Ubiquitin ligases of the Nedd4 family are important for axon and dendrite development, but little is known about their adaptor, Nedd4 family-interacting protein 1 (Ndfip1), that is responsible for their enzymatic activation. To study the function of Ndfip1 in cortical development, we generated a conditional knock-out (conditional KO) in neurons. The Ndfip1 conditional KO mice were viable; however, cortical neurons in the adult brain exhibited atrophic characteristics, including stunted dendritic arbors, blebbing of dendrites, and fewer dendritic spines. In electron micrographs, these neurons appeared shrunken with compacted somata and involutions of the nuclear membrane. In culture, Ndfip1 KO neurons exhibited exuberant sprouting suggesting loss of developmental control. Biochemical analysis of postsynaptic density (PSD) fractions from Ndfip1 KO cortical and hippocampal neurons showed that the postsynaptic proteins (Arc and PSD-95) were reduced compared with wild-type controls. In addition, the PI3 kinase/Akt signaling pathway was altered. These results indicate that Ndfip1, through its Nedd4 effectors, is important for the development of dendrites and dendritic spines in the cortex.
Appropriate development of dendritic arbors and excitatory synaptic connections, which form on dendritic spines, is crucial for normal cognition. Specifically, decreases in the number of spine or increases in the number of spines with an “immature” morphology (thin protrusions lacking a well-developed spine “head”) correlate strongly with deficits in excitatory synaptic function and mental retardation (Fiala et al. 2002). These disease-associated changes in spine density and shape are frequently caused by mutations in genes that regulate dendrite growth and synapse formation, but in many cases the underlying biochemical mechanisms are unknown (Blanpied and Ehlers 2004; van Galen and Ramakers 2005).
Recently, ubiquitination has emerged as a key process in regulating the spatial and temporal distribution of proteins involved in establishing neuronal polarity, branching patterns, and synaptic connectivity (Yi and Ehlers 2007). A major consequence of protein modification by the addition of ubiquitin (Ub) chains is degradation by the Ub–proteasome system, thereby removing proteins that are no longer required (Glickman and Ciechanover 2002). Ubiquitination of proteins is a tightly regulated process, requiring the combined activity of 3 enzymes: An E1 Ub-activating enzyme, an E2 Ub-conjugating enzyme, and an E3 Ub-ligase (Hershko and Ciechanover 1998). Although the E3 Ub ligases may directly recognize substrates via specific binding motifs, in many cases they require the additional participation of adaptors for target recruitment (Shearwin-Whyatt et al. 2006).
The Nedd4 family-interacting protein 1 (Ndfip1) is an adaptor protein for the Nedd4 family of E3 Ub ligases that consists of 9 members (Yang and Kumar 2010). Of these, Ndfip1 has been shown to interact with Nedd4 family proteins such as Nedd4-1, Nedd4-2, Itch, and WWP2 (Yang and Kumar 2010). In addition to its adaptor function, Ndfip1 is an activator, releasing Nedd4 from its autoinhibitory conformation (Mund and Pelham 2009). The adaptor function of Ndfip1 serves to broaden the range of substrates, in particular proteins lacking the amino acid motif PPxY (PY), necessary for binding to the WW domains of Nedd4 (Yang and Kumar 2010). The importance of the Nedd4 system for nervous system development is demonstrated by 2 recent reports. In nonmammalian neurons, Nedd4-mediated ubiquitination promotes axonal branching through ubiquitination and downregulation of phosphatase and tensin homology deleted on chromosome 10 (PTEN) (Drinjakovic et al. 2010). Nedd4-1 has also been reported to promote dendritic branching in mouse neurons by ubiquitinating a negative regulator of branching from the Rho family of small GTPases (Kawabe et al. 2010).
Our recent studies have shown that Ndfip1 is an important partner for Nedd4 activity in mouse and human neurons, specifically for neuroprotection during stress and brain injury (Sang et al. 2006; Howitt et al. 2009, 2012; Lackovic et al. 2012). As these proteins are also expressed in the developing brain, we hypothesize that Ndfip1 has important functions during neuronal development. In the present report, we show that Ndfip1 is highly expressed in the embryonic and early postnatal cortex. Using a conditional knock-out (conditional KO) strategy to inactivate the Ndfip1 gene, we assessed the consequences of the lack of Ndfip1 for developing cortical neurons. In pyramidal neurons, the soma, dendritic branches, and dendritic spines exhibited signs of atrophy, with branches being shorter and spines reduced in number. Direct observations of cultured neurons suggest that an initial period of exuberant dendritic sprouting may precede the atrophy and loss of branches and spines. This finding, together with the detection of lower levels of postsynaptic proteins (PSD-95 and Arc) in Ndfip1 mutant mouse brains, supports a functional role for Ndfip1 in the development of dendrites and spines. In addition, altered levels and staining patterns of phospho-Akt (pAkt) were detected in Ndfip1 mutant mouse brains, suggesting that misregulation of the PI3-kinase pathway may be responsible for the atrophic neuron phenotype.
