Stress deceleration theory – chronic adolescent stress exposure results in decelerated neurobehavioral maturation

Normative development in adolescence indicates that the prefrontal cortex is still under development thereby unable to exert efficient top-down inhibitory control on subcortical regions such as the basolateral amygdala and the nucleus accumbens. This imbalance in the developmental trajectory between cortical and subcortical regions is implicated in expression of the prototypical impulsive, compulsive, reward seeking and risk-taking adolescent behavior. Here we demonstrate that a chronic mild unpredictable stress procedure during adolescence in male Wistar rats arrests the normal behavioral maturation such that they continue to express adolescent-like impulsive, hyperactive, and compulsive behaviors into late adulthood. This arrest in behavioral maturation is associated with the hypoexcitability of prelimbic cortex (PLC) pyramidal neurons and reduced PLC-mediated synaptic glutamatergic control of BLA and nucleus accumbens core (NAcC) neurons that lasts late into adulthood. At the same time stress exposure in adolescence results in the hyperexcitability of the BLA pyramidal neurons sending stronger glutamatergic projections to the NAcC. Chemogenetic reversal of the PLC hypoexcitability decreased compulsivity and improved the expression of goal-directed behavior in rats exposed to stress during adolescence, suggesting a causal role for PLC hypoexcitability in this stress-induced arrested behavioral development.


Introduction:
The limbic circuitry of the mammalian brain is evolutionarily conserved to engender survival (Sokolowski and Corbin, 2012). Key subcortical structures of the limbic circuitry are the basolateral amygdala (BLA), extensively studied for its role in reward processing and assigning value to environmental stimuli (Wassum and Izquierdo, 2015), and the nucleus accumbens, a structure involved in reward and motivation and a primary target of midbrain dopamine projections (Kelley and Berridge, 2002). A later evolutionary addition to this circuit is the prefrontal cortex (PFC), which performs executive functions, such as attention and behavioral flexibility that are critical to problem solving (Yuan and Raz, 2014). The PFC sends substantial inputs to the BLA, and where it exerts top-down control over expression of emotional behavior (Sotres-Bayon and Quirk, 2010), and also to the nucleus accumbens (Phillipson and Griffiths, 1985), where it transmits information critical to goal-directed behavior (Gill et al., 2010). Mammalian brain development is characterized by protracted periods of excessive synaptogenesis under direct genetic control during the postnatal phase (Petanjek et al., 2008).
This excess of synaptic connectivity then undergoes pruning that is influenced by experience, thereby allowing the brain adaptation to the environment (Bick and Nelson, 2016). Importantly, brain development shows a hierarchical pattern with more complex and phylogenetically newer regions such as the PFC maturing later and exhibiting sensitivity to environmental influences into late adolescence (Bourgeois et al., 1994;Huttenlocher et al., 1982). The corticolimbic circuitry also undergoes functional maturation throughout early adolescence into late adulthood (Arruda-Carvalho et al., 2017;Caballero et al., 2014;Cunningham et al., 2002), which is environmentally driven by optimal experiences that promote normal development and facilitating the transition to adulthood (Arnett, 1999). Since subcortical regions mature earlier, compared to the PFC, and the corticolimbic circuitry is still 'under-construction' in adolescence, this imbalance is implicated in the expression of quintessential adolescent behaviors such as impulsivity and higher risk taking (Casey et al., 2016;Casey and Jones, 2010). Importantly, experiences during development that are considered outside the realm of "normal", particularly those involving increased stress, can also influence the pattern of corticolimbic development. Thus, stressful life-experiences, especially during the adolescent period, affect the dendritic structure of neurons in these key brain regions, as well as the functional connectivity among these areas (Leussis et al., 2008;Nooner et al., 2013;Thomason et al., 2015;Wei et al., 2018).
Here, we implement a chronic mild stress procedure during adolescence in male Wistar rats and find that this arrests their behavioral maturation such that they continue to express adolescent-like impulsive, hyperactive, and compulsive behaviors into late adulthood. This behavioral stasis is associated with the hypoexcitability of prelimbic cortex (PLC) pyramidal neurons (PN) and reduced PLC-mediated synaptic glutamatergic control of BLA and nucleus accumbens core (NAcC) neurons that lasts late into adulthood. Chemogenetic reversal of this PLC hypoexcitability decreased compulsivity and improved the expression of goal-directed behavior in rats exposed to stress during adolescence, suggesting a causal role for PLC hypoexcitability in this stressinduced arrested behavioral development.

