Mice with reduced NMDA receptor expression: more consistent with autism than schizophrenia?


Michael J. Gandal, PhD, Department of Psychiatry, University of Pennsylvania, Translational Research Laboratories, 125 S. 31st Street, Philadelphia, PA 19104, USA. E-mail: mgandal@mail.med.upenn.edu


Reduced NMDA-receptor (NMDAR) function has been implicated in the pathophysiology of neuropsychiatric disease, most strongly in schizophrenia but also recently in autism spectrum disorders (ASD). To determine the direct contribution of NMDAR dysfunction to disease phenotypes, a mouse model with constitutively reduced expression of the obligatory NR1 subunit has been developed and extensively investigated. Adult NR1neo−/− mice show multiple abnormal behaviors, including reduced social interactions, locomotor hyperactivity, self-injury, deficits in prepulse inhibition (PPI) and sensory hypersensitivity, among others. Whereas such phenotypes have largely been interpreted in the context of schizophrenia, these behavioral abnormalities are rather non-specific and are frequently present across models of diseases characterized by negative symptom domains. This study investigated auditory electrophysiological and behavioral paradigms relevant to autism, to determine whether NMDAR hypofunction may be more consistent with adult ASD-like phenotypes. Indeed, transgenic mice showed behavioral deficits relevant to all core ASD symptoms, including decreased social interactions, altered ultrasonic vocalizations and increased repetitive behaviors. NMDAR disruption recapitulated clinical endophenotypes including reduced PPI, auditory-evoked response N1 latency delay and reduced gamma synchrony. Auditory electrophysiological abnormalities more closely resembled those seen in clinical studies of autism than schizophrenia. These results suggest that NMDAR hypofunction may be associated with a continuum of neuropsychiatric diseases, including schizophrenia and autism. Neural synchrony abnormalities suggest an imbalance of glutamatergic and GABAergic coupling and may provide a target, along with behavioral phenotypes, for preclinical screening of novel therapeutics.

Reduced NMDA-receptor (NMDAR) mediated glutamatergic signaling has been strongly linked to the pathophysiology of schizophrenia (Tsai & Coyle, 2002). A number of animal models have been created to investigate this connection, including acute administration of NMDAR antagonists (Jentsch & Roth, 1999) and NR1neo−/− mice, which were generated to have constitutive ∼85% downregulation of the obligate NMDA-NR1 subunit (Mohn et al., 1999). NR1neo−/− mice have extensive behavioral abnormalities, including reduced social interactions and mating, locomotor hyperactivity, deficits in sensorimotor gating, self-injury and cognitive inflexibility (Duncan et al., 2004, Duncan et al., 2006, Dzirasa et al., 2009, Mohn et al., 1999, Moy et al., 2008). Although these results have been interpreted in the context of schizophrenia, such behavioral deficits are largely non-specific and the specificity of these findings for schizophrenia has been recently questioned (Barkus et al., 2012). To more specifically examine endophenotypic deficits in this model, we previously investigated auditory and visual event-related potentials (ERPs), which show a complex pattern of disruption in schizophrenia (Amann et al., 2010, Bodarky et al., 2009, Halene et al., 2009a). However, NMDA-NR1 deficiency caused significant sensory disinhibition, a pattern not closely resembling that seen in the clinical population. Such findings suggested that NR1neo−/− mice might be a more robust model for disorders characterized by negative symptoms and sensory hypersensitivity, such as autism (Tannan et al., 2008).

Emerging evidence has indicated that disrupted NMDAR signaling may contribute to the pathogenesis of idiopathic autism (Carlson, 2012, Carlsson, 1998, Gai et al., 2011, Moskal et al., 2011, Myers et al., 2011, O'roak et al., 2011) and related clinical syndromes, including Rett and Fragile X Syndrome (Krueger et al., 2011, Maliszewska-Cyna et al., 2010, Yun & Trommer, 2011). Autism is a highly disabling neurodevelopmental disorder defined by a triad of core symptoms, including impairments in reciprocal social interactions, deficits in expressive and receptive language function and repetitive or restricted interests and behaviors. In addition, autism is associated with a number of comorbid symptoms and disorders, including cognitive disability, attentional problems, disrupted sensory processing, obsessive-compulsive disorder, self-injury, epilepsy and sleep dysfunction (Levy et al., 2009). Auditory processing is disrupted in autism, with a delayed profile and reduced gamma synchrony (Gandal et al., 2010) – a pattern distinct from schizophrenia.

