Autism-associated gene Dlgap2 mutant mice demonstrate exacerbated aggressive behaviors and orbitofrontal cortex deficits

We used the method of targeting vector construction to generate Dlgap2 knockout (KO) mice as previously described by Liu, Jenkins, and Copeland [28]. In summary, exon 6 of Dlgap2 (Gene ID: 244310) in embryonic stem (ES) cells from the 129S1/Sv mouse strain was replaced by a construct containing Dlgap2 exon 6 interposed between two loxP sites and a NEO cassette via spontaneous homologous recombination. Exon 6 of Dlgap2 was knocked out by Cre-induced homologous recombination (Figure 1A). The ES cell clones containing the mutant allele were rechecked by Southern blot DNA fragment analysis to ensure the correctness of gene targeting. The targeted ES cells were microinjected into the blastocysts from the C57BL/6 mouse strain and the injected blastocysts were transferred into ICR foster mothers to produce chimeric mice. F1 founders were produced by mating male chimeric mice with wild-type (WT) C57BL/6 females. The mice were backcrossed for more than eight generations to make the congenic strain in a 99.9% C57BL/6 genetic background. Combinations of forward primer F1 (GCCACATTCATAACATAGCTAC), reverse primer 1 (R1) (ACCTC TGCTACATACCCACTC) and reverse primer 2 (R2) (ACACATGGGATGCTGTACGC) were used to determine the genotype of each mouse. The amplicon of mutant allele was 800 bp and the amplicon of the WT allele was 600 bp (Figure 1B). The congenic strain Dlgap2 KO mice and its WT littermates were used for analysis in this study.

Adult male mice were allowed to explore freely for 30 minutes in a square (16 × 16 inches) open-field arena (San Diego Instruments, San Diego, CA USA). Locomotion activity was monitored using a 16 × 16 photo-beam sensor. The numbers of beam-breaking events were computed as an index of locomotion activity.

Social approach and social novelty tests were performed as previously described with minor modifications [29‐31]. A Plexiglas cage (40 × 60 × 22 cm) was divided into three equal regions (40 × 20 × 22 cm). At the beginning of the test, a mouse was allowed to habituate in the central chamber for 10 minutes. We then removed the doorway between the chambers to let the mouse acclimate to the environment for another 10 minutes. For the sociability test, we first confined the test mouse in the central chamber again, and then put one small Plexiglas cylinder to constrain a target mouse in one side chamber and an empty Plexiglas cylinder in the third chamber. The Plexiglas cylinders have numerous holes to facilitate social communications. To enhance the behavioral measure sensitivity, we defined the small square space (20 × 20 cm) containing the Plexiglas cylinder with a mouse as the social region, and the space with the empty Plexiglas cylinder as the object region. After the removal of the doorway, the time spent in each region by the test mouse was automatically recorded for 5 minutes. When the test was over, the mice were allowed to explore the chambers freely for another 5 minutes. For the social novelty test, a novel mouse was then put in the previously empty Plexiglas cylinder. The time spent in each region was automatically recorded for 5 minutes again. Group-housed adult male mice (8 to 9 weeks of age) were used in these experiments.

Adult male mice (8 weeks of age) from both genotypes were single-housed in their home cages for 3 weeks. On the test day, an adult male intruder with similar body weight was placed into the cage. For the initial 5 minutes of social contact, aggressive behaviors were scored by a rater blinded to the genotypes. The aggressive behavioral indexes, include biting, wrestling, tail-rattling, aggressive grooming and chasing, were as previously described [32].

The tube test was performed as previously described with minor modifications [33, 34]. We designed transparent Plexiglas tubes with a length of 30 cm and inside diameter of 3.2 cm. This narrow space is just sufficient for a mouse to walk through without being able to reverse its direction. Mice were trained to walk through the tube before testing. On the test day, mice from both genotypes were released at opposite ends of the Plexiglas tube, and we made sure that they met at the middle of the tube. The mouse that retreated first from the tube was defined as the loser and the other mouse, which stayed in the tube, was the winner.

The reversal learning in water T-maze test [35] was modified from the method described by Gusariglia and Chadman [36]. The water T-maze consists of a white Plexiglas T-maze, which is placed in a circular pool with a diameter of 100 cm. The pool was filled with water to a depth of about 33.5 cm at 22 ± 0.5°C. The Plexiglas T-maze consists of a start arm (30 × 10 cm) and two goal arms (30 × 10 cm) with walls (20 cm high). A platform (Plexiglas, 5 × 5 cm) was placed at the end of one goal arm and submerged 1 cm below the surface of the water. The apparatus was set in a room without visual cues. Data were collected manually by a single observer.

