NMDA receptor agonists reverse impaired psychomotor and cognitive functions associated with hippocampal Hbegf-deficiency in mice

Background Structural and functional changes of the hippocampus are correlated with psychiatric disorders and cognitive dysfunctions. Genetic deletion of heparin-binding epidermal growth factor-like growth factor (HB-EGF), which is predominantly expressed in cortex and hippocampus, also causes similar psychiatric and cognitive dysfunctions, accompanying down-regulated NMDA receptor signaling. However, little is known of such dysfunctions in hippocampus-specific Hbegf cKO mice. Results We successfully developed hippocampus-specific cKO mice by crossbreeding floxed Hbegf and Gng7-Cre knock-in mice, as Gng7 promoter-driven Cre is highly expressed in hippocampal neurons as well as striatal medium spiny neurons. In mice lacking hippocampus Hbegf gene, there was a decreased neurogenesis in the subgranular zone (SGZ) of the dentate gyrus as well as down-regulation of PSD-95/NMDA-receptor-NR1/NR2B subunits and related NMDA receptor signaling. Psychiatric, social-behavioral and cognitive abnormalities were also observed in hippocampal cKO mice. Interestingly, D-cycloserine and nefiracetam, positive allosteric modulators (PAMs) of NMDA receptor reversed the apparent reduction in NMDA receptor signaling and most behavioral abnormalities. Furthermore, decreased SGZ neurogenesis in hippocampal cKO mice was reversed by nefiracetam. Conclusions The present study demonstrates that PAMs of NMDA receptor have pharmacotherapeutic potentials to reverse down-regulated NMDA receptor signaling, neuro-socio-cognitive abnormalities and decreased neurogenesis in hippocampal cKO mice. Electronic supplementary material The online version of this article (doi:10.1186/s13041-015-0176-0) contains supplementary material, which is available to authorized users.


Background
According to the National Comorbidity Survey Replication, more than one-quarter of adult Americans would be diagnosed with DSM-IV mental disorders based on a fully structured diagnostic interview [1]. Structural and functional changes of the hippocampus are correlated with neuropsychiatric disorders, including major depression, which is associated with a reduced hippocampal volume and consequent functional deficits [2,3]. Hippocampal dysfunction is also closely associated with cognitive dysfunction, as seen in schizophrenia, attention-deficit hyperactivity disorder (ADHD) and Alzheimer's disease [4][5][6].
Heparin-binding EGF-like growth factor (HB-EGF) is an endogenous ligand for EGF receptors, as described for ErbB1 and ErbB4 [7,8]. Moreover, ErbB4 displays cross talk with postsynaptic density-95 (PSD-95) and NMDA receptor signaling, which are closely related to behavioral abnormalities [9,10]. Accumulating evidence has suggested that tight coupling of the ErbB4-PSD-95-NMDA receptor complex may underlie the pathophysiological molecular signature of psychiatric and cognitive disorders [11][12][13]. Conditional knockout (cKO) mice lacking Hbegf in the ventral forebrain showed some cognitive and neuropsychiatric abnormalities [12,13]. There are many reports of cKO mice showing psychological disorders and most of these cKO mice are deficient in specific molecules in both the cerebral cortex and hippocampus. However, very few descriptions of psychological and learning disorders in hippocampus-specific Hbegf cKO mice exist. The present study focused on behavioral phenotypes and therapeutic specificity in mice with a hippocampal deficiency of the Hbegf gene, as the hippocampus is known to be closely related to depression [2,3], psychiatric disorders [4], cognition [5,6] and neurogenesis [14][15][16]. Therefore, we developed hippocampus-specific cKO mice by crossbreeding floxed Hbegf [17] and Gng7-Cre knock-in mice [18], as Gng7 promoter-driven Cre is highly expressed in hippocampal neurons as well as striatal medium spiny neurons, while Hbegf is highly expressed in cerebral cortex and hippocampal neurons [19,20]. The resultant hippocampusspecific cKO mice retained the neuropsychiatric disorders as well as decreased learning potential, and also showed down-regulation of NMDA receptor signaling molecules. Additionally, these mice showed marked defects in neurogenesis. In the present study, we characterized the pathophysiological features of these hippocampus-specific Hbegf cKO mice in relation to ADHD and obsessive-compulsive disorder (OCD), and propose therapeutic avenues for inhibiting these features.