Results:
Given the central role of the hypothalamo-pituitary-adrenal (HPA) axis in stress regulation (Smith and Vale, 2006), adaptations in corticosterone response to acute stress was assessed in control and CMUS rats. CMUS rats demonstrated lower corticosterone levels on exposure to acute stress, indicating either downregulation or reduced sensitivity to activation of the HPA axis following adolescent chronic stress exposure (Rich and Romero, 2005) (Fig1A) Aversive experiences may produce neurological changes that can be detected by unconditioned spontaneous behavior. Therefore, we evaluated the effect of adolescent CMUS by testing the performance of these rats in adulthood on the elevated plus maze (EPM)  Preclinical and human literature suggests that stressful experiences often increase impulsivity (Torregrossa et al., 2012), and adverse childhood experiences are associated with higher impulse control deficits (Barch et al., 2018). Therefore, rats subjected to CMUS in their adolescence were evaluated in the 5-choice serial reaction time task (5CSRTT) in adulthood. In this task, rats must withhold a prepotent motor response to gain access to a reward during a signaled waiting period.
While both groups demonstrated similar accuracy (Fig1G), indicating similar learning, the CMUS rats exhibited a lower proportion of omissions (Fig1H), and a significantly higher proportion of premature responses, specifically at the 5s and 7s waiting periods (Control = 23.01 ± 2.72; CMUS = 35.04 ± 3.04; Control = 32.02 ± 4.10; CMUS = 47.64 ± 3.22; % premature responses respectively; Fig1I). A similar accuracy and fewer omissions by the CMUS rats indicated that while the control rats prefer to not respond if they miss the cue light (which potentially reflects a good strategy given the time out as a punishment following an incorrect response), the CMUS rats continue to respond, indicating cognitive inflexibility, and an impulsive phenotype.
The rats were also tested on an attention deficit (AD) task in which they had to maintain attention to a cue light to express an appropriate response. The cue light signaling the correct expected response was decreased in duration over sessions increasing the level of difficulty in detecting the signal. Although both groups exhibited a reduction in the number of correct responses (Fig1J) as the signal duration was decreased, significantly fewer correct responses at the 1s signal duration were observed in the CMUS rats, consistent with impaired attention. Analyzing the incorrect responses in this task (Fig1K) confirmed that adolescent CMUS resulted in lower attention as they consistently display significantly higher incorrect responses at all signal durations (e.g. at 3s signal duration, Control = 7.95 ± 1.26, CMUS = 22.88 ± 4.28; % incorrect responses). Finally, the CMUS rats had a significantly lower proportion of omissions (Fig1L), with this difference being most evident at 0.5s signal duration. Thus, the AD task also demonstrated that, not only do lower correct responses and higher incorrect responses reflect an attention deficit in the CMUS rats, but fewer omissions observed for this group indicated that CMUS rats adopt a higher responding strategy that is non-optimal since it contributes to increased incorrect responses that are punished by a time out period.
The aforementioned behavioral tests investigated motor impulsivity, whereas the delayed discounting task (DDT) measures cognitive impulsivity where we trained rats to choose between a small immediate, or a larger delayed reward with the delay increased within session. Whereas a larger reward was preferred when there was no delay, the choice of larger reward decreased in both the groups as the delay increased (Fig2A). However, the choice of delayed large reward differed between control and CMUS groups, with the latter showing a significant reduction in this choice at a 10s delay (Control = 73.75 ± 6.28; CMUS = 51.67 ± 5.79, % choice for larger reward) ( Fig. 2A). This reduced preference for larger reward at the 10 second delay in CMUS rats indicates that they are less able to wait for the larger rewards upon introduction of the delay. However, delays longer than 10s might represent similar difficulty for both groups.
Together, the inability to inhibit a prepotent motor response in the 5CSRTT, the lower proportion of omissions during the AD paradigm, and the inability to choose a larger delayed reward in the DDT, indicate an impulsive phenotype that resulted in inefficient strategies to obtain reward in CMUS rats. However, the motivation to obtain reward was not explicitly assessed in these tasks, and this became the focus of subsequent experiments.
In a test of motivation for reward, the CMUS rats exhibited a significantly higher preference for saccharine and much lower variability on this measure than controls (Control = 82.25 ± 7.14, CMUS = 98.20 ± 0.19, % saccharine preference) (Fig2B). CMUS rats also consumed higher volumes of saccharine, compared to water (Control = 31.95 ± 10.64; CMUS = 67.68 ± 6.43, saccharine to water ratio) (Fig2C). The increased saccharine preference was also associated with an increased motivation for reward, as CMUS rats had significantly higher break points in the progressive ratio task (Control = 79.29 ± 12.48; CMUS = 132.8 ± 17.78, lever presses until rats stop responding) (Fig2D). The CMUS rats were also more persistent at reward taking when lever presses for saccharine reward were followed by mild foot shocks (Control = 11.44 ± 4.91; CMUS = 87.67 ± 33.94, lever presses) (Fig2E) indicating increased compulsive reward taking in the CMUS rats.
Finally, compulsivity is also defined by the inability to adapt responding to the changing value of reward, reflecting a shift toward habitual, rather than goal-directed behavior. Therefore, we used a behavioral task to assess responding to two distinct rewards, one that was devalued and the other non-devalued, followed by extinction. Rats demonstrating normal goal directed behavior focus their responding on a lever previously associated with the non-devalued reward (valued lever) and ignore the lever associated with the devalued reward (Balleine and Dickinson, 1998 Having identified these behavioral changes caused by stress exposure during adolescence, we next sought to identify changes in neuronal function that may underlie these effects. For these electrophysiological experiments, rats included in the CMUS group underwent the stress experience as explained in the methods, followed by evaluation in the EPM. Like our previous experiment, the CMUS rats spent significantly more time in the open arms of the EPM, compared the controls (Suppl Fig1A).
Further, when the contribution of small conductance calcium-activated potassium channel (Sk) function to after-hyperpolarization (AHP) amplitudes was measured, we found no difference in the sAHP amplitude (Control = 0.56 ± 0.14mV, CMUS = 0.84 ± 0.16mV), and a significant increase in the mAHP amplitude (Control = 4.02 ± 0.25mV, CMUS = 5.05 ± 0.35mV) in the CMUS rats PLC PNs . This increase in mAHP could act to decrease AP discharge frequency, thereby contributing to increased inhibition of these cells. We also found that the frequency of spontaneous synaptic glutamate EPSCs (sEPSCs) was lower in neurons from CMUS rats (Control = 2.98 ± 0.41Hz, CMUS = 1.77 ± 0.35Hz) (Suppl Fig1B 2,3), suggesting a decrease in excitatory synaptic drive of these cells. Together, these data identify several cellular adaptations that underlie L5 PLC neuron hypoexcitability in rats exposed to the adolescent stress procedure.
