|Home | About | Journals | Submit | Contact Us | Français|
Neuregulin 1 (NRG1) signaling is critical to various aspects of neuronal development and function. Among different NRG1 isoforms, the Type III isoforms of NRG1 are unique in their ability to signal via the intracellular domain following γ-secretase-dependent intramembranous processing. However, the functional consequences of Type III NRG1 signaling via its intracellular domain are largely unknown. In this study, we have identified mutations within Type III NRG1 that disrupt intramembranous proteolytic processing and abolish intracellular domain signaling. In particular, substitutions at valine 321, previously linked to schizophrenia risks, result in NRG1 proteins that fail to undergo γ-secretase-mediated nuclear localization and transcriptional activation. Using processing-defective mutants of Type III NRG1, we demonstrate that the intracellular domain signaling is specifically required for NRG1 regulation of the growth and branching of cortical dendrites but not axons. Consistent with the role of Type III NRG1 signaling via the intracellular domain in the initial patterning of cortical dendrites, our findings from pharmacological and genetic studies indicate that Type III NRG1 functions in dendritic development independent of ERBB kinase activity. Taken together, these results support the proposal that aberrant intracellular processing and defective signaling via the intracellular domain of Type III NRG1 impair a subset of NRG1 functions in cortical development and contribute to abnormal neuroconnectivity implicated in schizophrenia.
The neuregulin 1 (NRG1) encodes a family of growth factors that interact with ERBB receptor tyrosine kinases via the epidermal growth factor (EGF)-like domain to initiate intracellular signaling cascades critical to neuronal functions (Kwon et al., 2005; Li et al., 2007; Woo et al., 2007; Mei and Xiong, 2008; Zhong et al., 2008; Barros et al., 2009; Fazzari et al., 2010; Wen et al., 2010). The NRG1 proteins are divided into 6 types, Types I to VI. Most of NRG1 isoforms are single-pass transmembrane proproteins with the transmembrane domain (TMc) located at the C-terminus. In contrast, Type III NRG1 proproteins span the membrane twice due to the hydrophobic cysteine-rich domain (CRD). NRG1 proproteins undergo extracellular proteolytic processing at the juxtamembrane region to release EGF-like domain, which then participates in paracrine signaling involving ERBB receptors (Falls, 2003). For Type III NRG1, the cleaved ectodomain containing the EGF-like domain remains tethered to the membrane by the CRD. Thus, Type III NRG1 is believed to signal via ERBB receptors in a juxtacrine manner (Wolpowitz et al., 2000; Leimeroth et al., 2002; Falls, 2003). Synaptic activation, membrane depolarization or ERBB binding triggers a γ-secretase-dependent cleavage of Type III NRG1 to release the intracellular domain (NRG1-ICD) from the membrane which translocates to the nucleus to regulate gene expression (Bao et al., 2003; Bao et al., 2004). Type III NRG1-ERBB complexes therefore signal bidirectionally, i.e. via activation of ERBB and via the NRG1-ICD (Wolpowitz et al., 2000; Leimeroth et al., 2002; Bao et al., 2003; Falls, 2003; Bao et al., 2004; Hancock et al., 2008; Zhong et al., 2008).
NRG1 is a candidate schizophrenia susceptibility gene (Stefansson et al., 2002). Studies of adult Type III Nrg1 heterozygous mice have revealed defects reminiscent of schizophrenia-associated endophenotypes (Chen et al., 2008). One schizophrenia-associated NRG1 single nucleotide polymorphism (SNP) changes a valine to leucine within the TMc (Walss-Bass et al., 2006). How this valine-to-leucine substitution affects schizophrenia susceptibility is unknown. We have begun addressing this issue by studying the effects of mutations within the TMc on the growth and branching pattern of cortical dendrites and axons because abnormal cortical connectivity has been implicated in schizophrenia (Rajkowska et al., 1998; Kalus et al., 2000).
Here, we have shown differential requirements for Type III NRG1-mediated signaling pathways in the development of dendrites vs. axons of cortical neurons. Cortical neurons of Type III Nrg1 knockout embryos exhibit defects in the growth and branching of dendrites, a phenotype that is rescued by re-expression of wild type Type III NRG1 but not by NRG1 mutants that are defective in γ-secretase-dependent signaling, including the mutant containing the valine-to-leucine substitution within the TMc. Whereas NRG1 signaling via the intracellular domain plays a crucial role in dendritic development, NRG1 regulates axon extension via distinct mechanisms that do not involve intramembranous proteolytic cleavage and nuclear targeting. Together, our study provides evidence for differential requirements of NRG1 signaling in the development of cortical axons vs. dendrites, and insights into how the schizophrenia-associated valine-to-leucine mutation might contribute to the disease.
Dispersed cortical neurons were prepared from C57/BL6 mice following previously described methods (Ghosh and Greenberg, 1995). Neuronal cultures were maintained in L-glutamine free Basal Media Eagle Media supplemented with 1% N2 supplement, 1 mM L-glutamine, 5% fetal bovine serum, 100 units/ml of penicillin and 100 ug/ml of streptomycin. All reagents were purchased from Invitrogen.
