NF-κB regulates neuronal ankyrin-G via a negative feedback loop

The axon initial segment (AIS) is a neuronal compartment defined by ankyrin-G expression. We here demonstrate that the IKK-complex co-localizes and interacts with the cytoskeletal anchor protein ankyrin-G in immunoprecipitation and proximity-ligation experiments in cortical neurons. Overexpression of the 270 kDa variant of ankyrin-G suppressed, while gene-silencing of ankyrin-G expression increased nuclear factor-κB (NF-κB) activity in primary neurons, suggesting that ankyrin-G sequesters the transcription factor in the AIS. We also found that p65 bound to the ank3 (ankyrin-G) promoter sequence in chromatin immunoprecipitation analyses thereby increasing ank3 expression and ankyrin-G levels at the AIS. Gene-silencing of p65 or ankyrin-G overexpression suppressed ank3 reporter activity. Collectively these data demonstrate that p65/NF-κB controls ankyrin-G levels via a negative feedback loop, thereby linking NF-κB signaling with neuronal polarity and axonal plasticity.


Results
The IKK-complex co-localizes and interacts with ankyrin-G in vitro and in vivo. To determine subcellular accumulation of NF-κB pathway proteins in the central nervous system, we used gradient centrifugation to separate synaptosomal and lipid fractions derived from the adult mouse cortex. Anti-IKKα/β (inhibitor of kappa light polypeptide gene enhancer in B-cells, kinase) co-localized with anti-NEMO (NF-κB essential modulator) and anti-p65/NF-κB reactivity in ankyrin-G-positive lipid fractions in addition to synaptophysin-positive synaptosomes (Fig. 1A,B). We next determined whether anti-IKKα/β immunoreactivity co-immunoprecipitates with anti-ankyrin-G (AnkG). Following pull-down of ankyrin-G proteins from mouse cortical extracts using anti-ankyrin-G monoclonal and polyclonal antibodies, anti-IKKα/β immunoreactivities were strongly detected, but not found in the lanes of negative control precipitates (Fig. 1C, Suppl. Fig. 1A,C). These findings were validated following reverse immunoprecipitation of IKKβ as well as ankyrinG-eGFP (ankyrinG::enhanced green fluorescent protein) pull-down, which co-immunoprecipitated IKKβ-Flag as detected by anti-Flag antibodies ( Fig. 1D-G, Suppl. Fig. 1B). We then investigated how ankyrin-G and IKK-signalosome components co-localized to the cytoskeleton in the axon initial segment following detergent extraction. IKKγ/NEMO and IKKα/β immunoreactivities were retained at the proximal AIS cytoskeleton following Triton X-100-or saponin-mediated extraction of cytosolic proteins ( Fig. 2A-D). Following overexpression of Flag-tagged IKKβ we found the protein was also present in the AIS when detected by anti-Flag antibodies. This localization was preserved following detergent extraction, where anti-Flag immunoreactivity co-localized to co-overexpressed ankyrinG-eGFP in the cells' soma and in the AIS (Fig. 2E,F).

Interaction of IKKα/β and ankyrin-G in situ.
To further analyze whether anti-IKKα/β epitopes were found in molecular proximity (<  40 nm) to ankyrin-G epitopes, we employed proximity-ligation assays (PLA, Schematic Fig. 3A). Following plasmid transfection, we found significantly more anti-ankyrin-G & anti-IKKα/β PLA-derived spots in ankG-eGFP plus IKKβ overexpressing cortical neurons than we found in control transfected neurons, and notably more than we found in ankG-eGFP or IKKβ plus eGFP expressing neurons or neurons transfected with the unrelated EB1-eGFP vector ( Fig. 3B-G), overall suggesting antigen-specific reactions.
