Cortical Abnormalities and Non-Spatial Learning Deficits in a Mouse Model of CranioFrontoNasal Syndrome

Eph receptors and their ephrin ligands play critical roles in the development of the nervous system, however, less is known about their functions in the adult brain. Here, we investigated the function of ephrinB1, an ephrinB family member that is mutated in CranioFrontoNasal Syndrome. We show that ephrinB1 deficient mice (EfnB1Y/−) demonstrate spared spatial learning and memory but exhibit exclusive impairment in non-spatial learning and memory tasks. We established that ephrinB1 does not control learning and memory through direct modulation of synaptic plasticity in adults, since it is not expressed in the adult brain. Rather we show that the cortex of EfnB1Y/− mice displayed supernumerary neurons, with a particular increase in calretinin-positive interneurons. Further, the increased neuron number in EfnB1Y/− mutants correlated with shorter dendritic arborization and decreased spine densities of cortical pyramidal neurons. Our findings indicate that ephrinB1 plays an important role in cortical maturation and that its loss has deleterious consequences on selective cognitive functions in the adult.


Introduction
Eph receptor/ephrin signaling is a cell-cell communication pathway that plays a crucial role in embryonic development [1], yet, increasing evidence shows a role for Eph/ephrin signaling in adult physiology [2]. There are two classes of Eph receptors and ephrins (A and B) and in the adult brain, both classes have been shown to control synaptic development and plasticity in the hippocampus [3,4,5,6,7]. To date, the EphB receptors and their ligands the ephrinBs have been implicated in hippocampal LTP and in spatial learning paradigms. For example, double and triple mutant mice lacking EphB1, EphB2, and EphB3 showed significant deficits in dendritic spine formation and clustering of AMPA and NMDA receptors [8]. EphrinB3 mutant mice have impairments in hippocampal mossy fibre LTP and in hippocampal-based learning tasks [9], while ephrinB2 conditional knockout mice showed severe deficits in both LTP and in long-term depression [10,11].
EphrinB1, which is the third member of the ephrinB family is expressed predominantly in apical progenitors in the developing cortex [12] where it has been shown to control the switch between progenitor maintenance and neuronal differentiation by regulating levels of miR-124, a pro-neuronal miRNA [13,14]. More recently we showed that ephrinB1 is required in cortical progenitors to maintain their apical adhesion thereby ensuring the structural integrity of the developing cortex [15]. Interestingly, ephrinB1 is encoded by EfnB1, an X-linked gene associated with the human CranioFrontoNasal Syndrome (CFNS) [16,17]. This syndrome is characterized by severe hypertelorism, frontonasal dysplasia, craniosynostosis and developmental delays [18,19]. In contrast to ephrinB2 and ephrinB3, very little is known for the role of ephrinB1 in the adult brain. Previous work in cell culture has shown that ephrinB1 increases the number of dendritic spines on hippocampal rat neurons [20], regulates EphB-dependent presynaptic development on cortical rat neurons [21], and undergoes time-dependent up-regulation in lesion induced plasticity in the adult mouse hippocampus [22]. Collectively these findings suggest that ephrinB1 plays a role in synaptic development and plasticity, yet the role of ephrinB1 in cognitive function remains unknown.
Here we show that EfnB1 deficient mice (EfnB1 Y/2 ) display normal learning and memory in spatial learning paradigms. No deficits in locomotion or anxiety were detected in EfnB1 Y/2 mice; however, they exhibited a specific impairment in non-spatial learning and memory tasks. Given that we observed no impairment in spatial learning, which strongly implicates the hippocampus (reviewed in [23,24]) and rather observed learning impairments that are strongly associated with cortical functioning [25,26,27,28], emphasis was placed on analyzing the cortex, specifically the perirhinal cortex, in these mutants. Surprisingly, we established that ephrinB1 does not control learning and memory through modulation of synaptic plasticity in adults, since it is not expressed in the adult brain. Further scrutiny of the adult cortex of EfnB1 Y/2 mice showed supernumerary neurons, with increased interneuron number, and decreased dendritic complexity in cortical pyramidal neurons as compared to wild-type (EfnB1 Y/+ ) littermates. Altogether, our findings indicate that the absence of ephrinB1 results in increased neuron numbers with changes in dendritic morphology in the postnatal cortex that correlate with a selective learning and memory deficit.

