Lack of kinase-independent activity of PI3Kγ in locus coeruleus induces ADHD symptoms through increased CREB signaling

Although PI3Kγ has been extensively investigated in inflammatory and cardiovascular diseases, the exploration of its functions in the brain is just at dawning. It is known that PI3Kγ is present in neurons and that the lack of PI3Kγ in mice leads to impaired synaptic plasticity, suggestive of a role in behavioral flexibility. Several neuropsychiatric disorders, such as attention-deficit/hyperactivity disorder (ADHD), involve an impairment of behavioral flexibility. Here, we found a previously unreported expression of PI3Kγ throughout the noradrenergic neurons of the locus coeruleus (LC) in the brainstem, serving as a mechanism that regulates its activity of control on attention, locomotion and sociality. In particular, we show an unprecedented phenotype of PI3Kγ KO mice resembling ADHD symptoms. PI3Kγ KO mice exhibit deficits in the attentive and mnemonic domains, typical hyperactivity, as well as social dysfunctions. Moreover, we demonstrate that the ADHD phenotype depends on a dysregulation of CREB signaling exerted by a kinase-independent PI3Kγ-PDE4D interaction in the noradrenergic neurons of the locus coeruleus, thus uncovering new tools for mechanistic and therapeutic research in ADHD.


Introduction
Phosphoinositide 3-kinases (PI3Ks) are intracellular signaling enzymes activated by various cell-surface receptors, modulating important cellular functions, such as cell survival, proliferation, migration and adhesion (Toker & Cantley, 1997;Bondeva et al, 1998). PI3Ks are grouped into three classes on the basis of their structural and biochemical features, and among these, the class I, further divided into IA and IB, has been the most extensively studied. The class IB (PI3Kc) is composed by the single p110c catalytic subunit, which is linked to the regulation of G protein-coupled receptor signaling (Stoyanov et al, 1995;Patrucco et al, 2004) and acting with a dual mechanism involving both kinase-dependent and kinaseindependent activities (Carnevale & Lembo, 2012). It has been well demonstrated that PI3Kc is expressed in the immune and cardiovascular systems, and more recently in the brain (Toker & Cantley, 1997;Oudit et al, 2004;Viard et al, 2004;Kok et al, 2009). However, although PI3Kc has been extensively investigated in inflammatory and cardiovascular diseases, the exploration of its functions in the brain is just at dawning. On this issue, many years ago, it has been shown that PI3Kc is present in neurons and that its Akt/PKB signaling is required for the potentiation of L-type channels and for changes in neuronal excitability (Viard et al, 2004). More recently, it has been shown that the lack of PI3Kc in mice leads to an impairment in synaptic plasticity, associated with alterations suggestive of a role in mediating behavioral flexibility .
Several neuropsychiatric disorders, among which schizophrenia, attention-deficit/hyperactivity disorder (ADHD) and behavior addictions are the most common, involve an impairment of behavioral flexibility (Laughlin et al, 2011), with related symptoms principally attributed to combined deficits in alerting executive control, involving the prefrontal cortex and the striatum. The resulting poor inhibitory control seems to depend on variations in the monoaminergic system dynamics and particularly on the reciprocal interactions between catecholamines, particularly dopamine (DA) and noradrenaline (NA) (Thapar et al, 2005).
Here, we found a previously unreported expression of PI3Kc throughout the noradrenergic neurons of the locus coeruleus (LC) in the brainstem, serving as a mechanism that regulates its activity of control on attention, locomotion and sociality. Mice with genetic ablation of PI3Kc exhibit an unprecedented phenotype that mimics the core symptoms of ADHD: They display deficits in attentive domain and are hyperactive, together with an unbalanced catecholaminergic activity in the fronto-striatal areas, receiving projections from the LC. Finally, we propose a novel molecular mechanism that regulates the LC function, mediated by PI3Kc through a kinaseindependent activity and controlling PDE4D to keep homeostatic levels of cAMP. Interestingly, it is conceivable that this mechanism may be relevant for other psychiatric conditions linked to a malfunctioning of the LC, such as chronic stress, opiate addiction and depression.

Results
PI3Kc is expressed in the noradrenergic neurons of the LC We observed a previously unreported strong expression of PI3Kc throughout the locus coeruleus (LC) in the brainstem ( Fig 1A). In particular, tyrosine hydroxylase (TH)-and dopamine-b-hydroxylase (DBH)-positive cells, characterizing noradrenergic neurons, express PI3Kc ( Fig 1A). The noradrenergic LC mediates a control function on attention, sociality, arousal and activity, which are behavioral states having also a profound impact on cognition, involving the prefrontal cortex (PFC) and the striatum (STR) (Biederman & Spencer, 1999;Berridge & Waterhouse, 2003). Thus, we aimed at assessing the role of PI3Kc signaling on the behavioral states regulated by the LC and, in order to accomplish this issue, we performed a series of behavioral tests that engage this pathway (Supplementary  Table S1).

