIPOPS hyper-responsive period (P30-36), which by itself would not generate detectable long-term consequences on anxiety-like behavior but would produce a susceptibility to develop anxiety-like behavior. This susceptibility would then require a second hit of chronic stress to become detectable at the behavioral level (Fig.5a).
We first verified that 7 days of chronic stress applied from P30 to P36 would not result in behavioral abnormalities by testing general motility and anxiety levels 2 weeks after the end of the period of stress. No differences could be found between controls and stressed animals ( Fig.5b-e). In sharp contrast, several weeks after a second period of stress applied during early adulthood (P60-P66), animals that had received the early stress presented higher anxiety-like behavior based on distance travelled, number of entries and time spent in an anxiogenic zone ( Fig.5f,h-i). These differences could not be explained by motor dysfunctions as evidenced by a similar mean speed of the animals. In contrast to anxiety, depressive-like behavior was not significantly affected ( Fig.5j), as measured by the Porsolt swim test. Interestingly, though, curiosity toward a new object was reduced (Fig.5k). However, this reduction in curiosity could be explained by potential anxiety toward the new object. Indeed, stressed mice were not more indifferent to the new object but avoided it, compared to unstressed mice.
These results show how a chronic stress during the pre-adolescent period can induce susceptibility to anxiety-like behavior revealed at the adult stage upon stressful events. This time window may thus represent a critical period of HIPOPS development particularly involved in setting long-term anxiety levels.
4. Deletion of Otx2 restricted to the HIPOPS protects from susceptibility to develop anxiety-like behavior on the long-term.
To directly implicate the HIPOPS in the mechanisms leading to increased anxiety-like behavior, we tested whether interfering with the normal development of the HIPOPS would change the outcome of individuals that experienced the early stress. Since Otx2 is a master gene of HIPOPS development, we designed a strategy to delete Otx2, specifically in these circuits. We used the Ngn1 CreERT2 line, which expresses an activable form of Cre driven by the Ngn1 promoter (Koundakjian, Appler et al. 2007, Kim, Hori et al. 2011. In this line, CreERT2 expression becomes gradually restricted to a few sites as mice age. The lineage of CreERT2 expressing cells was followed from P0 to P30. Otx2 + cells belonging to this lineage were almost exclusively found in the dMHb ( Fig.6 and Supplemental Fig.S3). Only one other site could be found; it corresponded to rare photoreceptors in the retina (Supplemental Fig.S3).
Otx2 invalidation was triggered by tamoxifen injection at P0. At P30, Otx2 flox/flox ;Ngn1 CreERT2 animals were compared to Otx2 flox/flox or to Otx2 +/+ ;Ngn1 CreERT2 mice. All received tamoxifen. The habenula in these mice appeared normal. Projections from the neurons that had lost Otx2 expression were maintained, surrounding their lat-IPN targets (Fig.6b-e'), as in control mice. The size of the MHb was unchanged further supporting that the postnatal ablation of Otx2 did not affect their survival (Fig.6f). We next assessed the behavior of these mice at the adult stage at P100. No abnormalities were detected in term of anxiety levels or motor parameters measured by the mean speed or distance travelled in the open field ( Fig.7a-b).
Thus, postnatal Otx2 deletion in the dMHb and in rare cells in the retina does not result in anatomical or behavioral abnormalities under standard mouse rearing conditions.
The dMHb is particularly active during chronic stress and especially during the critical period of HIPOPS development. We therefore tested whether Otx2 deletion would affect stress-induced anxietylike behavior. We first assessed behavior after an initial stress hit. As in control mice, we did not detect behavioral abnormalities induced by the first stress ( Fig.7c-e). We next analyzed behaviors after the two-hit stress protocol. As expected, in two different controls Otx2 flox/flox and Otx2 +/+ ;Ngn1 CreERT2 , the two-hit stress protocol increased anxiety levels; however, this stress effect could not be observed in the Otx2 flox/flox; Ngn1 CreERT2 mice (Fig.7f). In fact, all the parameters that were affected in control animals (see Fig.5), after the early + late stress compared to the late stress alone, were unaffected in the Otx2 cKO mice (Fig.7g-i). Comparison of a global score of anxiety level based on both Elevated Plus Maze and open field parameters confirmed that only