This study investigated the relationship between NMDAR downregulation and autism spectrum disorders (ASD)-like phenotypes in adult mice, to more rigorously establish a connection between NMDAR-dysfunction and disease pathophysiology. Using established behavioral paradigms, we show that NR1neo−/− mice have deficits in all core domains of autism, including sociability, communication and repetitive behaviors (Silverman et al., 2010). Likewise, these mice recapitulate electrophysiological endophenotypes seen in autism more closely than in schizophrenia. Finally, these findings suggest that NMDAR dysregulation could contribute to core ASD symptoms and clinical biomarkers.

Materials and methods


The NR1-transgenic mice were bred and genotyped in house, as previously published (Bodarky et al., 2009, Halene et al., 2009a, Mohn et al., 1999). Mice were maintained on a first-generation (F1) C57BL/6J × 129S6/SvEvTac hybrid background. The line was maintained on the two separate backgrounds (B6 and 129) and was crossed to generate F1 test subjects. Only F1 subjects were used in this study (not backcrossed). Only adult male NR1-transgenic mice and wildtype (WT) littermates were used in this study. Three separate sets of mice were tested. The first group (age 5.4 ± 1.3 months) was assessed for male–female social interactions and vocalizations, then tail suspension and finally 5-day locomotor activity (LMA). The second group (age 4.1 ± 0.1 months) was measured for auditory startle followed by auditory electrophysiology. The final group (age 10.1 ± 0.3 months) was assessed for male–male social interactions followed by T-maze performance. Tests in the same mice were separated by at least 6 days. Animals were housed four to five per cage on a 12-h light/dark cycle in a temperature-controlled facility with food and water available ad libitum. After electrode implantation surgery, mice were housed individually. All testing, with the exception of 5-day LMA, was performed exclusively in the light cycle. All protocols were approved by the University of Pennsylvania Institutional Animal Care and Use Committees and were conducted in accordance with National Institutes of Health guidelines. Throughout the study, all possible effort was taken to minimize pain and discomfort in experimental subjects.

Social approach/avoidance paradigm

Sociability was measured in NR1 (n= 7) and WT mice (n= 8) using a standard social choice task as previously published (Gandal et al., 2010, Halene et al., 2009a). Briefly, test mice were placed in a large arena (50 × 35 × 40 cm) covered with 2 cm of bedding in a dimly lit room. In phase 1 of the experiment, two identical Plexiglas cylinders were placed at opposite ends of the chamber. The test mouse was allowed to freely explore the arena and cylinders for 10 min, tracked by an overhead camera interfaced with TopScan software (Cleversys, Inc, Reston, VA, USA). In phase 2, an unfamiliar WT A/J stimulus mouse was placed under one of the two cylinders. The test mouse was reintroduced to the arena for an additional 10 min. The cylinders had multiple holes (1 cm in diameter) to allow for auditory, visual and olfactory investigation between mice. Equipment was thoroughly cleaned between tests, and the ‘social cylinder’ was counterbalanced among mice. NR1neo−/− and WT mice were assessed for the total time spent sniffing social and non-social chambers. Sniffing time was measured offline by a single-blinded experimenter using two stopwatches and was defined as the time a mouse spent with its nose oriented toward the cylinder at a distance of less than 1 cm. Significance was assessed with a group × condition repeated measures anova, followed by Bonferroni posttests where appropriate.

Male–female social interaction and vocalizations

Mice were well handled and acclimated to the testing room for 1 h in low light. A single male mouse (WT or NR1neo−/− in alternating order, n= 8/group) was then removed from its home cage and placed into a large, clean testing chamber (46 × 24 × 20 cm) with no bedding. An ultrasonic microphone (Pettersson Elektronik AB, Uppsala, Sweden; Model D940) was placed approximately 10 cm above the setup. The microphone is sensitive to frequencies 10–200 kHz and was interfaced with Spike2 software through Power1401 hardware at a 250 kHz sampling rate (CED; Cambridge, UK). Male mice were allowed to acclimate to the testing chamber for 2 min, during that time no ultrasonic vocalizations (USVs) were detected. Next, an unfamiliar receptive WT female mouse was placed in the cage. Social interactions initiated by the male were timed with a silent stopwatch and USVs emitted by the male were recorded during a 5-min testing session. At no point was any mounting behavior observed. Mice were then placed in a holding cage until all males from the same home cage had been tested. The testing chamber was cleaned with 70% ethanol and allowed to dry for at least 5 min between sessions. Sonograms were created in Spike2 using the fast Fourier Transform (FFT; length 1024 points, Hanning window). Vocalizations were hand counted by a single, blinded experimenter offline. Peak frequency was measured by FFT (2048 point, 0.1 kHz resolution) for 20 randomly sampled calls per animal. Significance was assessed by unpaired t-tests.