We used wild type and knockout littermates of a mouse line with genetic disruption of Type III Nrg1 (Nrg1tm1/Lwr) (Wolpowitz et al., 2000). To label neurons, female Type III Nrg1 heterozygous mice were bred with male transgenic mice overexpressing Yellow Fluorescent Protein (YFP) from the Thy-1 promoter (line YFP-H, (Feng et al., 2000), Jackson Laboratory) in a subset of neurons. Animals from the resulting mouse line show YFP expression in the cortex. Male and female YFP-expressing, Type III Nrg1 heterozygous mice were bred and embryos generated from the crosses were used in the study. For studies involving ErbB4 knockout animals, we used ErbB4 knockout mice that were rescued from embryonic lethality by re-expression of ErbB4 in the heart (Tidcombe et al., 2003). The ErbB4 knockout animals were kind gifts from Dr. G. Corfas (Harvard University). The use of the animals was approved by the Institutional Animal Care and Use Committee of SUNY at Stony Brook.
A 622 bp BamHI and HindIII fragment containing murine Type III Nrgβ1a sequences (position 378–1000, Genbank Accession #AY648975) was subcloned into the pAlter-Ex1 vector (Promega). Mutant forms of Type III Nrg1 were generated using the following primers and Altered Sites II in vitro Mutagenesis System (Promega) according to the manufacturer’s instruction:
Inserts containing mutated sequences were sequenced and reinserted into full-length Nrg1 sequences using BamHI and HindIII sites. To generate Gal4DBD fusion proteins, the Gal4 DNA binding domain was amplified by PCR from the pGBKT7 vector (BD Biosciences) using the following primers:
Gal4DBD was subcloned in frame into the full-length Type III Nrgβ1a sequences at the HindIII site (position 1000). Type III Nrgβ1a-Gal4DBD was inserted into pSP72βact (Cho et al., 1998) under the transcription control of β-actin promoter.
Rat Type III Nrgβ1a sequences were cloned in pcDNA3.1 (D. L. Falls, Emory University, Atlanta, Georgia; (Wang et al., 2001)). This plasmid was used as the template for mutagenesis. The valine residue at codon 322 of rat Type III Nrgβ1a, which corresponds to valine 321 of murine Type III Nrgβ1a, was substituted to leucine by mutating the nucleic acid sequence from GTG to TTG (custom mutagenesis by Retrogen, Inc.). This valine to leucine substitution was verified by DNA sequencing of the entire rat Type III Nrgβ1a insert cloned into pcDNA3.1 (DNA sequencing by Retrogen, Inc.). All codon numbering is based on murine Type III Nrgβ1a sequences.
Media from HEK293 cells expressing the extracellular domain of ERBB4 fused to the human IgG Fc domain (B4-ECD; (Fitzpatrick et al., 1998)) were concentrated using Microcon centrifugal filter devices (Amicon). B4-ECD fusion proteins were purified using Protein A agarose columns and detected by immunoblot analysis using an antibody specific for the human IgG Fc domain (Sigma). The integrity of the recombinant protein was confirmed by SDS-polyacrylamide gel separation. Dot blot analyses of purified ERBB4-ECD-Fc protein preparations with comparisons made to a dose curve of human IgG were performed; blots were probed with anti-Fc antibodies and quantified using a LiCor Odyssey infrared imager to determine fusion protein levels. B4-ECD lacks the kinase domain. B4-ECD binds with high affinity to the EGF-like domain of NRG1 (Fitzpatrick et al., 1998) and has been used to induce NRG1 processing and NRG1-ICD release from the membrane (Bao et al., 2003).
Cortical cultures were fixed in 4% formaldehyde/4% sucrose/1X PBS for 15 minutes. Following fixation, cells were washed for 5 minutes with 1X PBS for 3 times. Subsequently, cells were blocked for 30 minutes with a blocking solution containing 10% normal donkey serum or 3% goat serum, depending on the secondary antibodies used, 0.25–0.3% Triton X-100, 3% BSA and 1X PBS. Cells were incubated at 4°C overnight with primary antibodies diluted in the blocking solution. Cells were washed 3 times and then incubated with secondary antibodies for 45 minutes. All fluorophore-conjugated secondary antibodies were purchased from Molecular Probes as 2μg/ml stock solutions and used at a dilution of 1:1000. Cells were washed 3 times before mounting in ProlongGold (Molecular Probe).