Gene silencing of ankyrin-G increases, while overexpression of ankyrin-G decreases NF-κB dependent reporter-gene activity. IKKα/β is a positive regulator of p65-phosphorylation and NF-κB-dependent transcription. Surprisingly, we found that ankyrin-G overexpression depressed NF-κB activity in mature cortical neurons (Fig. 4A), while ankyrin-G small-hairpin vector-mediated depletion significantly increased NF-κB reporter gene activity over a period of five days (Fig. 4B,C). We also tested the validity of the reporter construct for p65/NF-κB-dependent transactivation by overexpression or knockdown of the transcription factor (Fig. 4A,C). Notably, comparable to IKK-complex immunreactivity (Fig. 2), anti-p65 immunoreactivity was preserved following detergent extraction using two different antibodies ( Fig. 5A-D). We additionally determined that an overexpressed p65 photoactivatable-GFP (p65-paGFP) fusion protein exhibits significantly decreased mobility in the AIS when compared to paGFP control. Dynein-dependent microtubule-based transport of p65-eGFP in dendrites was previously reported 27,28 . Despite the cytoplasmic diffusion barrier in the AIS 11,29 , p65-paGFP mobility in the AIS was comparable to the distal axon and dendrites, suggesting similar retention and active transport mechanisms in all three compartments. paGFP mobility was significantly lower in the neurites compared to the AIS (Fig. 5E-I). Overall, these results suggested that a pool of neuronal p65/NF-κB protein is sequestered by the detergent-insoluble cytoskeleton at the AIS. p65/NF-κB binds the ankyrin-3 gene promoter sequence and regulates neuronal ankyrin-G expression via a negative feedback. Next, we decided to test whether NF-κB transcriptional feedback controls ankyrin-G expression levels. We focused on a stretch of about 2000 base pairs around the transcription start site (TSS) upstream of exon 1e, expressed in many tissues including the frontal cortex, hippocampus and caudate putamen 5 , of the mouse ankyrin-G (ank3) gene. The region was identified on ENSEMBL (ENSMUSR00000028730) and the eukaryotic promoter database ('Ank3_1' , epd.vital-it.ch, ref. 30). We examined IKK-complex co-immunoprecipitate from adult mouse cortical lysates. Mouse cortices were homogenized and minced in Na-Sucrose buffer. Equivalent amounts were used for immunoprecipitation (IP) using antiankyrin-G, control anti-IgG, anti-MAP2 or anti-α-Tubulin antisera, followed by immuno-blotting (IB) using anti-ankyrin-G (463), anti-pan-IKKα/β (H-470) and anti-MAP2 antibodies. Full-blot views including spliced supernatant control bands in Suppl. Fig. 1A. (D) For the reverse experiment, cortical tissue was processed and immunoprecipitated using a rabbit polyclonal antibody to IKKα/β (H-470, SCBT), a rabbit polyclonal anti-IgG antiserum served as control. Rabbit-polyclonal antibodies against ankyrin-G (S.E. Lux) and an alternative commercial antibody raised against ankyrin-G (H-215) were applied to additional samples. Following electrophoresis and immunoblotting, anti-ankyrin-G (S.E. Lux) and anti-pan-IKKα/β (H-470) antibodies were applied in Western-blotting. Full-blot views included in Suppl. Fig. 1B. (E) The relative OD of anti-IKKα/β immunoreactivity of all three Ankyrin-G IP's ( Fig. 1C,D and Suppl. Fig. 1C) was plotted against respective IgG-control IPs. (F) Anti-GFP antibodies pulled Flag-tagged IKKβ down from ankG-eGFP co-transfected, but not eGFP co-transfected cells. AnkyrinG-eGFP pull-down experiments were conducted following a 2-day transfection of SH-SY5Y cells with ankG-eGFP and IKKβ-FLAG vectors. Tag-specific antibodies were used to detect the proteins following Western-blotting procedure. (G) Quantification of anti-FLAG immunoreactivity (ODs representing IKK-β) depicted in (F) and three additional, similarly conducted pull-down experiments was determined.