Hbegf knockout mice causes dysfunctional learning behavior and impulsivity
To investigate whether disruption of the complex infers on the learning behavior in male mice, contextual and cued fear conditioning tests were performed. Hbegf cKO mice showed decreased basal and foot shock (with tone)-induced freezing behavior in the training session in the square chamber. Data were analyzed by repeated measures one-way ANOVA, F (1, 17) = 65.81, p < 0.0001 (Fig. 5a). In the retention test after 24 h, assessing freezing behavior in the square chamber to evaluate contextual memory, there was a significant decrease in the performance of cKO mice (repeated measures one-way ANOVA: F (1, 17) = 13.90, p = 0.0017, Fig. 5b). In the experiment to evaluate cue-dependent freezing behavior, mice were put in the rectangular chamber and a cue was given at 140 s after the start of the experiment, because the difference in the initial freezing between control and cKO mice disappeared at 100-140 s. Cue-dependent freezing during the period 140-200 s after the start of the experiment was significantly lower in cKO mice than in control mice (repeated measures one-way ANOVA: F (1, 17) = 18.25, p = 0.0005, Fig. 5c). We therefore calculated activity suppression ratio [31,32]  Hbegf deficiency decreases hippocampal NMDA-R expression and plasticity. a, Altered protein expression. Immunoblotting for NMDA receptor subunits (NR1, NR2A and NR2B), PSD-95 and β-tubulin in the hippocampi of control and cKO mice. n = 8 (Control), n = 8 (cKO).b, Micrograph and schematic illustration of a hippocampal slice placed onto the multi-electrode dish. c, Representative traces of fEPSPs before and after the theta burst stimulation. The fEPSPs before and 30 min after the theta burst stimuli are indicated by black and red lines, respectively. d, Time course of normalized fEPSP amplitude after theta burst stimulation. n = 9 (Control), n = 7 (cKO). All results are presented as means ± s.e.m testing, and also cued testing. There were significant increases in the activity suppression ratio in the contextual testing (Cont: 0.2069 ± 0.02146; cKO: 0.2899 ± 0.01623; Student's t test: t = 3.121, df = 17, p = 0.0062), and in the cued testing (Cont: 0.1686 ± 0.03371; cKO: 0.3093 ± 0.02577; Student's t test: t = 3.356, df = 17, p = 0.0037) (Fig. 5d). Learning/memory in Hbegf cKO mice was also observed in the step-through type passive avoidance (PA) test (Fig. 5e, Fig. 5e).

in contextual
Next, we determined whether dysfunctional learning behavior in male Hbegf cKO mice has associations with anxiety and impulsivity, as tampering with chemical communication in the hippocampus is reported to interfere with motivational, emotional and cognitive processes in animals [33]. Locomotor activity was tested in the subjects using the square open-field (OF) test. The center area was defined as the dark field (#) in the square chamber, as depicted in Fig. 5g. Hbegf cKO mice showed increased spontaneous activity in the marginal Learning and psychomotor disability in Hbegf cKO mice. a-d, Impaired learning and memory in the context and cued fear conditioning test. Freezing behavior was observed and shown as the percentage of freezing time. a For conditioning, mice were placed in a conditioning chamber and given three sets of CS and co-terminating US. n = 9 (Control), n = 10 (cKO). b Twenty-four hours after conditioning, contextual memory was evaluated for 3 min in the chamber used for conditioning. c In the cued test, mice were placed in the novel chamber for 2 min, then exposed to a CS for another 2 min. d Activity suppression ratio in contextual and cued testing. e and f, Impaired learning and memory in the step through type passive avoidance (PA) test. e In the training session, learning and memory functions in control and cKO mice were assessed based on escape latencies in the PA test. n = 15 (Control), n = 16 (cKO). f In the retention test 24 h after training, maximum latency was set as 600 s. g-q Hyperactivity was evaluated in Open-field test.   Fig. 5o) and total center time (%) (Student's t test: Cont: 13.27 ± 1.442; cKO: 9.436 ± 0.9503; t = 2.179, df = 46, p = 0.0345, Fig. 5p), no change in moving velocity (Student's t test: Cont: 21.89 ± 0.3297; cKO: 23.14 ± 0.6017; t = 1.874, df = 46, p = 0.0673, Fig. 5q) was recorded. These data suggest that Hbegf cKO mice exhibited hyperactivity and anxiety-like behavioral phenotypes as previously reported [13]. Furthermore, cliff avoidance (CA) test was used to evaluate impulsivity [34]. Most of the control mice (n = 12) showed approximately 1 min-long exploratory behavior on the platform of the transparent plexiglas cylinder, followed by longer staying in the center for 406.0 ± 9.893 s, while cKO mice (n = 10) showed a shorter latency time to fall or jump from the platform (154.8 ± 47.77 s, Student's t test: p = 0.0001, Fig. 5r, s). There is a possibility that the shorter latency of cKO mice to fall or jump from the platform in the cliff avoidance test is attributed to their hyperactivity. However, as cKO mice always first look into the floor and fall from the platform, they unlikely fall by excessive moving. This view may be supported by the evidence that no significant change in the moving velocity of cKO mice, despite significant increases in total moving time, total distance and number of turns were observed.