control mice were statistically affected by the two-Hit stress protocol (Fig.7k). The time of inactivity in the FST test was similar in Otx2 flox/flox ;Ngn1 CreERT2 mice with or without the early stress (Fig.7j). This result was expected as the stress-protocol did not induce depressive-like symptoms, even in the controls. In the curiosity test, the proportion of curious, anxious and indifferent individuals facing a new object were similar between the no stress and the twohit stress protocol groups (Fig.7l). This is dramatically different from what we observed in control animals (see Fig5k and Fig.7l). Interestingly, the effect of adult stress alone followed the same trend in controls ( Thus, Otx2 deletion in the dMHb leads to a relative protection against stress effects on the long-term regarding anxiety and curiosity. This effect of resilience seems to be the consequence of an altered interaction between Otx2 and stress specifically during the juvenile period.
5. Deletion of Otx2 restricted to the HIPOPS leads to a bimodal response to juvenile chronic stress and protects from stress-hypersensitivity of the circuit later on.
To understand the mechanisms of resilience to stress observed in Otx2 flox/flox ;Ngn1 CreERT2 animals, we analyzed the response of their HIPOPS to juvenile stress. Interestingly, in Otx2 flox/flox ;Ngn1 CreERT2 mice, although the dMHb was even more active than in controls at the end of the juvenile chronic stress, IPN neurons did not respond (Fig.8a-d and graph Fig.8g). One possibility is that Otx2 deletion leads to abnormal connectivity between the dMHb and its IPN output. However, at earlier time points during the protocol of stress, Otx2 + cIPN and C-lat-IPN neurons were readily activated and this activation correlated with a relatively strong dMHb activation, even after an acute stress ( Fig.8e-f). This was surprising as an acute stress is normally not strong enough to significantly activate the dMHb at any of the time points tested from P20 to adulthood (see supplemental Fig.S2).
We then tested whether this biphasic response to the early stress would have an impact on the long-term response of the circuit when confronted again to stress. We examined the status of the circuit at the adult stage, in controls and Otx2 cKO that had or had not undergone the two-hit protocol, when subjected to a final acute stress, 2-hour before sacrifice. No activity was detected in the dMHb in any of the conditions (Fig.9a,c). Higher activity in IPN neurons could be detected in control animals with previous two-hit stress (Fig.9d), compared with the activity observed in Otx2 cKO with prior two-hit stress (Fig.9e) or in controls with no previous stress (Fig.9f). As expected, following an acute stress in controls (see Fig3i), the most activated Otx2+ neurons corresponded to neurons located in the L-lat-IPN, including the PNN + population (Fig.9f1,f1'). Many more cells were strongly Egr1-immunoreactive in control animals with previous two-hit stress protocol ( Fig.9d and graph in Fig.9g). These cells belonged to both PNN+ (Fig.9d1, d1') and PNN-populations (Fig.9d2). In sharp contrast, both populations were weakly Egr1-immunoreactive in Otx2 cKO animals ( Fig.9e-e2). Interestingly, many other regions known to be involved in stress response followed the same pattern as the HIPOPS. Indeed, regions such as the paraventricular nucleus of the thalamus (PV) or the LHb (Reviewed in (Hikosaka 2010, Rowson andPleil 2021) and Fig.9a-c) presented a higher activity in control mice subjected to previous two-hit stress, compared to the activity found in both control and Otx2 cKO animals without or with prior two-hit stress, respectively.
Thus, the deletion of Otx2 in the dMHb changes the normal response of the HIPOPS to a byphasic response with hyperactivation in the beginning of the stress followed by inhibition of the IPN targets as the stress becomes more chronic. In the long term, Otx2 ablation in the dMHb counteracts the hyperresponsiveness of HIPOPS neurons and stress-related circuits, correlated with apparent protection against the development of stress-induced anxiety. This suggests that genetic factors, such as Otx2, interact with environmental stress at the level of the HIPOPS, during a critical period of development. This also suggests that blocking this interaction, through deletion of Otx2 in the dMHb, leads to a form of resilience to develop anxiety in adults. Finally, it suggests that HIPOPS neurons can drive this longterm anxiety and that the experience of stress by the HIPOPS, during the juvenile period, conditions the response of stress-related circuits to later stress-experiences (Fig.10).