Materials and Methods
Mice were housed in the Florey Institute of Neuroscience and Mental Health animal house under standard conditions with ad libitum access to food and water. All animal procedures were approved by the Institutes Animal Ethics Committee. For the collection of embryonic (E) brain tissue from timed-mated pregnant females, the day of plug detection was designated E0.5. Ndfip1 total KO mice were produced by breeding the Ndfip1flox/flox mouse line (Howitt et al. 2012) with the Cre deleter mouse (Schwenk et al. 1995). Ndfip1 conditional KO mice were produced by breeding either the Emx1-Cre transgene (Iwasato et al. 2000) or the Nestin-Cre transgene (Tronche et al. 1999) into the Ndfip1flox/flox line. Conditional KO mice used for this study were therefore heterozygous for the Cre transgene and homozygous for the floxed Ndfip1 allele. As controls for normal levels of Ndfip1 expression, we used either wild-type C57Bl6/J, Nestin-Cretg/wt, Emx1-Cretg/wt, or Ndfip1flox/flox Cre–ve on a C57Bl6/J background.
Immunostaining of Neocortex
Newborn (postnatal day 1; P1) 1-week-old pups or adult mice were transcardially perfused for 5 min (after sodium pentobarbitone anesthesia, 80 mg/kg) with 4% paraformaldehyde (PFA) in Sorensen's phosphate buffer (PB: 0.1 M phosphate buffer, pH 7.4). Dissected brains were then postfixed in 4% PFA and cryoprotected (20% sucrose in PB, overnight at 4 °C) before being embedded in OCT (Tissue-Tek, Miles, Torrance, CA, USA) for cryosectioning. Coronal sections (14 μm) of the cortex were prepared and mounted on Superfrost® Plus glass slides (Menzel-Gläser, Braunschweig, Germany). For Ndfip1 immunostaining, sections were incubated overnight at room temperature (RT) in primary antibody (rat anti-Ndfip1 monoclonal antibody 1G5; generated in-house, 1:250 in blocking buffer containing 10% fetal bovine serum and 0.1% v/v Triton X-100) and then in an Alexa Fluor 594-congugated goat anti-rat (1:500 Invitrogen, Carlsbad, CA, USA). Additional antibodies used for fluorescent immunostaining were: for Nedd4-2 staining, rabbit polyclonal anti-Nedd4-2 (Abcam, Cambridge, MA, USA; 1:100); for PTEN staining, rabbit monoclonal anti-PTEN (clone Y184, Abcam, Cambridge, UK 1:100); for pAkt S473 (pAkt), rabbit monoclonal anti-pAkt S473 (Cell Signaling, Temecula, CA, USA 1:100). Following washes, sections were mounted in 10% Mowial (Hoechst, Melbourne, Australia) in 25% glycerol, 0.1 M Tris–HCl, pH 8.5, and 2.5% (w/v) 1,4-diazobicyclo-[2.2.2]-octane (Sigma, St Louis, MO, USA). For labeling of neuronal subtypes, we used the following primary antibodies (diluted in 0.1 M PBS with 0.3% Triton X-100): a mouse monoclonal to NeuN (Chemicon, Temecula, CA, USA; 1:200), a rabbit polyclonal to Emx1 (1:200, gift from Alain Trembleau), a rat monoclonal to Ctip2 (Abcam, UK; 1:500), a rabbit polyclonal to γ-aminobutyric acid (GABA) (Sigma, USA; 1:500), and a rabbit polyclonal to Cux1/CDP (Santa Cruz Biotechnology, Inc., Santa Cruz, CA, USA; 1:500). Secondary antibodies used were an Alexa Fluor 594-conjugated goat anti-mouse IgG (Molecular Probes, Eugene, OR, USA; 1:500) and an Alexa Fluor 488-conjugated goat anti-rabbit IgG (Molecular Probes, 1:500). For pAkt S473 immunohistochemistry, sections (8 μm) from wax-embedded brains were incubated with a pAkt S473 antibody (Cell Signaling rabbit monoclonal, 1:50) overnight and for the detection of a positive signal, a VECTASTAIN ABC kit (Vector Laboratories, Burlingame, CA, USA) was used according to the manufacturer's instructions. Sections were counterstained with hematoxylin (Vector Laboratories) for 4 min.