Mice were placed in the water T-maze without the platform to acclimate them to the task environment for 60 s as pre-training. In the habit acquisition session, a platform was located at the end of only one goal arm. Each mouse was placed in the starting arm and allowed up to 60 s to find the submerged platform. When a mouse failed to find the platform in 60 s, it was picked up by the experimenter and placed onto the platform for 10 s. Mice underwent ten trials per day. On each trial, an error was recorded when the mouse entered the arm without the platform or entered the arm with the platform and left that arm. Only when the mouse entered the arm and climbed onto the platform directly was this recorded as a correct choice. When a mouse had been trained to perform eight out of ten trials correctly for four consecutive days, it was subjected to training for reversal learning. In the reversal learning session, the platform was switched to the opposite arm of the T-maze. The training procedure was the same as the acquisition session. The entry into the arm where the platform was originally located was recorded as an error. The total numbers of errors per day were measured throughout the 4-day reversal learning session.

Crude synaptosomal fractions of mouse cerebral cortex were prepared as previously described in Schmeisser et al.’s work [25]. In brief, a cortex was quickly dissected out and homogenized in HEPES-buffered sucrose (320 mM sucrose, 5 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES), pH 7.4) containing proteinase inhibitor cocktail. The homogenized cortices were then centrifuged at 1,000 g for 10 minutes at 4˚C to remove cell debris and nuclei. Then, the supernatant was spun at 12,000 g for 20 minutes to yield the soluble and the crude synaptosomal fractions. Equal amounts of protein per lane from both genotypes were separated by SDS-PAGE. Primary antibodies used in immunoblot experiments included NR1 (Abcam, Cambridge, MA USA), NR2A (Abcam), NR2B (Abcam), GluR1 (Millipore, Billerica, MA USA), GluR2 (Abcam), Shank3 (Abcam), PSD95 (Abcam), Homer-1b/c (Santa Cruz, Dallas, Texas USA), αCaMKII (Santa Cruz), βCaMKII (Abcam), β-actin (Cell Signaling, Danvers, MA USA) and Dlgap2 (GeneTex, Hsinchu Taiwan). After incubation with the primary antibodies, peroxidase-conjugated secondary antibodies were applied to the blots and incubated for 1 hour at room temperature. The blots were visualized with a chemiluminescence reagent using Biospectrum Imaging System (UVP, Upland CA). The quantity of each protein was normalized for the quantity of the corresponding β-actin detected in the sample.

Dendritic spines were counted in layer II/III pyramidal cells in the OFC using the Golgi-Cox stain. In brief, brain tissues were taken and placed in the impregnation solution (solutions A and B in FD Rapid GolgiStain kit, FD NeuroTechnologies, Ellicott City, MD, USA) for 17 days. After several washes, brain tissues were cut into coronal sections with a vibratome to a thickness of 150 μm. Brain sections were collected and reacted with the mixture of developer and fixer in FD Rapid GolgiStain kit (solutions C and D, 1:1) for 2 minutes. Dendritic spines were examined with a 100× objective oil immersion lens and captured with the Stereoinvestigator system (Microbrightfield, Williston, VT USA). For quantitative analysis of spine density, the spines were counted along dendritic segments chosen from secondary and tertiary dendrites using the Image J software. The experimenter was blind to genotype and selected one to three segments for each cell.