Behavioural Testing
All experiments were performed during the light phase. Threeto 4-month old mice (n = 12 per genotype) were familiarized to the experimenter and mice were subject to two behavioral tasks; one non-stressing test (object recognition or object location) and one aversive task (Morris water maze, passive avoidance or fear conditioning). The order of experiments was pseudo randomized to avoid any influence of test order. Locomotor activity was analyzed during the 10 min familiarization phase of the object recognition or object location tasks. Anxiety was evaluated based on percent time spent in the open arms of the elevated plus maze.
Morris water maze. Spatial memory testing was conducted as described [30,31]. Briefly, mice were introduced to a circular pool (110 cm in diameter) filled with water made opaque. Subjects were trained to locate the hidden platform, which was submerged 0.5 cm below the water. One mass-training procedure was performed to unambiguously separate acquisition from consolidation processes. With this massed procedure, we have the same hippocampal involvement during acquisition and consolidation as found in the distributed procedure [32,33]. The procedure included one training session composed of 4 blocks, each consisting of three consecutive trials. The phase between consecutive blocks was 15 to 20 min long, during which the mouse was returned to its home cage. Twenty-four hours post acquisition, memory was assessed during a single one minute probe test in the absence of the platform.
Object recognition. The procedure consisted of three different phases as described [34]. Briefly, a familiarization phase in which each mouse was placed in the empty square open-field for 10 min. A sample phase, 24 h later, in which two identical metallic objects were placed in the middle of the open-field. A test phase, 24 h later whereby mice were reintroduced into the arena and exposed to two objects, a familiar object and a novel object, to test recognition memory. The percent time spent exploring the novel object was calculated as a preference index to measure novel object recognition Performance in this non-spatial task, in contrast to the object location task, is not affected by hippocampal lesions [35,36] unless conducted in a complex spatial environment [36]. To avoid this bias, we used the original procedure described by Ennaceur and Delacour [35], which we have previously demonstrated does not require an intact hippocampus [37].
Object location. The same open-field with the same environment was used as that of the object recognition task. A similar procedure was employed except that one of the two identical objects was moved to a novel location [34]. The percent time spent exploring the displaced object was calculated as a preference index to measure spatial memory.
Fear conditioning. Mice were placed in the fear-conditioning apparatus chamber for 5 min and 30 sec. After a 2-min exploration period, a sound (CS) was emitted for 30 sec and a footshock of 0.7 mA (US) was superposed to the tone during the last 2 sec. After 2 min, the paired CS-US was repeated. Twenty four hours after conditioning, context-dependent freezing rate was measured in the conditioning chamber for 4 min. Three hours later, context-independent (tone dependent) freezing rate was measured. Specifically, mice were placed in a modified context; 2 min after their introduction in the modified chamber, mice received a 2-min tone presentation. Freezing rates to CS or US were calculated for EfnB1 Y/+ and EfnB1 Y/2 mice as described [38].
Passive avoidance. The apparatus consisted of a rectangular box divided into an illuminated safe compartment and a dark shock compartment. Subjects were individually placed into the illuminated compartment and when subjects entered completely inside the dark box, a foot shock (280 mA for 2 sec) was administered. Subsequently, the mouse was placed in its home cage. Retention was measured 24 h later by placing the mouse in the light compartment and measuring the latency to enter the dark box.

Brain Sample Preparations
Four to five month old EfnB1 Y/+ and EfnB1 Y/2 mice month old mice, following behavioural testing, were anesthetized with an injection cocktail of 3:3:1 ketamine (100 mg/ml)/xylazine (20 mg/ml)/acepromazine (10 mg/ml) at a dose of 0.01 ml injection cocktail/g body weight and intra-cardially perfused with 0.9% saline followed by 4% formaldehyde in phosphate buffered saline (PBS). The dissected brains post-fixed in 4% PFA at 4uC. All sample brains were removed from PFA and either equilibrated in 70% ethanol and embedded in paraffin or equilibrated in sucrose and embedded in OCT blocks for cryostat sectioning. Alternatively, EfnB1 Y/+ and EfnB1 Y/2 mice, at specified ages were sacrificed by cervical dislocation and quickly weighed. The brains were removed, weighed and harvested in RNA LaterH (Ambion) until use. Coronal sections (7 mM) were obtained and placed on Superfrost microscope slides (Fisher Scientific) and stored at room temperature until use.

Cresyl Violet Staining
Multiple serial sections were Nissl-stained to provide a qualitative view of the cortical and hippocampal structures. The slides were immersed in 25uC cresyl violet for 10 min, washed vigorously in rapid exchanges of distilled water to remove the excess cresyl violet, dehydrated in graded ethanol, cleared in xylenes, and coverslipped. Images were acquired using a Nikon Eclipse 80i.