PI3Kc KO mice display poor ability in set-shifting attention
In the first place, we investigated the cognitive responses of PI3Kc KO mice in the attentional domain, through attentional set-shifting (ASS) test. ASS, adapted to animal models from the original human version, requires attending to relevant stimuli while ignoring irrelevant ones and subsequently shifting the allocation of attention, either within 'dimensions' or between 'dimensions' (Supplementary Tables S2 and S3) (Colacicco et al, 2002). Accordingly, mice from all groups required a smaller number of trials to learn simple (SD) and compound (CD) discrimination, compared to other tasks ( Fig 1B). Conversely, when both intra-and extradimensional shift tasks (respectively, IDS and EDS) were required, PI3Kc KO mice needed a significantly higher number of trials to reach the criterion, as compared to WT, suggesting poorer ability in set-shifting attention ( Fig 1B). This conclusion was also confirmed by the higher number of perseverative errors, considered as a measure of disengaging from an old rule to a new one ( Fig 1C). Moreover, PI3Kc KO required significantly more trials in all the reversal tasks (Fig 1B), thus suggesting an impaired behavioral flexibility.

PI3Kc KO mice are hyperactive
Because the LC neurons that project to the PFC may influence spontaneous physical activity, we hypothesized that the presence of PI3Kc in this brainstem region may also serve as a regulator of this behavioral state. When we tested PI3Kc KO and WT mice in the open field (OF), PI3Kc KO displayed significantly greater locomotor activity, manifesting a clear hyperactive phenotype, as shown by distance moved, crossing and vertical activity (Fig 1D-F). The increased activity was also evident in the movement trace patterns: While WT mice moved in a circular fashion following the arena diameter, PI3Kc KO tended to move with many short stops, frequently changing direction ( Fig 1G). No differences emerged in the anxiogenic thigmotaxis behavior (i.e. time spent close to the wall; Supplementary Fig S1).
Attention deficit and hyperactivity of PI3Kc KO mice are rescued by MPH These behavioral phenotypes of PI3Kc KO mice mimic the core symptoms of ADHD, one of the most prevalent human psychiatric disorders, more recently attributed also to an impairment of LC function (Biederman & Spencer, 1999;Berridge & Waterhouse, 2003). Thus, we hypothesized that methylphenidate (MPH), the drug of choice for the symptomatic treatment of ADHD (Mehta et al, 2004) that rescues as well the corresponding behavior in animal models of ADHD (Gray et al, 2007;Cao et al, 2012) would normalize the phenotype of PI3Kc KO mice. We found that MPH significantly improved the attention deficit of PI3Kc KO mice, as evaluated in the ASS test (Fig 1B and C). In particular, PI3Kc KO mice treated with MPH needed a reduced number of trials to reach the criterion

EMBO Molecular Medicine
PI3Kc/CREB signaling in the LC is crucial for ADHD Ivana D'Andrea et al ( Fig 1B), with less perseverative errors ( Fig 1C). WT mice did not show significant alterations after MPH treatment, according to previous studies using similar chronic dosage (Koda et al, 2010). Moreover, the MPH treatment suppressed the hyperactivity of PI3Kc KO, restoring locomotion to levels comparable to that of WT mice (Fig 1D-G).

Impaired memory in PI3Kc KO mice
Attention-deficit/hyperactivity disorder patients usually show cognitive deficits in different behavioral patterns. Accordingly, mouse models of ADHD also display impaired memory (Won et al, 2011). When we evaluated PI3Kc KO mice in the Morris water maze (MWM), which tests the spatial memory, we found that, during the acquisition phase, all mice significantly reduced the latency to find the platform (Fig 2A), indicating that PI3Kc did not affect the normal processes of learning. However, in the probe phase, performed the day after the last acquisition trial, PI3Kc KO spent significantly less time than WT mice in the quadrant where the platform was located before, indicating an impaired memory of platform location ( Fig 2B). With regard to the other parameters considered in this analysis, PI3Kc KO covered more distance and moved faster, as compared to WT mice ( Supplementary Fig S2A and B), further supporting their hyperactive phenotype. During both the acquisition and the probe phases, no difference was found in the anxiogenic thigmotaxis behavior among the experimental groups (Supplementary Fig S2C and D).

PI3Kc KO mice show poor social skills with conspecifics
The ADHD psychopathology also unveils poor social adjustment, mainly as a consequence of stigmatization by peers. Thus, we investigated the spontaneous social behavior (SSB) in the home cage, in mice housed each one with a familiar cage mate, during two consecutive days exploiting a procedure known to challenge the social hierarchy (D'Andrea et al, 2007). As expected, WT mice modified their behavior according to the change in social hierarchy (Fig 2C and D;Supplementary Fig S3A and B). By contrast, PI3Kc KO mice spent less time interacting with a conspecific (Fig 2C; Supplementary Fig S3A) and showed an abnormal aggressive grooming ( Fig 2D; Supplementary Fig S3B), independently from the environmental challenge. Moreover, also in this test, PI3Kc KO mice showed a significant increase in the locomotor activity, as compared to WT mice ( Supplementary Fig S4A and B). Interestingly, when mice were tested in the elevated plus maze (EPM) test to specifically assess anxiety-like behavior, no difference emerged between two groups ( Supplementary Fig S5), suggesting that the higher locomotion observed in PI3Kc KO mice does not reflex an alteration in the emotional response, but is suggestive of a hyperactive phenotype.