Scientific REPORTS | 7:42006 | DOI: 10.1038/srep42006 the sequence for potential p65/NF-κB binding sequences using an established promoter search algorithm (<15% dissimilarity margin, TESS, Fig. 6A). Contained in this area we found 4 potential binding sites for p65/ NF-κB. Following brief stimulation of a mouse cell line with tumor necrosis factor-α for appropriate NF-κB induction, we used primer pairs surrounding the previously identified areas in chromatin immunoprecipitation were cultured until maturity in vitro and were cytosol-extracted for the visualization of cytoskeletal-associated proteins. They were incubated for 5 minutes in 1% (m/V) Triton-X-100 (A-C) or 0.5% saponin (D) in 10 mM Na 3 PO 4 -buffer (pH 7.4), 1 mM MgCl 2 , 3 mM CaCl 2 , 150 mM NaCl. Extractions were performed at 4 °C, prior to fixation and immunostaining. Anti-IKKγ/NEMO immunoreactivity is highlighted in red (A) or green (B), the AIS-specific marker 44 antibody 14D4 (green,A) or anti-AnkG (red, B) were used to identify the AIS (green). (C,D) Extractions were similarly performed followed by immunolabeling using pan-IKKα/β -specific antibodies (red, rabbit polyclonal, H-470, SCBT) with the AIS labeled by anti-ankyrinG in green. Color overlays are depicted following background substraction and sharpening, arrows highlight AISs, scale bars, 10 μm. (E) IKKβ-Flag and eGFP vectors were transfected into mouse cortical neurons on DIV13 using Ca-precipitation. 2 days following transfection, neurons were fixed and immunolabeled using a rabbit polyclonal anti-Flag-specific antibody (Sigma, red), as well as a mouse monoclonal ankyrin-G-specific antibody (Zymed). Note the Flag-tag-specific immunoreactivity within the ankyrin-G positive AIS stretch (arrow, yellow). (F) IKKβ-Flag and ankG-eGFP vectors were transfected into mouse cortical neurons as above. 2 days following transfection neurons were cytosol-extracted for visualization of cytoskeletal-associated proteins by incubation in 1% (m/V) Triton-X-100 buffer for 5 minutes. Extractions were performed using pre-cooled solutions and performed at 4 °C, prior to fixation and immunostaining. Note that overfixation often obliterates AIS-immunoreactivity. The inset shows a magnified view of the anti-Flag positive AIS. Note that ankG-eGFP and anti-Flag reactivity remained associated with the non-soluble cytoskeleton in both, in the AIS (arrow), but also in spots within the soma following Triton-X100 extraction. The ankyrin-G antibody showed a notably stronger immunolabeling in the ankG-eGFP transfected cell than in the surrounding untransfected control cells. Scale bars, 10 μm. assays (ChIP) using a rabbit monoclonal antibody raised against p65/NF-κB tested for application in ChIP (D14E12). We obtained pronounced amplification from two regions containing potential NF-κB binding sites proximal to the TSS ('4&6' , '32&33' , Fig. 6B). NF-κB/p65 dependent regulation of ank3 transcripts was determined by depletion of p65 in PC12 cells. Abundance of transcripts was decreased by small-hairpin mediated reduction of p65 protein levels, while ankG-eGFP overexpression confirmed primer specificity (Fig. 6C). We then inserted the ank3 promoter sequence upstream of a firefly-luciferase reporter gene in order to test whether p65/NF-κB induced ank3-promoter dependent reporter activity. p65-cotransfected PC12 cells showed a pronounced and significant up-regulation of luciferase activity from the above reporter construct which was not observed following eGFP transfection or p65 reduction (Fig. 6D). Upon further split into proximal and distal promoter constructs (Schematic Fig. 6A), we found pronounced p65-dependent and IKKβ-potentiated activity from the proximal region mainly in primary cortical neurons, in agreement with the results from the ChIP analyses (Fig. 6E). Additional experiments using the full-length promoter construct in primary cortical neurons confirmed the previous results obtained from PC12 cells (Fig. 6F), and showed that basal ank3-reporter gene activity was diminished in p65-depleted cells (Fig. 6G). We also tested whether transfection of phospho-site mimetics potentiates ankyrin-G protein expression at the AIS of primary neurons. Interestingly, following a 3-day period of expression, we found a significant increase of the ankyrin-G intensities by transfection of p65-S536D mutants (Fig. 6H,I). Consistent with decreased p65/NF-κB activity following overexpression of ankyrin-G, we found notably decreased ank3-dependent reporter activity following overexpression of the scaffolding protein (Fig. 6J).

