Abstract
Schizophrenia (SCZ) is a severe mental disorder characterized by delusion, hallucination, and cognitive deficits. We have previously identified from schizophrenia patients a loss-of-function mutation Arg215 → His215 (R215H) of neuroligin 2 (NLGN2) gene, which encodes a cell adhesion molecule critical for GABAergic synapse formation and function. Here, we generated a novel transgenic mouse line with neuroligin-2 (NL2) R215H mutation, which showed a significant loss of NL2 protein, reduced GABAergic transmission, and impaired hippocampal activation. Importantly, R215H KI mice displayed anxiety-like behaviors, impaired pre-pulse inhibition (PPI), cognition deficits and abnormal stress responses, recapitulating several key aspects of schizophrenia-like behavior. Our results demonstrate a significant impact of a single point mutation NL2 R215H on brain functions, providing a novel animal model for the study of schizophrenia and neuropsychiatric disorders.
Schizophrenia (SCZ) is a chronic neuropsychiatric disorder caused by both genetic and environmental factors. It is featured by long-standing delusion and hallucination (psychosis), and cognitive deficits (Freedman, 2003; Insel, 2010; Lewis and Lieberman, 2000). SCZ is a highly heritable disorder (Sullivan et al., 2003) with a complex genetic basis. Recent genomic studies identified a number of genetic variants associated with SCZ, including a group of variants resided in the genes encoding synaptic adhesion molecules that promoting synaptic function and development such as IGSF9B, and NLGN4X (Schizophrenia Working Group of the Psychiatric Genomics, 2014).
Neuroligins (NLGNs) are a family of synaptic adhesion molecules highly expressed in the brain and are ligands for another group of cell adhesion molecules neurexins (NRXNs) (Ichtchenko et al., 1995). There are five neuroligin genes (neuroligin-1, -2, -3, -4, and -5) in humans and four in mice (neuroligin 1-4). Neuroligin-1, -2, and -3 are close homologs between human and mice. Neuroligin-1 and neuroligin-2 differentially locate to excitatory and inhibitory synapses and are critical for the excitatory and inhibitory synapse formation and function, respectively (Chubykin et al., 2007; Dong et al., 2007; Levinson et al., 2005; Nam and Chen, 2005; Scheiffele et al., 2000; Song et al., 1999; Varoqueaux et al., 2004). Neuroligin-3 locates at both type of synapses and contributes to both neurotransmission (Budreck and Scheiffele, 2007; Etherton et al., 2011). In recent years, several genetic variants of neuroligin-3 and neuroligin-4 have been identified in autism patients (Chih et al., 2004; Jamain et al., 2003). Mutations in proteins interacting with neuroligins such as Neurexin1, SHANK and MDGA have also been associated with autism and schizophrenia patients (Bucan et al., 2009; Durand et al., 2007; Kim et al., 2008; Kirov et al., 2008). Genetic mouse models based on these findings recapitulate some aspects of patient symptoms as well (Baudouin et al., 2012; Connor et al., 2016; Etherton et al., 2011; Etherton et al., 2009; Jamain et al., 2008; Peça et al., 2011; Rothwell et al., 2014; Südhof, 2008; Tabuchi et al., 2007; Zhou et al., 2016).
We have previously reported several novel mutations of NLGN2 from schizophrenia patients (Sun et al., 2011). Among the NL2 mutants, we found that the R215H mutant protein was retained in the endoplasmic reticulum (ER) and could not be transported to the cell membrane, resulting in a failure to interact with presynaptic neurexin and a loss of function in GABAergic synapse assembly (Sun et al., 2011). Based on these studies, we have now generated a transgenic mouse line carrying the same NL2 R215H mutation to test its functional consequence in vivo. We demonstrate that the R215H knock-in (KI) mice show severe GABAergic deficits and display anxiety-like behavior, impaired pre-pulse inhibition, cognitive deficits, and abnormal stress responses. These deficits are more severe than reported NL2 KO mice. Our results suggest that a single-point mutation R215H of NL2 can result in significant GABAergic deficits and contribute to SCZ-like behaviors. This newly generated NL2 R215H KI mouse may provide a useful animal model for the studies of neuropsychiatric disorders including SCZ.