Given the central influence of the BLA on the adolescent corticolimbic circuitry (Walker et al., 2017), the effects of stress exposure during adolescence on BLA PNs was investigated (Fig3E).
Finally, since CMUS rats exhibited enhanced reward sensitivity in our behavioral tasks, and the NAcC is involved in motivation and receives glutamatergic projections from both the BLA and PLC, we investigated the electrophysiological properties of medium spiny neurons (MSN) in this brain region (Fig3I). We found that there were no differences in the properties of NAcC MSNs in between the CMUS and control groups. Thus, input resistance (Fig3J), rheobase (Fig3K), AP frequency, Sk channel function (Suppl Fig1D 4,5) and sEPSC frequency and amplitude were all unchanged by CMUS (Suppl Fig1D 2,3).
To determine the duration of the effects of stress exposure in adolescence, additional cohorts of rats underwent the CMUS procedures and the neuronal properties were evaluated at PND90. All post-synaptic differences observed between CMUS and control rats at PND50 were observed at PND90, indicating that these changes are not transient but long-term and potentially irreversible. Having established that the stressful life experiences in adolescence alters the excitability of the PLC and BLA PNs and that NAcC MSNs are unchanged, we asked whether communication among these brain regions was affected by stress. To do this, we used optogenetics (Suppl Fig3) combined with invitro electrophysiology to examine changes in excitatory synaptic pathways among these regions. There was a tendency towards smaller optically elicited EPSCs (oEPSCs) following CMUS (Fig4A). Thus, the amplitudes of oEPSCs in both groups increased with increased 473nm light intensity (input-output relationship, I-O) but at different rates and optogenetic stimulation activated significantly weaker oEPSCs in CMUS rat BLA PNs at 10 mW intensity (Control = 400.9 ± 56.56 pA, CMUS = 234.1 ± 38.66 pA) (Fig 4B,C). Additionally, a significantly smaller proportion of BLA PNs from the CMUS rats fired APs driven by light-activated Having investigated the strength of the synaptic connectivity between the three key regions of the corticolimbic circuitry, we next investigated the effects of adolescent stress on short-and longterm synaptic plasticity. All of the glutamatergic projections demonstrated similar short-term plasticity, as evaluated using paired pulse activation of oEPSCs (Fig5E-H). Therefore, we next evaluated long-term spike-timing dependent plasticity (STDP) in which synaptic glutamate release is paired with firing of the postsynaptic neuron. In BLA PNs, the STDP protocol resulted in long-term depression (LTD) of PLC to BLA glutamatergic oEPSCs in both CMUS and control groups. However, the magnitude of LTD was significantly larger in the CMUS group when measured 30 min after the STDP protocol was initiated (Two way repeated measures ANOVA [TW-RM-ANOVA], group effect, F(1,22)= 5.96, p=0.023) (Fig5I, Fig5M). Similarly, spike-timing-dependent LTD of the glutamatergic projection from PLC to NAcC MSNs was significantly larger in the CMUS group 30 min after initiating the STDP protocol Control = 72.61 ± 6.14 pA, CMUS = 53.46 ± 4.25pA) (Fig5J, Fig5N). Further, there was stronger LTD of the glutamatergic projection from the BLA to PLC PN in the CMUS group (TW-RM-ANOVA, group effect, F(1,21)= 8.36, p=0.0087) over the last 5 minutes of the experiment (Control = 81.01 ± 14.25pA, CMUS = 46.34 ± 4.5pA) (Fig5K-O). Finally, LTD in the projections from BLA to NAcC was not affected by adolescent stress (Fig5L, 5P). These data suggest that CMUS is associated with an increased susceptibility to long-term depression of PLC to BLA, BLA to PLC, and PLC to NAcC projections, but not in the BLA to NAcC projection.
The electrophysiology experiments demonstrated that stress exposure during adolescence results in the hypoexcitability of PLC PNs, combined with weaker top-down control from the PLC to the BLA and NAcC. To determine whether this weakened PLC output is involved in the altered behaviors observed after adolescent stress, we used chemogenetics to reverse the PLC hypoexcitability during behavioral tasks. Prior to surgical infusion of the AAV8-hM3D-Gq in the PLC, we confirmed the disinhibited behavior on the EPM, hyperactivity on the OFT, and motor impulsivity on the 5CSRTT in the CMUS rats (Suppl Fig4A-E). Following the CMUS procedure, rats were assigned to one of 4 groups (i) AAV-hM3D-Gq in PLC + vehicle (ii) AAV-hm3D-Gq in PLC + clozapine, (iii) Sham surgery in PLC + vehicle, (iv) sham surgery in PLC + clozapine (Suppl Fig4F-J). After training on a FR1 schedule for saccharine, rats from each group were injected with 0.1mg/kg clozapine, or an equivalent volume of vehicle, 30 minutes before the compulsivity testing and active lever presses for 0.2% saccharine during foot shock were measured. CMUS rats infused with AAV8-hM3D-Gq in the PLC and injected with clozapine had fewer active lever presses during foot shock, compared to CMUS rats receiving inactive sham injection or those receiving virus injection in PLC and vehicle injection (active lever presses per group (i) 79 ± 22.82, (ii) 19.75 ± 3.54, (iii) 95.57 ± 30.02, (iv) 137.85 ± 33.35) ( Fig 6C). Moreover, the number of active lever presses was not affected by virus infusion or clozapine injection in CMUS rats (Suppl Fig4K), suggesting no changes in general motor behavior. Together the data suggest that the PLC is recruited during conflictual situations and that stress during adolescence suppresses PLC PN excitability resulting in ineffectual compulsive behavior.
Finally, these rats were tested in a goal-directed devaluation task using a Latin square design in which they each received vehicle or clozapine injections in different sessions. Thus, each rat underwent 5-minutes of extinction after 1hr of sensory specific devaluation and was entered into the next phase of the study. They then received 0.1mg/kg of clozapine, or in a separate session, an equivalent volume vehicle injection. A two-way repeated measures ANOVA, with clozapine injection and devaluation effect as two factors, was run separately for the control rats (n=7), CMUS inactive sham rats (n=10) and CMUS Gq virus injected rats (n=11). Three control rats, 4 CMUS inactive virus surgery rats and 5 CMUS Gq virus injected rats were deleted from analysis due to either low baseline lever pressing behavior or biased consumption test ( 50.60 ± 5.21; Devalued: 51.45 ± 9.79, % of baseline lever press/min) (Fig6D3). These data suggest that stress exposure in adolescence renders PLC PNs hypoexcitable and this leads to aberrant goal directed behavior that can be reversed by chemogenetically increasing the activity of these neurons. Therefore, the data indicate that PLC neuron hypofunction is likely causal in the behavioral deficits caused by adolescent stress.