Discussion
Neuronal plasticity allows the nervous system to adapt to intrinsic and extrinsic cues. Recent evidence suggests a novel form of plasticity owing to alterations of positioning and length of the axon initial segment (AIS) 31,32 . The AIS integrates information from somatodendritic, axo-axonic and inhibitory input, and computes axonal output based on the molecular make-up and positioning of this compartment 8,31,33 . These properties may explain its recently emerging involvement in the pathophysiology of neurodevelopmental disorders such as Angelman syndrome 34 , bipolar disorder [35][36][37] , intellectual disability 38 and posttraumatic stress disorder 39 (reviewed in ref. 40).
The transcription factor NF-κB was found in the axoplasm [41][42][43] . Previous studies described an accumulation of phosphorylated NF-κB pathway proteins at the AIS of rodents and humans in vitro and in vivo 44 . Immunoreactivity to phosphorylated IκBα and activated IKK has been found at nodes of Ranvier 45 and both proteins were implicated in axon outgrowth 46 . Despite the proposed key role of the AIS in neuronal physiology and pathophysiolgy, little is known about the transcriptional regulation of its main components. Here, we found that ankyrin-G and IKKα/β co-localized in membrane and synaptosomal fractions of mouse cortical extracts and co-immunoprecipitated from cortical lysates, also evidenced in the reverse co-immunoprecipitation and pull-down experiments. Membranes were extracted using the mild detergent IGEPAL CA-630, as previously employed in co-immunoprecipitation work on the membrane-associated ankyrin-spectrin network 47 . In IκBα knockout mice persistent staining for phosphorylated IκBα 48 was shown, raising concerns regarding phosphorylation-specific antibody specificity. A potential compensatory up-regulation or cross-reactivity to phosphorylated IκB isoforms have not been determined in the cited study. Whole protein, pan-specific polyclonal antibodies are considered less promiscuous compared to phosphorylation-specific antibodies. Here we found that the IκBα-upstream kinase complex associated with detergent-insoluble cytoskeletal compartments and demonstrate that pan-IKKγ and pan-IKKα/β co-localized with ankyrin-G. Previous studies suggested that IKKγ/NEMO also localized to the AIS cytoskeleton following detergent extraction, and demonstrated that it interacted with AIS-localized FHF1/FGF12 (fibroblast growth factor homologous factor 1) 26 . Furthermore, IKK-complex inhibitors also inhibited the interaction of sodium channels with FHF4/FGF14, both also localized at the AIS 49,50 . Collectively, these data suggest that IKK signaling components localize to the AIS.