Discussion
In this study, we have shown that chronic restraint stress during the P30-P36 period, corresponding to the sexual maturation period in mice (pre-adolescent period), specifically induces the HIPOPS circuit within the HIPS and makes animals more prone to develop high levels of anxiety but not depression. However, a role of the HIPOPS in the latter may have gone unnoticed because the stress used did not induce depression. Thus, while we cannot conclude on a potential role for the HIPOPS in depression, this study clearly shows its involvement in stress-induced anxiety. The two-hit stress protocol that we developed will certainly become a valuable tool to study mechanisms of susceptibility or resilience. Indeed, one can study neuronal circuit abnormalities that may be the basis of susceptibility or resilience without confounding effects of a pathological state only triggered after the second hit of stress. These early changes may be at the origin of the stress hypersensitivity that seemed to finally spread to the whole stress-related circuit. Also, we had previously shown that Otx2 is important for HIPOPS early development and now suggest that Otx2 is a master gene of HIPOPS development and function We showed that from P36 onwards, HIPOPS neurons are not responsive to acute stress, except for PV + /PNN + neurons. PNN structures revealed by WFA-imunoreactivity are rare before P30 (see Fig.4). Interestingly, before P30, neurons in the prospective PNN + region are not activated upon an acute stress (Fjerdingstad et al; unpublished observation). Thus, the maturation of PNN + cells may well coincide with the maturation of acute stress response. We have no direct evidence that the maturation of these cells is related to the changes in circuit activity and behavior that we observed. However, this may be an interesting avenue to pursue. Indeed, PV + cell maturation has been shown to control the opening and closing of critical periods in the brain and spinal cord (Wingert and Sorg 2021). It is therefore particularly interesting that this period corresponds also to the most sensitive period of HIPOPS activation. Environmental stress experienced at this period may, therefore, set the final sensitivity of the HIPOPS to stress in the long term (Fig.10), through habenula activation and PNN + cell maturation.
To explain what we observed, one possibility would be the existence of a bidirectional connectivity between the two HIPOPS sub-populations, PNN + /Otx2 + and PNN -/Otx2 + cells. When PNN + neurons, become mature, they can inhibit other HIPOPS cells during an acute stress. If the stress becomes chronic, Otx2 High neurons of the habenula are strongly activated and overcome the inhibition from PNN + cells on Otx2 + /PNNneurons, leading to the activation of the latter. In turn, these newly activated Otx2 + cells, also GABAergic, inhibit the PNN + cells. Although, still speculative this crosstalk deserves investigation as it would provide a straightforward explanation of our observations of the differential response to acute or chronic stress.
Another major result of this study is that the Otx2 protein, essential for the development of HIPOPS, also plays an important role during the P30-P36 period in setting the stress response of this circuit for the entire life. In control animals, the progressive activation of Otx2 High habenular neurons as the stress becomes chronic, leads to a strong activation of Otx2 + IPN neurons, until the end of the stress, correlated with a stress hyper-response of the HIPOPS in adults. In contrast, in Otx2 cKO animals, there is a relative stronger activation of Otx2 High habenular neurons, correlated with a paradoxical inhibition of Otx2 + IPN neurons towards the end of the stress period that may be due to long term depression upon overstimulation of the IPN or a gradual loss of function of Otx2 high neurons, leading to stress hyporesponse of Otx2 + IPN neurons, in adults. These observations suggest that the level of HIPOPS activation, at the end of the stress period at P36, determines the intensity of HIPOPS response to further stress episodes after the closure of the hypothetical critical period (Fig.10). Speculating on reciprocal connection between Otx2 + /PNN + and Otx2 + /PNNneurons of the IPN, a strong activation of Otx2 + /PNNcells at the end of the stress period at P36 would lead to weak reciprocal inhibition in the long-term leading to HIPOPS hyperactivity. On the contrary, an inhibition of Otx2 + /PNNcells at the end of the stress period at P36 would lead to a strong reciprocal inhibition in the long-term, leading to a hypo-reactive HIPOPS.
Whether or not reciprocal connections exist, hypo-responsivity due to a relative too early or too strong habenular activation could explain why individuals with a certain genetic background develop resistance rather that susceptibility to disease after trauma. Along this lines, lower levels of Otx2 could be an important factor in promoting resilience to stress-induced anxiety, at least at these developmental periods. This fits well with the hypo-anxiety that has been recently described in Otx2 heterozygote mice in which Otx2 is expressed at a twofold lower level than wild-type mice (Vincent, Gilabert-Juan et al. 2021). Such protection against stress-induced anxiety observed in Otx2 cKO mice cannot be explained by the few recombined photoreceptors. Indeed, deletion of Otx2 in all photoreceptors does not affect anxiety levels (Pensieri et al.; unpublished observation).
To demonstrate the sensitivity of the HIPOPS to stress, we did not directly record electrical activity, but monitored the expression of immediate early genes such as Egr1 and c-Fos that integrate more cellular activities. For instance, Egr1 is regulated by a wide variety of environmental stimuli, including stress, and in turn can regulate transcriptional program related to neuronal activity and synaptic plasticity (Gallo, Katche et al. 2018). Therefore, Egr1 represents a key integrating factor between environmental perception and translation to the appropriate response. It is therefore not surprising to find Egr1 associated with neuropsychiatric disorders in which neuronal plasticity and activity are altered, such as stress-induced anxiety (Gallo, Katche et al. 2018). Whether or not they reflected direct neuronal activity, the changes in Egr1 levels we found are significant and may be even more valuable than direct recording, as they reflect neuronal activity or plasticity of neurons in response to stress. A thorough electrophysiological correlate of these changes, however, would be helpful.
Animal models for reproducing stress-induced mood and anxiety disorders have largely focused on the pre-weaning or adult period. Here we found that the period that particularly involves the HIPOPS in the stress response is the period of sexual maturation. This period of development is a known window of vulnerability to anxiety, depression, schizophrenia, and substance abuse (Dahl 2004, Patton andViner 2007), and exposure to stress plays an important role (Grant, Compas et al. 2004). It is therefore of utmost importance to further investigate the role of the HIPOPS during this period as it may hold the key to understanding susceptibility versus resilience to stress-induced anxiety.