Modified tail suspension test

A modified tail suspension task was performed on adult mice (n= 10–11/group). Mice were held upside down 20 cm from a clean table surface. Mice were videotaped for 2 min and measured for repetitive, forepaw handwringing behavior. Handwringing was defined as repeated clasping of forepaws while shaking them up and down in front of the face (see Video S1 & S2). Wildtype mice generally spread their forepaws during tail suspension. Again, a single-blinded experimenter rated the videos as to whether handwringing behavior was present at any time during the 2-min testing session. Significance was assessed using a chi-square test.

Auditory electrophysiology

Animals (n= 8/group) were anesthetized with isoflurane and underwent stereotaxic implantation of tripolar electrode assemblies (PlasticsOne, Roanoke, VA, USA) for non-anesthetized recording of auditory event-related potentials (ERPs), according to previously published methods (Ehrlichman et al., 2009a, Gandal et al., 2008, Halene et al., 2009a, Lazarewicz et al., 2010). A low-impedance stainless steel macroelectrode (<5 kΩ, 1000 Hz) was placed in the right CA3 hippocampal region (1.8 mm posterior, 2.65 mm right lateral and 2.75 mm deep) referenced to the ipsilateral frontal sinus. The pedestal was secured to the skull with cyanoacrylate and dental cement. This low-impedance, differential recording configuration captures both early and late components of the auditory-evoked potential, including the acoustic brainstem response, mid-latency P1 (e.g. human P50/M50) and N1 (e.g. human N100/M100), as well as the late P2 and P3a peaks, as published (Connolly et al., 2003, Connolly et al., 2004, Siegel et al., 2003) with strong analogy to human scalp electroencephalogram (EEG) (Gandal et al., 2008, Gandal et al., 2010).

Recording of auditory ERPs was performed at least 1 week after electrode implantation in a home cage environment, as previously described (Ehrlichman et al., 2009a, Gandal et al., 2008, Halene et al., 2009a, Lazarewicz et al., 2010). Cages were placed in a sound attenuated recording chamber inside a Faraday electrical isolation cage. Electrode cables were connected to a high-impedance differential AC amplifier (A-M Systems; Seattle, WA, USA). Stimuli were generated by Micro1401 hardware with Spike2 software (CED; Cambridge, UK) and were delivered through speakers attached to the cage top. The sequencer file consisted of a series of 1000 white-noise bursts presented at 80 dB sound pressure level (SPL) (10 milliseconds in duration) with a 1-second interstimulus (ISI) interval. Recording sessions were preceded by a 15-min acclimation phase. Raw EEG was filtered between 1 and 500 Hz. Individual sweeps were rejected for movement artifact based on a criterion of two times the root mean squared amplitude per mouse.

Grand-average waveforms were created for each mouse and N1 latency was measured as the local minimum with zero slope between 20 and 60 milliseconds. Waveforms that did not meet these criteria were not included (one mouse).Time-frequency decomposition of EEG signal was performed with the EEGLab toolbox in Matlab (Delorme & Makeig, 2004). Single trial epochs between −0.3 and 0.8 seconds relative were extracted from the continuous data sampled at 1667 Hz. Phase-locking (e.g. intertrial coherence) was calculated using Morlet wavelets in 100 linearly spaced frequency bins between 5 and 100 Hz, with wavelet cycles linearly increasing from 3 to 6, as published (Gandal et al., 2010). For each subject, phase-locking was averaged over the peak response from 0 to 100 milliseconds post-stimulus in the gamma frequency range (30–80 Hz).

Delayed alternating T-maze paradigm

Spatial cognition was measured using a discrete trial, spontaneous alternation T-maze task with a 5-second delay, according to published protocols (Deacon & Rawlins, 2006). The maze consisted of three identical arms (30 cm long, 10 cm wide and 12 cm tall), with a central partition and guillotine doors for each arm. Wildtype mice have a natural tendency to alternate when placed in such a chamber, achieving spontaneous alternation rates up to 90% (Deacon & Rawlins, 2006). Mice were well handled and acclimated to the room and apparatus prior to testing, which occurred in low-lux red light to minimize anxiety. To begin the testing session, mice were placed in the central arm of the partition with the guillotine door closed. The door was then opened and mice were allowed to choose one of two side arms. Once a mouse had entered the side arm, the guillotine door was closed and mice were confined to that arm for 30 seconds. Mice were then returned to the central partition, which was closed off for 5 seconds (delay period), the guillotine door was opened and mice were allowed to make a second choice of goal arms. Mice (n= 7–9/group) were tested for a total of 10 trials.