NMDA mimetics reverse behavioral abnormalities in
Hbegf cKO mice From the foregoing, Hbegf cKO mice have shown clear signs of altered social, parental, anxiety-like and locomotor behaviors. To examine whether these phenotypes are related to altered HB-EGF/ErbB4-PSD-95-NMDA receptor signaling, we tested two positive allosteric modulators (PAMs) of NMDA receptor, D-cycloserine [39] and nefiracetam [40,41], since D-cycloserine has been reported to prevent relational memory deficits and suppression of LTP in the hippocampus [42] and nefiracetam administration is known to improve depressive behaviors [43]. In addition, we also examined the effects of atomoxetine (representative ADHD medicine), which exhibits dual norepinephrine transporter (NET) inhibition [44] and NMDA receptor antagonism [45]. All drugs were administered once daily for seven consecutive days, and behavioral tests were performed 24 h after the last administration (Fig. 7a).
In the PA test, all three compounds had no effect on latency in the training task (ANOVA: F (4, 63) = 1.592, p = 0.1874, Fig. 7i). At the test session, atomoxetine further shortened the latency, suggesting this action is independent of the inhibition of hyperactivity (ANOVA followed by a post hoc Tukey's test: F (4, 63) = 6.511, *p < 0.05 and ***p < 0.001 versus Control-Saline, Fig. 7j), on the other hand, D-cycloserine and nefiracetam showed an insignificant tendency to reverse the decreased latency in male cKO mice. In the MB test, both D-cycloserine and nefiracetam but not atomoxetine reversed this abnormal behavior in male Hbegf cKO mice (ANOVA followed by a post hoc Tukey's test: F (4, 74) = 29.24, ***p < 0.001 versus Control-Saline, ###p < 0.001 versus cKO-Saline, Fig. 7k, l). In the NB test, D-cycloserine and nefiracetam abolished the abnormal behavior in terms of the levels of nest rating (ANOVA followed by a post hoc Tukey's test: F (4, 42) = 19.92) and unused Nestlets (ANOVA followed by a post hoc Tukey's test: F (4, 42) = 20.09), whereas atomoxetine worsened this behavior (Fig. 7m-o). ***P < 0.01 versus Control-Saline, #p < 0.05, ##p < 0.01, and ###p < 0.001 versus cKO-Saline. From the foregoing, atomoxetine had no effect on abnormal MB activity whilst worsening nestbuilding activity. In contrast, D-cycloserine and nefiracetam significantly reversed these abnormal behaviors, thus, indicating that the pathophysiological mechanisms and pharmacological actions in the MB and NB tests appear to be independent of hyperactivity and its regulation.

Neurobiology of therapeutic actions
We have demonstrated that the alteration in social, parental, and locomotor behaviors and anxiety-like phenotypes in male Hbegf cKO mice has links with changes in neurobiological assembly of PSD-95 [27]/NMDA-receptor NR2 subunit [28] in the hippocampus. Next, we sought to examine whether the observed reversal of behavioral phenotypes in Hbegf cKO mice by D-cycloserine and nefiracetam has direct implication on the ErbB/PSD-95/NMDA receptor signaling. Indeed, hippocampus samples derived from male Hbegf cKO mice showed a marked decrease in the levels of phosphorylation of ERK 1/2 and NR1, and in the protein levels of NR1, NR2B and PSD-95 (Fig. 8a). As shown in Fig. 8b, both D-cycloserine and nefiracetam [46] significantly increased the phosphorylation level of ERK 1 , ERK 2 and NR1 over saline-treated Hbegf cKO group (ANOVA followed by a post hoc Tukey's test: p-ERK 1  cycloserine and nefiracetam mainly reversed the decrease in phosphorylation of ERK 1/2 at the level of mossy fiber of the hippocampus (Fig. 8c).
Chronic treatments with nefiracetam, but not Dcycloserine (Fig. 9a), reversed the decrease in neurogenesis in terms of the numbers of BrdU-positive cells in the SGZ of male Hbegf cKO mice (Fig. 9b, c).  e, Effects on impulsive behavior. Immobility times on the platform (s) were evaluated in the CA test. n = 23, 27, 19, 18, 16, respectively, for the groups mentioned above. f-h, Effects on dysfunction in contextual-: (f) and cue-exposed freezing: (g), and activity suppression ratio: (h) using the fear conditioning test. All data are shown as the percent of total freezing. n = 8, 9, 10, 16, 10, respectively, for the groups mentioned above. i and j, Effects on memory deficit using the step-through passive avoidance (PA) test. In the training session, each chronic drug administration did not change the PA latency: (i). Atomoxetine further decreased the PA latency in Hbegf cKO mice, while the decrease induced by D-cycloserine and nefiracetam was slight, but not significant: (j). The numbers of animals used were 10, 14, 10, 16, and 18, respectively. k and l, Effects on the obsessive behavior in Hbegf cKO mice. Both D-cycloserine and nefiracetam, but not atomoxetine reversed obsessive behavior in the MB test. Representative pictures of the results of the MB test 30 min after the start of the experiment: (k). Quantitative comparison of the numbers of buried marbles: (l). The numbers of animals used were 12, 18, 18, 15, and 16, respectively. m-o, Effects on decreased social activity. Typical pictures of differential pharmacological effects on nest building behavior: (m). Quantitative comparison on the rating scale: (n) and the levels of unused Nestlets: (o). The numbers of animals used were 8,9,8,16, and 6, respectively. Values are the means ± s.e.m respectively (ANOVA followed by a post hoc Tukey's test: F (3, 15) = 7.322, *p < 0.05 versus Control-Saline, ##p < 0.01 versus cKO-Saline, Fig. 9d, e).