Discussion:
The theory of developmental programming posits that environmental influences during sensitive periods of development can influence the structure and function of physiological systems, and potentially contribute to persistent neural changes (Padmanabhan et al., 2016). Thus, the consequences of stressful experiences beyond the boundaries of normal expectancy can be magnified during a sensitive period like adolescence because it represents a time of critical neural development marked by high cellular plasticity (Spear, 2000) and maturation of brain circuits (Giedd, 2004).
It is well-established in both clinical (Cauffman et al., 2010;Figner et al., 2009) and preclinical studies (Andrzejewski et al., 2011) (Adriani and Laviola, 2003) that adolescents exhibit impulsive behavior. Moreover, preclinical studies show that stressful adolescent experiences are associated with impulsive behaviors even in adulthood (Baarendse et al., 2013;Torregrossa et al., 2012), and this is confirmed by clinical observations (Barch et al., 2018;Schalinski et al., 2018). Studies in rodents also show that chronic stress results in decreased attention in adulthood (Novick et al., 2013), and that perturbation of social context results in a disinhibited behavior in adulthood (Shao et al., 2009;Watt et al., 2009). In addition to these consequences of stress, adolescents are known to engage in risky behavior characterized by sensation seeking (Doremus-Fitzwater and Spear, 2016), and in the absence of intervening challenge this behavior declines during maturation (Romer et al., 2017). The increased impulsive behavior in adulthood following adolescent stress suggests that this experience might slows the normal neurobehavioral maturational process. Therefore, we suggest that the high impulsivity, hyperactivity, disinhibited behavior, and increased reward seeking we report in rats undergoing CMUS in our study is a consequence of this altered developmental process. More formally, we hypothesize that stress experienced during adolescence slows down the behavioral maturation process resulting in expression of behavior reminiscent of adolescence in adulthood ( Figure 7). This 'behavioral syndrome' consists of higher emotionality, diminished forethought and impaired executive control, all characteristics previously associated with PFC and BLA function. In our study, stress during adolescence weakened glutamatergic projections to PLC PNs and decreased intrinsic excitability of neurons in this structure. This likely contributes to impaired information integration in these neurons and reduced propagation of this activity to subcortical regions.
However, we also found that adolescent CMUS had an opposite effect on BLA PNs. Thus, BLA cells received stronger glutamatergic input, and exhibited higher intrinsic excitability, potentially contributing to increased propagation of information to downstream brain regions, such as the NAcC (Figure 7). This opposing pattern of adolescent stress effects on the PLC and BLN in CMUS rats is similar that reported in humans , where amygdala activity is increased and PFC activity decreased during emotional regulation in adults that experienced peri-adolescent stress (Kim et al., 2013;McLaughlin et al., 2015).
In contrast to the effects of adolescent stress on BLA and PLC neurons, we observed no changes in properties of NAcC MSNs. This may be due to MSNs reaching functional maturity before the CMUS procedure. It has been suggested that the NAcC acts to integrate information coming from the PFC and BLA, with the BLA transmitting emotionally salient information (Stuber et al., 2011) and the PFC input promoting executive control over inappropriate action (Kalivas et al., 2005). In our study the glutamatergic projection from PLC to NAcC was weakened, and that from BLA to NAcC strengthened following adolescent stress, suggesting that this loss of cortical control would shift more toward BLA-mediated emotional control of the NAcC output. Human brain imaging research also shows decreased functional coupling between PFC and ventral striatum during risky decision making in adolescent humans (Telzer et al., 2015), as well as stronger resting state connectivity between amygdala and accumbens in adolescents that have experienced trauma (Nooner et al., 2013). In general, our findings agree with these changes and provide putative mechanisms for changes observed in these imaging studies. We propose based on these findings that stress-induced corticolimbic imbalance tips the scale towards a greater influence of BLA glutamatergic projections on NAcC MSNs, potentially increasing the emotional bias of information impinging on the NAcC. This, together with the weakened PFC driven inhibitory control, likely contributes to the hyperactive, disinhibited, impulsive phenotype observed in our rats. Further, as the PFC is necessary for goal-directed behavior expression (Buschman and Miller, 2014), and decreased top-down control likely affects this function, perhaps not surprisingly we also demonstrate that goal-directed behavior is reduced in the CMUS rats.
It has recently been shown that model-based control, which is a reflection of goal directed behavioral control, improves with age during adolescence and this is associated with higher functional coupling between the PFC and ventral striatum in humans (Vaghi et al., 2020).
However, adolescents who do not exhibit normative maturation of PFC-ventral striatum connectivity exhibit higher compulsivity (Vaghi et al., 2020). In our study, stress exposure in adolescence resulted in impaired goal-directed behavior and increased compulsive behavior that was associated with reduced PLC-NAcC connectivity and reduced glutamatergic synaptic control.
Thus, the deceleration of corticolimbic circuit development after adolescent stress resulted in the expression of adolescent like behavior in adulthood in our CMUS rats. Other evidence supporting this idea comes from the finding that the trajectory of development of goal-directed behavioral control is steeper in early human adolescence (Vaghi et al., 2020), which is consistent with the developmental stage at which our CMUS paradigm was implemented in rats (PND21). Thus, our results support the idea that there is a critical developmental window when exposure to stress has a large impact upon adult behavior.