Prepulse inhibition

Prepulse inhibition (PPI) of acoustic startle was assessed in mice (n= 8/group) using a four-chamber system (San Diego Instruments, San Diego, CA, USA) as published (Gandal et al., 2010). Briefly, 10 trials of startle responses were measured to 40-millisecond pulses at 0 and 90–120 dB sound pressure (5 dB increments), randomly presented five times each with an ISI interval from 10 to 20 seconds. Five total prepulse trials consisted of a 20-millisecond pulse at 69, 73 or 81 dB followed by a 40-millisecond pulse of 120 dB 100 milliseconds later. Prepulse inhibition was calculated as the percent difference in startle units following the prepulse/startle pair as compared to the startle tone alone. Significance was assessed with a group × intensity repeated measures anova.

Locomotor activity

The effect of constitutive NMDAR hypofunction was assessed on LMA, as previously published (Ehrlichman et al., 2009b). Mice (n= 4–5/group) were allowed to acclimate to the testing room for 30+ min on two consecutive days prior to testing. After acclimation, mice were placed in a new home cage environment with clean bedding inside an automated LMA photobeam frame (30 × 24 × 8 cm) with sensors arranged in an eight-beam array strip with 3.175 cm spacing (MED Associates, St. Albans, VT, USA). For 120 h, LMA was recorded in 120-min bins by a computer using med pc software set to detect the number of times that horizontal light beams were broken. The testing room was set to continue the standard 12-h light/dark cycle. Data were analyzed with a group × time-period repeated measures anova.

Statistical analysis

Significance was assessed using the statistical tests as described above(t-test, chi-square or anova) with GraphPad Prism software (v5, La Jolla, CA, USA). Normality was assessed with a Shapiro–Wilkes test and non-normally distributed data was log-transformed prior to analysis. Significance was set at P < 0.05 for all measures.


Behavioral phenotypes relevant to the core symptoms of autism

Social deficits

Social function was assessed in two separate paradigms. In a mating paradigm, transgenic male mice and WT littermates were paired with an unfamiliar, WT female. In this task, NR1neo−/− mice initiated significantly less social contact, as measured by total interaction time (Fig. 1; t14= 6.03, P < 0.0001). In addition, sociability to a same-sex, unfamiliar WT mouse was measured with a standard three-chambered social approach/avoidance paradigm. In this task, NR1neo−/− mice again showed significantly reduced sociability (Fig. 1; group × condition interaction: F3,68= 4.667, P= 0.004; Bonferroni posttest P < 0.001). These results are consistent with previously published studies showing social deficits in mice with reduced NMDAR signaling (Duncan et al., 2004, Halene et al., 2009a, Mohn et al., 1999).

Figure 1.

Constitutive NMDAR downregulation impairs sociability in mice. (a) Adult male WT or NR1neo−/− mice were paired with a receptive, WT female for 5 min. Mating vocalizations, emitted by the male mouse, were recorded with an ultrasonic-range microphone placed above the testing chamber. (b) Transgenic mice showed significantly reduced interest in the female mouse, as measured by the total time of male initiated interactions. (c) Sociability to a gonadectomized, same-sex WT stimulus mouse was measured using a three-chambered approach/avoidance paradigm. Adult NR1-transgenic mice showed significantly reduced social interest in this task as well. Plots show mean ± SEM (***P< 0.001).

Communication impairments

Mice emit USVs in several contexts that have communicative significance (Panksepp et al., 2007, Scattoni et al., 2009). When paired with a receptive female, male mice will reliably emit 70 kHz ‘pre-mating’ vocalizations prior to mount followed by 40 kHz ‘mating’ calls, thought to facilitate copulation (Ricceri et al., 2007, White et al., 1998). Such pre-mating USVs were recorded during the male–female social interaction described above. Over the 5-min testing session, NR1neo−/− mice emitted significantly fewer vocalizations compared with WT littermates (Fig. 2; t14= 2.271, P < 0.0001). Interestingly, the number of calls emitted by WT males was highly correlated with the amount of social interaction time, indicating the social significance of USVs (Fig. 2C; R2= 0.8, P < 0.003). This association did not hold for NR1-transgenic mice, indicating a lack of appropriate social communication function (R2= 0.003, P= 0.9). In addition, the peak frequency of calls emitted by NR1neo/− mice was significantly lower than that of the WT group (NR1: 40.0 ± 5.4 kHz, WT: 73.5 ± 5.5 kHz, t9= 4.06, P= 0.003). These findings indicate that transgenic mice fail to emit vocalizations in socially appropriate contexts.

