Hyperactivity is a Core Endophenotype of Elevated Neuregulin-1 Signaling in Embryonic Glutamatergic Networks

Abstract The neuregulin 1 (NRG1) ErbB4 module is at the core of an “at risk” signaling pathway in schizophrenia. Several human studies suggest hyperstimulation of NRG1-ErbB4 signaling as a plausible pathomechanism; however, little is known about the significance of stage-, brain area-, or neural cell type-specific NRG1-ErbB4 hyperactivity for disease-relevant brain endophenotypes. To address these spatiotemporal aspects, we generated transgenic mice for Cre recombinase-mediated overexpression of cystein-rich domain (CRD) NRG1, the most prominent NRG1 isoform in the brain. A comparison of “brain-wide” vs cell type-specific CRD-NRG1 overexpressing mice revealed that pathogenic CRD-NRG1 signals for ventricular enlargement and neuroinflammation originate outside glutamatergic neurons and suggests a subcortical function of CRD-NRG1 in the control of body weight. Embryonic onset of CRD-NRG1 in glutamatergic cortical networks resulted in reduced inhibitory neurotransmission and locomotor hyperactivity. Our findings identify ventricular enlargement and locomotor hyperactivity, 2 main endophenotypes of schizophrenia, as specific consequences of spatiotemporally distinct expression profiles of hyperactivated CRD-NRG1 signaling.


Introduction
Schizophrenia is a neuropsychiatric disorder with a significant but complex genetic component. 1 According to the neurodevelopmental hypothesis of schizophrenia, genetically imparted perturbations of brain development contribute to disease manifestation in young adults. 2,3 However, stage-, area-, or neural cell typespecific pathomechanisms that link genetic risk factors with disease-relevant brain endophenotypes are not well defined.
Variants of the human NRG1 and ERBB4 genes are plausible genetic risk factors for schizophrenia. 4-6 NRG1 encodes a family of alternatively spliced isoforms that serve as epidermal growth factor (EGF) like ligands for the receptor tyrosine kinase ErbB4. Human NRG1 isoforms belong to 6 different groups (I-VI), based on distinct N-terminal protein domains. NRG1 isoforms from group III contain an N-terminal cysteine-rich domain (CRD) that serves as a second transmembrane domain. As a consequence, CRD-NRG1 isoforms mediate juxtacrine signaling, even after proteolytic processing. 7 NRG1/ErbB4 signaling regulates various neurodevelopmental processes, in particular, the migration of cortical interneurons, synapse formation, and myelination, 8 consistent with a NRG1/ErbB4 signalopathy as a plausible component of schizophrenia etiopathophysiology. 9 Some insight into the nature of this signalopathy came from human studies using post mortem brain tissue, which found evidence for increased NRG1 and ErbB4 expression, as well as ErbB4 receptor hyperactivation in schizophrenia patients. [10][11][12] Notably, individuals inheriting a schizophrenia risk haplotype located in the regulatory region of the NRG1 gene (HapICE) exhibit an increased mRNA expression for CRD-NRG1 13 , suggesting a pathogenic role for hyperactive juxtacrine CRD-NRG1/ErbB4 signaling in schizophrenia.
To test this concept, we and others have investigated transgenic mouse lines with constitutive overexpression of the CRD-NRG1 isoform under control of the Thy1.2 or CamKII promoter and found several brain endophenotypes with relevance for schizophrenia, including ventricular enlargement, impaired prepulse inhibition (PPI), anxiety-like behavior, and deficits in working memory. [14][15][16] Together, these studies imply hyperactive CRD-NRG1/ErbB4 signaling as an attractive candidate for a major signalopathy in schizophrenia. However, considering complex and dynamic expression patterns in the brain, 17 constitutive overexpression in available CRD-NRG1 transgenic mice has impeded the precise mapping of brain endophenotypes to spatiotemporally distinct CRD-NRG1 signaling dysfunctions. Thus, in support of informed future therapeutic strategies, we generated a mouse model for the conditional activation of juxtacrine CRD-NRG1 signaling in the brain.
Images were processed with FIJI software. 21 Fluorescent immunostaining of paraplast embedded brain was described. 18 Chromogenic immunostaining was conducted as above, except for peroxidase inactivation for 5 min with 3% H 2 0 2 prior to blocking. Cryotome and vibratome sections were immunostained following permeabilization and blocking with 4% horse serum, 0.1 % Triton X-100 in 1× PBS for 30 min at RT. For antibodies and visualization of immunoreactivity, see supplementary material. Images were acquired using a Zeiss AxioZ brightfield microscope (Carl Zeiss Jena, Germany). Z-stacks were acquired using a Leica TCS SP2 microscope (Leica Microsystems GmbH, Wetzlar, Germany). Images were processed using Zeiss Zen, ImageJ, 22 and Adobe Photoshop. For the quantification of immunostainings, see 23 and supplementary material.
µCT Scans Mice were sacrificed by cervical dislocation and decapitated. Dissected brains were rinsed 5× in water and transferred to 35% and 70% ethanol for 1 h each. Brains were fixed and stained at RT in 4% PFA in PBS, pH 7.4 and 0.7% PTA in 70% ethanol for 8 days under constant agitation and stored in 70% ethanol. For µCT analysis, brains were embedded in 4% agarose. Samples were scanned using QuantumFX in vivo µCT (Perkin Elmer), nanotome µCT (Phoenix GE), or eXplore Locus SP µCT (Trivoil), resulting in 3D data sets with isotropic voxel sizes ranging from 40 µm to 5 µm. Brain and ventricle sizes were quantified in 3D using a region growing algorithm implemented in SCRY v4.0 software (Kuchel & Sautter GbR). The region growing function was used for volume tracing in virtual Z-stacks in 3D. Data were analyzed (GraphPad Prism software) by 2-tailed t-tests or one-way ANOVA. For immunostaining and quantifications of post µCT brains see supplementary material.