Discussion
Previous study [13] has reported that there was a psychiatric disorder in conditional Hbegf KO mice using Six3-Cre mice, whose gene expression is widely distributed in retina and ventral forebrain including basal ganglia, medial amygdaloid area, hypothalamus, olfactory bulb and septum [47]. The ventral forebrain cKO mice showing Hbegf gene disruption in prefrontal cortex and hippocampus manifest hyperactivity and decreased prepulse inhibition, which were reversed by typical and atypical antipsychotics in behavioral studies [13]. In addition, the ventral forebrain cKO mice showed disorders in social activity and working memory, which is related to forebrain dysfunction. The mice also showed decreased levels of NMDA receptor-related signaling only in the prefrontal cortex, though it remained elusive whether these changes occur in the hippocampus. In the present study, we successfully generated hippocampal Hbegf cKO mice, by crossbreeding floxed Hbegf and Gng7-Cre knock-in mice. The specificity was observed in the transcription study, which shows a selective decrease of Hbegf gene expression in hippocampus, but not in olfactory bulb, striatum, cortex or cerebellum.
To characterize the phenotypes of hippocampal cKO mice we performed various cellular and molecular biological studies. First, we examined the NMDA receptor signaling, which is augmented through ErbB4 receptor, the target for HB-EGF [7]. Significant decrease was observed in the protein expression of NR1, NR2B and PSD-95, but not NR2A in the hippocampus, being in contrast to the case with ventral forebrain Hbegf cKO mice, where the decrease was observed only in NR1 and PSD-95, but not NR2A or NR2B in the prefrontal cortex [13]. Hippocampal Hbegf cKO mice also showed significant decreases in the phosphorylation levels of ERK 1/2 , NR1, which are closely associated to the NMDA signaling. Lack of Hbegf gene in hippocampal neurons induced a reduced LTP based on Schaffer collateral-CA1 pathway, which is closely related to NMDA receptor functions, in the experiment using multi-electrode arrays. However, by using this apparatus we could not detect the LTP based on perforant path-DG and mossy fiber- CA3 pathways, which are related to NMDA receptor and metabotropic glutamate receptors function, respectively [48][49][50]. Known electrophysiological studies using microelectrodes would be necessary to see LTP based on these pathways as the next subjects. It should be noted that hippocampal cKO mice showed a significant reduction in neurogenesis evaluated by BrdU-labelling only in the SGZ of DG, but not in the SVZ, RMS or OLF, compared with control mice. Furthermore DG of hippocampal cKO mice also showed a mis-positioning of immature neurons stained with DCX, being consistent with the previous report that HB-EGF plays roles in the maturation of neurons as a growth factor, and enhances neurogenesis in the SGZ [14].
In the behavioral studies, the cKO mice showed a dysfunction in learning behavior in the contextual and cued fear conditioning tests, where the hippocampus plays key roles. The Hbegf cKO mice showed significant decreases in baseline freezing without a foot shock or sound, and in freezing in the contextual task after training. An important argument arises that lower basal freezing level in hippocampal cKO mice is responsible for the decreased retention activity in fear conditioning test. However, the increase of activity suppression ratio in the contextual and cue-dependent testing was both significant after normalization. Furthermore, as these mice did not show any significant difference in the EPW test, the possibility that cKO mice are less responsive to electrical footshock would not be supported. Taking into account that passive avoidance activity in the PA test was decreased in hippocampal cKO mice, they apparently manifest the loss of learning activity. Significant decrease in the total center time may indicate that cKO mice show anxiety-like behavioral phenotype. However, the possibility cannot be excluded that hyperactivity may affect this phenotype. Impulsivity or the reduction of immobility time due to fall or jump does not seem to be simply attributed to the hyperactivity, since no significant change in the velocity in the OF test was observed and cKO mice always first look into the floor before falling from the platform. Thus, the cKO mice appear to show both hyperactivity and impulsivity, which are major symptoms observed in ADHD patients, along with cognitive dysfunction [51].
OCD, characterized by obsessive symptoms, on the other hand, was also demonstrated in the cKO mice. In the present study, the cKO mice showed excessive marble burying activity and lower nest-building ability, which represent obsessive-like behavior and social activity, respectively, for which normal hippocampal function is required [52][53][54]. As these abnormalities were not improved by atomoxetine, which inhibited hyperactivity and impulsivity, excessive marble burying activity and lower nest-building ability may not be attributed to hyperactivity. Additionally, the cKO mice also showed lower levels of maternal behaviors, such as retrieving and nursing (lactation behavior), which is defined as "pup neglect".