Figure 2.

GRIN1 mutant mice demonstrate abnormalities in social communication, as measured by USVs. (a) Male mice emit 70 kHz ‘pre-mating’ vocalizations and 40 kHz ‘mating’ USVs when paired with a receptive female. This sonogram shows four pre-mating calls, recorded from a WT mouse. (b) NR1neo−/− emit significantly fewer ultrasonic pre-mating calls than do WT littermates. (c) In WT mice, there was a significant, linear association(on a log–log plot) between the number of pre-mating vocalizations emitted and male initiated social interactions, highlighting the relevance of such ultrasonic calls for murine social communication. This relationship was absent for NR1neo−/− mice. Plots show mean ± SEM (**P< 0.01, ***P< 0.001).

Repetitive, restricted interests and behaviors

The third core symptom of autism is characterized by motor stereotypies, repetitive or restricted interests, self-injury and resistance to change. Stereotyped hand movements (e.g. hand-flapping, ringing or rocking) are frequently observed in patients with autism and Rett Syndrome (RTT), and related phenotypes of abnormal forepaw clasping have been reported in mouse models of RTT (Carter et al., 2010, Chao et al., 2010, Pierce & Courchesne, 2001). During tail suspension, NR1-transgenic mice exhibited a similar forepaw clasping/wringing behavior at a significantly greater rate than WT littermates (Fig. 3A; χ2= 17.33, P < 0.0001). Likewise, NR1neo−/− mice were found to have significantly more skin lesions caused by self-injurious, over-grooming behavior (Fig. 3B; χ2= 5.193, P < 0.02). Along similar lines, previous studies have identified cognitive inflexibility in NR1neo−/− mice, as measured by repetitive nose poke responses in a novel hole board paradigm (Moy et al., 2008) and perseverative errors in an eight-arm radial maze task (Dzirasa et al., 2009).

Figure 3.

NMDAR hypofunction leads to increased repetitive behaviors and self-injury. (a) During a tail suspension test, 90% of NR1 mutant mice show a repetitive, handwringing stereotypy that was only observed in 10% of WT littermates (see Video S1 & S2). (b) Approximately 25% of NR1neo−/− mice developed skin lesions due to self-injurious over-grooming behavior. Such lesions were not observed in any WT littermates. Plots show mean ± SEM (*P< 0.05, ***P< 0.001).

Behavioral phenotypes relevant to ASD comorbidities

Cognitive deficits

Intellectual disability is a common comorbidity associated with ASD (Levy et al., 2009). We assessed cognitive function using a discrete T-maze paradigm with a 5-second delay, which measures spatial working memory (Deacon & Rawlins, 2006). NR1-transgenic mice showed significantly reduced alternation, consistent with a deficit in cognitive function (WT: 85 ± 6%, n= 9, NR1: 64 ± 5%, n= 7; t14= 2.36, P < 0.05). Similar results were observed in an eight-arm radial maze task (Dzirasa et al., 2009).

Abnormal activity patterns

Clinical studies have reported various activity level disturbances in subjects with autism using actigraphs and polysomnography (Miano et al., 2007, Souders et al., 2009). In our study, we monitored the activity patterns inNR1neo−/− mice across the light–dark cycle over a 120-h period. A main effect of genotype (Fig. 4; F1,7= 6.52, P < 0.04) indicated increased LMA in NR1-transgenic mice, an effect that was most apparent during the light (inactive) cycle. Elevated LMA has been reported in other mouse models of autism (Penagarikano et al., 2011) and is frequently observed in patients with autism, along with other symptoms of attention-deficit hyperactivity disorder (Lee & Ousley, 2006).

Figure 4.

NR1 mutant mice show increased LMA during light cycle. Group averages of activity are shown over 120-h testing session.