Neuronal Cell Counts
The mouse brain atlas was used to identify different regions of the cortex, hippocampus, amydgala and striatum. Serial coronal sections (7 mm) were obtained. NeuN staining was performed on every 20 th mm -section from postnatal and adult brains (n = 3/ genotype/age). For quantification of cells, images were taken with a 406 objective using a confocal microscope (Leica SP2 or SP5).
Cell counts included all fully visible cells within the 100 mm 2 counting frame using the cell counter plugin for ImageJ. For each cortical area the number of NeuN+ cells was quantified in 3nonoverlapping 100 mm 2 counting boxes starting at the pial surface. Emphasis was placed on sections containing perihinal cortex. For the hippocampal CA1 and CA3 a 100 mm 2 counting box was placed over each region and NeuN-positive cells were counted manually using the cell counter plug-in from ImageJ. For the amygda two non-overlapping 100 mm 2 counting boxes were placed in the amygdala region at the level of 2.34 mm to 1.62 mm posterior to the interaural plane. For the striatum, four coronal sections at the level of 3.94 mm posterior to the interaural plane were analyzed. Three non-overlapping 100 mm 2 counting frames were placed vertically starting at the corpus callosum. Interneuron staining was performed on every 10 th section between the interaural planes 3.10 mm to 1.10 mm (n = 6/genotype). A 100 mm 2 counting box was placed across three different counting regions of interest in the frontal, somatosensory or perirhinal cortices. Triplicate cell counts were performed in each region. Furthermore, the cortical thickness was measured as the shortest distance between the grey/white matter boundary and pial surface. Triplicate measurements were performed on 10 sections per sample; n = 3 per genotype.

Golgi Cox Impregnation
Brain tissues were removed and processed by Golgi-Cox staining, using procedures described previously [39]. Briefly, brains were first stored in the dark for 14 days in Golgi-Cox solution followed by 3 days in 30% sucrose. The brains were sectioned 200-mm thick on the coronal plane using a vibratome. Sections were collected on cleaned, gelatin-coated slides and stained in ammonium hydroxide for 30 min, followed by Kodak Fix for film for another 30 min and finally were washed with water, dehydrated, cleared and mounted using Permount (Fischer Scientific). Slides were coverslipped and allowed to dry before quantitative analysis. Several pyramidal neurons impregnated with the Golgi solution were readily identified in the cortex by their characteristic triangular soma shape and numerous dendritic spines. Four to five neurons per animal were reconstructed by ImageJ software. For spine quantification, a 1006 oil-immersion objective was used to identify spines in dendrites longer than 10 mm. Quantification of the number of primary dendrites (defined as dendrites longer than 21 mm emanating directly from the soma) and total dendrites (defined as the amount of all dendritic branches) was done on images acquired with 256 objective, in which a circle was drawn around the cell body and the number of dendrites crossing each circle (primary dendrites) and the total branches were manually counted. Quantitative analysis for spines was performed using NeuronStudio and ImageJ in images acquired with 1006 objective. The dendritic branching patterns were analyzed using Sholl's method of regularly spaced concentric circles centred on the neuronal soma [40]. The number of dendritic intersections crossing each 20 mm-radius circle progressively more distal from the soma was counted. For each of these parameters, an average was calculated for each animal. Means for the EfnB1 Y/+ and EfnB1 Y/2 groups were then obtained from these individual values. Spine densities were calculated as mean numbers of spines per micrometer per dendrite per neuron in individual mice per group.

Quantitative RT-PCR
RNA from embryonic and adult cortices was harvested in RNA LaterH (Ambion) and total RNA was extracted using TrizolH (Invitrogen) according to the manufacturer's instructions. Messen-ger RNA expression was measured by qRT-PCR using a Bio-Rad thermo-cycler. Reverse transcription (RT) was perfomed using SuperscriptH III (Invitrogen) with 1 mg of total RNA per reaction. For the qRT-PCR reaction the resultant cDNA was diluted 1:1000. Each RT step was performed in duplicate and the qRT-PCR in triplicate for each RT reaction. RT-PCR was performed using SYBR Green JumpStartTM (Sigma-Aldrich). Relative values were calculated by the 2-DdCT method. Values were normalized to b-Actin or GAPDH mRNA levels. Results in adult tissues are standardized to EfnB1 expression values at E16.5. Primer sequences were as follows: EfnB1F:TGAAGGTTGGGCAAGATCC, R:GGTTCA-CAGTCTCATGCTTGC; EfnB1F:GACAAGCTGGACATCATCTGC, R:GCCGCCCGCTGCTGTGT; EfnB1F:TTGGCCAAGAACCTGGAG, R:GCCCTTCCCACTTAGGAACT.
Note that different primer sequences for EfnB1 were used to confirm expression levels in adult cortices.