Western Blotting Analysis of Brain Lysates and Subcellular Fractions
Protein lysates were prepared in RIPA buffer containing Complete Mini Protease inhibitors (Roche Diagnostics, Mannheim, Germany) and PhosSTOP phosphatase inhibitors (Roche Diagnostics). For subcellular fractionation, synaptosomes were prepared as described (Srivastava et al. 1998) and further fractionated (Phillips et al. 2001). Cortical and hippocampal tissue lysates/fractions were separated on 7.5–10% polyacrylamide gels and transferred onto supported nitrocellulose (Hybond™-C, Amersham Biosciences, UK) or Biotrace PVDF membranes (Pall Corporation, Port Washington, NY, USA). Primary antibodies (and dilutions) used to probe western blots were as follows: Ndfip1 (rat monoclonal antibody 1G5, 1:1000), β-actin (mouse monoclonal, Sigma, USA; 1:5000), Arc (activity-regulated cytoskeleton-associated protein; rabbit polyclonal, Synaptic Systems, Göttingen, Germany, 1:1000), p-Akt S473 (rabbit monoclonal, Cell Signaling; 1:1000), and PTEN (rabbit monoclonal, clone Y184, Abcam, UK; 1:2000). Blots of subcellular fractions were probed with antibodies to PSD-95 (mouse monoclonal Millipore, Billerica, MA, USA; 1:1000) and synaptophysin (mouse monoclonal Clone SV-38, Sigma, Australia; 1:1000). The secondary antibody was either horseradish peroxidase (HRP)-conjugated goat anti-rat (Upstate Cell Signaling; 1:5000), HRP-conjugated goat anti-rabbit Ig (Upstate, 1:5000), or HRP-conjugated goat anti-mouse IgG (Upstate, 1:5000). Proteins were detected using ECL reagent according to the manufacturer's instructions (GE Healthcare, Little Chalfont, Buckinghamshire, UK) and visualized by exposure to X-ray film.
Mice were perfused with 4% PFA and the brains were collected. Coronal slices (1 mm) were made using a mouse brain matrix (Ted Pella, Inc., Redding, CA, USA), and sections corresponding to −0.10 to −1.10 of the bregma line (Franklin and Paxinos 1997) were collected. Small pieces (∼1.0 mm × 1.0 mm) containing deep-layer neurons were postfixed overnight in Karnovsky's fixative (4% PFA, 2.5% glutaraldehyde in Na+ cacodylate buffer pH 7.4), washed in Na+ cacodylate buffer (3 × 5 min), and stored at 4 °C for resin embedding.
Sections were prepared using standard electron microscopy protocols.
Brains were processed using the Rapid Golgi Stain™ kit from FD NeuroTechnologies (Columbia, MD, USA) according to the manufacturer's instructions, with the modification that the mounting solution (Solution C) was diluted 1:2 with water. Sections (100 μm) were prepared and basal dendritic arbors of pyramidal LV neurons were traced using the Neurolucida™ software.
Primary Neuron Culture, Transfection, and Immunostaining
Cortices from wild-type or Ndfip1 KO mouse embryos (E15) were dissected into cold PBS. Tissue was digested with the Papain Dissociation Kit (Worthington Biochemical Corporation, Lakewood, NJ, USA). Neurons were plated at 1–2 × 105 on poly-dl-ornithine/laminin-coated coverslips in Neurobasal medium + 2% B27 supplement (Invitrogen, VIC, Australia), 0.5 mM glutamine, and antibiotics (50 IU/mL penicillin and 50 μg/mL streptomycin). For dendritic measurements, cultures at 19 days in vitro (DIV) were lipofectamine-transfected with Venus-YFP at a concentration of 1 μg DNA/12 well coverslip. At 25 DIV, the cultures were fixed with 4% PFA for 3 min and then mounted in 10% Mowial (Hoechst) in 25% glycerol, 0.1 M Tris–HCl, pH 8.5, and 2.5% (w/v) 1,4-diazobicyclo-[2.2.2]-octane (Sigma, USA).
Rat hippocampal neurons (from E18 rat embryos) were cultured and transfected as described (Kennedy et al. 2010). Neurons were transfected (after 18 DIV) with a plasmid expressing a fusion protein of Ndfip1-enhanced green fluorescent protein (Ndfip1-EGFP) using Lipofectamine 2000 (Invitrogen, Australia). The ratio of DNA:Lipofectamine used was 1 μg DNA:1 μL Lipofectamine 2000. Lipofectamine/DNA complexes were incubated on the neurons for 45 min before replacement of the conditioned medium. Neurons were fixed after 21 DIV (4% PFA, 5 min at RT). Untransfected neurons were immunostained with primary antibodies against Ndfip1 (rat anti-Ndfip1 monoclonal antibody 1G5; 1:1000) and Bassoon, a presynaptic marker (guinea pig polyclonal antibody; Synaptic Systems; 1:1000).
Image Analysis and Statistics
Sholl analysis was carried out in ImageJ (National Institutes of Health), and the mean number of dendritic branch crossings at a given radius from the center of the cell soma was plotted for neurons of each genotype. The distribution curves were segmented and modeled by linear regression for each segment. The slope and elevation of the lines of best fit were compared using an analysis of covariance (ANCOVA). Dendritic branching parameters and spine numbers from Golgi-Cox impregnated neurons were compared between genotypes using Student's t-test.
Neurons positive for NeuN (all neurons), Cux1/CDP (upper layer neurons), and Ctip2 (lower layer neurons) in P30 and P60 cortices were counted using ImageJ, and numbers were compared between genotypes using Student's t-test. Cortical thickness was measured using ImageJ on NeuN-stained sections, and the values were compared using Student's t-test. Values are expressed as mean ± SEM.