Results
Generation of neuroligin-2 R215H mutant mice
Following our original discovery of a loss-of-function mutation R215H of NL2 in SCZ patients (Sun et al., 2011), we generated the NL2 R215H mutant mice by introducing the same R215H mutation into the exon 4 of Nlgn2 gene in the mouse genome via homologous recombination (Figure 1a). NL2 R215H heterozygotes were mated to obtain wild type (WT), heterozygotes (referred here as Het mice), and homozygotes (referred here as KI mice) (Figure S1a). Sequencing analysis confirmed the R215H mutation in the NL2 KI mice (Figure S1b). Mice carrying R215H mutation were born at a normal Mendelian rate (Male mice: WT = 26.5%, Het = 52.9%, KI = 20.6%; Female mice: WT = 24.1%, Het = 52.8%, KI = 23.1%). Both R215H Het and KI mice were viable and fertile and did not exhibit premature mortality. The weight of R215H Het and KI mice was not significantly different from WT mice (Figure S2). The mouse colony was maintained on a hybrid genetic background to avoid the artificial phenotype contributed by other homozygous genetic variants in a homozygous inbred background.
Reduction of neuroligin-2 protein level in NL2 R215H KI mice
After obtaining the NL2 R215H Het and KI mice, we first analyzed the NL2 protein expression level in the brain. We found that NL2 expression level reduced 90% in NL2 R215H homozygotes, and 40% in heterozygotes (Figures 1b and c), which was in contrast to a complete absence of NL2 in the NL2 KO mice (Figure 1b). Notably, the residual NL2 R215H protein band only showed a very weak lower band compared to the thick WT NL2 protein bands, suggesting that the residual NL2 R215H proteins were likely immature NL2 without glycosylation (Sun et al., 2011; Zhang et al., 2009).
To investigate the localization of NL2 R215H proteins inside the brain, we performed immunohistochemistry with NL2-specific antibodies and found a significant reduction of NL2 puncta in R215H Het mice and almost absence of NL2 puncta in homozygous R215H KI mice (Figure 1d-f). In WT mouse brains, NL2 formed numerous postsynaptic puncta on cell soma and dendrites opposing presynaptic vGAT puncta (Figure 1d-f, puncta density 15.0 ± 0.9 per 100 μm2, puncta size = 0.27 ± 0.01 μm2).
The number and size of NL2 puncta were significantly reduced in the NL2 R215H Het mouse brains (Figure 1d-f, puncta density, 9.4 ± 1.6 per 100 μm2, p < 0.01, One way ANOVA followed with Tukey post hoc test; puncta size, 0.21 ± 0.02 μm2, p = 0.02). Interestingly, in the homozygous NL2 R215H KI mouse brains, only faint NL2 signal was observed inside cell soma (Figure 1d, right columns) and not colocalized with vGAT, indicating that the NL2 R215H proteins could not be transported to the cell membrane, consistent with our previous observation in cell cultures (Sun et al., 2011). To get a clear understanding of the physiological role of NL2 R215H mutation in vivo, we focused our studies on the homozygous NL2 R215H KI mice in this study.
Reduced GABAergic synapse density in NL2 R215H KI mice
NL2 has been reported to form complex with gephyrin and collybistin at postsynaptic sites to recruit GABAA receptors (Poulopoulos et al., 2009). Consistent with a substantial reduction of NL2 puncta in the KI mice, we detected a remarkable decrease of postsynaptic GABAA receptor γ2 subunit and the scaffold protein gephyrin around cell soma in hippocampal regions (Figure 2a). Quantitative analysis revealed that both the puncta number and size of postsynaptic γ2 subunit and gephyrin decreased significantly in homozygous R215H KI mice (Figure 2b-e), consistent with previous findings in NL2 KO mice (Babaev et al., 2016; Gibson et al., 2009; Jedlicka et al., 2010; Poulopoulos et al., 2009). In addition to postsynaptic changes, we also observed a significant reduction of presynaptic PV and vGAT puncta (vesicular GABA transporter) in the hippocampal region of KI mice (Figure 3, Figure S3). The number of PV neurons was not changed in the KI mice (Figure 3b-d). However, both PV and vGAT puncta number and size were significantly reduced in the dentate granule cells (Figure 3e-i), as well as in the CA1/CA3 pyramidal cells in KI mice (Figure S3a-j). These results suggest that NL2 R215H mutation impaired both pre- and post-synaptic GABAergic components.