Statistical Analysis
Mean, standard deviation and p values were calculated using Excel software. For behavioural analyses SYSTAT 11.0 statistical software was used. The results were analyzed using one or twoway ANOVAs, or a repeated measure ANOVA when appropriate. Post-hoc multiple comparisons were carried out when allowed, using Tukey's Honestly Significant Distance (HSD) test. Student ttest was performed to determine significant differences between samples. A value of p,0.05 was considered statistically significant. For dendritic complexity and spine density, the results are presented as mean 6 S.E.M. Statistical differences were determined by Student's t test for two-group comparisons.

Spatial Memory is Unaffected in EfnB1 Mutant Mice
The general behaviour of EfnB1 Y/2 mice is comparable to that of EfnB1 Y/+ mice, as we observed no significant differences in the level of locomotor activity or anxiety between EfnB1 Y/2 mice and control mice (Fig. 1A). Spatial learning and memory competences of EfnB1 Y/2 males were assessed using the Morris Water Maze (MWM). EfnB1 Y/+ and EfnB1 Y/2 mice performed equally during the acquisition phase (Fig. 1B) and during the probe test 24 h after acquisition (Fig. 1C). These findings demonstrate that EfnB1 Y/ 2 mice showed no deficits in spatial learning and memory as determined by the MWM.
To validate the notion that loss of ephrinB1 does not influence spatial memory, EfnB1 Y/+ and EfnB1 Y/2 male mice were further assessed in the object location task. During the retention phase, both EfnB1 Y/+ and EfnB1 Y/2 mice spent significantly more time exploring the displaced object, showing a preference index significantly different from chance level (50%; Fig. 1D). Therefore, our findings indicate that spatial learning and memory, which are hippocampus-dependent, were normal in absence of ephrinB1.

Non-spatial Memory Deficits in EfnB1 Mutant Mice
We next studied non-spatial learning and memory competencies in EfnB1 Y/2 mice. In the object recognition test, mutant and control males spent an equal amount of time exploring the two identical objects during the training phase (data not shown).
During the retention test EfnB1 Y/+ mice spent significantly more time exploring the novel object as opposed to the familiar one, so that the preference index was significantly different from chance level (50%). In contrast, the preference index for EfnB1 Y/2 mice was not different from chance level indicating that these animals were not able to recognize the novel object ( Fig. 2A). These results suggest that EfnB1 Y/2 mice exhibited a non-spatial memory deficit. To further explore non-spatial learning and memory in EfnB1 mutants, the subjects were examined in the passive avoidance and fear conditioning paradigms. In the passive avoidance paradigm, EfnB1 Y/2 mice exhibited shorter latencies to enter the dark chamber during the retention test than did the corresponding EfnB1 Y/+ controls (Fig. 2B) therefore indicating that mutant males displayed a significant memory retention deficit in this aversive task. We further tested these mice in the fear conditioning paradigm. Interestingly, we found that EfnB1 Y/2 mice demonstrate impairments both in contextual and cued conditioning tests (Fig. 2C, D), thus the deficit is not specific to hippocampus-dependent (context) memories [41]. Altogether these findings indicate that loss of ephrinB1 leads to impairments in non-spatial learning and memory that preferentially implicate the cortex.