Ndfip1 Is Expressed in Embryonic and Postnatal Cortex
To investigate the expression profile of Ndfip1 in the developing neocortex, protein extracts prepared at different time points during pre- and postnatal development were analysed by western blotting and probed with a monoclonal antibody raised against Ndfip1 (Howitt et al. 2009). While Ndfip1 was detected in all samples tested (from E13 to P28; Fig. 1A), the highest levels of Ndfip1 expression were observed at P1 and P7 (relative to the control protein, β-actin). Immunostaining of brain sections at these 2 ages revealed that Ndfip1 is expressed in developing neurons of the cortical plate (Fig. 1B,C) and in all neurons positive for the pyramidal neuron marker Emx1 (Fig. 1C). Ndfip1 expression was observed in the neuronal soma, in which it exhibited a perinuclear distribution with strong expression in apical processes and dendrites (Fig. 1C inset, arrowheads). We previously showed that Ndfip1 in the adult brain functionally interacts with Nedd4-2 and PTEN, thereby acutely altering levels of p-Akt (pAkt S473) (Howitt et al. 2012; Putz et al. 2012). To determine whether the expression pattern of Ndfip1 overlaps with those of its putative interacting partners in the developing cortex, we performed dual immunohistochemistry on P1 brain sections (see Supplementary Fig. 1). Both PTEN and Nedd4-2 were detected in neurons in the cortical plate at P1 with distributions overlapping that of Ndfip1 (see Supplementary Fig. 1A–F). A similar, but much weaker, staining pattern was observed for pAkt S473 (see Supplementary Fig. 1G–I), indicating that cortical levels of pAkt S473 are low at birth.
Conditional Deletion of Ndfip1 in Cortical Pyramidal Neurons Causes Degenerative Changes in Neuronal Morphology
The time course of Ndfip1 expression in the developing cortex led us to hypothesize that Ndfip1 is involved in neuronal maturation and, therefore, important for the development of neuronal circuitry. To test this hypothesis, we adopted a conditional KO deletion strategy in order to circumvent the inflammatory phenotype of a total Ndfip1 gene KO, as previously reported (Oliver et al. 2006). We generated a loxP-flanked Ndfip1 mutant allele and bred homozygous Ndfip1flox/flox littermates carrying the Emx1-Cre transgene for comparison with Emx1-Cre negative or wild-type controls. Emx1 has previously been shown to be specifically localized to pyramidal cortical neurons (Chan et al. 2001).
Immunostaining of control and Ndfip1 conditional KO (Emx1-Cre) brain sections confirmed that Ndfip1 was deleted in all cortical pyramidal neurons, and that the remaining scattered Ndfip1-positive neurons were, without exception, GABA-positive interneurons (see insets: Fig. 2A,C). We then used immunostaining for NeuN to gain an overview of the organization of neurons in the conditional KO cortex. Interestingly, while no gross morphological abnormalities were observed in the conditional KO cortex at P30 (Fig. 2C, Cre +ve) compared with the control (Fig. 2A, Cre −ve), closer examination revealed that neurons in the deep layers of the conditional KO cortex were disorganized (compare Fig. 2B with D). Staining with NeuN showed that neuronal somata in the KO cortex were irregularly arranged compared with control neurons (arrowheads, Fig. 2B) and exhibited lacunae (arrowheads, Fig. 2D). Furthermore, the normal orientation of the pyramidal apex toward the pial surface present in control pyramidal neurons was less defined in the conditional KO neurons. Further investigations of neuronal morphology were performed using transmission electron microscopy of deep-layer pyramidal neurons following conditional Ndfip1 deletion with Nestin-Cre (Fig. 2E,F). These studies revealed that conditional KO pyramidal neurons (at P60) possessed large indentations in their nuclear membranes (arrowheads, Fig. 2F), which were not seen in control neurons (Fig. 2E) and ruffling of the nuclear membrane (arrow, Fig. 2F). In spite of these abnormal features of neuronal morphology and ultrastructure, there was no difference in the total number of neurons (NeuN positive) in layers II/III (Cux1/CDP positive) and V (Ctip2 positive) in the cortices of conditional KO mice and controls (Table 1). Furthermore, differences in cortical thickness were not observed, either at P30 (cortical thickness, control 1.27 ± 0.04 mm; conditional KO 1.27 ± 0.01 mm, n = 5, P = 0.92) or at P60 (cortical thickness, control 1.25 ± 0.01 mm; conditional KO 1.24 ± 0.03 mm, n = 5, P = 0.82).
|Control||1249 ± 132||1386 ± 52||2072 ± 129|
|Nestin-Cre CKO||1301 ± 40 (P = 0.74)||1476 ± 34 (P = 0.23)||2103 ± 137 (P = 0.88)|
|Control||1250 ± 70||1274 ± 83||2101 ± 9|
|Nestin-Cre CKO||1065 ± 45 (P = 0.10)||1182 ± 73 (P = 0.45)||2072 ± 77 (P = 0.74)|
|Control||1249 ± 132||1386 ± 52||2072 ± 129|
|Nestin-Cre CKO||1301 ± 40 (P = 0.74)||1476 ± 34 (P = 0.23)||2103 ± 137 (P = 0.88)|
|Control||1250 ± 70||1274 ± 83||2101 ± 9|
|Nestin-Cre CKO||1065 ± 45 (P = 0.10)||1182 ± 73 (P = 0.45)||2072 ± 77 (P = 0.74)|
Note: Values are mean ± SEM (n = 3).