The idea that stress decreases top-down control of behavior is well supported in our investigation
where it is demonstrated that there is weaker glutamatergic control of BLA PNs that originates from the PLC. As the maturation of this neural circuitry is associated with increased connectivity during the transition from adolescence to adulthood (Arruda-Carvalho et al., 2017;Hare et al., 2008), and exposure to trauma in adolescent humans results in lower PFC-amygdala functional connectivity (Thomason et al., 2015), we suggest that our findings support the stress-induced deceleration of corticolimbic circuitry proposed in this model (Figure 7).
While it is often stated that the PFC is modulated by the environment, at the neural level, one of the major sources of influence over the PFC is the amygdala, which takes in environmental information and transmits it to the PFC, thus shaping its development especially during adolescence (Tottenham and Gabard-Durnam, 2017). This is supported by previous literature which shows that the connectivity between BLA, and PFC continues to develop from adolescence to adulthood (Caballero et al., 2014;Cunningham et al., 2002;Nooner et al., 2013). Thus, the decreased strength of synaptic glutamatergic connections between BLA and PFC observed in our study supports the proposal that stress during adolescence also decreases the functional maturation of this circuit (Figure 7).
Finally, our study also provides causal evidence that the hypoexcitability of PLC PNs caused by adolescent stress is involved in the expression of the compulsive and diminished goal-directed behavior. Thus, when the hypoactive PLC was chemogenetically activated, these behavioral changes were normalized. Our control experiments also confirm the specificity of this manipulation because clozapine alone did not affect appropriate baseline lever press behavior, indicating that the drug did not alter motor behavior. These data also show that the imposition of a conflictual state, such as that seen with the presence of shock during reward seeking in the compulsivity test, or in a value assessment seen after devaluation in goal-directed behavioral test, is necessary for the recruitment of the PLC and to observe these behavioral deficits. Thus, the hypoexcitability the PLC following adolescent stress was causally associated with this behavioral expression since chemogenetic activation of this brain structure increased behavioral control as shown by reduced compulsivity and improved goal-directed behavior.