The animal-use protocol was approved by the Animal Ethics Committee of the National Taiwan University, Taiwan. Briefly, animals were sacrificed by decapitation and the fresh brains were quickly removed into chilled (0 to 4°C) cutting solution containing: 0.5 mMCaCl₂, 110 mM choline chloride, 25 mM glucose, 2 mM KCl, 7 mM MgSO₄, 26 mM NaHCO₃, 1.25 mM NaH₂PO₄, 11.6 mM sodium ascorbate and 3.1 mM sodium pyruvate. Coronal brain slices (250 to 300 μm in thickness) containing the OFC were cut by a microslicer (Dosaka DTK-1000, Ted Pella, Altadena, CA USA) then transferred to a holding chamber with artificial cerebrospinal fluid consisting of: 2 mM CaCl₂, 10 mM glucose, 3 mM KCl, 1 mM MgCl₂, 125 mM NaCl, 26 mM NaHCO₃ and 1.25 mM NaH₂PO₄. Slices were then maintained at room temperature for at least 1 hour before recording and bubbled with 95% O₂ and 5% CO₂. A brain slice was transferred to the recording chamber and held in position using a stainless steel grid with a nylon mesh and submerged in a solution at 1 to 2 ml/min. Whole-cell recordings were made from layer II/III OFC pyramidal neurons using a patch-clamp amplifier (Axopatch 200B, Molecular Devices, Sunnyvale, CA USA). Neurons were visualized under infrared-differential interference contrast lens and digital camera by an upright microscope (BX51WI, Olympus, Tokyo, Japan). Patch electrodes (3 to 8 MΩ) were pulled from borosilicate glass capillaries by a micropipette puller (P97, Sutter Instrument, Novato, CA USA) and filled with a solution containing: 140 mM K-Glu, 5 mM KCl, 10 mM HEPES, 0.2 mM EGTA (ethylene glycol tetra-acetic acid), 2 mM MgCl₂, 4 mM MgATP, 0.3 mM Na₂GTP (guanosine-5'-triphosphate, disodium salt) and 10 mM Na₂-phosphocreatine, pH 7.2 (with KOH). Data acquisition was performed using a digitizer and pClamp 10 software (DigiData 1440A, Molecular Devices). A presynaptic stimulation was carried out with a glass pipette filled with 3 M NaCl solution placed in layer I of the prelimbic region of the recording site using a stimulator (S-48, Grass-Telefactor, Warwick, RI USA) and an isolation unit (ISO-Flex, AMPI, Jerusalem, Israel). To calculate the paired-pulse ratio (PPR), two stimuli pulses were delivered at 25-, 50- or 100-ms intervals and the PPR was measured by dividing the second excitatory postsynaptic current (EPSC) by the first. Data were excluded from the analysis if the input resistance varied >20% throughout the experiment. To isolate the α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor-miniature excitatory postsynaptic currents (AMPAR-mEPSCs), a cell was clamped at -70 mV in the presence of tetrodotoxin (1 μM), picrotoxin (100 μM) and APV (amino phosphonovalerate, 50 μM) and analyzed using MiniAnalysis Program (Synaptosoft, Decatur, GA USA). Signals were filtered at 2 Hz and digitized at 10 Hz.

Adult male mouse brains were transcardially perfused with fixative (2% paraformaldehyde and 2% glutaraldehyde in 0.1 M phosphate buffer). Subsequently, the brains were dissected out and post-fixed in the same fixative as described above overnight at 4°C. We then used a vibratome to section the mouse frontal cortex into 300 μm thicknesses, and isolated the region of interest, the OFC. These samples were further post-fixed in 1% osmium tetroxide for 1 hour at 25°C. The samples were then dehydrated and embedded in epoxy or spurr resin. Ultrathin sections of 70 nm were cut using an ultramicrotome. After lead citrate staining, these sections were examined by electron microscopy. Images were acquired at 40,000× magnification, and PSD thickness and length were measured with ImageJ.

We generated Dlgap2 ^−/− mice (Figure 1A,B) by deleting exon 6 from the gene, the largest coding exon of the Dlgap2 gene. The deletion generated a frame-shift mutation that resulted in complete functional loss of DLGAP2 protein, due to inadequate translation of the C-terminal. With Western blotting analysis, we further confirmed that mutant mice lacked DLGPA2 protein products (Figure 1C). It is known that Dlgap2 is highly expressed in the cortex, striatum, hippocampus and olfactory bulbs [14, 37], as a result, we could only detect its protein products in these brain regions of WT mice but not those of the mutants (Figure 1D). Dlgap2 ^−/− mice had similar body shapes and weights as their WT littermates (Additional file 1: Figure S1). To assess their basic locomotion ability, we used the open-field test (Figure 1E) and found that Dlgap2 ^−/− mice demonstrated intact locomotion ability since their performances were quite comparable to the WT mice. However, Dlgap2 ^−/− mice displayed novelty-induced hyperactivity (Figure 1E) in the first 5 minutes of the test specifically, which was likely due to impulsivity [38].

Because Dlgap2 has been linked to human social dysfunction, we performed a modified version of the three-chamber social approach test to investigate the social behaviors of the mutant mice [29, 30]. Initially, a test mouse was left to explore and interact with a strange mouse held in a plastic cage or with an empty plastic cage without any mice inside. Both genotypes manifested clear preference for social stimulus, indicating normal sociability (Figure 2A). However, on examining the data more carefully, we can see that Dlgap2 ^−/− mice seemed to display enhanced social approach behavior relative to the WT littermates since they spent significantly more time in the social chamber (Figure 2A). In the subsequent experimental trial, a new stimulus mouse was introduced into the previously empty cage to see if the test mice would interact with the familiar or novel mouse. Since mice have a natural tendency to explore novelty, they prefer to investigate the novel mouse if they are able to distinguish and recognize it. Our results demonstrated that both genotypes showed an intact and similar social recognition ability in the social novelty test (Figure 2B).