Animals
All mice used were generated and maintained in the animal house of the Institut de Biologie Valrose, Nice, France. All mouse lines were kept in the 129/Sv background. Only males were used in this study to avoid having effects due to sexual dimorphism. Otx2 CreERT2/+ and Otx2 flox/flox mice were generated as described previously (Fossat, Chatelain et al. 2006). The Ngn1 CreERT2/+ ;Ai14/Ai14 line was kindly provided by Lisa V. Goodrich (Koundakjian, Appler et al. 2007). The Ai14 allele was used as the reporter of Ngn1 CreERT2/+ expression. Adult mice were used for viral injection of DREADDs. Behavioral tests were done during the day cycle of the 12h day/night cycle. Postnatal deletion Otx2 was achieved injecting intraperitoneally 15-30µl of tamoxifen in neonates at P0 at a concentration of 2,5mg/ml with 300µl insulin syringes. Care and handling of the animals prior or during the experimental procedures followed European Union rules and were approved by the French Animal Care and local ethic Committees.

DREADDs microinjections
Animals were anesthetized by intraperitoneal injection (IP injection) with a mix of Tiletamine-Buprenorphine-Zolazepam-Xylazine (TBZX) (60 mg/kg). Xylocaïne gel was applied on the skull prior to the incision and in the ear to lessen the pain perception. After disinfection of the area of incision, each mouse was successively placed and maintained on the stereotaxic apparatus. The eyes were protected from drying with ocry-gel. The cannula was implanted unilaterally from the bregma based on Paxinos et Franklin (2004). The implantation was successively made on the left and then on the right side, in the MHb according to these coordinates : A/P= -1,5 ; D/V= -2,75 ; lateral= +/-0,2. 500 nl of the activating Gq-coupled DREADDs hM3D receptor (pAAV9-hSyn-DIO-hM3D(Gq)-mCherry, 100 µL from titer ≥ 1×10¹³ vg/mL) (Cat No. 44361-AAV9; Addgene, USA) were injected at a speed of 100 nl/ms in Otx2 CreERT2/+ or Otx2 flox/flox and Otx2 +/+ control mice bilaterally. At the end of the injection, the cannula was left in place for infusion for 5 min at the site of injection before withdrawal. At the end of the surgery, an anti-inflammatory (Metacam; ,5 mg/ml) was administered intraperitoneally. The antiinflammatory treatment was repeated for 3 days post-surgery. Ten days after the transduction, tamoxifen (Sigma-Aldrich) diluted in corn oil at a concentration of 10mg/ml was IP injected at 5 μl/g of body weight to induce the activation of Cre and, thus, the expression of DREADDs in the neurons of the MHb. Ten days after the tamoxifen injection, Clozapine N-oxide hydrochloride (CNO) (SML2304, Sigma-Aldrich) diluted in DMSO was IP injected at a concentration of 0,3 mg/kg. The mice were left undisturbed, for 7h, to let Egr1 expression return to baseline, as it is induced by the stress of the injection, and then they were rapidly sacrificed.