For subcellular localization of Type III NRG1, dispersed cortical neurons (0.5 × 106 cells per well) were plated on 24-well plates containing 12-mm polylysine-laminin-coated glass coverslips (BD Biosciences). At indicated day in vitro (DIV), cells were processed for immunofluorescence studies using sheep anti-CRD (raised against a synthetic peptide derived from human Type III NRG1 amino acid residues 92–122; (Yang et al., 1998); 1:1000, purified IgG) as the primary antibody and Alexa Fluor 488 donkey anti-sheep IgG as the secondary antibody. Images were captured with a Zeiss LSM 510 META scanning confocal microscope as 0.5-μm stacks at 63X magnification. For colocalization with MAP2, the following primary and secondary antibody pairs were used: sheep anti-CRD and Alexa Fluor 568 donkey anti-sheep IgG; mouse anti-MAP2 ascites (1:3000, Sigma) and Alexa Fluor 488 donkey anti-mouse IgG. Neurons were counterstained with DAPI and imaged with a Zeiss LSM 510 multiphoton confocal microscope at 40X magnification, 1.5X digital zoom and 0.5-μm interval. For colocalization with VGLUT1 and MAP2, the following primary and secondary antibody pairs were used: sheep anti-CRD and Alexa Fluor 488 donkey anti-sheep IgG; mouse anti-MAP2 ascites and Alexa Fluor 647 donkey anti-mouse IgG; rabbit anti-VGLUT1 (1:250, Synaptic Systems) and Alexa Fluor 568 donkey anti-rabbit IgG. Neurons were imaged with a Zeiss LSM 510 META confocal microscope at 100X magnification, 1.0X digital zoom and 0.5-μm interval.
Dispersed E18/E19 cortical neurons (0.5 × 106 cells per well) from wild type and Type III Nrg1 knockout animals were plated on 24-well plates containing 12-mm polylysine-laminin-coated glass coverslips. A few hours after plating, cells were transfected with pEGFP-C3 encoding Enhanced Green Fluorescent Protein (EGFP; 0.15 μg/well; Clontech) using Lipofectamine 2000 (Invitrogen). For the rescue experiments, cortical neurons were transfected with pEGFP-C3 (0.15 μg/well) alone or together with a plasmid (0.45 μg/well) encoding Gal4DBD-tagged NRG1-WT, NRG1-T307A/G308A, NRG1-V321A/V322A, NRG1-K329A/Q330A or NRG1-WT or NRG1-V321L with no Gal4DBD tag. Two days post-transfection, cells were processed for immunofluorescence studies. For ERBB kinase inhibitor experiments, cells were treated with DMSO (1:1000) or PD158780 (1μM, Calbiochem) for ~20–24 hours prior to fixation. Transfected neurons were double labeled with rabbit anti-green fluorescent protein (GFP) antibodies (1:3000, Molecular Probes) to outline the entire neuron including processes and mouse anti-MAP2 antibodies (1:3000, Sigma) to identify the somatodendritic regions of neurons. The fluorophore-conjugated secondary antibodies used were Alexa Fluor 555 donkey anti-mouse IgG and Alexa Fluor 488 donkey anti-rabbit IgG. Cells were mounted in ProlongGold containing DAPI (Molecular Probes).
Image stacks of GFP and MAP2 positive neurons with intact nuclei showing no signs of nuclear fragmentation were captured at 40X magnification, 1.5X digital zoom and 0.5-μm interval using a Zeiss LSM 510 META scanning confocal microscope. Only neurons with identifiable axons were included in the analysis. Axons do not stain for MAP2 except in the region very proximal to the cell body. Axons were defined as the longest process extended from the cell body exhibiting a taper appearance and maintaining about the same thickness throughout the entire length. These features are reminiscent of those of neurons in the cortical plate, i.e. the cell dense zone just beneath the molecular layer, in vivo. Axons and dendrites were analyzed for the total length and number of branch points using Zeiss LSM software.
Brains of E19 YFP-expressing, wild type and Type III Nrg1 heterozygous and knockout embryos were immersion fixed in 3% paraformaldehyde/1X PBS for 1 hour at room temperature and sectioned into coronal sections (300 um) on a vibratome. Slices were immunostained with rabbit anti-GFP as the primary antibody and Alexa Fluor 488 donkey anti-rabbit IgG as the secondary antibody. Brain sections were mounted in ProlongGold. Image stacks of YFP-labeled neurons were captured at 40X magnification, 1.5X digital zoom and 0.5-μm interval using a Zeiss LSM 510 META scanning confocal microscope. Basal dendrites of the individual neuron were traced through the entire image stack. Total length of all basal dendrites and number of basal dendritic branches of each neuron were quantified.
Dispersed P0/P1 cortical neurons (106 cells per well) from C57/BL6 mice were plated on polylysine-laminin-coated 12-well plates and transfected at 2 DIV using Lipofectamine 2000. Cells were co-transfected with 0.5 μg of constructs encoding mouse wild type NRG1 (NRG1-WT) or different mutants of NRG1 fused with Gal4DBD (i.e. NRG1-T307A/G308A, NRG1-V321A/V322A or NRG1-K329A/Q330A) together with 0.5 μg of Gal4-UAS-luciferase in p4Luc (Bao et al., 2003). In order to control for transfection efficiency, cells were also transfected with phRL-CMV (Promega) encoding Renilla luciferase. Twenty-four hours post-transfection, cells were treated with 6.25 microliters per well of the vehicle (50% glycerol in 1X PBS) or 6.25 microliters per well of B4-ECD to a final concentration of 2 nM under normal growth condition. Cells were harvested for luciferase assays ~20 hours post-treatment. Luciferase assays were performed using Dual-Glo Luciferase Assay System (Promega) and a GloMax 96 Microplate Luminometer (Promega). Background values for firefly and Renilla luciferase were subtracted from all measurements. Following background corrections, values of firefly luciferase were normalized against those of Renilla luciferase to control for transfection efficiency.