Tourette syndrome is a neurodevelopmental condition characterized by multiple motor and vocal tics, often accompanied by behavioral symptoms, which may manifest as complex clinical features [55]. Large clinical studies indicate that many Tourette syndrome patients (estimated to be~60 %), have a psychiatric comorbidity such as ADHD and OCD [56]. Although the assessment of vocal and motor tics is important, Hbegf cKO mice may manifest comorbid psychiatric features of Tourette syndrome. ADHD, OCD and Tourette syndrome are considered to be developmental diseases, which often appear in childhood [57,58]. As the levels of HB-EGF are highly expressed in neurons in the embryonic and neonatal rodent brain, followed by a gradual decline with age in in situ studies [20], low HB-EGF levels in an early developmental stage may play key roles as a cause of psychiatric disorders. The findings that HB-EGF plays roles in the maturation of neurons as a growth factor, and enhances neurogenesis in the SGZ [12,14], suggest possible roles for this factor in normal psychiatric development and learning activity.
Social withdrawal, cognitive and psychiatric abnormalities exhibited by Hbegf cKO mice are similar to those observed in mice lacking neurexin-1α gene (well characterized in autism and schizophrenia) [59], synaptosomalassociated protein 25 (SNAP-25) gene (Coloboma mouse mutant) [60] and dopamine active transporter (DAT) KO mice (animal model of ADHD) [61]. This study provided evidence suggesting that psychiatric abnormalities in Hbegf cKO mice may be related to the alteration in HB-EGF/ErbB4-PSD-95-NMDA receptor signaling, since HB-EGF is known to activate NMDA receptors [62] concentrated and anchored at the synapse via PSD-95 scaffold [28] and PAMs of NMDA receptor such as D-cycloserine [39] and nefiracetam [40,41] aptly reversed NMDA receptor signaling modulation and all behavioral abnormalities tested except in the OF test. Specificity was observed in the finding that representative ADHD medicine atomoxetine inhibited only hyperactivity and impulsivity, but not learning/memory or social behavioral abnormalities. These pharmacotherapeutic differences support the view that learning/memory, impulsivity or social abnormalities are not simply attributed to the hyperactivity of cKO mice. Both PAMs significantly reversed NMDA receptor signaling, such as decreased phosphorylation of ERK 1/2 , NR1 and down-regulation of NR1, NR2B and PSD-95, though it remains elusive whether these pharmacological actions underlie the therapeutic actions against behavioral abnormalities. As PAMs markedly reversed the decreased ERK 1/ 2 phosphorylation in the mossy fiber of cKO mice, further pharmacological studies on LTP and NMDA receptor signaling using sub regions of hippocampus should be necessary as the next subjects.
D-cycloserine is known to bind the glycine-binding site on the NMDA receptor NR1 subunit [63], and to play a partial agonist role [39,42]. The choice of Dcycloserine was informed by previous findings that this compound showed preclinical benefits in Neuroligin 1 KO mice (human autism and mental retardation) [64] and Tourette syndrome with accompanying ADHD and/or OCD [65]. The present study revealed that D-cycloserine significantly inhibited the ADHD-and OCD-like behavioral phenotypes as well as the learning disabilities in Hbegf cKO mice. Similar results were observed with nefiracetam, another PAM of NMDA receptors via the glycinebinding site [40,41,46]. Recently we have clarified the mode of nefiracetam binding to NR1 subunit by a molecular dynamics simulation study, in which nefiracetam first binds to a novel site of NR1 and causes a release of glycine, followed by the direct binding (replacement) to the glycine pocket [41].
The reduction in SGZ neurogenesis was selectively reversed by nefiracetam, but not by D-cycloserine. The mechanisms underlying the nefiracetam-selective reversal may be in part attributed to the differential modes of action to NR1, as above-mentioned. However, the possibility of non-NMDA receptor-based mechanisms cannot be excluded, and there is a report that nefiracetam increases the number of immature neurons expressing polysialylated neural cell adhesion molecule (PSA-NCAM) in the DG, possibly through an action on nicotinic acetylcholine receptors [66]. Nevertheless, nefiracetam significantly, and D-cycloserine partially reversed the abnormal expression of DCX-positive neurons in the SGZ and DG of Hbegf cKO mice, where some DCX-positive neurons were observed at the inner part of DG, suggesting both PAMs reverse the abnormal maturation of newly born neurons. The further studies are necessary whether these pharmacological actions are mediated though NMDA receptor signaling or whether pharmacologically improved neurogenesis and neuronal maturation are involved in the pharmacotherapeutic actions against behavioral abnormalities.