Endophenotypes relevant to autism

Auditory startle

Autism is characterized by hypersensitivity to auditory stimuli coupled with reduced PPI of the acoustic startle response, an endophenotype which has been correlated with core symptoms of the disorder (Levy et al., 2009, Mcalonan et al., 2002, Perry et al., 2007, Yuhas et al., 2011). NR1neo−/− mice showed enhanced startle amplitudes, indicative of increased sensory-motor reactivity (Fig. 5A; main effect of group, F1,14= 5.2, P < 0.04). Conversely, transgenic mice exhibited a significant reduction in PPI (Fig. 5B; F1,14= 8.1; P < 0.02), recapitulating clinical observations.

Figure 5.

NMDAR disruption recapitulates several auditory endophenotypes observed in autism. (a) Transgenic mice show elevated acoustic startle responses and (b) reduced PPI. (c) Developmental NMDAR downregulation leads to delayed latency of the N1 component of the auditory-evoked response, as observed in children with autism. (d) Grand-average auditory-evoked response waveforms showing P1 and N1 peaks. (e, f) Transient auditory-evoked phase-locking (PLF; e.g. intertrial coherence) was assessed in the gamma frequency range (30–80 Hz). NR1neo−/− mice showed significantly reduced gamma synchrony, mirroring results from the clinical population. Plots show mean ± SEM (*P< 0.05).

Auditory-evoked responses

Clinical studies measuring auditory ERPs have shown that early auditory encoding processes are abnormal in autism (Jeste & Nelson, 2009, Roberts et al., 2008). Using pure-tone or click paradigms, recent work has reported delayed profiles of auditory processing in subjects with autism (Bruneau et al., 1999, Gage et al., 2003, Oram Cardy et al., 2005b, Roberts et al., 2010, Sokhadze et al., 2009) as well as in unaffected relatives to a lesser degree (Maziade et al., 2000). Such findings have been shown to be predictive of communication deficits in ASD (Oram Cardy et al., 2005a, Oram Cardy et al., 2008, Roberts et al., 2011). NR1neo−/− mice were evaluated in an analogous auditory-evoked response paradigm. Auditory-evoked responses are highly analogous between mice and humans, with corresponding P1 (mouse P20, human P50) and N1 (mouse N40, human N100) peaks (Amann et al., 2010). Transgenic mice showed a significant, 10–15% delay in the latency of the N1 component of the auditory-evoked response (Fig. 5C; t14= 2.2, P < 0.05), identical to that observed in clinical studies of autistic children (Gandal et al., 2010, Roberts et al., 2010). The amplitudes of P1 and N1 peaks did not differ between groups, consistent with the clinical phenotype in ASD (Lincoln et al., 1995, Roberts et al., 2010).

Gamma synchrony

Recent work has shown abnormal neural synchrony in autism, which may contribute to core phenotypes (Uhlhaas & Singer, 2006, Uhlhaas & Singer, 2007, Welsh et al., 2005). In particular, high-frequency, gamma (30–80 Hz) oscillations have been shown to mediate a host of cognitive and sensory functions, including sensory encoding, selective attention and memory (Herrmann et al., 2010). Several studies have identified reduced transient auditory-evoked gamma-band responses in autism, thought to reflect abnormal cognitive and perceptual functioning (Braeutigam et al., 2008, Gandal et al., 2010, Rojas et al., 2008, Wilson et al., 2007). Likewise, the phase-locked early auditory gamma-band response was significantly reduced in NR1neo−/− mice (Fig. 5; t14= 2.28, P < 0.05), recapitulating the clinical endophenotype. Such gamma frequency abnormalities are in accordance with previous reports of disrupted selective attention and gamma-theta phase coherence following NMDAR disruption (Bickel et al., 2008, Dzirasa et al., 2009).


Model validity

This study has shown that constitutive NMDAR hypofunction in mice is associated with deficits in social and communicative functioning as well as elevated repetitive and self-injurious behaviors. This pattern of behavioral dysfunction is consistent with the negative symptoms of schizophrenia as well as autism-like phenotypes. Results in the social interaction paradigms complement and replicate previously published findings (Duncan et al., 2004, Halene et al., 2009a, Mohn et al., 1999). The strong reduction of mating vocalizations fits with reports of sexual dysfunction in NR1neo−/− mice (Mohn et al., 1999) and levels of self-injury were highly consistent with previous work (Moy et al., 2008). In addition, transgenic mice showed several comorbid symptoms and endophenotypes relevant to autism, including deficits in PPI, which have been widely reported in the clinical population (Mcalonan et al., 2002, Perry et al., 2007, Yuhas et al., 2011). As such, adult NR1neo−/− mice show strong face validity for autism-like phenotypes (Crawley, 2007). The emergence of clinical and preclinical evidence implicating reduced NMDAR signaling as a pathophysiological deficit in autism bolsters the construct validity of this model for the disease. Predictive validity is harder to assess, given the lack of viable treatments currently available for autism. The antipsychotic risperidone has shown efficacy for some symptoms of autism, including the repetitive, restricted interests and behaviors (Mcdougle et al., 2005). Similarly, antipsychotics have been shown to reduce motor stereotypies in NR1-hypomorphic mice as well as reverse PPI deficits (Duncan et al., 2006, Mohn et al., 1999).