Impaired GABAergic neurotransmission in NL2 R215H KI mice
We next investigated the function of inhibitory neurotransmission in the R215H KI mice. Whole-cell patch-clamp recordings were performed on dentate granule cells in acute brain slices of adult WT and homozygous R215H KI mice. We found that both the frequency and amplitude of miniature inhibitory postsynaptic currents (mIPSCs) were significantly decreased in the granule cells of R215H KI mice (Figure 4a-d; Frequency: WT = 8.28 ± 2.21 Hz, KI = 3.98 ± 0.78 Hz, p = 0.041; Median amplitude: WT = 41.3 ± 2.9 pA, KI = 32.7 ± 1.8 pA, p = 0.019; Student’s t-test). In contrast, there was no significant change of miniature excitatory postsynaptic currents (mEPSCs) in the dentate granule cells of R215H KI mice compared to WT mice (Figure 4e-h), indicating that NL2 R215H mutation is primarily affecting inhibitory neurotransmission.
Behavioral deficits in NL2 R215H KI mice
The significant reduction of inhibitory neurotransmission in the NL2 R215H mutant mice prompted us to further investigate whether such severe GABAergic deficits will result in any behavioral deficits. We first performed open field test and elevated plus maze test to measure the mouse anxiety level. In the open field test, we found that the KI mice spent significantly less time in the center region, although the total distance traveled was similar to the WT mice (Figure 5a-d). Consistently, in the elevated plus maze test, the KI mice spent much less time in the open arm compared to the WT mice, while the total travel distance was also similar between the KI and WT mice (Figure 5e-h). These results suggest that the R215H KI mice display an increased level of anxiety while their locomotion activity is relatively normal.
We next examined in R215H KI mice the acoustic startle response and pre-pulse inhibition, a standard test for the sensory motor gating function often assessed in schizophrenia patients (Braff et al., 1992). R215H KI mice showed a significant reduction in the startle response when stimulated at 100 – 120 dB (Figure 5i). Furthermore, the pre-pulse inhibition was significantly impaired in the KI mice compared to the WT mice (Figure 5j). Together, these deficits of R215H KI mice suggest that this new transgenic mouse model may have symptoms of schizophrenia-like behavior.
To further characterize the R215H KI mice, we investigated their cognitive functions by conducting contextual fear conditioning test, a hippocampal dependent fear-learning test. We found that while the KI mice were capable to associate the conditioning chamber with foot-shock in the initial training, indicated by an increase of freezing state after foot-shock, they failed to retain the fear context memory in the following days when tested (Figure 5k), indicating an impaired cognition function. Importantly, the behavioral data shown above was all obtained from male mice, the female mice were also tested and exhibited the same trend (Figure S4).
Impaired hippocampal activation toward acute stress in NL2 R215H KI mice
Schizophrenia is associated with abnormal response to stress (Walker and Diforio, 1997). Stress is known to activate the hypothalamic-pituitary-adrenal axis (HPA axis) and induce the hormone release of corticosterone (CORT) into circulation (Koob, 1999; McGill et al., 2006). To investigate the stress response of R215H KI mice, we put the WT and R215H KI mice into restraining tubes for one hour as an acute stress test. We found that R215H KI mice struggled much more intensively for a long time and excreted much more than the WT mice during the restraining test. After restraining, KI mice were more dirty and stinky than the WT mice (Figure 6a). In accordance, R215H KI mice showed a much higher level of CORT (384 ± 53 ng/ml) after restraining compared to the WT mice (215 ± 20 ng/ml). The baseline level of CORT was similar between WT (49 ± 4 ng/ml) and KI mice (36 ± 4 ng/ml) (Figure 6b; p = 0.0035 after restraint, Two way ANOVA followed with Sidak’s post hoc test). These results suggest that R215H KI mice have hyperactive HPA response toward stress.
Following the activation of HPA axis, hippocampus will be activated as a negative feedback regulator and control the CORT level within normal range (Herman et al., 2012; Ulrich-Lai and Herman, 2009). To test whether hippocampus in R215H KI mice were activated following the acute stress, we used a naïve cohort of mice to perform the restraining test again. R215H KI and WT mice were subjected to restraint for half an hour and then sacrificed after 2 hours. Hippocampal activation was examined by assessing the expression level of an immediate early gene cFos (Morgan et al., 1987; Ramirez et al., 2013). At the baseline level, very few cFos-positive neurons were detected in the hippocampal regions in both WT and KI mice (Figure 6c, top row). After stress, we observed a significant increase of cFos-positive cells in the DG and CA2/3 regions of the hippocampus in WT mice (Figure 6c, bottom left). In contrast, the R215H KI mice showed much reduced cFos-positive cells in the same regions of hippocampus (Figure 6c, bottom right). This is better illustrated in the enlarged images showing the CA2/3 and DG regions of WT mice (Figure 6d, top row) and KI mice (Figure 6d, bottom row). Quantitative analysis confirmed the reduction of cFos-positive cells in both CA2/3 (Figure 6e) and DG (Figure 6f) regions in the KI mice. These results suggest that NL2 R215H KI mice had impaired hippocampal activation during acute stress.