ADHD-like behaviors of PI3Kc KO mice depend on a kinaseindependent mechanism
During the past years, PI3Kc has been characterized as an enzyme that can act in a dual manner with kinase-dependent and kinase-independent activities (Patrucco et al, 2004;Viard et al, 2004;Perino et al, 2014). The kinase activity of PI3Kc is mainly linked to the generation of phosphoinositide (3,4,5)-trisphosphate (PIP3), whereas its kinase-independent actions principally affect cAMP levels through the modulation of specific phosphodiesterases (PDEs), the only means of degrading cAMP and maintaining its homeostasis (Patrucco et al, 2004;Damilano et al, 2011;Perino et al, 2011;Ghigo et al, 2012). In order to identify the contribution of kinase-dependent and kinase-independent activities of PI3Kc, we tested the knock-in mouse model expressing a kinase-dead form of PI3Kc (PI3Kc KD) (Supplementary Fig S6). Interestingly, PI3Kc KD exhibited no alterations of the behavioral phenotype both in the attention domain ( Supplementary Fig S6A and B) and in the locomotor activity ( Supplementary Fig S6C-E), which were clearly overlapping to that of WT mice, suggesting that the ADHD-like behaviors showed by PI3Kc KO mice should be mediated by a kinase-independent mechanism.
PI3Kc in the LC control PDE4D activity Interestingly, cAMP levels are the main regulator of the LC function (Nestler & Aghajanian, 1997;Han et al, 2006). Since the lack of PI3Kc, but not of its kinase activity, affects the brain domains organized and regulated by the noradrenergic LC, we reasoned that a mechanism affecting cAMP metabolism in the LC could be engaged in our model. Several isoforms of PDEs are known, but PDE4 is the cAMP-specific PDE that is prevalent in the brain (Kleppisch, 2009). Interestingly, PI3Kc is also a critical activator of membrane-bound PDE4 in the myocardium (Ghigo et al, 2012). Thus, we examined whether PI3Kc controls PDE4 in the LC. The major PDE4 isoenzymes of the mouse brain were immunoprecipitated from membrane and cytosolic fractions of PI3Kc KO, KD and WT LC punches, and IP pellets were assayed for cAMP PDE activity ( Fig 3A). The catalytic activity of particulate and soluble PDE4D was significantly lower in PI3Kc KO than in WT tissues, whereas no difference was found in PI3Kc KD as compared to WT ( Fig 3A). PDE4A and PDE4B activities were unchanged in PI3Kc KO ( Fig 3A). Thus, the presence of PI3Kc is required to selectively control PDE4D activity in both membrane and cytosolic compartments of the LC ( Fig 3A). The finding that PDE4D physically interacts with PI3Kc when we immunoprecipitated this latter from LC punches ( Fig 3B) further strengthens the evidence that suggest that PI3Kc signaling in the LC controls cAMP levels and homeostasis through PDE4D.

Lack of PI3Kc turns into an increased CREB signaling in the LC and catecholaminergic activity in projection areas
The LC neuronal excitability is regulated by the activation status of the transcription factor cAMP response element binding protein (CREB) and its binding to the promoter of target genes (Nestler & Aghajanian, 1997;Han et al, 2006;Cao et al, 2010). Thus, we conceived that the reduced PDE4D activity observed in PI3Kc KO could turn into dysregulated cAMP and increased CREB activation. Indeed, we found and increased pCREB/CREB in the LC of PI3Kc KO ( Fig 3C-E), suggestive of an increased activation of the noradrenergic activity of the LC, possibly determining an inhibitory effect on the dopaminergic transmission in the projecting brain areas. Accordingly, we measured NA and DA in PFC and STR, finding a significant increase in NA, with a concomitant reduction in DA ( Fig 3F). Conversely, both NA and DA remained unchanged in the LC of PI3Kc KO mice ( Fig 3F). In order to exclude the possibility that the NA/DA balance in projecting areas could be due to other areas controlling the monoaminergic system, we evaluated the presence of PI3Kc in the substantia nigra (SN) and in the ventral tegmental area (VTA), the major sources of dopaminergic neurons. As shown in Supplementary Fig S7A, we did not find expression of PI3Kc in dopaminergic neurons of both SN and VTA (Supplementary Fig S7A).