Dispersed P0 cortical neurons (0.6 × 106 cells per well) from wild type C57/BL6 mice were plated on 24-well plates containing 12-mm polylysine-laminin-coated glass coverslips. Cells were transfected at 1 DIV with a plasmid (0.75 μg/well) encoding NRG1-WT or NRG1-V321A/V322A using Lipofectamine 2000. Two days post-transfection, cells were treated with 10% DMSO in water (1:2000) or the γ-secretase inhibitor L685458 (1 μM; Sigma) for 1 hour followed by a 20-minute treatment with 3.75 microliters per well of 50% glycerol/1X PBS or 3.75 microliters per well of B4-ECD to a final concentration of 2 nM. Subsequently, cells were processed for immunofluorescence studies using the following primary and secondary antibody pairs: mouse anti-Gal4 (1:200, Abcam or 2 μg/ml, Zymed) together with Alexa Fluor 488 donkey anti-mouse IgG; sheep anti-CRD (1:500, purified IgG) together with Alexa Fluor 568 donkey anti-sheep IgG. Confocal image stacks were captured with a Zeiss LSM 510 multiphoton confocal microscope at 100X magnification and 1-μm interval. The single optical section through the nucleus of individual neuron was selected for analysis using Image J (NIH). To determine the amount of nuclear Gal4DBD-tagged NRG1-ICD immunoreactivity, we measured the mean pixel value of Gal4DBD fluorescence intensity within the DAPI-stained nucleus. All measurements were normalized against those of vehicle-treated (i.e. 10% DMSO in water followed by 50% glycerol/1X PBS) NRG1-WT-expressing cells.
The results were analyzed by performing the nonparametric Mann-Whitney Rank Sum test for comparisons between two groups or Kruskal-Wallis test for multi-group comparisons using SigmaStat (SPSS Inc.). Statistically significant differences between control and experimental conditions are specified in the text.
To investigate functions of Type III NRG1 signaling in the development of neuroconnectivity in cortex, we began by studying the growth and branching of basal dendrites of cortical neurons from Yellow Fluorescent Protein (YFP)-expressing, Type III Nrg1 knockout (KO/YFP+) mice as well as their wild type (WT/YFP+) littermates. The expression of YFP from Thy1-YFP transgene has been shown to label only pyramidal neurons in the mouse cortex (Feng et al., 2000); also see Supplementary Fig. 1). In addition, the morphology of cortical pyramidal neurons is outlined by YFP localizing to cell bodies and projections (Wang et al., 2006). Confocal image stacks of YFP-labeled cortical pyramidal neurons were analyzed for the total length and number of branches of basal dendrites. We studied the basal dendrites of 60 WT/YFP+ and 55 KO/YFP+ cortical pyramidal neurons. Our results revealed that basal dendrites of cortical neurons from KO/YFP+ mice (i.e. KO/YFP+ neurons) showed a significant 52% reduction (P < 0.001) in the total length and a 44% decrease (P < 0.001) in the number of branches (Fig. 1). In contrast, KO/YFP+ mice show no apparent defect in the cortical plate thickness (Supplementary Fig. 2) and this is consistent with normal development of cortical cell layers at postnatal age in mice lacking ERBB2/ERBB4 (Barros et al., 2009). In sum, Type III NRG1 signaling is required for full elaboration of the basal dendritic arbors of cortical pyramidal neurons.
To delineate the molecular signaling mechanisms involved in Type III NRG1 function in the development of cortical connectivity, we performed additional experiments in cortical neuronal cultures. We began by examining the subcellular distribution of Type III NRG1 in cortical neurons in culture using an antibody recognizing the CRD which is present only in Type III NRG1 ((Yang et al., 1998); Supplementary Fig. 3). Type III NRG1 was detected throughout the soma and along both dendrites (MAP2 positive processes) and glutamatergic axons (VGLUT1 positive but MAP2 negative processes) of cortical neurons (Figs. 2A–2C). These results are consistent with the somatodendritic and presynaptic localization patterns of NRG1 isoforms in human prefrontal cortical neurons (Law et al., 2004) as well as Type III NRG1 function in the development of cortical connectivity in vivo.