Dendritic Complexity Is Increased in Embryonic Ndfip1 KO Neurons in Culture
Since the somata of pyramidal neurons without Ndfip1 exhibited clear signs of degeneration, we reasoned that this could be a downstream consequence of abnormal neuron development, because Ndfip1 is most strongly expressed in the developing cortex prior to, and soon after, birth. We tested whether Ndfip1 is important for the initial stages of dendritic arbor development using a neuron culture model. Embryonic (E15) cortical neurons from Ndfip1 total KO mice (in which the Ndfip1 gene was deleted from conception using a total Cre-deleter) were transfected at 19 DIV with a Venus (EYFP, enhanced yellow fluorescent protein) expression plasmid and the dendritic arbors compared at 25 DIV with those of wild-type neurons using Sholl analysis. Concentric circles centered on the neuronal soma (with radii increasing in 1 μm steps) were drawn and the number of branch intersections at each radius plotted. Neurons lacking Ndfip1 exhibited a “bushy” appearance with excessive sprouting around the neuron soma (compare Fig. 3A,B—control neurons with Fig. 3C,D—Ndfip1 KO neurons). Sholl distribution plots were analyzed using ANCOVA, revealing that the observed distributions (partitioned into subregions) were best modeled as 2 separate lines with different slopes and/or elevations (Fig. 3E; F-test results for subregions: 30–100, 100–170, and 240–310 μm from soma: slope d.f. 1,106 P < 0.05, elevation d.f. 1, 107, P < 0.001; 170–240 and 310–400 μm from soma: slope N.S., elevation d.f. 1,109 or 1, 139, respectively, P < 0.001). Thus, embryonic cortical neurons cultured from Ndfip1 KO mouse brains exhibited significantly more complex dendritic arbors than control wild-type neurons, with excess dendritic branches giving the KO neurons a “bushy” appearance.
Adult Ndfip1 Conditional KO Neurons Displayed Stunted Dendrites and Dendritic Spines
Although we had shown that Ndfip1 is important for regulating dendritic patterning and neurite outgrowth in vitro, it was unclear how to reconcile this exuberant dendritic sprouting in vitro with somatic and nuclear degenerative changes observed in conditional KO neurons in vivo. To clarify this, we used Golgi-Cox impregnation of intact brains at various postnatal time points to investigate the effects of Ndfip1 deletion on neuron morphology. Initial observations of Golgi-Cox impregnated conditional KO and control brain sections at approximately 1 month of age (Fig. 4) revealed some obvious differences in neuronal morphology. Layer V pyramidal neurons of conditional KO cortices (Ndfip1loxP/loxP; Emx1-Cre+) exhibited dendritic blebbing (marked with arrowheads in Fig. 4G), which was not seen in control neurons (Fig. 4C). In addition, neurons in the conditional KO cortex appeared to possess smaller dendritic arbors (Fig. 4E–G) when compared with controls (Fig. 4A–C). This was borne out by tracing of Golgi-Cox impregnated neurons (Fig. 4D,H). More extensive tracing of layer V pyramidal neurons from mouse brains at 2 different age ranges revealed that conditional KO neurons consistently displayed reductions in the combined length of their basal dendritic arbors (at P37: total basal dendritic length conditional KO neurons—331.4 ± 62.9 μm, control neurons—887.1 ± 177.7 μm, n = 8, P = 0.02; at P98–120: total basal dendritic length conditional KO neurons—1010.6 ± 111.5 μm, control neurons—1483.1 ± 160.1 μm; n = 16, P = 0.02; Fig. 5A). The reduction in overall length of the dendritic arbor at the older age was at least partially due to a reduction in higher-order (tertiary and above) dendritic branch numbers (Fig. 5B), which also correlated with a lower dendritic spine density on the distal dendritic branches (dendritic spine number/30 μm length distal dendrite: conditional KO: 11.78 ± 0.48, control: 14.14 ± 0.53, P < 0.001; Fig. 5C). To investigate whether the smaller dendritic arbors of neurons lacking Ndfip1 were the end result of excessive dendritic pruning earlier in the postnatal period, we traced Golgi-Cox impregnated neurons from Ndfip1 KO brains at P9–21. Over this time period, dendritic spine density (spine number/μm basal dendritic arbor length) increased in both control neurons and neurons lacking Ndfip1, as expected. Interestingly, however, at both P9 and P21, KO neurons exhibited a significantly lower spine density than control neurons (Fig. 5D). No dramatic change in total dendritic length was observed in either genotype from P9 to P21; however, the KO neurons did exhibit significantly shorter basal dendritic length than control neurons at P14 (Fig. 5E). Collectively, these results confirmed that KO neurons in the postnatal cortex exhibited reduced basal dendritic arbor complexity and reduced dendritic spine densities.