Selective inhibition of CREB in the LC rescues the ADHD-like behavior of PI3Kc KO mice
It has been reported that in vivo blockade of CREB activity in the LC prevents its neuronal hyperexcitability (Han et al, 2006). In order to better understand the role of the PI3Kc-mediated regulation of cAMP-CREB signaling in the LC, we selectively injected the LC of PI3Kc KO mice with an adeno-associated viral vector carrying dnCREB or GFP alone as a control, by bilateral stereotactic surgery ( Fig 4A; Supplementary Fig S8). Three weeks later, we obtained an effective expression of the viral vector ( Fig 4A). We found that, by antagonizing CREB in the LC, we were able to rescue the attention deficit (Fig 4B and C) and the hyperactivity (Fig 4D-F), displayed by PI3Kc KO mice, while ameliorating their poor social skills (Fig 4G and H; Supplementary Fig S9). Moreover, we found a restoration of the NA/DA unbalance in the PFC and STR in PI3Kc KO mice injected with AAVdnCREB, as compared to PI3Kc KO AAVGFP control mice (Supplementary Fig S10). A Kinase-independent activity of PI3Kc selectively regulates PDE4D activity in both membrane (mem) and cytosolic (cyt) compartments of LC (*P = 0.015 and *P = 0.046 in mem and cyt vs WT and # P = 0.0037 and # P = 0.030 in mem and cyt vs KO; ANOVA followed by Bonferroni's post hoc test, n = 5). Data are means AE SEM. B PI3Kc is constitutively associated with PDE4D, as shown by the co-IP of the two proteins in homogenates from bilateral LC punches (n = 5 for each experimental group). C, D CREB (C, red) expression and pCREB (D, red) expression in TH (yellow)-positive noradrenergic neurons of the LC in PI3Kc KO and WT mice (n = 5 for each experimental group  Constitutive activation of CREB in the LC of WT mice resembles ADHD-like phenotype In order to assess further the relevance of the cAMP-CREB signaling in the behavioral alterations related to ADHD, we overexpressed a constitutively active CREB in the LC of WT mice, by an adenoassociated viral vector carrying caCREB (or GFP as a control), as described above. Interestingly, WT mice with AAVcaCREB induced behavioral alterations overlapping those observed in PI3Kc KO mice. In particular, they needed more trials to reach the criterion ( Fig 5A) and made more perseverative errors (Fig 5B) in the ASS compared to AAVGFP control mice. Moreover, AAVcaCREB mice were hyperactive (Fig 5C-E) and displayed poor social skills (Fig 5F and G; Supplementary Fig S11). Further supporting the ADHD-like phenotype, we found that the injection of AAVcaCREB in WT mice induced an alteration of NA and DA in the PFC and STR (Supplementary Fig S12) similar to that observed in PI3Kc KO mice.
Overall, these results clearly associate the ADHD phenotype, previously characterized in PI3Kc KO mice, to a dysfunction of the cAMP-CREB pathway in LC.