Biochemistry
Protein extract preparation was performed as described. 14 Proteins were separated on 8-15% SDS-polyacrylamide gels, blotted onto polyvinylidene difluoride (PVDF) membranes (Hybond-P, Invitrogen) and visualized according to manufacturer's instructions (Western Lightning Plus-ECL, Perkin Elmer Life Science, Inc.). For details and antibodies, see supplementary material. Membranes were scanned using an Intas ChemoCam Imager. Densitometric analysis was conducted using FIJI software. Statistical significance was tested by one-way ANOVA with Bonferroni's multiple comparison test and t-test (GraphPad Prism 5.0).

Electrophysiology
Recordings were performed on hippocampal slices in standard recording solutions. Data acquisition and analysis were done as described 14,15 (see supplementary methods for details).

Behavior
Behavioral testing and statistical analysis were conducted as described 14,15,24,25 (see supplementary methods for detail).

A Transgenic Mouse Model for Conditional
Hyperstimulation of CRD-NRG1/ErbB4 Signaling CRD-NRG1 harbors 2 transmembrane domains ( figure 1A) and signals to ErbB4, its main receptor in the brain. Here, we generated conditional transgenic mice (Stop-Nrg1) for Cre-mediated expression of N-terminally hemagglutinin (HA) epitope-tagged CRD-NRG1 26 (figure 1B; supplementary figures 1A and 1B). Based on "Stop" element-encoded GFP, we observed β-actin promoter-driven transgene expression in the brain and various organs (supplementary figure 1C; data not shown). Immunostaining in Stop-Nrg1 mice (in the absence of a Cre "driver") demonstrated transgene (GFP) expression in glutamatergic neurons, oligodendrocytes, and spinal cord α−motoneurons. However, only a small subset of GABAergic interneurons and astrocytes expressed the Stop-Nrg1 transgene (supplementary figure 1D; data not shown).
Cortical CRD-NRG1 expression peaks during embryonic and postnatal stages. 17 To model CRD-NRG1 overexpression in the embryonic cortex, we crossbred Stop-Nrg1 mice to NEXCre driver mice, 19 in which Cremediated recombination is initiated at embryonic day (E) 12 in postmitotic glutamatergic neurons (figure 1B). Immunostaining identified moderate transgenic CRD-NRG1 expression in the cortical plate of NEXCre*Stop-Nrg1 double transgenic mice (referred to as NC-Nrg1) at E16 (figure 1C). We also produced a model for CRD-NRG1 overexpression in glutamatergic neurons starting at postnatal stages by breeding to CamKII-Cre driver mice 20 (supplementary figure 2A). Resulting double transgenic mice (CC-Nrg1) and NC-Nrg1 mice displayed glutamatergic neuron-specific HA-CRD-NRG1 expression in the brain (figure 1D; supplementary figures 2B and 2C). Western blotting demonstrated CRD-NRG1 overexpression in the hippocampus and neocortex of NC-Nrg1 and CC-Nrg1 mice, which was associated with hyperphosphorylated ErbB4 receptor (figures 1E and 1F; supplementary figures 2D and 2E). In conclusion, the expression profile of CRD-NRG1 in NC-Nrg1 mice (figure 1G) allows to address the impact of a developmentally hyperstimulated schizophrenia risk pathway on adult brain functions.