Ndfip1 Is Detected in all Subcellular Fractions of Postnatal Cortical Neurons
The somatic and dendritic localization of Ndfip1 suggests that its subcellular distribution is important for its function during neuron differentiation. To study the subcellular distribution of Ndfip1, western blot analysis of subcellular fractions was performed. The results showed that Ndfip1 was present in the synaptosome fraction (Fig. 6A: lane a). Further fractionation revealed that Ndfip1 was present in the extra-, pre-, and postsynaptic fractions (Fig. 6A—lanes b, c, and d). Significantly, the postsynaptic fraction was free of presynaptic vesicle membrane contamination as demonstrated by the absence of the 38-kDa synaptophysin band (Fig. 6A: lane d), and the postsynaptic fraction was enriched for PSD-95 (compare Fig. 6A: lane d with c). No Ndfip1 was detected in extracts prepared from the Ndfip1 KO cortex (Fig. 6A: lane f). The presence of Ndfip1 in the detergent-extractable postsynaptic fraction is consistent with the observed phenotype in dendritic spines.
To visualize the subcellular localization of Ndfip1 in developing neurons, we immunostained cultured rat hippocampal neurons (21 DIV) for Ndfip1 together with the presynaptic marker Bassoon. Ndfip1 staining throughout the dendrites is punctate; however, no appreciable overlap with the punctate staining for the presynaptic marker Bassoon was observed (Fig. 6D, main panel and inset). Ndfip1-positive puncta were seen in the vicinity of dendritic branch points (Fig. 6B,D and insets, arrowheads) and, to investigate this further, we over-expressed Ndfip1 in cultured rat hippocampal neurons using an EGFP-tagged Ndfip1 construct (transfected at 18 DIV imaged 3 days later at 21 DIV). In neurons over-expressing Ndfip1-EGFP, the fusion protein was distributed throughout the soma and dendritic arbor (Fig. 7A,D). Interestingly, fluorescently tagged Ndfip1 appeared to accumulate at dendritic branch points and at the base of dendritic spines (Fig. 7C,F, arrowheads).
Synaptic Protein Levels Are Reduced in Ndfip1 Knock-out Brain Protein Extracts
The presence of Ndfip1 in the postsynaptic fraction and at the base of dendritic spines, together with the reduced spine densities observed in conditional KO neurons, indicated a likely role in regulating spine composition and function. First, we compared the levels of the scaffolding protein PSD-95 in the PSD of excitatory spine synapses. In KO brain, PSD-95 levels were lower in both neocortical and hippocampal extracts compared with the wild-type (Fig. 7G). This result was consistent with data from dendritic spine counts. Secondly, we compared levels of the activity-regulated cytoskeleton-associated protein Arc, since Arc expression is upregulated by synaptic activity and Arc acts as a homeostatic regulator of synaptic α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor levels (Shepherd et al. 2006). In 2 separate cortical protein extracts from 4-month-old KO mice, the levels of Arc protein were reduced relative to those of wild-type control (Fig. 7H). However, this effect was not seen in a third sample at this age or in samples taken from younger mice (7 weeks of age; Fig. 7H).
Loss of Ndfip1 Affects Steady-State Levels of PAkt S473
As Ndfip1 has multiple known targets (Mund and Pelham 2009), we searched for signaling pathways that could contribute to the neuronal phenotypes observed. Our recent data suggested the involvement of Ndfip1 in the monoubiquitination of PTEN leading to its trafficking into the cell nucleus and into the extracellular environment (Howitt et al. 2012; Putz et al. 2012). Since PTEN is the central regulator of the PI3K pathway, we predicted that the lack of Ndfip1 would result in altered PI3K signaling leading to changes in pAkt abundance in neurons, and this was observed in 62% of the 21 paired Ndfip1 KO and control age-matched extracts we analyzed. In spite of pAkt levels being highly variable over the first postnatal months, increased Akt activity (phosphorylated S473) was seen in 38% of cortical extracts tested. For example, a 1.4-fold increase in cortical pAkt S473 and a 1.6-fold increase in the hippocampal extract prepared from the same P30 animal are shown in Figure 8A. Interestingly, increased pAkt was observed even when the steady-state level of PTEN did not differ between Ndfip1 KO extracts and controls (Fig. 8B). To investigate whether the evidence for altered pAkt could be correlated with an altered staining pattern in neurons, immunostaining for pAkt (S473) was performed. In sections from nestin conditional Ndfip1 KO mice (7-month old), pAkt staining in cortical neurons displayed a distinct pattern, with prominent perinuclear staining observed in some deep-layer neurons, compared with control (compare Fig. 8E,F, arrowheads in inset). Furthermore, in the aging (10 month) cortex, pAkt staining in large projection neurons was clearly different between the Emx1 conditional Ndfip1 KO cortex and control. More intense somatodendritic staining was observed so that the apical dendrites could be more clearly discerned (Fig. 8G,H). Taken together, these western blot and immunostaining results indicate that Ndfip1 regulates PI3K-Akt signaling in cortical neurons.