Discussion
In the present work, we have identified that PI3Kc in the brainstem is expressed in the LC and that its absence reproduces typical ADHD-like behaviors. In particular, PI3Kc KO mice exhibit deficits in sustained attentional shifting capacity and are hyperactive. In addition, PI3Kc KO mice show impaired spatial memory and social dysfunctions characterized by poor skills and competences with conspecifics. This behavioral phenotype is accompanied with neurochemical alterations, reflected by an unbalanced catecholaminergic activity in the PFC and STR, the brain areas receiving projections from the LC. The diagnostic criteria in ADHD are currently based on behavioral symptoms rather than on molecular or neuroanatomical indicators. Thus, the development of mouse models recapitulating all the behavioral features of such disease is crucial to study the molecular bases underlying ADHD etiology. However, since ADHD is a heterogeneous disorder, suggested to result from combinations of genetic and environmental factors, we should be careful when translating behavioral data from animal models to humans. Murine models can mimic only certain aspect of the complex symptomatology of ADHD, but may still provide feasible hypotheses regarding the underlying causes of specific ADHD symptoms. Novel insights into disease mechanisms could be the only means for the design of targeted therapeutic treatments that indeed, so far, are mainly based on drugs that mildly relieve from some of the symptoms.
Catecholamines have an essential role in the emergence of ADHD. Despite several works focused on DA, it is known that also NE strongly affects cognitive abilities via postsynaptic a2A-adrenoreceptors (a2A-ARs) (Arnsten & Goldman-Rakic, 1985). Indeed, NA activation is known to profoundly affect the performance of attention, especially the maintenance of arousal, a cognitive function known to be poor in ADHD (Biederman & Spencer, 1999). In the present study, the molecular analysis of PI3Kc KO mice revealed a main mechanism, involving LC-NA system leading to an increased NA transmission in fronto-striatal structures, predominantly regulated by NA neurons, able to exert an inhibitory control over DA transmission. This neurobiological perspective fits very well with the current neuropsychological, genetic, imaging and pharmacological data emerging in ADHD research, providing compelling support for a noradrenergic hypothesis of ADHD and demonstrating that very high levels of catecholamine release disrupt cognitive functions of the PFC (Arnsten & Goldman-Rakic, 1985;Franowicz et al, 2002). Our findings show that this mechanism is mediated by the constitutive presence of PI3Kc in the LC where, through a kinase-independent mechanism, it controls PDE4D activity to keep homeostatic levels of cAMP. A balanced cAMP-CREB signaling pathway is well known to be needed for a proper working of the LC.
The ultimate goal of understanding the pathophysiology of the disorder is to develop therapeutic interventions, which would improve or restore physiological brain activity and, ultimately, lead to a rescue of the behavioral phenotype. Here, we have showed that a tonic upregulation of cAMP-CREB activity in the LC of PI3Kc KO mice turns into a behavioral phenotype having face validity with ADHD symptoms. Interestingly, a selective genetic intervention obtained by a stereotaxis-driven inoculation of AAV carrying dnCREB in the LC rescued such behavioral alterations in PI3Kc KO mice. Finally, the same stereotaxis manipulation, applied for inoculation of AAV overexpressing CREB in WT LC, reproduces the ADHD-like phenotype observed in PI3Kc KO mice. These findings sustain the hypothesis whereby PI3Kc controls NA transmission of the LC, by regulating the homeostasis of cAMP levels and the Overall, our results propose a novel molecular mechanism that regulates the LC function and is mediated by the constitutive presence of PI3Kc through a kinase-independent activity that controls PDE4D to keep homeostatic levels of cAMP. Besides the relevance of this pathway in ADHD and the perspective to use PI3Kc KO mice as a model of ADHD, our findings put perspective into a novel molecular pathway that balances the homeostasis of cAMP in the LC, allowing its relay activity, known to be important also in chronic perturbations, such as chronic stress, opiate addiction and depression (Nestler & Aghajanian, 1997;Mazei-Robison & Nestler, 2012).
Further supporting an essential role of PI3Kc in the phenotypes related to behavioral flexibility, it has also been reported a genetic link between PI3Kc dysfunction and mental disorders, particularly autism. Indeed, the gene encoding PI3Kc, PI3KCG, is located within the autism susceptibility locus AUTS1 on chromosome 7q22 (Kratz et al, 2002;Serajee et al, 2003). On this issue, it is important to notice that, for long time, autism and, more in general, autism spectrum disorders (ASD) have been considered somewhat completely distinct from ADHD. However, this belief has been reevaluated when many observational data brought to light a frequent co-occurrence of the two conditions, opening the possibility that some symptoms found in ASD meet also the criteria of ADHD. With our findings that strongly suggest PI3Kc as a crucial player of ADHD, we propose that the relationship between this psychiatric syndrome and ASD needs a more profound molecular and mechanistic investigation in mouse models, in order to open novel therapeutic perspectives. Vertical activity in OF: 5-min blocks 1: *P = 0.000008; 5-min blocks 2: *P = 0.016, WT AAVcaCREB vs WT AAVGFP. F, G SSB showing duration of (F) social investigation (*P = 0.000001, WT AAVcaCREB vs WT AAVGFP; # P = 0.000007 day 2 vs day 1) and duration of (G) aggressive grooming (*P < 0.000001 and *P = 0.017, WT AAVcaCREB vs WT AAVGFP on day 1 and day 2 respectively; # P = 0.000022 day 2 vs day 1). control mice were generated as previously described (Patrucco et al, 2004;Kok et al, 2009;Perino et al, 2011;Carnevale & Lembo, 2012;Ghigo et al, 2012). The main phenotype of PI3Kc KO mice, that is, ADHD, was present and comparable in both strains. Accordingly, we used PI3Kc KO mice in the C57Bl/6J background as matched experimental group of PI3Kc KD, whereas we carried out all the other experiments on 129Sv/Pas background. The animals, aged 6-8 weeks, were housed in an air-conditioned room (temperature 21 AE 1°C, relative humidity 60 AE 10%) with lights on from 08.00 to 20.00 h, in same-sex groups of 4 individuals in 42 × 17 × 14 cm Plexiglas boxes with a metal top and sawdust as bedding, and with pellet food (Enriched standard diet, Riper, Vandoies, BZ, Italy) and tap water ad libitum. Regarding the number of mice used for each experimental setting, we referred to sample sizes commonly used for the same kind of studies. All animal handling and experimental procedures were performed according to the European Community guidelines (EC Council Directive 2010/63) and the Italian legislation on animal experimentation (Decreto L.vo 116/92). The protocol was approved by the Italian Ministry of Health (Permit number 58/2012-B). All efforts were made in order to minimize suffering. In particular, the principles of Replacement, Reduction and Refinement (i.e., the 'three Rs') have been applied to all experiments.