To study whether Type III NRG1 regulates the growth and complexity of dendrites and axons of developing cortical neurons, we compared the total length and extent of branching of dendrites and axons of cortical neurons cultured in vitro from wild type (WT) and Type III Nrg1 knockout (KO) mice. Cortical neurons were plated at equal densities and transfected with a plasmid encoding Enhanced Green Fluorescent Protein (EGFP) to outline the entire neuron including processes. Cells were immunostained for EGFP and MAP2, a somatodendritic marker of neurons. Cortical neurons maintained in vitro are polarized with distinguishable dendrites vs. axons (see Materials and Methods). Moreover, dendrites stain positive for MAP2 whereas axons do not stain for MAP2 except in the region very proximal to the cell body. Confocal image stacks of EGFP and MAP2 positive neurons were analyzed for the total length and number of branch points of dendrites and axons. This method of GFP-labeling of the neuronal processes of mouse cortical neurons has been previously used to study a number of morphological parameters including total dendritic length and the number of branch points per cell (Whitford et al., 2002).
We quantified total dendritic length and the number of dendritic branch points for 47 WT cortical neurons and 64 KO cortical neurons, KO neurons had a significant reduction in total dendritic length and number of branch points (length: 258.2 ± 22.5 um vs. 130.9 ± 9.8 um; branch point: 7.5 ± 1.3 branch points vs. 3.4 ± 0.5 branch points; P < 0.05; Figs. 2D–2F). We also quantified axonal length and number of axonal branch points for 39 WT cortical neurons and 88 KO cortical neurons, and found that the total axonal length of KO neurons was significantly decreased (668.2 ± 72.7 um vs. 443.2 ± 36.5 um; P < 0.05; Fig. 2G). The number of axonal branch points was not significantly different between WT neurons vs. KO neurons (19.9 ± 2.5 branch points vs. 18.2 ± 1.7 branch points; Fig. 2H). Collectively, our results indicate that Type III NRG1 signaling is critical to the elaboration of both dendrites and axons in the cortex.
Following synaptic activation, membrane depolarization or ERBB binding, Type III NRG1 undergoes γ-secretase-dependent intramembranous cleavage to release the intracellular domain (NRG1-ICD) from the membrane (Bao et al., 2003; Bao et al., 2004) (Fig. 3A). The released NRG1-ICD translocates to the nucleus to regulate gene expression (Bao et al., 2003; Bao et al., 2004) (Fig. 3A). To study whether NRG1 signaling via the intracellular domain contributes to the development of dendrites and axons, we began by identifying Type III NRG1 mutants defective in processing and therefore signaling to the nucleus. We generated alanine mutations in the C-terminal transmembrane domain and the nuclear localization signal (NLS) of Type III NRG1 (Fig. 3B). Subsequently, the ability of mutant forms of NRG1 to translocate to the nucleus was assayed by measuring transcriptional activation of a luciferase reporter gene in the nucleus. Cortical neurons were transiently co-transfected with constructs encoding full-length Type III NRG1 (NRG1-WT) or one of three mutant forms of NRG1 (i.e. NRG1-T307A/G308A, NRG1-V321A/V322A and NRG1-K329A/Q330A) fused to Gal4 DNA-binding domain (Gal4DBD), together with a luciferase reporter gene containing the Gal4-UAS promoter controlled by Gal4DBD. Under control conditions, neither wild type nor mutant forms of NRG1 fused with Gal4DBD induced the Gal4-UAS promoter and the transcription of the luciferase reporter (Fig. 3C). Previous studies demonstrated that treatment of cells with soluble ERBB4 containing only the extracellular domain fused to the Fc domain of human IgG (B4-ECD) (Fitzpatrick et al., 1998) leads to NRG1 proteolytic processing, nuclear translocation and transcriptional activation (Bao et al., 2003). Treatment of NRG1-WT-expressing cells with 2 nM B4-ECD induced luciferase expression ~3-fold indicating release and nuclear targeting of Gal4DBD-tagged NRG1-ICD (n=4 vector, 6 NRG1-WT + B4-ECD; P < 0.05; Fig. 3C). ERBB binding similarly induced luciferase expression in NRG1-T307A/G308A-expressing neurons (n=4 NRG1-T307A/G308A + B4-ECD; P < 0.05; Fig. 3C). Thus, mutations at residues 307/308 had no effect on NRG1 processing, nuclear translocation and transactivation. In contrast, in neurons expressing NRG1-V321A/V322A or NRG1-K329A/Q330A, the levels of luciferase activity following ERBB binding were similar to vector control (n=5 NRG1-V321A/V322A + B4-ECD, 5 NRG1-K329A/Q330A + B4-ECD; Fig. 3C) indicating that alanine substitutions at these residues interferes with B4-ECD induced NRG1 processing, nuclear targeting and transcriptional activation.