We previously generated "constitutive" transgenic mice, which overexpress HA-CRD-NRG1 under control of the Thy1.2 promoter 18 (T-Nrg1). Immunostaining confirmed cortical and subcortical CRD-NRG1 expression in T-Nrg1 transgenic mice, whereas overexpression in NC-Nrg1 mice was limited to cortical projection neurons (figure 2A). Western blotting at P10 revealed higher cortical CRD-NRG1 expression in T-Nrg1 compared to NC-Nrg1 mice (figures 2B and 2C). In line, cortical ErbB4 phosphorylation was increased in both lines; however, it was ~4-fold higher in T-Nrg1 compared to NC-Nrg1 mice (figures 2D-2F). Furthermore, we observed a corresponding reduction of total ErbB4 protein in NC-Nrg1 and T-Nrg1 mice (figures 2D and 2G). These findings indicate that a comparative analysis of these mouse lines could be employed to inform about expression level-, brain region-, and neural cell type-specific CRD-NRG1 functions.
Mice from the SMDA line display ventricular enlargement. 14 Ex vivo microcomputed tomography (µCT) confirmed a similar ventricular enlargement in T-Nrg1 mice at 3 months (figures 3A-3C) and P37 (figures 3D and 3E). These findings identify ventricular enlargement as a specific outcome of hyperactive CRD-NRG1 signaling and suggest that causative pathogenic signals originate outside glutamatergic neurons. Together, NC-Nrg1 mice provide a new mouse model to investigate the impact of moderate, glutamatergic network-specific CRD-NRG1/ ErbB4 hyperstimulation on brain functions in the absence of major confounding endophenotypes.

NC-Nrg1 Mice Display Reduced Inhibitory Neurotransmission, Impaired LTP, and Abnormal Theta Oscillations
Consistent with our findings in SMDA mice, 14 LTP at the Schaffer collateral (SC) CA1 synapse was reduced and inhibitory synaptic transmission in prefrontal cortex (PFC) was increased in T-Nrg1 mice (figures 4A, 4C, and 4D). Hippocampal LTP was also reduced in NC-Nrg1 mice (figure 4B), whereas fEPSP slopes and paired-pulse ratios in CA1, and glutamatergic neurotransmission   in PFC were unaltered in both lines (supplementary figure 4). In contrast to T-Nrg1 mice, inhibitory neurotransmission was reduced in the PFC of NC-Nrg1 mice (figures 4E and 4F). Furthermore, carbachol-induced theta oscillations were impaired and the relative theta power was increased in CA1 of both lines (figures 4G and 4H). These data show that a physiologically relevant chronic increase in glutamatergic CRD-NRG1 expression in NC-Nrg1 mice reduces, not increases, inhibitory neurotransmission in the PFC, and impairs synaptic plasticity and theta oscillations in CA1.
These synaptic and network dysfunctions could result from alterations in glutamatergic and/or GABAergic neurons. To identify structural changes in glutamatergic networks, we examined dendritic spines of CA1 pyramidal cells in stratum radiatum by STED microscopy but found no evidence for profoundly altered spine density or shape in T-Nrg1 mice at 10 weeks of age (supplementary figures 5A-5C). Consistent with our studies in SMDA and T-Nrg1 mice, 14