Ndfip1 is an adaptor and activator for E3 Ub ligases belonging to the Nedd4 family (Harvey et al. 2002; Shearwin-Whyatt et al. 2006). Binding of Ndfip1 to target proteins leads to Nedd4-mediated ubiquitination resulting in degradation or trafficking of Ndfip1-bound substrates. Ndfip1 has established roles in the regulation of inflammation (Oliver et al. 2006; Ramon et al. 2011) and cellular homeostasis (Konstas et al. 2002; Howitt et al. 2009), suggesting that Ndfip1-mediated ubiquitination has diverse physiological consequences. Such diversity of function is reflected in the varied nature of Ndfip1 substrates, which, to date, include the divalent metal transporter DMT1, transcription factor JunB, growth factor receptors, and the tumor suppressor PTEN (Oliver et al. 2006; Foot et al. 2008; Mund and Pelham 2009; Howitt et al. 2012; Putz et al. 2012). In the central nervous system, Ndfip1 is involved in neuronal responses to trauma (Sang et al. 2006) and cerebral ischemia (Howitt et al. 2012; Lackovic et al. 2012) and elevated Ndfip1 levels after injury are correlated with increased neuronal survival (Sang et al. 2006; Howitt et al. 2012). As Ndfip1 is clearly important for protecting neurons under threat, we reasoned that Ndfip1 regulation of cellular homeostasis and survival could be important for neuronal development and function.
Using a combination of in vitro and in vivo techniques to analyze the consequences of gene inactivation, we demonstrate here that Ndfip1 is required for dendritic arbor development. The pattern of Ndfip1 expression in cortical neurons during the latter part of prenatal development and the early postnatal period coincides with the growth and extension of the basic dendritic framework. Studies in rodents and humans have revealed that these processes begin prenatally, soon after the cessation of neuron migration. The extended human developmental trajectory relative to that of rodents has enabled the identification of several discrete periods of rapid dendritic arbor growth (Mrzljak et al. 1988, 1992; Petanjek et al. 2008). In the developing human neocortex, the basic scaffold of the apical and basal dendrites of pyramidal neurons develops relatively early (17–25 weeks of gestation), prior to the arrival of the thalamocortical afferents. In this first phase, basal dendrites of large layer IIIC and layer V pyramidal neurons slowly increase in number (Mrzljak et al. 1988, 1992). At around 27–32 weeks of human gestation, basal dendrites of LIIIC and LV pyramidal neurons extend rapidly, predominantly via an increase in the number of bifurcations and the growth of terminal segments (Mrzljak et al. 1992). This period of intensive dendritic arborization coincides with the growth of afferent fibers into the cortical plate and also with the appearance of dendritic spines on LIII and LV pyramidal neurons (Mrzljak et al. 1988). In rodents, synapses are relatively sparse in the cortical plate at prenatal stages, with the vast majority developing in the early postnatal period, correlating with sensory input (Valverde 1967; Miller 1981; Riccio and Matthews 1985).
Subsequent growth of primary dendrites and extension of a complex branched arbor occurs through environmental cues provided by cell adhesion molecules, neurotrophins, and neuronal activity (McAllister et al. 1995; Whitford et al. 2002; Gunnersen et al. 2007; McAllister 2007; Cline and Haas 2008; Lohmann and Bonhoeffer 2008; Kerschensteiner et al. 2009). An important role for Ndfip1 in regulating dendritic growth in the early postnatal rapid-growth phase is indicated by the appearance of morphological abnormalities in Ndfip1 KO neurons (e.g., reduced spine densities) from P9, which may have resulted from delayed thalamocortical afferent innervation and/or reduced afferent activity in Ndfip1 KO mice. In addition, reduced dendritic arbor lengths were evident by P37 and fewer higher-order branches and a lower spine density by 3–4 months. These data indicate that Ndfip1 may also play a role in the elongation, stabilization, and maintenance of newly formed/rearranging dendrites, similar to that which occurs from 2 to 16 months in the human cortex (Vukšić et al. 2002; Petanjek et al. 2008).
At later time points, Ndfip1 KO neurons exhibited dendritic blebbing, akin to that described in ischemic brains (Li and Murphy 2008; Tran et al. 2012) or the “vacuolar change” described by Marin-Padilla in pyramidal neuron dendrites of Down syndrome (DS) motor cortex (Marin-Padilla 1976). Furthermore, in Ndfip1 KO cortex, stunting of dendritic arbors and reduced dendritic spine numbers on large layer V pyramidal neurons was observed, reminiscent of the morphological abnormalities observed in DS children (reviewed by Rueda et al. 2012). In DS, smaller dendritic spines were observed on pyramidal neurons from the newborn visual cortex (Takashima et al. 1981), although progressive dendritic arbor atrophy was not observed until 4 months of age (Takashima et al. 1981; Becker et al. 1986; Vukšić et al. 2002). In the DS brains studied by Becker and colleagues (4 months to 7 years of age), excessive early outgrowth of dendritic branches (similar to our observations in cultured Ndfip1 KO neurons) was followed by atrophy (Becker et al. 1986), and it was suggested that the excessive dendrite branching may have represented a compensatory response to the decreased number of spines and synapses. However, despite signs of degeneration in Ndfip1 KO pyramidal neurons across the cortical depth, layering order was intact, as was neuronal number and cortical thickness.