As with any preclinical model of a neuropsychiatric disease, it is also important to establish the specificity of phenotypic deficits (Silverman et al., 2010).Global neurologic or functional impairment could confound observed results and would look less like autism, which is characterized by deficits in specific domains. Here, we show that auditory function, exploratory ability and the capacity to vocalize are all intact. Likewise, we and others have previously reported that visual and olfactory function is unaltered in NR1neo−/− mice, using behavioral and electrophysiological measures (Duncan et al., 2004, Halene et al., 2009b). Pain sensation was shown to be present, and perhaps even enhanced, in a previous study (Bickel et al., 2007). Finally, no deficit in motor coordination was found in either rotarod or swim tests (Duncan et al., 2004, Fradley et al., 2005). These results indicate that developmental NMDAR disruption does not cause global neurological impairment. While these findings suggest that gross sensory and motor function are intact, future studies employing more complex cognitive and sensory testing are necessary to show true disease specificity.

Model and disease specificity

NR1neo−/− mice were originally developed as a model for the glutamate hypothesis of schizophrenia (Mohn et al., 1999) and have been extensively investigated as such (Duncan et al., 2004, Duncan et al., 2006, Halene et al., 2009a, Moy et al., 2006). Our results indicate that these mice recapitulate autism-like phenotypic deficits in core and several associated symptom domains. Indeed, the most severe phenotypes were profound reductions in social interactions and social vocalizations, deficits more pronounced in autism than schizophrenia (Solomon et al., 2011). This highlights the challenge modeling a complex human behavioral disorder in rodents using behavioral phenotypes alone. To overcome this difficulty, we also investigated auditory-evoked response endophenotypes, which have distinct profiles in each disorder. For example, delayed latencies in the early components of the auditory-evoked response (as observed here) have been reported in ASD patients and their first-degree unaffected relatives and have been associated with language impairment in autism (Maziade et al., 2000, Oram Cardy et al., 2008, Roberts et al., 2010). In contrast, latencies of these peaks are not different in schizophrenia (Hanlon et al., 2005). Reduced peak amplitudes of the auditory ERP (not observed here) are well replicated in schizophrenia, with a strong genetic contribution (Turetsky et al., 2008), and have been observed in multiple mouse models of the disorder (Carlson et al., 2011, Lazarewicz et al., 2010). Subjects with ASD, however, do not show reduced amplitudes of these ERP peaks (Roberts et al., 2010). Indeed, we previously showed increased ERP amplitudes in NR1neo−/− mice, likely elicited by longer ISI intervals. This study employed a shorter ISI to more closely match parameters used in clinical studies (Lincoln et al., 1995, Roberts et al., 2010). Similar deficits in auditory-evoked gamma synchrony have been reported across both disorders and may reflect shared neural abnormalities, as recently suggested (Gandal et al., 2010, Gandal et al., 2012). The continued translation of such endophenotypes to preclinical settings will provide additional measures with which to investigate disease-specific pathophysiology.