Behavioral tests
Ten mice per genotype were evaluated for each behavioral test. All tests were carried out between 09.30 and 16.30 h, by blinded investigators. Animals were transferred to the experimental room at least 45 min before the tests in order to let them acclimatize to the test environment. After testing each animal, each apparatus was thoroughly cleaned with cotton pads wetted with 30% ethanol. Behavior was videorecorded using a digital video camera (ICD-49E, Ikegami, Japan). The behavioral scoring was carried out using a computer-based video-tracking system Ethovision XT (Noldus, Wageningen, the Netherlands) and the commercial software Observer XT (Noldus). All scores were assigned from the same observer who was unaware of animal genotype and/or treatment.
Attentional set-shifting We adopted the attentional set-shifting task, developed by Colacicco et al (2002). Apparatus and general procedure: The apparatus was an opaque PVC U-shaped box with a grid floor and a transparent Plexiglas lid (45 cm × 30 cm × 15 cm). Two identical choice compartments (15 cm × 15 cm) at one end of the apparatus could be accessed through sliding doors from a starting compartment (30 cm × 30 cm). A cylindrical food cup (40 mm diameter, 35 mm high) in each choice compartment could be baited with a small piece of cereal (30 mg; Honey Nut Loop, Kellogg's). The food was then covered with a layer of medium-scented digging (20 mm). The presence or absence of food reward in a cup was indicated by either tactile (type of digging medium) or olfactory stimuli (scent of the digging medium). The entire procedure took 5 days, a 1-day habituation period and a 4-day testing period. Habituation: On the day before testing, mice were given access to the apparatus for 30 min. Following this preliminary exposure, mice were trained to dig into food-baited bowls during a series of 9 consecutive trials.
Testing: In order to avoid potential biases, the odor presentation was randomized across cage mates. A stage was considered complete when the mouse achieved a criterion of 8 correct trials out of 10. Trials were concluded after 5 min if no choice was made, and marked as incomplete. A session would continue until the animal ceased responding. However, one hour of responding resulted in an average of approximately 30 trials per day per subject. Briefly, a trial was initiated by raising the sliding wall to give the mouse access to the two digging bowls, only one of which was baited. Mice were required to dig into a rewarded bowl in order to obtain highly palatable food pellets. Digging bowls varied across two dimensions (digging medium and scent). During simple discrimination (SD), mice had to learn to discriminate between two different odors or digging media. After this stage, mice were required to perform a compound discrimination (CD), during which the baited stimulus of the previous stage was presented together with another, newly introduced, irrelevant stimulus of the other dimension. Despite the presence of the new stimulus, the correct and incorrect exemplars remained constant. At the end of this stage, mice had to perform CD reversal (CDR) learning. For the reversal, the exemplars and the relevant dimensions were unchanged: The mouse had to learn that the previously correct stimulus was now incorrect. For both intra-dimensional shift (IDS) and extradimensional shift (EDS), new exemplars of both dimensions were used. In the IDS, the relevant dimension is the same as before, whereas in the EDS, the mouse had to shift attention to the previous irrelevant dimension. The order of discriminations was always the same, but dimensions the pairs of exemplars were equally represented within groups and counterbalanced between groups as far as possible. Errors during testing were broken down into perseverative errors and never-reinforced errors. Perseverative errors were scored when a mouse continued to use the previously relevant but currently irrelevant rule. In 5 out of every 10 consecutive trials, the mouse was allowed to respond this way. Never-reinforced errors were scored when mice made incorrect choice using a rule that has never been rewarded previously.

Open field
Mice were individually placed in the center of a cubic arena (40 × 40 × 40 cm) made of black Plexiglas and allowed to freely explore for 15 min. The open field box was ideally divided into 25 squares and ideally partitioned into a central portion (26.4 × 26.4 cm) and a peripheral one (i.e., thigmotaxis-time spent near the walls), identified as the remaining parts of the arena. When data were analyzed, each session was subdivided in 3 5-min blocks and time spent in each portion of the arena, total distance moved as well as latency, frequency and duration of locomotion (crossings of squares, vertical activity, i.e., rearing and wall rearing), and self-directed behaviors (self-grooming) were scored.

Morris water maze
The water maze apparatus consisted of a white Plexiglas circular pool 88 cm in diameter and 33 cm in height, filled with water and kept at a temperature of 26 AE 1°C. A plastic transparent platform (8 cm in diameter) was placed 0.5 cm below the water surface and 10 cm from the edge of the pool. The ability of experimental subjects to identify and reach a visible platform was tested in one visual cued version of the task which preceded the 3 days of Ivana D'Andrea et al PI3Kc/CREB signaling in the LC is crucial for ADHD EMBO Molecular Medicine acquisition phase. Each day, mice underwent 3 trials during which they were allowed to freely swim for 60 s or until they found and climbed onto the platform; each trial was spaced from the other by a 40-min inter-trial interval. In the visual phase of the task, a visual cue was always located on the platform and the platform position was changed in each trial, whereas during the acquisition phase, the platform position remained fixed for all the trials in a 'target quadrant'. Platform finding was defined as staying for at least 3 s on it. During the acquisition phase, mice that did not find the platform were trained in locating it by placing them on the platform for 10 s at the end of the trial. On the last day (probe phase), each mouse was tested for memory retention in a 60-s probe trial during which the platform was removed from the pool and the time spent in the area of the maze where the platform was located was scored.
For the acquisition phases, the variables recorded were latency to reach the platform, mean swimming speed, distance moved and thigmotaxis. For the probe phases, the variables recorded were time spent in each quadrant, mean swimming speed, distance moved and thigmotaxis.