CRD-NRG1 Overexpression in Glutamatergic Cortical Networks Induces Hyperactivity
T-Nrg1 mice display increased anxiety, impaired PPI, and working memory deficits. 14,15 To obtain a corresponding behavioral profile for NC-Nrg1 mice, we performed a behavioral test battery. A basic physical exam (including testing of eyelid, whisker, and ear reflexes and grip strength) revealed no obvious deficits in parental lines and NC-Nrg1 mice (supplementary table 1). Also pain sensitivity in the hot plate test was not affected in NC-Nrg1 mice (supplementary figure 6A) in line with intact sensory functions (that critically depend on NRG1 signaling 30,31 ).
In contrast to T-Nrg1 mice, NC-Nrg1 mice showed no signs of anxiety in the light-dark preference test (supplementary figure 6B) and open field test, with time spent in the center of the arena and number of rearings similar to controls (supplementary figures 6C and 6D). Furthermore, while NC-Nrg1 mice displayed a subtle reduction in the startle response prior conditioning (supplementary figure 6F), PPI, and working memory, as assessed in the Y-maze, were not altered (supplementary figures 6G and 6H).
Of note was moderately reduced freezing of NC-Nrg1 mice when exposed to context and auditory cue in fear conditioning (figure 5A). However, reduced freezing behavior in NC-Nrg1 mice may be attributed to increased motor activity. Whereas in the hole board test, the number of holes visited and exploration time were unaltered (figure 5B; data not shown), NC-Nrg1 mice showed an increased distance traveled compared to controls ( figure 5C). This was particularly visible in the first minutes of the hole board test ( figure 5D). We also observed hyperactivity in the open field test, where NC-Nrg1 mice showed an increased time active (supplementary figure 6E), number of corner visits (figure 5E), and distance traveled ( figure 5F). Again, NC-Nrg1 mice were more active during early stages, consistent with novelty-induced locomotor hyperactivity ( figure 5G). Finally, NC-Nrg1 mice exhibited increased time active in the tail suspension test compared to controls (figure 5H), most likely reflecting general hyperactivity, instead of increased motivation to escape from the aversive situation. These findings suggest locomotor hyperactivity as a core endophenotype of moderately increased CRD-NRG1 signaling in cortical glutamatergic networks.