Conclusion:
Stress is ubiquitous in life and short-term neurocircuit adaptations are essential for adjustment to temporary stressors. Whereas these adaptations can be beneficial in dealing with short-term stressors, they can be detrimental to developing brain circuits and behavior when the stress is sustained over a longer period of time. Here, we demonstrate that chronic and sustained mild stressful experiences in adolescence can result in the development of neurobehavioral adaptations that could increase the long-term risk of detrimental behaviors and psychiatric disorders. Therefore, understanding the neural mechanisms contributing to circuit adaptations resulting in the development of this hyperactive, impulsive, compulsive phenotype is important to developing treatments and interventions. Future experiments should focus on potential preventive and curative strategies to counter the deleterious effects of adolescent stress.   (F(1,30) 5,6,7,8,p<0.

Animals:
Male Wistar rats were bred in-house at the Center for Psychiatric Neuroscience animal facility (breeders ordered from Charles River, France). They were at PND21 and weighed 50-60 g at the beginning of the experiment and they were randomly assigned to the CMUS or the control group.
All experiments were performed in accordance with the Swiss Federal Act on Animal Protection and the Swiss Animal Ordinance and were approved by the cantonal veterinary office (authorization 3047 to Benjamin Boutrel).
All electrophysiological experiments were approved by the Institutional Care and Use Committee of the National Institute on Drug Abuse Intramural Research Program (NIDA-IRP), National Institutes of Health (NIH), and conducted in accordance with the Guide for the Care and Use of Laboratory Animals provided by the NIH and adopted by the NIDA-IRP. The Male Wistar pups for these experiments were ordered from Charles-River Lab when they were 14 days old along with a dam (6 pups per dam). After one-week (PND21) (acclimatization period), pups were weaned, and group housed (3 per cage). Then they were randomly assigned to the CMUS or the control group.
In both settings the rats were kept in reversed 12-h light/dark cycle (lights off at 0830 hours) and housed in controlled temperature and humidity conditions.

Chronic Mild Unpredictable Stress (CMUS) procedure:
Rats assigned to the CMUS groups were subjected to the CMUS procedure as mentioned in Table 1.
Control rats were handled regularly but not subjected to the CMUS procedure.
Corticosterone response to acute stress: The baseline blood sample was collected from the tail vein of the rats 1 day prior to the acute stress paradigm at 1200 hours. Next day, the rats were subjected to the elevated platform (12 X 12 X 100 cm above the ground) at 1200 hours for a period of 30 mins followed immediately by tail vein blood collection. The rats were then returned to their home cage and the final tail vein blood sample was collected 1 hour after the end of the acute stress procedure.  and locomotor activity was further assessed. The video tracking system AnyMaze 6.0 (Stoelting Europe, Dublin, Ireland) was used to track the activity of the rats.

Open Field Test:
Three open field arenas were used for this procedure. Each arena was a cylinder made of grey Plexiglas with a diameter of 70cm and a height of 30cm. The three cylinders were put on a large grey Plexiglas table. The luminosity was approximately 24 lux at the center of the arena. The three arenas were recorded simultaneously by a vertically mounted video camera linked to a computer in the same room. The video tracking system AnyMaze 6.0 (Stoelting Europe, Dublin, Ireland) was used to track the activity of the rats. Three animals from the same cage were placed each in a different arena against the wall. The animal freely explored the arena for 30 minutes. At the end of the experiment the rats were put back in their home cage and the arenas were cleaned with Deconex and dried before bringing in the next set of animals.