Behavioural Deficits Indirectly Correlate with EphrinB1 Function
To establish the neuroanatomical basis of the observed behavioural deficits in EfnB1 mutants, we performed a histological analysis of adult brains from EfnB1 Y/+ and EfnB1 Y/2 mice. EfnB1 mutants displayed similar brain:body mass indices compared to littermate controls (EfnB1 Y/+ : 0.21160.05, EfnB1 Y/2 0.20860.07). The gross morphology of EfnB1 Y/2 brains was normal (Fig. 3A, B); however, 28% (4/14) mutants demonstrated a slight enlargement of the ventricles (Fig. 3B). We did not observe agenesis of the corpus callosum in EfnB1 Y/2 mice, even though it has been previously reported by others [42]. This discrepancy is likely due to the different genetic backgrounds of the mice under study.
The EfnB1 Y/2 mice displayed no deficit in any of the spatial learning paradigms that are known to strongly involve the hippocampus [23,24]; but rather showed impairment in nonspatial learning tasks that have been shown to be largely independent of hippocampal involvement but involve the cortex [35,36]. To better understand the putative role of ephrinB1 in non-spatial learning and memory, we performed immunohistochemical staining to establish the expression pattern of ephrinB1 in the adult brain. Surprisingly, we could not detect ephrinB1 in the adult WT cortex or in the adult WT hippocampus at the stage where behavioural studies were performed (Fig. 3). These findings indicate that the learning and memory deficit observed may be independent of ephrinB1 expression in the adult brain. To further assess the spatio-temporal expression pattern of ephrinB1 in the postnatal brain we performed qRT-PCR and immunohistochemical analyses and found that while ephrinB1 is expressed at P5 in the cortex, its levels decrease by P10 and it is no longer detected in the P15-P150 brain (Fig. 3E). Similar to our findings for the Graph presenting swim path length as a function of individual training sessions (S1-S4). Throughout the training sessions, EfnB1 Y/2 mice learned equally well to locate the hidden platform and exhibited decreasing swim distances over blocks of trials (F 3,45 = 12.58, p,0.001). C. Number of annulus crossings during probe tests. All groups of mice showed similar preference for the target zone where the platform was located during training sessions as compared to the adjacent (Adj1 and Adj2) and opposite zones (target vs others, **p,0.01; ***p,0.001). D. Performances in the object location task are expressed as the group mean (6 SEM) preference index. The horizontal line represents equal exploration of the two objects. EfnB1 Y/2 and WT mice presented the same level of reaction to the object displacement (F 1,17 = 0.139, p = 0.714) and spent significantly more time exploring the displaced object than the non-displaced one ( ### p,0.001; index vs chance level (50%)). NS: not significant. doi:10.1371/journal.pone.0088325.g001 Role of EphrinB1 in the Postnatal Cortex PLOS ONE | www.plosone.org cortex, we found that ephrinB1 is expressed throughout the hippocampus in the early postnatal brain and not expressed in the adult (Fig. 3F, G). Therefore, ephrinB1 expression coincides with early postnatal brain maturation.

Supernumerary Neurons in the Cortex of Adult EfnB1 Y/2 Mutant
Given that ephrinB1 is not expressed in the adult brain we asked whether the cortex exhibited anomalies at the cellular level that may correlate with the non-spatial learning and memory deficit. We performed immunohistochemistry with the pan-neuronal marker (NeuN) at various postnatal stages and observed that Figure 2. EfnB1 mutant mice present a deficit in non-spatial memory. A. Performances in the object recognition task are expressed as the group mean (6 SEM) preference index. The horizontal line represents equal exploration of the two objects. EfnB1 Y/2 male mice exhibited severe deficits in this task (F 1,14 = 17.890 p,0.001) and were not able to distinguish the new object. B. In the passive avoidance paradigm, EfnB1 Y/2 male mice exhibit decreased memory performance relative to wild-type littermates (F 1,15 ; 6.002 p = 0.027). C. Acquisition curve for the percentage freezing rate in the fear conditioning task. EfnB1 Y/2 mice in the 2 groups showed similar levels of freezing before the presentation of any tones (F 3,45 = 0.272 p = 0.845). During the two tone presentation, the freezing performances are not different between the two groups (first tone: (F 1,15 = 2.282 p = 0.110;second tone: F 1,15 = 1.472 p = 0.244) and subsequent to the first tone presentation, all mice showed increased freezing (F 6.90 = 3.781 p = 0.002). Even though lower levels of freezing behaviors were observed in EfnB1 Y/2 mice, this difference was not significant (F 1,15 = 3.48 p = 0.082). Solid lines represent tone and arrows indicate the presentation of a foot shock D. 24 h following conditioning, male mutant mice were impaired in contextual fear conditioning (F 1,15 = 13.694 p = 0.002) and in cued fear conditioning (F 1,15 = 7.921 p = 0.013). E During the two tests, the dynamic of the freezing response is not different between the two groups indicating that the deficit in performances reflects more a global decrease in the quantity of freezing than a deficit in delayed recall (contextual:  starting at P10, a significantly greater number of NeuN positive cells were present in the cortex of EfnB1 Y/2 compared to EfnB1 Y/+ mice (Fig. 4A-C). This difference in neuron number was still present at p150, where counts were conducted in all cortical regions; the frontal cortex, the somatosensory 1 cortex (S1), the somatosensory 2 cortex (S2), and the perirhinal cortex (Phr). Importantly, this increase in cell number did not correlate with an increased thickness of the mutant cortex (Fig. 4D). To determine whether the increase in neuron number was specific to the cortex, we quantified NeuN positive cells in the striatum, amygdala and hippocampal CA1 and CA3 regions. We observed no statistical difference in NeuN positive cells in any of these structures in EfnB1 Y/2 compared to EfnB1 Y/+ mice (Fig. 4E). Altogether, these results indicate that loss of ephrinB1 leads to a postnatal increase in cortical neuron number.