Although Dlgap2 ^−/− mice seemed to display normal sociability, the motivations behind the approaching behaviors could not be easily discerned by the three-chamber task. For example, social affiliation might be a positive motivation for mice to approach others; however, aggression may be another factor in motivating the mice to approach. We thus used the resident–intruder task to evaluate the aggressiveness of the Dlgap2 ^−/− mice. Briefly, mice were single-housed in their home cage for 3 weeks to establish their territory. On the test day, intruders were put in to induce aggressive behaviors in the resident mice. As the results show, Dlgap2 ^−/− mice displayed an obviously shorter attack latency (Figure 2C). Additionally, the duration and frequency of aggressive behaviors also dramatically increased compared to the WT mice (Figure 2D,E).

To provide further evidence of elevated aggression, we used the tube dominance test. Mice from each genotype were released into the opposite ends of a narrow tube. The more dominant mouse aggressively forces the opponent mouse out of the tube. The mouse to step out of the tube first was identified as the loser. Our results clearly manifested that the Dlgap2 ^−/− mice had a much higher probability of winning, indicating elevated aggression (Figure 2F).

We used performance in the water T-maze to determine the repetitive behaviors of Dlgap2 ^−/− mice. As shown in Figure 3A, there was no significant difference in the number of days to reach the 80% right position criterion between WT and Dlgap2 ^−/− mice (5.5 and 5.4 days, respectively). All mice reached the criterion on Day 6 and were moved into training session of reversal learning test (Figure 3B). In the learning session of reversal learning test, we found that Dlgap2 ^−/− mice made a significantly greater number of errors on the first day of learning session in comparison with WT mice (P < 0.05) (Figure 3C). Figure 3D shows the number of errors on each day throughout the 4-day reversal learning session. During reversal learning, both strains improved across days.

The cerebral cortex plays a critical role in social cognition and inhibition of aggressive drives [39, 40]. Given that DLGAP2 is a synaptic protein highly expressed in the cortex, and the integrity of Dlg4-DLGAP-SHANKs may play critical roles in synaptic structure and functions, we isolated crude synaptosomal fractions from the cerebral cortex and analyzed the protein composition. There was a clear downregulation of synaptic receptors GluR1 and NR1 in mutants, which may indicate that synaptic transmission is impaired (Figure 4A). Moreover, the levels of synaptic scaffold proteins Homer-1b/c and αCaMKII were also lower in the cortical synaptosomal fractions, which may imply a disruption to synaptic structures (Figure 4B). Based on these results, we reasoned that there are severe synaptopathies within the cerebral cortex of Dlgap2 ^−/− mice.

To further provide a link between cortical synaptopathies and elevated aggressive behaviors, we used the Golgi stain to analyze the spine density of pyramidal neurons in OFC (Figure 5A), which is known to play a crucial role in inhibition of aggressive drives [40]. In human clinical studies and animal studies on primates and rodents, dysfunction of the OFC has been show to cause exacerbated aggression [40]. Furthermore, since DLGAP2 is one of the main protein components of dendritic spines [11], it may regulate synaptogenesis in the OFC. As our experimental results showed, Dlgap2 ^−/− mice demonstrated a clear reduction of spine density (Figure 5B), indicating the pyramidal neurons received less excitable inputs in the OFC.

Aside from synaptic structure abnormality, electrophysiology experiments were performed to analyze the synaptic functions of Dlgap2 ^−/− mice. We recorded AMPAR-mEPSCs in acute slices of the OFC. Clearly, our results with a lower peak mEPSC amplitude for Dlgap2 ^−/− mice (Figure 6A,B) indicate that the postsynaptic responses of available synapses was weaker functionally. In addition, there was a trend towards a reduction in the frequency of mEPSCs, although this was not statistically significant (Additional file 1: Figure S2). We wondered whether this phenomenon was due to a presynaptic deficit. We thus measured the paired-pulse ratio (PPR). Surprisingly, we found that there was an obvious enhancement of PPR in mutants (Figure 6C,D). Because the PPR of the synaptic response is inversely correlated with presynaptic release probability, our results indicate the release probability of presynaptic vesicles was downregulated. Since DLGAP2 is a scaffold protein located in the PSD, this presynaptic deficit was a surprise. We reason that it was probably mediated by trans-synaptic adhesion molecules [41] or retrograde synaptic signaling [42].

Furthermore, we used transmission electron microscopy to investigate the ultra-structure of synapses in the OFC. Since the morphological integrity of the PSD, which contains vital receptors, scaffolds and signaling molecules, is highly relevant to proper synaptic function, we measured PSD thickness and length (Figure 7). Strikingly, our results demonstrated a clear reduction in the length of the PSD in the OFC (Figure 7B). Additionally, the thickness of the PSD also had dramatic downregulation (Figure 7C). These results indicate the obvious deficits of postsynaptic ultra-structures in Dlgap2 ^−/− mice.