Conclusions
In the present study, we demonstrate that the hippocampal Hbegf cKO mouse is a unique model manifesting experimental ADHD/OCD-like behavioral phenotypes. Additionally, PAM of NMDA receptor would be one of promising candidates to cure related pathologies.
It would be highly desirable that Gng7 wt/wt ; Hbegf flox/ flox , Gng7 wt/cre ; Hbegf wt/wt , and Gng7 wt/wt ; Hbegf wt/wt are used as the littermate control of Gng7 wt/cre ; Hbegf flox/flox (Hbegf cKO) mice. Due to the practical difficulty of use of all these littermate controls, male Gng7 wt/cre mice were used as controls for male Hbegf cKO mice to rule out a possible influence of genetic background, since former mice showed no significant change in psychomotor activity, compared with wild-type mice [26].
All mice were used at 10-16 weeks of age and kept in a room with a temperature of 21 ± 2°C with ad libitum access to a standard laboratory diet and tap water in standard animal cages in a 12-h light/dark cycle (lights on at 08:00).

Histology and immunohistochemistry
Under deep pentobarbital anaesthesia (50 mg/kg, i.p.), male mice were perfused transcardially with 20 ml of potassium-free PBS (K + -free PBS, pH 7.4), followed by 50 ml of a 4 % paraformaldehyde (PFA) solution in potassium-free PBS. Brains were isolated, post-fixed for 3 h, and cryoprotected overnight in a 25 % sucrose solution. Tissues were fast frozen in cryoembedding compound in a mixture of ethanol and dry ice and stored at −80°C until use. Brain tissue was cut on a cryostat at a thickness of 30 μm, and then brain sections were thawed into the 0.1 % sodium azide stock solution at 4°C until use. β-Galactosidase (lacZ) staining was conducted overnight at 37°C in PBS containing 5 mM potassium hexacyanoferrate (III), 5 mM potassium hexacyanoferrate (II), 2 mM magnesium chloride and 1 mg/ml 5-bromo-4-chloro-3-indolyl-β-D-galactopyranoside. Immunohistochemistry was conducted using brain sections that were washed with TBST (0.1 % Triton X-100 in TBS). The sections were incubated with blocking buffer containing 3 % bovine serum albumin in TBST and subsequently reacted with rabbit anti-phospho-ERK 1/2 (1:300) overnight at 4°C. After washing, the sections were incubated with second antibody, Alexa 488-conjugated anti-rabbit IgG (1:300; Invitrogen, CA, USA) for 2 h at room temperature. After further washing, the sections were mounted with Pristine Mount, and analysed using a structured illumination microscopy (BZ-X700, Keyence, Osaka, Japan).

Design of behavioral experiments
In all experiments, the order of behavioral tests was designed according to the expected degree of stress induced. To avoid cross-test interaction, each test was performed after a habituation period of 1 week. Only one behavioral test was performed each day. Furthermore, the same male mice were used for the following tests; assessment of general health, OF test, rotarod test, stationary horizontal thin-rod test, CA test, MB test, NB test, contextual and cued fear conditioning test, and PA test. In the initial behavioral battery experiments, we started with n = 6 per each group and performed the statistical comparison between control and Hbegf cKO mice. To clarify the reliable significance of the difference, we performed additional experiments in some tests. In this case we confirmed that there is no significant difference in respective data between the initial and additional mice, which have been used over a week after arrival. A different group of female mice was used for the tests for parity and maternal behavior. Also, independent groups of male mice were used for the experiments to determine the effects of drugs.

Assessment of general health
Body weight and rectal temperature (as body temperature) of 10-16-week-old male control and Hbegf cKO mice were measured. In the wire hang test, the male mouse was first placed on a 1 × 1 cm wire mesh. The wire mesh was then inverted and waved gently, so that the mouse gripped the wire. Latency to fall was recorded, with a 180 s cutoff time. The nociception threshold was evaluated by the latency of paw withdrawal upon a thermal stimulus, using a thermal stimulator (IITC Life Science, CA, USA), as described previously [68]. A cutoff time of 20 s was set to avoid tissue damage. The electrical stimulation-induced paw withdrawal (EPW) test was performed as described previously [68,69]. Briefly, electrodes were fastened to the plantar surfaces and insteps of male mice. Transcutaneous nerve stimuli with each of the three sine-wave pulses (5, 250, and 2000 Hz) were applied using a Neurometer® CPT® system (Neurotron, MD, USA). The minimum intensity (μA) at which each mouse withdrew its paw was defined as the current stimulus threshold.