Decreased Dendritic Arborisation and Synaptic Density in EfnB1 Mutants
We next sought to determine whether changes in neuronal number correlated with modification of synaptic contacts in the cortex of EfnB1 Y/2 mice. The dendritic complexity of the layer II/ III pyramidal neurons from three different regions of the mouse cortex: the frontal cortex, S1, S2, and Phr cortical neurons was analysed in adult EfnB1 Y/+ and EfnB1 Y/2 mice (Fig. 6A, B). Sholl analysis showed a significant difference in the number of intersections at longer distances from the cell soma (Fig. 6C). Further, the dendritic branching of neurons from EfnB1 Y/2 mice was less complex than that observed in EfnB1 Y/+ animals. Moreover, the majority of dendritic intersections in the cortex of EfnB1 Y/2 mice were found proximal to the soma, at a distance of ,20-50 mm compared to EfnB1 Y/+ (Fig. 6C). Quantitative morphological analysis of neurons in the cortex of EfnB1 Y/+ and EfnB1 Y/2 mice showed that the length of dendritic branches was significantly reduced in EfnB1 Y/2 samples (88 mm612) compared to EfnB1 Y/+ controls (125 mm67) (Fig. 6D). In addition, we observed varicosities on dendrites from EfnB1 Y/2 samples (Fig. 6B, F) that are not seen in EfnB1 Y/+ controls (Fig. 6E). Dendritic spines were then analyzed on the basilar tree of the pyramidal neurons. The number of morphologically distinct dendritic spines on the basilar tree of cortical layer II/III pyramidal neurons was significantly different between EfnB1 Y/+ and EfnB1 Y/2 samples (Fig. 6E, F). Long, thin spines were prevalent on dendritic varicosities of neurons from EfnB1 Y/2 mice (Fig. 6F). Analysis of spine morphology revealed that neurons in EfnB1 Y/2 cortices have a significantly higher number of thin spines and reduced numbers of mushroom or stubby shaped spines compared to neurons in EfnB1 Y/+ controls (Fig. 6G). Analysis of the total number of spines along a dendrite showed fewer spines throughout the entire length of the basal dendrites of neurons from EfnB1 Y/2 samples (Fig. 6H).