DISCUSSION
In the present study, we generated a unique mouse model carrying a single point mutation R215H of NLGN2 gene that was originally identified from human schizophrenia patients. The NL2 R215H KI mice have impaired GABAergic synapse development, reduced inhibitory synaptic transmission, and decreased hippocampal activation in response to stress. Moreover, the R215H KI mice display anxiety-like behavior, impaired pre-pulse inhibition, cognitive deficits and abnormal stress response, partially recapitulating some of the core symptoms of schizophrenia patients. These results suggest that this newly generated R215H KI mouse line may provide a unique animal model for studying molecular mechanisms underlying schizophrenia and related neuropsychiatric disorders.
GABAergic and behavioral deficits in NL2 R215H KI mice
NL2 plays important roles in regulating perisomatic GABAergic synapse development, phasic GABAergic transmission, and neural excitability (Babaev et al., 2016; Blundell et al., 2009; Chubykin et al., 2007; Gibson et al., 2009; Hines et al., 2008; Hoon et al., 2009; Jedlicka et al., 2010; Liang et al., 2015; Poulopoulos et al., 2009; Varoqueaux et al., 2004; Wohr et al., 2013). Consistent with our previous in vitro studies, the current in vivo work demonstrates that R215H mutation disrupts GABAergic synapse development. Functionally, NL2 R215H mutation caused a reduction of both frequency and amplitude of inhibitory neurotransmission. These results suggest that the R215H KI mice display more severe GABAergic deficits than the NL2 KO mice (Babaev et al., 2016; Chubykin et al., 2007; Gibson et al., 2009; Jedlicka et al., 2010; Poulopoulos et al., 2009). The more severe GABAergic deficits in our R215H KI mice might explain why they display more severe behavioral deficits than NL2 KO mice, such as PPI impairment, cognitive deficits, and abnormal stress response. Coincidentally, previous studies reported that NL3 R451C KI mouse also displayed stronger phenotypes than the NL3 KO mice (Etherton et al., 2011; Foldy et al., 2013; Tabuchi et al., 2007; Zhang et al., 2016). These evidences suggest that genetic mouse models based on mutations identified from patients may be more suitable than the germline KO mouse models for studying pathologic mechanisms of human diseases.
Behaviorally, NL2 R215H KI mice display an anxiety phenotype, which may be the result of decreased GABAergic inhibition (Blundell et al., 2009; Dalvi and Rodgers, 1996; Zarrindast et al., 2001). Interestingly, R215H KI mice also show impaired startle responses and deficits in pre-pulse inhibition (PPI). Previous study in rats has reported that disturbance of PV neuron development in the hippocampal DG region may cause reduction of PPI (Guo et al., 2013). A recent study also demonstrates that specific inhibition of PV neurons in the ventral hippocampus results in a reduction of both startle response and PPI (Nguyen et al., 2014). Consistent with these findings, we demonstrate here that our R215H KI mice display a significant reduction of PV innervation in the hippocampus, which may underlie the deficits of PPI. In contrast, the NL2 KO mice lack PPI deficit, which might be related to an insufficient loss of PV innervation at hippocampal regions (Wohr et al., 2013).
Another interesting observation is that the R215H KI mice are hyperactive after acute stress and are associated with impaired hippocampal activation. It has been reported that robust neuron activation requires low background activity before stimulus (Koistinaho et al., 1993; Rao et al., 2006). However, due to the reduction of GABAergic inhibition in our R215H KI mice, the baseline activity of hippocampal neurons may be elevated and thus a further activation of hippocampus by external stimulation will be dampened. The impaired activation of hippocampal neurons in R215H KI mice may contribute to the abnormal stress response we observed, as hippocampus acts like a “brake” during acute stress to prevent HPA axis from over activation (Hariri, 2015).