Rotarod test
The rotarod test was performed as previously reported [26]. The rotarod apparatus (MK0610A, Muromachi KIKAI) was used to measure fore-and hind-limb motor coordination. During the training period, each male mouse was placed on the rotarod revolving at a constant speed (20 rpm) for a maximum of 60 s, and the latency to fall off the rotarod within this time period was recorded. Mice were trained for 3 consecutive days, receiving four trials per day with a 1 h inter-trial interval.
During the test period on day 4, mice were tested in four consecutive 60 s trials at constant speeds of 10, 20, 30 and 40 rpm. The mean latency to fall from the rotarod (for the four trials at each speed level) was recorded.

Accelerating rotarod test
The accelerating rotarod test was used to measure motor learning as previously reported [70]. During the training period, each male mouse was placed on the rotarod at a constant speed (4.5 rpm), which was accelerated to 45 rpm. The maximum observation time was 5 min. Mice were trained for 3 consecutive days, receiving four trials per day with a 1 h inter-trial interval. During the test period on day 4, mice were tested for seven consecutive 60 s trials at constant speeds of 5, 12, 18, 25, 31, 40 and 45 rpm. During the constant speed rotarod experiment, mice were given three trials per day. The mean latency to fall from the rotarod (for the three trials at each speed level) was recorded.

Stationary horizontal thin-rod test
The stationary horizontal thin-rod test was used to measure motor coordination and balance as previously reported [71]. Each male mouse was placed on the thinrod (15 mm in diameter, 50 cm long and placed 40 cm high to discourage jumping). The thin-rod was flanked at both ends by a large acryl wall to prevent falling at the end of the thin-rod. The maximum observation time was 1 min. Mice were trained for three consecutive days, receiving four trials per day with a 1 h inter-trial interval. The mean latency to fall from the thin-rod was recorded.
Step through-type passive avoidance (PA) test The PA test was performed as described previously [72]. Briefly, the apparatus consisted of an illuminated compartment (light room) and a dark one (dark room) connected by a guillotine door. In the training trial, the male mouse was first placed into the light room and after 10 s, the partitioning door was opened. When the mouse entered the dark room, a 2 s footshock at 0.8 mA was given through the grid floor. Twenty-four hours after the training trial, the mouse was tested for retention time by being placed mice into the light room and the step through latency was measured (maximum latency, 600 s).

Contextual and cued fear conditioning test
The contextual and cued fear conditioning test was performed in a soundproof behavioral apparatus (Muromachi Kikai, Japan). Male mice were placed in a conditioning chamber, with a plexiglas front, gray side-and back-walls and electrical grid floors, and allowed to move freely for 2 min. They then received three pairings of a tone (20 s, 4000 Hz, 80 dB) as a conditioned stimulus (CS), and a coterminating footshock (2 s, 0.5 mA) as the unconditioned stimulus (US), with an inter-stimulus interval of 2 min. After the last footshock, the mice remained in the conditioning chamber for 2 min. Twenty-four hours after the conditioning, mice were tested for contextual memory in the same conditioning chamber for 3 min. The cued testing was performed in a triangular chamber, with a plexiglas front, black and white stripe side-and back-walls and flat floors covered by paper towels. Mice were allowed to move freely for 2 min, and then received a tone (120 s, 4000 Hz, 80 dB). All experiments were observed using a video tracking system (Muromachi Kikai, Japan) to record the moving and freezing behavior of the mice. Freezing was defined as complete resting, when a mouse does not move for more than 1 s [73]. The loss of retention activity was evaluated by the activity suppression ratio in the contextual testing using the activity levels during the first 2 min conditioning before a shock delivery and during the first 2 min contextual testing, according to the formula: activity test /(activity train + activity test ), as previously reported [31,32]. Similarly, in the cue-dependent testing, the suppression ratio was calculated using the activity levels during the first 3 min cued testing before cue-exposure and during the last 3 min after the exposure.

Open field (OF) test
In the OF test, locomotor activity was measured for 60 min in a square chamber (70 × 70 × 30 cm) with a video tracking system (Muromachi Kikai, Japan). Each mouse was placed in the corner of a square chamber. The test was illuminated at 100 lux. Moving time, distance travelled, and time spent in the center area were recorded.

Cliff avoidance (CA) test
In the CA test, cliff avoidance and jumping were measured for 7 min using a round platform (a transparent plexiglas cylinder with a diameter of 13 cm and a height of 20 cm). Each male mouse was placed on the platform, and the jumping behavior was defined as positive performing, while falling from the top was defined as passive performing [34].

Marble burying (MB) test
In the MB test, each male mouse was placed in an open plexiglas cage (28 × 45 × 20 cm) with a 5 cm layer of fine bedding material and 25 equally spaced glass marbles (17 mm in diameter). The number of buried marbles was counted after 30 min. When 2/3 of a marble was in the fine bedding material, it was defined as buried [52].