Spontaneous social behavior
The spontaneous social behavior was recorded in the home cage according to the paradigm previously described (D'Andrea et al, 2007). Ten WT and 10 KO cages (two mice in each cage) were observed for a single 30-min session on two consecutive days. Since data collected concerning 20 mice per experimental group have been found to be dependent (significant effect of random cage factor in the mixed model ANOVA), only one mouse, randomly selected from each cage, has been considered in the analysis (n = 10). Seven days before the first day of observation, the cages were cleaned and one of the two experimental subjects was marked with a blue, scentless and nontoxic felt pen in order to discriminate the two animals during data collection. During the first day of observation, no manipulation was performed. On the second day, the cage sawdust was changed with clean sawdust just before the session of observation. This procedure is known to elicit aggressive behavior and to challenge social hierarchy. Social behavior was scored during a single 30-min session each day. The behavioral categories and elements scored for both frequency and duration were social investigation (sniffing and grooming the partner in all body regions indicative of affiliative behavior) and aggressive grooming (violent grooming of the animal on the back of the partner).

Drug administration
Mice were randomized to receive methylphenidate (3 mg/kg) or vehicle, daily administrated by gavage for 21 consecutive days.

Protein extraction
Total membranes and cytosolic fractions were prepared by homogenization of liquid nitrogen-frozen LC in 120 mmol/l NaCl, 50 mmol/l Tris-HCl (pH 8.0), protease inhibitor complete (Roche Applied Science, Indianapolis, IN) and phosphatase inhibitors (50 mmol/l sodium fluoride, 1 mmol/l sodium orthovanadate and 10 mmol/l sodium pyrophosphate). Lysates were incubated on ice for 30 min and then centrifuged at 1,500 g for 5 min at 4°C. Membrane and cytosolic fractions were then separated by centrifugation of supernatants at 38,000 rpm for 1 h at 4°C in a SW55Ti rotor (Beckman Coulter). Supernatants (cytosolic fraction) were collected, and pellets (membrane fraction) were solubilized in icecold lysis buffer, supplemented with 1% Triton X-100. Protein concentration was determined by the Bradford method, and extracts were used for immunoprecipitation and then assayed for PDE activity.

Immunoprecipitation
For immunoprecipitation assays, 150 lg of pre-cleared extracts was incubated with 20 ll of a 1:1 slurry of protein A-or G-Sepharose (Amersham Biosciences, Buckinghamshire, UK) and 1 lg of antibody/mg of protein for 2 h at 4°C. Immunocomplexes were then extensively washed with lysis buffer for PDE activity assay.

PDE assay
Phosphodiesterase activity in immunoprecipitates was measured according to the two-step method of Thompson and Appleman. In brief, immunoprecipitations were assayed in a total volume of 200 ll of reaction mixture containing 40 mmol/l Tris-HCl (pH 8.0), 1 mmol/l MgCl 2 , 1.4 mmol/l 2-mercapto-ethanol, 1 lmol/l cAMP (Sigma-Aldrich, Saint Louis, MO) and 0.1 lCi of [3H]cAMP (Amersham Bioscience, Buckinghamshire, UK) for 45 min at 33°C. In order to stop the reaction, samples were boiled at 95°C for 3 min. The PDE reaction product 5 0 -AMP was then hydrolyzed by incubation of the assay mixture with 50 lg of Crotalus atrox snake venom for 15 min at 37°C (Sigma-Aldrich, Saint Louis, MO). The resulting adenosine was separated by anion exchange chromatography using 400 ll of a 30% (w/v) suspension of Dowex AG1-X8 resin (Bio-Rad, Segrate, Milano, Italy). The amount of radiolabelled adenosine in the supernatant was quantified by scintillation counting (Ultima Gold scintillation liquid from Perkin Elmer, Waltham, MA).

Immunofluorescence
Mice were anesthetized with farmotal and transcardially perfused with saline containing 4% paraformaldehyde. Brains were removed and immersed in fixative for six hours at 4°C and then cryoprotected in 20% sucrose overnight at 4°C. Thirty-micrometer-thick floating frozen section were cut using a cryostat (Leica) and then sequentially incubated with primary antibody: rabbit anti-Creb 1:1,600 (Cell Signaling), rabbit anti p-Creb 1:500 (Cell Signaling), mouse anti-PI3Kc produced as previously described (