Discussion
Here we show that loss of ephrinB1 leads to specific cognitive deficits in adult mice. EfnB1 deficient mice exhibit defects in nonspatial learning and memory but performed equally to the EfnB1 Y/ + controls in all spatial learning tasks. Strikingly, while mutant mice were able to detect spatial differences in the object location task, all subjects showed impairment in the object recognition task. Mutant mice also displayed a significant memory retention deficit in the passive avoidance task and in the fear conditioning paradigm. While ephrinBs were recently implicated in neuropathic pain responses [44], we found no difference in the response of wild type and EfnB1 mutant mice to the shock in the fear conditioning experiment, indicating that all subjects were able to sense pain in these learning paradigms. These findings demonstrate that the absence of ephrinB1 does not lead to a general cognitive impairment, but rather to a highly selective deficit. To the best of our knowledge, this is a type of cognitive phenotype affecting specifically non-spatial learning and memory has not been reported to date.
Non-spatial learning and memory employs a number of different structures including cortical regions such as the entorhinal and perirhinal cortices, sub-cortical structures such as the amygdala and striatum, and occasionally the hippocampus [45,46]. Using a number of different learning paradigms with distinctive methodologies, we aimed to identify the most probable brain structure impaired in absence of ephrinB1. For example, the cortex, amygdala and striatum are involved in passive avoidance [46] and fear conditioning [27,28,47,48] and so are potential candidate structures to explain the performance deficit in EfnB1 mutant mice. However, the striatum and amygdala are not implicated in the object recognition task suggesting that a different structure may be involved in the behavioural deficits observed in our mutant mice. The non-spatial learning tasks are not completely independent of hippocampal functioning as these tests are conducted in a specific context. It is possible, however, to reduce hippocampal involvement by reducing the complexity of the contextual environment [36]. For this reason, we used the object recognition task with the simplest possible context; the passive avoidance task was performed with a very small light chamber in a dark experimental room; and the fear conditioning task employed two different procedures, one in the same context and the second in a different context. As we obtained a deficit in all these procedures, this strongly suggested that the deficit is not context dependant and therefore is independent of hippocampal functioning. Given that the EfnB1 mutant mice displayed no impairment in either of the spatial learning paradigms employed, which strongly involve the hippocampus, we concluded that the most likely structure implicated in learning and memory deficit in EfnB1 mutants is the cortex and we thus focused our neuroanatomical analyses on the cortex.
Interestingly, we could not detect ephrinB1 expression in the adult cortex or hippocampus at the time the behaviour studies were performed. While an earlier study has shown widespread diffuse ephrinB1 expression in adult brain in mice [49], we show here, using the EfnB1 knockout mouse line as a control, that ephrinB1 is not expressed in the adult cortex or hippocampus. compared to control embryonic (E) day 16.5 cortical extracts that express EfnB1. n = 3/age. F-G'. Immunohistochemistry for ephrinB1 in EfnB1 Y/+ and EfnB1 Y/2 hippocampi at P5 and P150. EphrinB1 is not detected in EfnB1 Y/+ adult hippocampi (F), but observed in the CA1 region at P5 (F'; inlet). doi:10.1371/journal.pone.0088325.g003 Therefore our findings suggest that ephrinB1 does not control learning and memory through direct modulation of synaptic transmission or synaptic plasticity, but instead plays a role in postnatal cortical maturation. Given that ephrinB2 and ephrinB3 have been shown to directly modulate hippocampal synaptic plasticity and learning and memory [50,51], our findings show that ephrinB1 impacts learning and memory through a mechanism that is distinct from other ephrinBs. Further analyses of EfnB1 Y/2 adult brain showed a significant increase in the number of neurons only in the cortex of EfnB1 Y/2 mice. The increase in the number of neurons was traced to a postnatal event since neuron numbers were not significantly different between wild type and EfnB1 mutant mice at birth (P0). How loss of ephrinB1 leads to an increased number of cortical neurons is unknown, yet the timing of appearance of this phenotype is suggestive of a decrease in programmed neuronal cell death that normally takes place in the cortex around P5. A number of studies have shown a potential role for Eph receptors and ephrins in brain region specific apoptosis (reviewed in [52]). Previous work has linked EphA/ ephrinA signalling to the control of cell number in the developing brain, via a direct effect on survival of embryonic neural progenitors [53]. While these studies all show a role for Eph/ ephrin signaling on apoptosis of neural progenitors, this is the first report to show that a member of the Eph/ephrin family, ephrinB1, is important to control neuron number in the postnatal cerebral cortex. Future work is required to understand the role of ephrinB1 in this apoptotic molecular cascade. Interestingly we noted no difference in cortical thickness from EfnB1 Y/+ and EfnB1 Y/2 cortices despite changes in cortical neuron numbers. This finding suggests that neuronal packing may be changed in absence of ephrinB1.