Discussion
To investigate the genetic component of the neurodevelopmental hypothesis of schizophrenia, we modeled moderate CRD-NRG1 overexpression with an embryonic onset in cortical glutamatergic neurons in NC-Nrg1 mice. Compared with constitutive T-Nrg1 mice, NC-Nrg1 mice provide a physiologically more relevant hyperstimulation of CRD-NRG1/ErbB4 signaling in the neocortex and hippocampus, supporting their employment to examine region-and cell type-specific CRD-NRG1 functions in the brain.
Intraperitoneal NRG1β injection reduces weight gain in mice, 27 raising the question of peripheral vs central NRG1 functions in body weight control. 32,33 Body weight was unaltered in NC-Nrg1 mice but reduced in T-Nrg1 mice and conditional CC-Nrg1 mice. This identifies a central, most likely subcortical, CRD-NRG1 function in the control of body weight, possibly via autocrine signaling to ErbB4 + neurons in the hypothalamus. 33,34 This hypothesis could be tested by breeding Stop-Nrg1 mice to a Credriver line for hypothalamic recombination. 35 Ventricular enlargement is a major neuroanatomical biomarker in schizophrenia. 36 It has been linked to NRG1 risk variants in first-episode schizophrenia patients 37 and in individuals at high risk for psychosis. 38 We reported enlarged lateral ventricles in transgenic mice  with CRD-NRG1 overexpression. 14 Here, we confirm increased ventricular size in T-Nrg1 mice, demonstrating ventricular enlargement as a specific outcome of hyperactive CRD-NRG1 signaling. Ventricular size was massively increased in DC-Nrg1 mice but unaltered in NC-Nrg1 mice. This indicates that pathogenic CRD-NRG1 signals affecting the ventricular system derive from cells other than glutamatergic neurons, eg, ependymal cells, 39 possibly causing abnormal autocrine NRG1/ErbB4 signaling, ciliopathy, and altered cerebrospinal fluid circulation. 40 Hyperactivity is a core symptom in schizophrenia, 41,42 but its neurobiological underpinnings are only partially understood. We found locomotor hyperactivity in NC-Nrg1 mice across several behavioral test paradigms. In particular, NC-Nrg1 mice displayed novelty induced hyperactivity, a rodent correlate of positive symptoms of schizophrenia 43 in the open field and hole board test. Unaltered working memory and PPI in conditional NC-Nrg1 mice (with moderate ErbB4 hyperstimulation) suggest that previously reported deficits in these domains in T-Nrg1 mice 14,15 result from more pronounced and/or spatially distinct (in particular subcortical) CRD-NRG1 dysfunctions in constitutive T-Nrg1 mice.
Hyperactivity in NC-Nrg1 mice is consistent with reduced motor activity in conditional NRG1 lossof-function mutants using the CKII-Cre driver line. 14 However, hyperactivity was also reported in heterozygous "global" NRG1 and ErbB4 loss-of-function mutants [44][45][46] but absent in constitutive transgenic mice with CRD-NRG1 overexpression under control of the CamKII promoter. 16 These discrepancies between our conditional and previous constitutive genetic approaches most likely emerge from different timing and/or the spatial extend of genetic manipulations in different models and underscore the importance of a precise isoform-, region-, and cell type-specific mapping of NRG1 dysfunctions in the brain.
How does hyperactivity emerge in NC-Nrg1 mice? We suggest synergistic reciprocal interactions between glutamatergic neuron-derived CRD-NRG1 overexpression and changes in GABAergic and dopaminergic neurotransmission via PV + /ErbB4 + interneurons. This notion is supported by our finding of reduced inhibitory neurotransmission in NC-Nrg1 mice and hyperactive motor behavior in several genetic and pharmacological models of GABAergic impairment, including mouse mutants with disrupted NMDA receptor function in interneurons. [47][48][49][50] Ultimately, GABAergic deficits could elicit secondary glutamatergic hyperexcitability, cortical network desynchronization, and hyperactivity in NC-Nrg1 mice. Preclinical animal studies also provide a strong link between locomotor hyperactivity and abnormal dopaminergic signaling, 51 and infusion of recombinant NRG1β into rat brains was shown to stimulate dopamine release in the hippocampus, presumably indirectly via D4 receptors on PV + interneurons. 52 Thus, chronically increased CRD-NRG1 signaling may promote altered dopaminergic signaling in the hippocampus and neocortex, which in concert with GABAergic changes may induce locomotor hyperactivity, as well as abnormal hippocampal LTP 52 and oscillatory activity. 53 Together, our data identify locomotor hyperactivity as a core outcome of cortical glutamatergic neuronderived hyperactive CRD-NRG1 signaling. ErbB4 hyperactivation-directed treatment studies, such as those performed in T-Nrg1 mice, 15 will allow to test the predictive validity of NC-Nrg1 mice and provide an entry point for the development of more potent treatment strategies that target specific endophenotypes in schizophrenia.

Supplementary Material
Supplementary material is available at Schizophrenia Bulletin.