Western blot and immunoprecipitation
Mouse brains were rapidly removed and chilled in ice-cold buffer containing (in mM) 126 NaCl, 5 KCl, 1.25 NaH 2 PO 4 , 2 CaCl 2 , 2 MgCl 2 and 10 D-glucose, pH 7.4. The LC was obtained from coronal brain slide of $ 1 mm, cut with McIlwain tissue chopper. Bilateral punches of the LC from two slices of the same animals were homogenized by dunce homogenizer with NE-PER nuclear and cytoplasmatic extraction kit (Thermo Scientific). Protein content was determined by the BCA method. Fifty micrograms of proteins from each sample was separated by SDS-polyacrylamide gel electrophoresis using a Tris/Glycine/SDS buffer (Bio-Rad) and transferred to a PVDF membrane (Bio-Rad). Membranes were blocked using BSA (Sigma) and then incubated overnight at 4°C with primary antibody pCreb (ser133; Cell Signaling) 1:400 and secondary antibody goat anti-rabbit (Santa Cruz). For immunoprecipitation, 250 lg of total proteins extracted from bilateral LC punches was incubated with 20 ll protein A/G agarose plus (Santa Cruz), 1 lg of anti-PI3kg antibody/mg of protein overnight at 4°C. Immunocomplexes were washed with lysis buffer and resolved by SDS-PAGE followed by Western blotting using the antibody PDE4D (Acris) 1:400.
Secondary antibodies were from Santa Cruz, and signal detection was performed with ECL Kit (Amersham) and analyzed with Chemidoc XRS Imaging System (Bio-Rad).

Noradrenaline and dopamine determination
Locus coeruleus was homogenized by sonication in lysis buffer 1 mM EDTA and 0.1 N HCl. Dopamine and Noradrenaline levels were measured with a high-sensitivity ELISA kit (IBL International), following the manufacturer's instructions.

Surgical procedure
Mice were anesthetized with an intraperitoneal injection of ketamine (90 mg/kg) and xylazine (10 mg/kg), and thermoregulation was provided through a thermostat-regulated heating pad (Harvard apparatus) and monitored through a rectal thermometer. The head was shaved and cleaned with iodine before incision. After making a skin incision of about 1 cm, mice were placed on the motorized stereotaxic apparatus (Stoelting). The point 0 (bregma) and subsequently the respective coordinates for injection into the LC (x-5.52 y-0.85 z-3.50) were established as described by Franklin and Paxinos (Franklin & Paxinos, 1996). Mice were randomized in experimental groups and injected with the following viral vectors: adeno-associated virus (AAV)-green fluorescent protein (GFP) as a control, AAV-dnCREB (a dominant-negative mutant of CREB) or AAV-caCREB (constitutively active mutant of CREB) as specified in each figure. Subsequently, skull and skin were closed with dental cement and a 6.0 (Ethicon) suture. After surgery, mice were monitored in an incubator at a controlled temperature and were treated with antibiotic for any post-operative infection.

Statistical analysis
Analyses were carried out using the SPSS 21.0 software (SPSS, Chicago, IL, USA). Unless otherwise indicated, results are presented as mean AE standard error of mean (SEM). All data normally distributed for more than two groups were examined by either one-or The paper explained Problem PI3Ks are enzymes characterized by both lipid and protein kinase activity, which once activated modulate important cellular functions, affecting cell survival, proliferation, migration and adhesion. Among these, PI3Kc, which is mainly linked to regulation of G protein-coupled receptor signaling, acts with a dual mechanism involving both kinasedependent and kinase-independent activity, the first being mainly linked to regulation of intracellular levels of phosphoinositide (3,4,5)trisphosphate (PIP3) and the second affecting cAMP levels through modulation of specific PDE isoforms activity. Although the role of PI3Kc in inflammation and in cardiac function has been extensively investigated, the exploration of its CNS function is just at dawning. On this issue, it has been recently shown that the lack of PI3Kc in mice leads to an impairment in synaptic plasticity associated with alterations suggestive of a role for PI3Kc in mediating behavioral flexibility. Interestingly, among the neuropsychiatric disorders, the syndrome called ADHD (attention-deficit/hyperactivity disorder) involves difficulties in suppressing aggressive behaviors and displays many traits ascribable to behavioral inflexibility.

Results
Here, we show that the absence of PI3Kc in mice reproduces the principal behavioral features observed in patients with ADHD. In particular, PI3Kc KO mice exhibit deficits in the attentive and mnemonic domains, coupled with typical hyperactivity, as well as social dysfunctions. This behavioral phenotype is accompanied with neurochemical alterations, reflected by an unbalanced catecholaminergic activity in brain areas receiving projections from LC, where PI3Kc is constitutively expressed. More importantly, we demonstrate, for the first time, that the observed behavioral phenotype, mediated by a kinase-independent mechanism, depends on a dysregulation of CREB signaling, exerted by PI3Kc-PDE4D in LC, the largest cluster of noradrenergic neurons in the brain.

Impact
With our findings that strongly suggest PI3Kc as a crucial player of ADHD, we propose an innovative molecular mechanism regulating behavioral flexibility in mouse models, which opens novel therapeutic perspectives and appropriate diagnostic strategies.