5-Choice serial reaction time task (5CSRTT): Impulsive action
The 5CSRTT is used to measure motor impulsivity or impulsive action in rats. Six operant chambers (30.5 x 24.1 x 29.2 cm) were used. Each chamber was individually enclosed in a wooden cubicle, which was equipped with an exhaust fan for ventilation as well as served as a white noise speaker for sound attenuation. Each operant chamber has a steel grid and a tray with bedding. On the left side of the chamber, five holes (25x25 mm) for nosepokes were present horizontally and separated by 25mm from each other. Opposite these five holess, on the right side of the chamber, was the food receptacle, located 20mm above the grid. Lights for the nosepokes, food receptacle and chamber ceiling were present. Sucrose pellets (Dustless precision pellet 45mg, rodent purified diet, Bioserv, Frenchtown, NJ) were used during the entire experiment as a reward delivered upon each correct response. The sucrose pellets were stored in a receptacle outside the cage and for each correct response, one sucrose pellet was released to the food receptacle.
Rats were maintained at 85% of their age matched body weight and were fed 1 hour after the end of the session each day. The 5CSRTT was carried out following five training steps.
Training 1: Rats underwent magazine training and learned to perform a nosepoke in the food receptacle to receive one sucrose pellet as a reward. The rats were considered to have reached the training criteria once they reached 50 pellets in 30 mins.
Training 2: Next, when a nosepoke in the food receptacle occurred, a light stimulus appeared randomly in one of the five holes and stayed lit until another nosepoke in this hole occurred, even if a nosepoke in any of the four other unlit holes was emitted. Once a nosepoke in the lit hole occurred, this light was extinguished, and the rat received a pellet in the food receptacle. Then another hole was illuminated, and the same process continued until the rat earned a maximum of 50 pellets in a 30-minutes training. The rats were trained until they reached this training criteria.
Training 3: Training 3 was similar to Training 2, except that this time, a delay of 5 seconds was introduced after the beginning of the trial and before one of the holes was illuminated. From Training 3 onwards, correct and incorrect responses were recorded. A correct response was a nosepoke in the illuminated hole and an incorrect response was a nosepoke in any of the other four unlit holes. After each correct or incorrect response, a new trial began with a nosepoke in the food receptacle, a 5 second waiting period and illumination of a random nosepoke hole. Rats had to learn this step until they reached 50 correct responses during the 30minute training.
Training 4: For Training 4, after the start of the trial by a nosepoke in the food receptacle, like training 3, a delay of 5 seconds was introduced but now it was additionally signified by an auditory tone. After the tone, one of the five holes was illuminated for a period of 2 seconds. A response during the tone and before receptacle illumination was measured as a premature response but there were no programmatic consequences. Correct and incorrect responses were recorded as well as omission responses, defined as the absence a nosepoke during the allowed 5 seconds after the cue light was illuminated. To advance to the final Training 5, the rats had to make at least usually required one day to learn this task. However, rats some required a second day of training to achieve the criterion. The rats had to reach this criterion two sessions in a row and the second session was considered as the baseline session.
The premature responses were calculated as premature responses X 100 / [correct + incorrect + omissions + premature].
After reaching the criteria in Training 5, the duration of waiting (and the tone) was increased to 7s and then 10s.
Attention Deficit (AD) task: The procedure was similar to Training 5 of the 5CSRTT, except that there was no tone during the waiting period and hence any response during this period had no programmatic consequences. Therefore, no responses were recorded as premature, even if the rats performed a nose poke in any of the five apertures before the light stimulus appeared in any holes. From session to session the time of illumination was decreased from 3s to 1s, 0.5s and finally 0.2s. The following parameters were evaluated.

Delay Discounting task:
The delayed discounting task was performed in operant chambers (32 x 24 x 25cm) (Med associates, Inc, St. Albans, VT , USA). Each chamber was individually enclosed in a wooden cubicle equipped with an exhaust fan and a white noise generator. Each operant chamber had a steel grid and a tray with bedding to collect the waste. On the right side of the chamber a food receptacle was present measuring 5 x 5 cm and 2 cm above the grid. Sucrose food pellets of 45mg were used as rewards. One retractable lever was available on either side of the food tray.
There were lights in the food receptacle, above each lever, as well as an overhead house light.
The rats were trained to lever press for one sucrose pellet (small reward) and to press the opposite lever to obtain four sucrose pellets (large reward). The small reward was always delivered immediately after the lever press. The larger reward was delivered after varying delays following the instrumental response. With this task the rat's willingness to wait for a larger reward rather than the immediate small reward is measured. In the first training stage rats were trained to lever press for a single pellet, followed by a training period during which the rats were required to wait increasing periods for small rewards. Following this, the delays and large reward were introduced.
The second step of the training was similar to the initial training except that a 40 second ITI was introduced irrespective of the animal's response. The trial started with the house light and stimulus light illuminated to signal a nose poke in the food receptacle. Then, the stimulus light was turned off and either the right or the left lever extended. Once pressed, the lever retracted, and a food pellet dropped in the food receptacle as a reward. In the last training phase, rats learned that one lever was associated with the immediate small reward (1 pellet) the other lever associated with the larger reward (4 pellets), delivered after varying delays. The training was divided into 5 blocks in which each block was composed of 10 trials at the same delay for the larger reward. Each delay was introduced in an increasing order: 0s, 10s, 20s, 40s, and 60s.
In between each block, one trial for each lever was run in which choice on this lever was rewarded without a delay. This was done to reinforce the memory of the reward values for each lever between different blocks. Each trial started with the house light and stimulus light ON indicating the appropriate lever and food receptacle for a nose poke. After the nose poke in the food receptacle, both levers extended, and the rat could choose between the left lever which always delivered the small immediate reward (1 pellet) and the right lever which gave the larger reward after varying delays. Upon pressing a lever both levers were retracted and the corresponding reward was delivered. From the nose poke at the beginning of each trial to the next trial the ITI was set to be 100 seconds. If the rats did not perform a nose poke in the 10 seconds following the beginning of the trial or if they did not press any of the two levers in the first 10 seconds after they were extended, the levers were retracted, and the rat did not receive reward. If the rat did not respond, it was counted as an omission. This training session was performed for 16 consecutive days and performance on the final four days was used as a measure of their impulsive choice.