In correlation with the increased neuronal numbers, we found aberrant dendritic morphology and arborisation, along with a significant decrease in spine density on pyramidal cortical neurons in EfnB1 Y/2 cortices. This reduced branching of the neurons from EfnB1 Y/2 mice could explain the maintenance of cortical thickness in EfnB1 Y/2 mice despite changes in cortical neuron numbers. Whether reduced branching is a consequence of increased neuron numbers is currently unknown. Neurons from EfnB1 Y/2 mice displayed dendritic varicosities, a peculiarity seen in some cases of infantile neurobehavioral failure, in addition to a number of progressive neurodegenerative disorders including Huntington's disease frontal lobe dementia and Alzheimer's disease (reviewed in [54]). The cause of the varicosities is not well known, yet varicosities are related to dendritic spine dysgenesis and loss. Closer scrutiny of dendritic spines revealed a significantly higher number of thin spines on neurons in EfnB1 Y/2 mutants with reduced mushroom or stubby shaped spines. Spines form a variety of shapes and sizes that correlate with their synaptic strength, motility, and structural plasticity [55]. Evidence shows that thin spines are abundant in the early postnatal brain and undergo refinement to promote dendritic spine maturation and synapse/ dendrite stability [2,54]. Previous studies have shown that stimulation of cultured neurons with soluble recombinant ephrinB1 protein induces dendritic spine formation and maturation [20,56]. These findings suggest a direct role of ephrinB1 in spine maturation and suggest that the abundance of thin spines in EfnB1 Y/2 cortices could be a manifestation for the privation of postnatal spine maturation in the absence of ephrinB1. An increase in thin spines has been associated with several hereditary forms of intellectual disability, and is often observed for many genes that underlie X-linked intellectual disability [57]. It is thus possible that the defective architectural structuring of synaptic connections in EfnB1 Y/2 cortical pyramidal neurons underlies cognitive impairment of these mutants. Alternatively, changes in the number of CR+ cells have been reported in various disorders [58], and we have observed an alteration of the CR+ interneuron subtype in the somatosensory cortex of EfnB1 mutant mice. To date the functional properties of CR+ interneurons have been primarily studied in the hippocampus where they are responsible in synchronizing dendritic inhibitory cells [59] to regulate input information into principal neurons [60]. Why CR+ interneurons were increased specifically in the S2 cortical region of EfnB1 Y/ 2 mutants and whether this increase contributes to the behavioural impairment observed in these mice remains unknown. Overall our findings suggest that in the absence of ephrinB1, postnatal neuronal cell death does not occur normally, thus leading to imbalance between neuronal subtypes and to improper maturation of dendritic arbors.
Mutations in EfnB1 are associated with a human syndrome called CranioFrontoNasal Syndrome (CFNS) [16,17]. CFNS is a rare X-linked disorder that shows paradoxically greater severity of craniofacial abnormalities in heterozygous females than in hemizygous males. This is due to the fact that these abnormalities are exacerbated by cellular mosaicism due to random X-  (A, B), CR (C, D) and NPY (E, F) The percent of interneuron numbers per 100 micro(m)m 2 is shown for the S2 region of EfnB1 Y/+ versus EfnB1 Y/2 samples. G. We observed an overall increase in PV+, CR+ and NPY+ neurons in EfnB1 Y/2 mutants as compared to EfnB1 Y/+ in all cortical regions studied. A significant increase was detected for CR+ neurons in the EfnB1 Y/2 S2 cortical region. * p,0.05. doi:10.1371/journal.pone.0088325.g005 Figure 6. Comprised dendritic arbors in ephrinB1 mutant cortical pyramidal cells. A, B. Representative photomicrographs of golgi-cox impregnated S2 pyramidal neurons from EfnB1 Y/+ (A) and EfnB1 Y/2 (B) cortices. Note the presence of dendritic varicosities on dendrites from EfnB1 Y/2 samples (arrows). C. Sholl analysis of the structure of dendritic arbors. Mean number of dendritic intersections of layer II/III cortical pyramidal neurons show a significant decrease in the number of intersections in pyramidal neurons from EfnB1 Y/2 cortex at longer distances from the cell soma. Note that majority of dendritic intersections in EfnB1 Y/2 cortex were found proximal to the soma, at a distance of ,20-50 mm compared to EfnB1 Y/+ cortex. D. Quantitative analysis of dendritic branch length shows a statistically significant decrease in the branch length of pyramidal neurons from EfnB1 Y/2 samples compared to controls (EfnB1 Y/+ = 125 mm67; EfnB1 Y/2 = 88 mm612; p,0.01). E, F. Higher magnified photomicrographs of EfnB1 Y/+ (E) and EfnB1 Y/2 (F) pyramidal neurons in the S2 field. Note the varicosities in dendrites from EfnB1 Y/2 samples. G. Quantitative analysis of the number of morphologically different spines on the basilar tree of cortical layer II/III pyramidal neurons in EfnB1 Y/+ and EfnB1 Y/2 samples. A significantly higher number of thin spines are observed on neurons from EfnB1 Y/2 mice. H. Analysis of the total density of spines along a dendrite showed fewer spines throughout the entire length of the basal dendrites. Each dot represents 4-5 neurons per individual. n = 5/genotype. * p,0.05, **p,0.001. doi:10.1371/journal.pone.0088325.g006 inactivation in female patients. This peculiar feature of CFNS is also observed in mice, since heterozygous females present more penetrant and generally stronger -yet similar-phenotypes than hemizygous males [61,62]. To date there is only sporadic information on the cognitive capacity of CFNS male or female patients; however, our study suggests that this syndrome may share a number of characteristic features with other neurodevelopmental disorders. Interestingly, cortical anomalies observed in EfnB1 Y/2 mutants also partially resemble those seen in genetic models for Fragile-X Syndrome, which exhibit a striking inhibitory/excitatory synapse imbalance and disrupted synaptic contacts [63]. Altogether, our findings demonstrate that ephrinB1 is involved in cortical maturation and in non-spatial learning and memory.