The clinical overlap between autism and schizophrenia further complicates the ability to discern the specificity of a given preclinical model. Although autism and schizophrenia are clearly distinct disorders, they share a significant number of common clinical characteristics, including behavioral phenotypes, comorbid symptoms, epidemiology, neuroimaging findings and genetics (Cheung et al., 2010, King & Lord, 2011). Both disorders disproportionately affect males, have an estimated population prevalence of 1–2%, are highly heritable but with complex non-Mendelian genetics, and are neurodevelopmental in origin (Levy et al., 2009, Van Os & Kapur, 2009). Many genes originally linked to autism have been recently associated with schizophrenia, including NRXN1 and SHANK3 (Gauthier et al., 2010, Levinson et al., 2011). The converse has been observed as well, for genes like DISC1 (Kilpinen et al., 2008). In addition, common copy number variants (e.g. 22q11 and 16p11.2) have been linked to both disorders with strong penetrance (Levinson et al., 2011, Sahoo et al., 2011). Common environmental risk factors include prenatal infection, maternal stress and perinatal hypoxia. Behaviorally, both disorders show impairments in social interactions, communicative function and stereotyped behaviors. Of note, recent studies that directly compared social cognition in subjects with ASD and schizophrenia have reported highly similar deficits and underlying neural substrates (Couture et al., 2010, Pinkham et al., 2008). Other shared phenotypes include sleep disruption, epilepsy, ADHD and cognitive abnormalities (Levy et al., 2009, Qin et al., 2005, Siegel & Ralph, 2011). Finally, recent evidence suggests that common phenotypic deficits may stem from shared circuit insults, such as interneuron dysfunction or disrupted excitatory/inhibitory balance, both of which can be caused by reduced NMDAR signaling (Levitt et al., 2004, Lewis et al., 2005, Rubenstein & Merzenich, 2003, Yizhar et al., 2011). Despite the aforementioned commonalities, a number of characteristics clearly distinguish these disorders, such as age of onset, psychosis and the degree of language impairment. As such, future work is needed to determine the biological mechanisms that dictate the pleiotropy of identified risk genes and ensuing phenotypic heterogeneity.

Conclusions and future directions

In contrast to the premise of this study, there has been some suggestion that autism is characterized by glutamatergic hyperfunction (Fatemi, 2008). This hypothesis is based on elevated blood levels of glutamate in high-functioning autism coupled with some mouse models showing enhanced long-term potentiation and/or elevated NMDAR levels (Etherton et al., 2011, Rinaldi et al., 2007, Shimmura et al., 2011). Overexpression of NMDARs has been shown to enhance some forms of learning and memory in mice and could account for the savant-like abilities reported in some individuals with autism (Tang et al., 1999). Like many neural pathways, it is likely that optimal NMDAR signaling follows an ‘inverted U’ profile, where deviations in either direction from a homeostatic set point would cause deteriorated behavioral function. This could also account for the significant degree of phenotypic heterogeneity observed in individuals with autism, some with severe intellectual disability and others with normal, or even elevated, cognitive function. To more rigorously test this hypothesis, future work should investigate the effect of NMDAR overexpression of target phenotypes, such as sociability and EEG biomarkers. A second potential limitation focuses on the age of the mice employed in this study. Only adults were used despite the fact that the age of disease onset is an important characteristic that separates autism from schizophrenia. Future work will focus on juvenile NR1neo−/− and WT mice to determine the onset of phenotypic and endophenotypic deficits.

In conclusion, this study shows that genetic disruption of NMDAR signaling leads to deficits in social and communicative function as well as increased repetitive and self-injurious behaviors – the core symptoms of autism. In addition, NMDAR disruption recapitulates auditory-evoked response endophenotypes characteristic of the ASD population, including a delayed profile of auditory processing and reduced gamma-band synchrony. Such electrophysiological biomarkers help to better characterize preclinical models of neuropsychiatric diseases and provide insight into circuit-level mechanisms that may underlie phenotypic deficits. Gamma synchrony is known to require tight coupling of excitatory and inhibitory balance as well as the appropriate function of fast-spiking, parvalbumin-expressing (PV+) interneurons (Sohal et al., 2009). As such, results implicate disrupted E/I balance and/or PV cell function as a potential mechanism underlying the behavioral deficits following NMDAR-hypofunction and suggest that restoration of such balance with glutamatergic or GABAergic agents would be a rational approach for pharmacologic intervention in neuropsychiatric diseases associated with impaired NMDAR signaling.


The authors work like to thank Tony Thieu, Mili Mehta, Gerald Jonak and Yuling Liang for assistance with the experiments presented in this manuscript. This study was funded by the National Institutes of Health Grants R01-DA023210 (S. J. S.), F30-MH087071 (M. J. G.) and R01-DC008871 (T. R.). The funders had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript. Dr Siegel reports having received grant support from Eli Lilly, AstraZeneca, NuPathe and Pfizer that is unrelated to the content of this article and consulting payments from NuPathe, Merck, Sanofi and Wyeth that are unrelated to this work. Dr Roberts would like to thank the Oberkircher family for the Oberkircher Family Chair in Pediatric Radiology at the Children's Hospital of Philadelphia. Dr Roberts also discloses a financial medical advisory board relationship with Prism Clinical Imaging, which had no overlap with this study. All other authors report no biomedical financial interests or potential conflicts of interest.