Saccharin preference test:
Rats were single housed in standard laboratory polycarbonate cages (370x260x180mm) where they were tested for saccharin preference. On the first day, animals were given access to one bottle containing 0.2% saccharin solution and one bottle containing water. On the next day the location of the two bottles was exchanged to avoid side preference. Consumption of water and saccharin solutions was recorded daily on each day at the same hour. The following two variables were measured: Percentage choice for saccharine = Amount of saccharin solution consumed X 100 /(Total amount of liquid consumed) Saccharin to water ratio = Amount of saccharin consumed / amount of water consumed

Progressive ratio test:
Pressing the right lever once delivered 0.1mL of 0.2% saccharin solution and the rats were trained for 15 days to self-administer the saccharin solution. Following this, the rats were trained on a progressive ratio schedule in which and increasing number of active lever presses were required based on the progression sequence given by the following formula: response ratio = (5e (reward× 0.2) ) − 560. Hence, the progressive-ratio schedule followed the progression: 1, 2, 4, 6, 9, 12, 15, 20, 25, 32, 40, 50, 62, 77, 95, 118, etc. Each session lasted 90 min, or was stopped following 30 consecutive minutes of inactivity on the active lever. The maximal number of active lever presses was defined as the breakpoint, reflecting the motivation for reward. The breakpoint for each rat was averaged across 3 consecutive daily sessions.

Compulsivity test:
Each lever press delivered 0.1 mL of 0.2% saccharin solution followed by a mild electric foot shock (0.22 mA for 0.5 second) through the grid of the SA chamber when the dipper retracted.

Goal-directed behavior:
Rats in the goal-directed behavioral task were food restricted to 85% of their normal body weight. This experiment was conducted in a self-administration box in which fluid reward could be delivered in a 2-well metallic drinking cup, placed between two operant levers, that allowed for up to 2 solutions to be administered upon the pressing of the appropriate corresponding lever.

Training
Step 1: Rats were trained on an FR1 schedule where only one lever was available in the self-administration box and one lever press delivered 0.1mL of 0.2% saccharine for a maximum of 50 rewards. This was done alternately for both the levers on separate days for 2 days on each lever.

Training
Step 2: In this training, lasting 30 mins, every 3 minutes one of two levers was inserted in the chamber and a single lever press (FR1) delivered 0.1 mL of liquid reward. At the end of this 3 min period, the alternate lever was inserted for 3 min, and this cycle repeated 5 times. The reward delivered was either 10% Sucrose or 10% Maltodextrin which was randomized between left and right lever. This step was done for 3 days

Training
Step 3: This was like step 2, expect that instead of FR1, reward was delivered at random ratio 5 schedule for 3 days

Training
Step 4: This was like Step 2, expect that instead of FR1, reward was delivered at random ratio 10 schedule for 3 days. The average number of lever presses per min on each lever was calculated and averaged over the last two sessions. If a rat showed excessive side preference by ignoring one lever (<2 lever presses/min) or showed very low baseline lever pressing on either lever (<2 lever presses/min) were excluded from the analysis.

Training
Step 5: Devaluation: Rats were isolated for 1 hour in a single cage and one of the rewards (sucrose/maltodextrin) was provided ad libitum for a period of 1 hr. The total amount of fluid consumed was measured.

Training
Step 6: Extinction: Rats were placed in the self-administration boxes and given access to both levers for 5 mins with lever presses having no consequence. Lever pressing behavior was measured as number of lever presses/min.

Training
Step 7: Consumption test: Rats were reintroduced to the single housed cage and then another bottle of the reward which was not devalued in step 5 was also introduced. Rats had access to 2 bottles for 1 hr and the total fluid consumed was measured.
For experiments using a Latin squares design, rats underwent Training Steps 1-6. They were returned to Training Step 4. Training Step 5 devaluation was done with the reward not used for devaluation previously. This was followed by a return to Training Step 6 and then, Training step 7 (consumption test).