Nest building (NB) test
In the NB test, each male mouse was placed into a clean cage containing fine wood-chip bedding material and single cotton batting Nestlets® nesting material (2.5 g of 5 cm squares, Ancare, NY, USA) approximately 1 h before the dark phase. The next morning, unused Nestlet pieces were weighed and nests were evaluated on a rating scale of 1-5 [38]. The threshold of unused Nestlet pieces was more than 0.1 g. The rating scale was defined as follows: 1, Nestlet not noticeably touched; 2, Nestlet partially shredded (50-90 % remaining intact); 3, Nestlet mostly shredded but no identified nest (<50 % of the Nestlet remaining intact); 4, Nestlet mostly shredded and a flat nest identified (>90 % of the Nestlet was shredded and used for the nest); and 5: almost all of the Nestlet was shredded and the nest was the shape of a crater with walls higher than mouse body height.

Parity and maternal behavior
Female Hbegf cKO mouse was mated with a male Hbegf flox/flox mouse. To rule out a possible influence of pups' genetic background for maternal behavior, female Hbegf flox/flox mouse was bred with male Hbegf cKO mouse as control mating. Pregnant mice were individually transferred to a new cage in the same environment and parturition was checked daily in the morning (09:00) and evening (19:00). The day when parturition was first detected was defined as postnatal day 0 (P0). Maternal behaviors were measured on P0 to P7. For the separation of a dam from her pups, the dam was put in one corner, while her pups were placed in the center of the same cage. After separation, maternal behavior, such as retrieving and suckling, were observed for 1 and 2 h. For this experiment, 10-12 weeks-old dams were used.

BrdU-and doublecortin-staining
BrdU (50 mg/kg in sterile saline) was injected intraperitoneally [15]. Injections were made once a day for three consecutive days. For brain tissue preparation, mice were deeply anaesthetized with sodium pentobarbital (50 mg/ kg, i.p.). Brains were quickly isolated, and washed with saline and 4 % PFA. Brains were post-fixed in 4 % PFA for 24 h and finally transferred to 25 % sucrose solution (in 0.1 M K + -free PBS) overnight for cryoprotection. Following freezing in cryoembedding compound, brain sections were prepared at 30 μm thickness and were stored in cryoprotectant solution (25 % ethylene glycol, 25 % glycerin, 0.05 M phosphate buffer) at −20°C until processing for immunohistochemistry [16]. For the detection of BrdU-positive cells in brain sections, DNA denaturation was performed prior to incubation with anti-BrdU antibody as follows: incubation in 50 % formamide/2× SSC (0.3 M NaCl, 0.03 M sodium citrate) for 2 h at 65°C, rinsing in 2× SSC for 5 min, incubation in 2 N HCl for 30 min at 37°C, and rinsing in 0.1 M boric acid (pH 8.5) for 10 min at room temperature. Floating brain sections were incubated in 0.3 % H 2 O 2 solution for 10 min, and rinsed with 0.1 % Triton X-100 in TBS (TBST). Brain sections were incubated with blocking buffer containing 3 % BSA and goat antiserum to mouse IgG (1:50; Cappel Laboratories, PA, USA) in TBST for 30 min. After thorough washing, the sections were reacted with mouse anti-BrdU monoclonal antibody (1:500: Roche, Basel, Switzerland), overnight at 4°C. After further washing, the sections were incubated with secondary antibody, biotin-conjugated secondary antibody (1:500; goat polyclonal anti-mouse IgG, Invitrogen) for 2 h, and subsequently treated with avidin-biotin peroxidase solution (ABC kit, Vectastain Vector, USA) for 1 h at room temperature. BrdU-positive cells were visualized by incubation with a solution containing 0.02 % 3,3′-diaminobenzidine tetrahydrochloride (DAB) (Dojindo, Kumamoto, Japan) and 0.005 % H 2 O 2 (Wako, Japan) in 0.05 M Tris-HCl buffer (pH 7.6), 1 % cobalt chloride (CoCl 2 ) and 1 % nickel sulphate (NiSO 4 ) solution until brown reaction products appeared. The brain sections were dehydrated through a series of ethanol solutions and xylene, and cover-slipped with Permount® (Fisher Scientific, Waltham, MA, USA).
BrdU-immunopositive cells in the olfactory bulb (OLF), rostral migratory stream (RMS), subventricular zone (SVZ), and subgranular zone (SGZ) were counted in every eighth section in a series of 30 μm brain sections using Image J software. The total number of BrdU-positive cells was estimated by multiplying the result by eight.
For the analysis of immature neuronal positioning, brain sections were incubated with guinea pig anti-doublecortin (DCX) polyclonal antibody (1:1000; Millipore, Billerica, MA, USA) overnight at 4°C. After washing, the brain sections were labelled with second antibody, Alexa Fluor 594-conjugated anti-guinea pig IgG (1:1000), for 2 h at room temperature. Localization patterns of cells were evaluated using a previously described method with slight modifications [74]. Images of DCX-positive neurons with Hoechst 33342 staining were used to determine cell localization.