To confirm that valine residues 321/322 are required for γ-secretase-dependent Type III NRG1 signaling in cortical neurons, we stimulated cortical neurons expressing either a construct encoding NRG1-WT or NRG1-V321A/V322A with B4-ECD in the absence or presence of a γ-secretase inhibitor (L685458) and quantified immunoreactive Gal4DBD-tagged NRG1-ICD in the DAPI-labeled nucleus. The following results are from the analysis of nuclear Gal4 staining in a total of 101 cells from two independent experiments. In vehicle-treated NRG1-WT-expressing cells, Gal4DBD (green) was detected in soma and processes and weakly in nuclei (blue), indicating low levels of nuclear targeting of NRG1-ICD in the basal state (Fig. 3D; each panel is from a single optical section through the center of individual nucleus). Stimulation of processing and translocation via B4-ECD treatment resulted in a 1.6-fold increase in nuclear Gal4 staining in NRG1-WT-expressing cells indicating enhanced nuclear localization of Gal4DBD-tagged NRG1-ICD (Figs. 3D and 3E). This B4-ECD-stimulated NRG1-ICD nuclear localization in NRG1-WT-expressing cells was blocked by inhibiting γ-secretase activity (Fig. 3E). This result is consistent with previously reported nuclear translocation of NRG1-ICD mediated by γ-secretase in 293T cells (Bao et al., 2003). However, B4-ECD failed to stimulate the nuclear translocation of NRG1-ICD in NRG1-V321A/V322A-expressing cells (Figs. 3D and 3E). Moreover, the γ-secretase inhibitor had no additional effect on the nuclear level of NRG1-ICD in NRG1-V321A/V322A-expressing cells further implicating valine residues 321 and 322 in γ-secretase-mediated NRG1 processing (Fig. 3E). In sum, the valine residues at positions 321/322 in the C-terminal transmembrane domain are essential for γ-secretase-dependent NRG1-ICD nuclear translocation and transcriptional activation.
Defects in the complexity of dendrites and the outgrowth of axons of cortical neurons in Type III Nrg1 mutant mice (Figs. 1 and and2)2) indicated that aspects of Type III NRG1 signaling contribute to the morphological development of cortical neurons. Type III NRG1-ERBB complexes can signal via activation of ERBB kinases and via the intracellular domain of NRG1 (Bao et al., 2003; Bao et al., 2004; Mei and Xiong, 2008). As such, we first tested whether NRG1 signaling via the intracellular domain contributes to the development of dendrites and axons. We analyzed the growth and branching pattern of dendrites and axons of Type III Nrg1 KO neurons expressing NRG1-WT or NRG1 mutants defective in processing and intracellular domain signaling. KO neurons expressing NRG1-WT had total dendritic length and number of dendritic branch points that were similar to that of WT neurons and were significantly higher than those of KO neurons (n=40 NRG1-WT neurons, Figs. 4A–4C; also see Figs. 4D and 4E, 35 NRG1-WT neurons; P < 0.05). Thus, restoring Type III NRG1 expression in KO neurons rescued dendritic defects in a cell autonomous manner. Expression of NRG1-T307A/G308A restored dendritic length and branch points of KO neurons (n=51 NRG1-T307A/G308A neurons; P < 0.05; Figs. 4B–4C). In contrast, the total dendritic length and number of branch points of NRG1-V321A/V322A expressing KO neurons were comparable to that of KO neurons indicating that the NRG1-V321A/V322A mutant failed to rescue the dendritic defects of KO neurons (n=43 NRG1-V321A/V322A neurons; Figs. 4A–4C). Therefore, the ability of Type III NRG1 to restore normal dendritic phenotypes when expressed in KO neurons requires specific valine residues within the C-terminal transmembrane domain, but not the nearby amino acids 307/308.
To directly demonstrate the importance of nuclear translocation of the NRG1-ICD in the regulation of dendritic development of cortical neurons, we assessed the effects of alanine substitutions of the NLS (Bao et al., 2003; Bao et al., 2004) on the ability of NRG1 to restore normal dendritic phenotypes of KO neurons. As expected, full-length NRG1 expression restored the total dendritic length and branch points in KO neurons (n=16 WT neurons, 48 KO neurons, 35 NRG1-WT neurons; P < 0.05; Figs. 4D and 4E). However, we found that the total dendritic length and number of branch points of NRG1-K329A/Q330A-expressing KO neurons were comparable to that of KO neurons demonstrating that the NLS mutant, NRG1-K329A/Q330A, failed to rescue the dendritic defects of KO neurons (n=37 NRG1-K329A/Q330A neurons; Figs. 4D and 4E). Collectively, our results indicate that NRG1 mutants defective in γ-secretase-dependent signaling are unable to promote dendritic growth and branching of cortical neurons.
Stimulated by our findings that the elaboration of dendrites requires Type III NRG1 signaling via the intracellular domain, we asked whether axonal growth also involves NRG1 signaling via the intracellular domain. To address this question, we assessed the ability of wild type vs. mutant forms of Type III NRG1 to rescue the axonal phenotypes of KO neurons. In KO neurons expressing full-length Type III NRG1, the total axonal length was similar to that of WT neurons demonstrating the ability of NRG1 to restore normal axonal growth of KO neurons (n=65 NRG1-WT neurons; Fig. 4F). Further, this ability of NRG1 to restore normal axonal growth was not affected by mutations of the transmembrane domain, or of the NLS; KO neurons expressing Type III NRG1 mutants showed comparable total axonal length as WT neurons (n=45 NRG1-T307A/G308A neurons, 34 NRG1-V321A/V322A neurons and 17 NRG1-K329A/Q330A neurons; Fig. 4F). Neither wild type nor mutant forms of Type III NRG1 affected the number of branch points of KO neurons (Fig. 4G). In contrast to dendritic development, Type III NRG1 therefore regulates axonal growth in a cell-autonomous fashion that does not require the transcriptional functions of the NRG1-ICD.