The subcellular distribution of endogenous Ndfip1 in developing neurons, which appears as punctate staining in dendrites and in the soma, is similar to that of other proteins associated with endosomal vesicles (Kim et al. 1999; Deane et al. 2013). When over-expressed, EYFP-labeled Ndfip1 was also found along dendrites and was seen to accumulate at dendritic branch points, in a similar location to that reported for endosomal huntingtin and the sodium/hydrogen exchanger NHE6 (Kim et al. 1999; Deane et al. 2013) and also that of Golgi outposts (Horton et al. 2005). Interestingly, septins are also enriched at dendritic branch points and at the base of dendritic spines (Tada et al. 2007), raising the possibility of a functional interaction between septins and Ndfip1 since knockdown of septins (Tada et al. 2007) or KO of Ndfip1 impaired dendritic complexity. Taken together, our data suggest that Ndfip1 is trafficked in membranous vesicles resembling endosomes, and these are probably trafficked through Golgi outposts at dendritic branch points (Horton et al. 2005).
Contrary to the in vivo data, Ndfip1 KO neurons in vitro displayed exuberant sprouting of primary/lower order dendrites. A likely reason for this discrepancy is that patterning and growth of the dendritic arbor in vivo are regulated by a complex interplay of cell intrinsic and cell extrinsic mechanisms, including the development of functional synapses. Thus, it is not surprising that disrupting this spatiotemporal organization of environmental cues by placing immature neurons into culture has affected the patterning of the basic dendritic scaffold. For example, cerebellar Purkinje neurons fail to develop terminal branches in vitro in the absence of contact with granule cells (Morrison and Mason 1998), while over-activation of brain-derived neurotrophic factor signaling also caused a relative loss of distal branches while promoting the addition of extra primary dendrites (Horch et al. 1999; Yacoubian and Lo 2000). Similarly, in vivo, the disorganized neuronal orientation, dystrophic dendrites, and fewer dendritic spines of Ndfip1 KO neurons are likely indicative of disturbed extrinsic regulatory cues including afferent inputs.
How could the lack of Ndfip1 lead to the degeneration of cortical principal neurons? One candidate pathway is the PI3K-Akt pathway that promotes growth and survival in diverse physiological systems (Engelman et al. 2006; Carracedo and Pandolfi 2008). Aberrant regulation of PI3K signaling could contribute to the observed dendritic abnormalities, since this pathway is important for regulating neuronal polarity, dendritic branching, and synaptic plasticity (Jaworski et al. 2005; Kumar et al. 2005). Hyperactivation of the PI3K pathway was seen to enhance the size and complexity of dendritic arbors and soma size, while pharmacological inhibition produced the opposite effects (Jaworski et al. 2005; Kumar et al. 2005). Similarly, deletion of PTEN, the central negative regulator of the PI3K pathway, resulted in neurons with hypertrophic and ectopic dendrites and axons along with increased synapses (Kwon et al. 2006). Our recent studies showed that Ndfip1 can regulate the mono- and polyubiquitination of PTEN resulting in PTEN trafficking into the cell nucleus and causing increased pAkt (Howitt et al. 2012). In the present study, changes in the steady-state levels of PTEN in KO cortical extracts were not observed in biochemical assays, although we cannot exclude dynamic perturbations to cellular pAkt caused by, for example, momentary depletion of PTEN from the cytoplasm due to nuclear trafficking or an increase in inactive phosphorylated PTEN. This local regulation and compartmentalization of Akt signaling has been shown to be important for neuron arbor growth (Yan et al. 2006; Henle et al. 2011). In line with the dynamic nature of this regulation, we found that the lack of Ndfip1 correlated with increased cortical pAkt levels in over one-third of the mice analyzed (n = 21) and more prominent perinuclear or somatodendritic pAkt immunostaining in pyramidal neurons. We suggest that the absence of Ndfip1 has caused misregulation of pAkt in neurons, resulting in pathological changes in pyramidal neuronal morphology. Work by others has shown that exogenous expression of constitutively active pAkt in primary hippocampal neurons can induce extra dendritic sprouting (Kumar et al. 2005), similar to the phenotype we observed in cultured Ndfip1 KO neurons, and increased dendritic arbor size (Jaworski et al. 2005).
In conclusion, our results show that Ndfip1 is required for the maintenance of cortical neurons and is an important regulator of the development and maintenance of neuronal connectivity. Ndfip1 has a known role in protecting neurons after injury, and we show here that it is also crucial for the maintenance of synapse numbers and neuronal integrity during development.
This work was supported by the National Health and Medical Research Council Project Grants 566620 and GNT1008787 and the Victorian Government through the Operational Infrastructure Scheme.
We thank Ms Alison Macintyre for helpful discussion and Dr Mark Habgood for assistance with statistical analyses. Conflict of Interest: None declared.