NL2 R215H mutation and schizophrenia
It is well documented that schizophrenia patients show impaired pre-pulse inhibition as an abnormal sensorimotor gating deficit (Braff et al., 1992; Grillon et al., 1992). Many patients also have emotional symptoms such as anxiety and depression (Lewis and Lieberman, 2000). Additionally, patients are hypersensitive toward stress and certain patients have been found with altered HPA axis function (Bradley and Dinan, 2010). Intriguingly, R215H KI mice recapitulated these SCZ-like behaviors, suggesting a potential role of NL2 R215H in the development of schizophrenia symptoms. Furthermore, reduction of PV expression and PV-positive synapses is a prominent phenotype observed in SCZ patients (Lewis et al., 2001; Lewis et al., 2012; Lewis et al., 2005; Woo et al., 1998). The R215H mutation KI mice also show a significant reduction of PV innervation, consistent with the pathogenic deficit of SCZ patients. These GABAergic deficits, together with cognition and PPI deficits manifested in the KI mice, support the hypothesis that GABA dysfunction makes an important contribution to the cognitive and attention deficits of SCZ. Taken together, NLGN2 R215H single point mutation has a significant impact on GABAergic synapse development and the pathogenesis of neuropsychiatric disorders. Our newly generated NL2 R215H KI mice may provide a useful mouse model for the study of molecular mechanisms and drug development of neuropsychiatric disorders including schizophrenia.
Materials and Methods
NL2 R215H knock-in mice
The NL2 R215H knock-in mice were generated by homologous recombination in embryonic stem cells by Dr. Siu-Pok Yee’s team at the University of Connecticut Health Center. The detailed procedures are described in the SI materials and methods.
All the experimental mice were group housed (2-3 mice per cage) in home cages and lived at a constant 25 °C in a 12h light/dark cycle. Mice were given ad libitum access to food and water. Littermate or age and gender matched mice were used for experiments. All animal care and experiments followed the Penn State University IACUC protocol and NIH guidelines.
Biochemical measurements
Protein levels were quantified using total brain homogenates from 3 groups of adult male littermates-WT, heterozygous and homozygous. The western blot system used was the standard Bio-Rad mini protein electrophoresis system and the procedure followed the system manual. LiCOR Odyssey Clx was used for protein signal detection. The antibodies used were Rb anti-Neuroligin 2 (1:1000, SYSY 129202), Rb anti-GAPDH (1:10000, Sigma G9545), and Gt anti-Rb 800 (1:15000, P/N 925-32210, P/N 925-32211). Detailed procedures are described in the SI materials and methods.
Immunohistochemistry, Image Acquisition, and Image Analysis
Mouse brain slices were prepared at 20-40 μM and reacted with the primary antibodies Rb anti-Neuroligin 2 (1:1000, SYSY129203), Ms anti-Parvalbumin (1:1000, MAB1572), GP anti-vGAT (1:1000, SYSY 131004), Gephyrin (1:1000, SYSY 147011), GABAaR γ2 (1:1000 SYSY 224003), and c-Fos (1:5000 Sigma F7799). The fluorescent secondary antibodies used were Gt anti-Rb 488, Gt anti-Ms Cy3, and Gt anti-GP 647. Images were taken with the Olympus FV1000 confocal microscope. The number of neurons and the density and size of synaptic puncta were analyzed with the NIH ImageJ software (NIH, Bethesda, MD, USA). A detailed description of the experimental procedures is in the SI materials and methods.
Slice electrophysiology
Horizontal acute hippocampal slices were used for whole-cell patch clamp recordings. Miniature inhibitory or excitatory postsynaptic currents (mIPSCs or mEPSCs) were pharmacologically isolated by including DNQX and APV or picrotoxin together with tetrodotoxin in artificial cerebrospinal fluid. Details are in the SI materials and methods.
Behavioral tests
Overview: The mice for behavior tests were group housed by genotype. All tests were performed during 1 pm to 6 pm. Three cohorts of mice were used: First cohort of mice was tested for open field and elevated plus maze at 2 to 3 month and tested for the startle response and prepulse inhibition test at 3.5 months. Second cohort of mice was used for contextual fear conditioning test at 2-3 month. Third cohort of mice was used for restraining and corticosteroid serum level test at 4 to 6 months. The open field test and elevated maze data were analyzed by Noldus Ethovision XT 8.0 software. PPI test and CORT test were analyzed with the researcher blind to genotype. Detailed procedures are in SI materials and methods.
Conflict of Interest
We declare no conflict of interest.
Acknowledgements
We would like to thank Dr. Thomas Fuchs for providing advices on behavioral tests, Yuting Bai for providing initial genotyping support. We thank all members from Chen lab for thoughtful suggestions. This study is supported by grants from NIH (MH092740 and MH083911) and Charles H. “Skip” Smith Brain Repair Endowment Fund to G.C.