NRG1 binding to ERBB receptor kinases initiates intracellular signaling cascades that are critical to neuronal functions (Mei and Xiong, 2008). Previous studies have implicated NRG1-ERBB signaling in the morphological development of hippocampal neurons and dendritic development of cerebellar granule cells (Gerecke et al., 2004; Rieff and Corfas, 2006). In cortical neurons, it remains unclear whether NRG1 signaling via ERBB kinase activation contributes to the growth and branching of dendrites. We therefore assessed the ability of full-length Type III NRG1 to restore normal dendritic phenotypes when expressed in KO neurons in the presence of an ERBB kinase inhibitor, PD158780. This inhibitor blocks ligand-induced activation of ERBB tyrosine kinases and prevents subsequent tyrosine phosphorylation of ERBB intracellular domain (Supplementary Fig. 4). We found that the total dendritic length and number of branch points were comparable between untreated and ERBB kinase inhibitor-treated NRG1-WT expressing KO neurons (n=48 ERBB inhibitor-treated NRG1-WT neurons; Figs. 4D and 4E; also see Figs. 5C and 5D, n=30 ERBB inhibitor-treated NRG1-WT neurons). Hence, the rescue of dendritic growth and branching by Type III NRG1 does not require ERBB kinase activity.
The results of our pharmacological studies indicate that Type III NRG1 might regulate the morphological development of cortical neurons independent of ERBB kinase activity in vivo. To test this possibility, we analyzed the growth and branching pattern of dendrites and axons of cortical neurons of ErbB4 knockout embryos rescued from early embryonic lethality by cardiac expression of ERBB4 (Tidcombe et al., 2003). The total dendritic length and number of dendritic branch points of ErbB4 knockout cortical neurons (B4KO neurons) were not significantly different from that of WT neurons (n=24 B4KO neurons; Figs. 5C and 5D). Our analysis also revealed that B4KO neurons had normal axonal length and branch points (n=24 B4KO neurons; Figs. 5E and 5F). Thus, the development of dendrites and axons of cortical neurons does not require ERBB4. Collectively, our results from pharmacological and genetic studies indicate that Type III NRG1 signaling regulates growth and branching of cortical dendrites and axons in a cell autonomous manner that does not require ERBB kinase activity.
Our results indicate that the ability of Type III NRG1 to restore normal dendritic phenotypes of KO neurons requires specific valine residues within the C-terminal transmembrane domain critical to intramembranous processing and nuclear translocation. A SNP that has been linked to increased risk for developing schizophrenia results in a leucine substitution for one of these valines, i.e. valine 321 (Walss-Bass et al., 2006). We therefore tested the effects of the comparable valine to leucine substitution of this residue (i.e. valine 321) on the ability of Type III NRG1 to support normal dendritic development in KO neurons (Fig. 5A). As before, the expression of wild type Type III NRG1 in KO neurons restored the dendritic length and branch point number to levels of WT neurons (n=42 WT neurons, 51 KO neurons, 29 NRG1-WT neurons; P < 0.05; Figs. 5C and 5D). However, the total dendritic lengths and branch points were similar between KO neurons and KO neurons expressing the NRG1-V321L mutant demonstrating that the mutant failed to rescue dendritic defects of KO neurons (n= 34 NRG1-V321L neurons; Figs. 5B–5D). Thus, the highly conserved valine residue in the C-terminal transmembrane region is critical to the ability of Type III NRG1 to regulate dendritic phenotypes in cortical KO neurons.
We also assessed the ability of NRG1-V321L mutant to support normal axonal growth. As before, the expression of full-length Type III NRG1 in KO neurons restored total axonal length to levels comparable to that of WT neurons (n=31 WT neurons, 54 KO neurons, 27 NRG1-WT neurons; Fig. 5E). NRG1-V321L mutant also restored axonal growth (n=35 NRG1-V321L neurons; Fig. 5E). In addition, WT neurons, KO neurons and KO neurons expressing either NRG1-WT or NRG1-V321L are not significantly different in axonal branch points (Fig. 5F). We conclude that the highly conserved valine 321 involved in Type III NRG1 signaling via the intracellular domain, plays an important role in NRG1 regulation of dendritic development, but is not required for NRG1 function in axonal growth.