Stress protocol
Mice were placed in pierced syringes suited for their size at each stage studied, to restrain their movements as much as possible. The restraint was used acutely (one unique 2h-restraint) or chronically (2h/day for 7 days). For the analysis related to the study of neuronal activity at the different developmental stages, the mice were immediately sacrificed at the end of the acute stress or at the end of the last chronic stress. The control mice that did not undergo any protocol of stress were quickly sacrificed at the proper age. At the end of the behavioral analysis, mice for each condition were divided into two groups: a group of mice that underwent a last restraint stress before the collection of the brains, and a group that was immediately sacrificed.

Acute sucrose consumption
The water in the bottle of the cage was replaced by a 30% sucrose solution for the night. The sacrifice occurred the following day.

Behavioral tests
The open field test and the elevated plus maze test were used to assess anxiety level during the Behavior 1 and Behavior 2 steps. We also assessed the curiosity level at both steps. Depression level was assessed using the forced swim test.
Mice were acclimated in the behavioral room 1h prior to the beginning of the tests. Videos were acquired and analyzed with the EthoVision XT software. For all the tests, the arena was cleaned with 70% ethanol and dried at the end of each trial.

Open field test
The open field test was used to measure the anxiety levels and the ambulatory behaviors of the mice. The test was recorded in an arena in which the walls and the floor are made of black PVC. The box dimension is (40 W x 40 L x 30 H cm) and was divided into the border area (50% of the total arena) and into the center area. The field was moderately illuminated at 60 lux. Mice were individually placed in a corner of the arena and the behavior was recorded for 5 min. The time spent, the latency to enter, the number of entries, the mean speed and the total distance travelled in the center, which is an anxiogenic area for the mice, were measured.

Elevated plus maze test
The Elevated Plus Maze test (EPM test) was used to assess the anxiety level. This apparatus is 50 cm elevated from the ground and composed of the open arms (50 x 5 cm) which are anxiogenic for the mice, and two closed arms (50 W x 5 L x 16 H cm) which spread from the central platform (5 x 5 cm).
Mice were placed at the center of the platform, head toward a closed arm and the video was recorded for 5 min. The latency to explore the open arms and the time spent in the open arms were measured.

Curosity test
This test allows the measurement of curiosity level of the mice, an aspect of anxiety. An exploratory zone, containing a transparent cylinder with holes, was designed in the middle of one side of the wall of the open field. Mice were placed facing that side and the video was recorded for 5 min (Habituation phase). Following the habituation phase, an object was placed inside the cylinder and the mouse was relocated at the opposite side of the object and the video was recorded for 5 more min (Exploration of the object phase). The percentage of time spent in the exploratory zone, during the exploration of the object phase, is calculated and the statistics were made on the values obtained. We also used these values to divide the mice into subgroups: mice that spent a percentage of time ≤ 40% in the exploratory zone are classified as anxious, those that spend a percentage of time ≥ 60% are classified as curious, the mice in between are classified as indifferent.

Forced swim test
The forced swim test was used to measure the depression level of mice. Mice were placed individually in a transparent cylinder (12 cm diameter; 30 cm of height) high enough to avoid escaping or touching the bottom filled at 25cm from the bottom with water (23-25°C) for 6 min. The time of immobility during the last 4 min was measured on the EthoVision software. The immobility was defined as the lack of movement, except those necessary to sustain the head outside of the water. The water was changed at each trial.

Immunohistochemistry and histological analysis
Mice used to identify critical periods of development were directly sacrificed at the end of the acute stress or at the end of the last stress in the two-hit stress paradigm. Brains were dissected and directly frozen in Tissue-Tek OCT compound at -80°C (Fisher Scientific, Waltham, MA, USA). In order to preserve endogenous fluorescence of the reporter, mice transduced with the DREADDs were subjected to intra-cardiac perfusion using PBS, then 4% paraformaldehyde (PFA). The same treatment was done to Otx2 +/+ ;Ngn1 CreERT2/+ ;Ai14/Ai14, Otx2 flox/flox ;Ngn1 CreERT2/+ ;Ai14/Ai14 and Otx2 flox/flox Ai14/Ai14 mice used to study the anatomical effect of Otx2 deletion in the MHb, the neuronal activity in response to stress between P30-36 and to mice that underwent the behavioral analysis. Brains were post-fixed overnight in 4% PFA at 4°C and cryoprotected in 30% PBS sucrose before freezing in Tissue-Tek OCT compound at -80°C. Eyes were collected from dislocated mice and placed in cold PBS. A small hole was done at the level of the ora serrata. The eyes were fixed in 4% PFA for 2h at room temperature, cryoprotected in 20% PBS sucrose overnight, then frozen in OCT and stored at -80°C.
Brains were cut in 16µm coronal sections using a Microm HM550 cryostat and sections were mounted on SuperFrost+ slides (Fisher Scientific). Eyes were cut in 14 µm sections collected on SuperFrost+ slides. The sections directly frozen were post-fixed in fresh 4% PFA. Before the blocking step, sections from perfused Otx2 +/+ ;Ngn1 CreERT2/+ ;Ai14/Ai14, Otx2 flox/flox ;Ngn1 CreERT2/+ ;Ai14/Ai14 and Otx2 flox/flox Ai14/Ai14 mice were subjected to an unmasking step to better reveal the Otx2 signal that is weaker in perfused brains. In this step, the slides were incubated in Citrate Buffer, pH 6, for 8 min at 95°C. A second treatment was performed with PBS-Glycine (100 mM glycine) to limit background noise. The rest of the steps were similar for all the samples. Briefly, the slides were incubated in a blocking solution (PBS with 0,2% dTriton 10% FBS) for 1h. The primary antibodies were incubated overnight at 4°C in a solution of PBS (0,1% Triton-10%FBS). The sections were incubated 10 min in DAPI (1µg/ml).

Microscopy
Images of the retina were acquired on a confocal microscope Zeiss 780 (Zeiss, Oberkochen, Germany). All other images were acquired on the wide-field microscope AxioObserver -Zeiss (2011) with the sCMOS ANDOR Neo camera and the stereomicroscope coupled with the digital camera (Zeiss axioplan2). Images were processed with the ImageJ software and figures were mounted on Adobe Photoshop.

Statistical analysis
To study the interaction between anxiety phenotype with genotype, rankings were made by grouping the animals with the same genotype (stressed and non-stressed mice). A rank was attributed to the mice according to the value of the variable (time spent in the open arms of the EPM), ordered increasingly. Statistical analyses were made using the R software. Results are presented as the mean ± SEM. The comparison of the mean of two groups was done by using the Student t-test when the variance was equal and with a Welch correction when it was not. Anova were calculated using the ANOVA function of the package "car" and corrected for the type III error for unbalanced datasets. Comparison of three groups were made with a one-way ANOVA. The interaction effects between the chronic stress and the Otx2 deletion were revealed by using a two-way ANOVA. Statistical details can be found in figure legends and in supplemental materials (Supplemental table S1).   ) and in response to a chronic stress between P30-36 (c and c'). d-j, Immunolabelling of coronal sections representing the area framed in (a) with anti-Egr1 (in red) and anti-Otx2 (in green) antibodies without stress P36 (d) and in response to a chronic stress between P30-36 (e), P37-43 (f), P14-20 (g), P60-66 (h) or an acute stress (i) or in response to sucrose consumption (j). The white arrows indicate a colocalization of Egr1 and Otx2 immunoreactivities. The magenta arrows indicate the same neurons as in Figure 4. k, Quantification of neuronal activity based on density of Egr1-immunoreactivity, after 7 days of stress or without stress as a control at different postnatal periods (chronic stress P14-20 n=3, P30-36 n=3, P37-43 n=3, P60-66 n=4 and no stress P20 n=4, P36 n=4, P43 n=4, P66 n=3) in the caudal IPN region (left graph) or specifically in Otx2 + and Otx2neurons of the caudal IPN (chronic stress P30-36 n=3 and no stress P36 n=4) after 7 days of stress between P30-36 (right graph). Central most part of the lat-IPN (C-lat-IPN). The error bars represent the SEM. * P < 0.05, ** P < 0.01, bilateral t-test. Scale bar 100 µm.    and Otx2 cKO mice that have endured juvenile (n=10, 6, 12, respectively) or juveline + adult stresses (n=10, 7, 11, respectively). g-j, Behavioral tests evaluating anxiety-like (g-i) and depression-like phenotype (j) by measuring the total distance traveled, the number of entries in the open field, the total time spent in the open arms of the elevated plus maze and the inactive time in the forced swim test for Otx2 cKO mice that have endured adult (n=6) or juveline + adult stress (n=12). k, global scoring for Ctrl1 and Otx2 cKO of anxiety-like levels based on mean ranking in both open field and elevated plus maze. l, Percentage of mice in the curious, indifferent and anxious subgroup in the curiosity test. m, Interaction plot representing the effect of Otx2 deletion and juvenile chronic stress associated with the adult stress on the time spent in the exploratory area in the presence of an object (F(1,38)= 6,7127, Interaction effect p=0, 01350). The error bars represent the SEM. *P < 0.05, **P < 0.01, one-way ANOVA with Holm-Šídák post-hoc test (a,b,k), bilateral t-test (c-j), two-way ANOVA (m).