Our results reveal that Type III NRG1 plays a crucial role in the normal development of both dendrites and axons of cortical neurons. Type III NRG1 regulation of dendritic complexity requires γ-secretase-dependent signaling via the NRG1 intracellular domain. Our pharmacological and genetic studies further indicate that Type III NRG1 signaling via the intracellular domain regulates cortical dendritic development independent of ERBB kinase activity and more specifically, ERBB4 function per se. These results not only reflect the ability of Type III NRG1 to function as a receptor but also demonstrate the importance of Type III NRG1 signaling via the intracellular domain in the development of cortical connectivity. In contrast to dendritic development, Type III NRG1 regulation of axon extension does not involve γ-secretase-dependent signaling via the intracellular domain and is likely to be related to γ-secretase-independent Type III NRG1 activation of PtdIns 3-kinase signaling in axons (Hancock et al., 2008). In sum, our results clearly demonstrate differential requirements for Type III NRG1-mediated signaling pathways in the growth and branching of dendrites vs. axons in the developing cortex.
A substitution of valine residue 321 to leucine is associated with schizophrenia (Walss-Bass et al., 2006). In cells expressing NRG1 mutant containing this valine to leucine substitution as well as in cells deficient in γ-secretase, there is an increase in the levels of partially processed NRG1 (Dejaegere et al., 2008). These findings demonstrate that the valine mutation or a lack of γ-secretase results in ineffective NRG1 processing thereby implicating a role of valine 321 in γ-secretase-mediated intramembranous cleavage. We show here that the valine substitution abolishes Type III NRG1 signaling via the intracellular domain and leads to dendritic defects of cortical neurons presumably by disrupting γ-secretase-dependent intramembranous processing.
Type III NRG1 rescue of the dendritic defects of Type III Nrg1 KO cortical neurons does not require ERBB kinase activity. In addition, cortical neurons from mice lacking ERBB4 do not have altered dendritic phenotypes. Therefore, Type III NRG1 appears to regulate cortical dendritic development independent of ERBB kinase activation, and ERBB4 function per se. Clearly, our results in cortical pyramidal neurons are distinct from demonstrations for NRG1 function via ERBB kinase signaling in the morphological development of hippocampal neurons and cerebellar granule cells (Gerecke et al., 2004; Rieff and Corfas, 2006) as well as in promoting the maturation of dendritic spines (Barros et al., 2009). It remains unclear as to why NRG1 signaling via ERBB kinases plays a role in the morphological development of hippocampal neurons and cerebellar granule cells but not in cortical neurons. The differences in NRG1 vs. ERBB requirement might reflect differences in their expression patterns. In the rodent hippocampus and cortex, ERBB4 expression has recently been shown to be selectively expressed in interneurons and not pyramidal neurons (Fazzari et al., 2010; Neddens and Buonanno, 2009; Vullhorst et al., 2009). In cerebellum, ErbB4 mRNA is expressed only in mature granule cells and not in other neuronal types (Elenius et al., 1997).
The lack of dendritic phenotypes in cortical neurons from the ErbB4 mutant embryos raises the question of how NRG1 signaling is stimulated. Synaptic activation and membrane depolarization have been shown to stimulate γ-secretase-dependent cleavage of Type III NRG1 as well as other substrates (Bao et al., 2003; Bao et al., 2004; Inoue et al., 2009). There is mounting evidence for activity-dependent regulation of dendritic development of neurons (Chen and Ghosh, 2005; Flavell and Greenberg, 2008) and the presence of spontaneous activity in cultured cortical neuronal networks (Kamioka et al., 1996; Lin et al., 2002). Neuronal activity may thus represent a critical signal for triggering Type III NRG1 proteolytic processing and signaling during cortical development.
In summary, we demonstrate that Type III NRG1 plays an essential role in the development of cortical connectivity. The current study also uncovers a signaling mechanism for Type III NRG1 that is specifically required for the development of dendrites but not axons of cortical neurons and provides insights into the ERBB kinase activity-independent mechanisms that control the early growth and patterning of cortical dendrites. Valine 321, which is required for NRG1 proteolytic cleavage, nuclear translocation and function in dendritic development, is associated with increased risk for developing schizophrenia (Walss-Bass et al., 2006). This raises the intriguing possibility that altered NRG1 intramembranous processing and defective signaling via the intracellular domain of NRG1 contribute to abnormal development of neuroconnections implicated in the etiology of schizophrenia.
We thank Ping Ting Chen and Adnaan Sheriff for technical support and image analysis. YC thanks David Lin for support and suggestions. This work was supported by grants from NARSAD/Baer Foundation (YC, DAT and LWR), McKnight Brain Disorders (LWR) and the National Institute of Health (DK07328 to MLH; NS29071 to LWR and DAT).
Author ContributionsY.C. designed and performed all experiments in cortical neurons, analyzed data, and prepared the manuscript and figures. M.L.H. generated Gal4DBD fusion proteins containing wild type and mutant forms of mouse Type III NRG1, performed the initial alanine scanning mutagenesis, prepared B4-ECD proteins and tested the specificity of anti-CRD antibody in the spinal cord. D.A. T. performed experiments testing for the effectiveness of the ERBB kinase inhibitor in T47D cells. D.A.T. and L.W.R. provided valuable suggestions in the design of experiments, in data analysis and edited the manuscript. The authors declare no conflict of interest.