Figure 8. Effects of Otx2 deletion in the MHb on the response to pre-adolescent stress. a-f,
Labelling of coronal sections with anti-Egr1 antibodies (in red) and DAPI staining (in blue) at the level of the MHb (a, c, e) and with anti-Egr1 (in red) and anti-Otx2 (green) antibodies at the level of the caudal IPN (b, d, f) in controls (n=3) (a-b) and in cKO mice (n=3) that underwent a chronic stress between P30-36 (c-d) and an acute stress at P30 (n=3) (e-f). a1-a2, b1-b2, c1-c2, d1-d2, e1-e2, f1-f2 Zoom of the area framed in a, b, c, d, e, f, respectively. g, Quantification of neuronal activity based on density of Egr1-immunoreactivity, after 7 days of stress between P30-36 (CS P30-36) in the MHb and the caudal IPN of control (n=3) and Otx2 cKO mice (n=3). The error bar represents the SEM. *P < 0.05, **P < 0.01,, bilateral t-test. Scale bars: 100 µm. Figure 9. Effects of Otx2 deletion in the dMHb on the neuronal activity of the HIPOPS in the longterm in response to a last acute stress. a-f, Labelling of coronal sections with anti-Egr1 antibodies (in magenta) at the level of the MHb (a-c) and the caudal IPN (d-f) after the "two-hit stress" protocol (juvenile+adult chronic restraint stress) followed by a last acute stress in controls (a,d; n=3) and Otx2 cKO mice (b,e; n=3), and in controls that only endured the acute stress before the sacrifice (c,f; n=3). a1-a2, b1-b2, c1-c2, d1-d1'-d2, e1-e1'-e2, f1-f1'-f2, Zoom on the area framed in a, b, c, d, e, f, respectively. PNN + cells are labeled with WFA (in green), and anti-Egr1 antibodies (in red) in d1', e1' and f1'. g, Quantification of neuronal activity based on the density of Egr1-immunoreactivity after a final acute stress before sacrifice in the caudal IPN of controls and Otx2 cKO that have or have not endured the two-hit stress protocol. The error bar represents the SEM. **P < 0.01, one-way ANOVA with Holm-Šídák post-hoc test. Scale bars: 100 µm. Figure 10. Proposed mechanism of susceptibility or resilience to juvenile-stress-induced anxiety.
The maturation of the stress response in HIPOPS circuits is taking place during the prepubertal period.
In control situation, no chronic stresses are endured, and the maturation is normal in the presence or absence of Otx2 (green curve). If a chronic stress is experienced during the critical period, the activity is higher by the time of closure (red curve). In the absence of Otx2 expression in the MHb, a chronic stress activates even more strongly the MHb, leading to abnormally low response of the IPN by the time of critical period closure (blue curve). In all cases, following closure of the critical period, the IPN activity in response to new stresses during later life is dependent on the activity it had at the end of the critical period. This activity reflects the susceptibility or resilient state of the individual to develop pathologies such as chronic anxiety.