Importantly, the co-localization of ankyrin-G and IKKα/β was consolidated in proximity-ligation assays, thus validating their proximity in the molecular range (smaller than about 40 nm 51 ). This interaction significantly increased following overexpression of both proteins. Ankyrin-G contains ankyrin-and death-domains (DDs) that may serve as scaffolds upstream of NF-κB activation. DD interactions form platforms that either result in apoptosis 52 or signal towards NF-κB. Depending on the nature of the scaffold they either decrease 53 or potentiate transcription factor activity 54 . Interestingly, we here show that ankyrin-G-eGFP expression  decreased NF-κB activity in cortical neurons, suggesting that p65 availability is reduced following sequestration by the ankyrin-repeats of ankyrin-G in a manner similar to IκBα-ankyrin-repeat domain sequestration of the p65-Rel-homology domain 55 . In turn we found that increased levels of p65 activated the NF-κB-reporter following silencing of ankyrin-G. We also found p65 to be associated with the AIS cytoskeleton. p65 immunoreactivity was previously described in the axoplasm [41][42][43] and was activated following sciatic nerve transection 56 . In further studies, we found that ank3 transcripts and reporter-gene expression were regulated in an NF-κB dependent manner, thus potentially allowing for a modulation of ankyrin-G length and position along the AIS 9,31,32,57 and its expression levels at spine heads 58 . Database queries suggest several transcription start sites for the rodent ank3 gene. Ankyrin-G isoform expression may be regulated not only through alternative splicing but also through alternative start sites. Notably, shorter variants occur post-synaptically and help in the organization of nanodomains at glutamatergic synapses 58 . It has previously been demonstrated that ischemic injury resulted in the disassembly of the ankyrin-G-related AIS-cytoskeleton, which began at early time-points following the insult, and was inhibited by a calpain inhibitor 17 . Interestingly, neurons that were completely devoid of the AIS diffusion barrier failed to re-establish a functional AIS. More recent work implicated P2X7 receptors in the calpain-dependent ankyrin-G depletion following CNS insults 59 . Our findings suggest that increasing the activity of the transcription factor NF-κB may up-regulate a set of ankyrin-G transcripts and protein under such pathophysiological conditions, thereby preventing injury-induced AIS disassembly and promoting spine maintenance.
In conclusion, we demonstrate that p65/NF-κB controls ankyrin-G levels via a negative feedback loop, thereby linking NF-κB signaling with neuronal polarity and axonal plasticity. Constitutive NF-κB activity may help to maintain ankyrin-G expression levels in differentiated neurons, while sequestration of the transcription factor by ankyrin-G may restrict NF-κB overactivity and overabundance of ankyrin-G. This feedback loop may also be appropriately positioned to maintain expression levels of ankyrin-G during pathophysiological conditions.
Gradient centrifugation for subcellular fractionation. Brains of adult C57BL/6 mice were dounced and minced in 5 ml STE-buffer (0.32 M Sucrose, 10 mM Tris, 1 mM EDTA, pH 7.2-7.4, including protease and phosphatase inhibitor cocktails (1:100, Sigma)) per brain. 10% and 7.5% Ficoll solutions (Sigma) were prepared in ice-cold STE-buffer from 20% Ficoll stock solution and carefully layered. The brain lysates were centrifuged at 1200 g for 5 minutes, the pellet labeled nuclei was immediately lysed in 5 M urea-SDS-lysis buffer. The supernatant of the above step was spun at 11,100 g, the pellet re-suspended and layered on top the Ficoll gradient. Gradient centrifugation followed at 100,000 g for 30 minutes using ultracentrifugation (Sorvall, Thermoscientific, UK) and a lipid fraction, a synaptosomal fraction as well as a mitochondrial and nuclear fractions were isolated and lysed in SDS-lysis or 5 M-urea buffer for Western-blot procedures.
Chromatin-Immunoprecipitation (ChIP) analysis. Cell culture of a murine BV-2 cell line was conducted using RPMI, 10% FBS, 2 mM Glutamine, 1% Penicllin/Streptomycin, the cells were stimulated using 100 ng/ml rmTNF-α (PeproTech, London, UK), 2 hours before fixation in 1% formaldehyde, left at 37 °C for 10 min and quenched with 2 mM Glycine. Following pellet washes, resuspension in sonication buffer followed (1% Triton X-100, 0.1% Na-deoxycholate, 50 mM Tris-Cl, pH 8.10, 150 mM NaCl, 5 mM EDTA, 0.1% PMSF, 1% SDS and protease inhibitor cocktail) and sonication on full power for 3 × 10 sec on ice (Branson Digital Sonifier 250, Danbury, USA). The cells were spun at 13,000 rpm for 2 min. Supernatants underwent agarose gel electrophoresis for fragment length check (400-600 bp) and were used for immunoprecipitation. To protein A/G-agarose, FBS and Fish Sperm DNA (Sigma) was added and incubated for 30 mins for preclearing, with 5 μl of rabbit monoclonal anti-p65 (D14E12, CST) for immunoprecipitation overnight at 4 °C. Samples were spun and underwent series of washing steps before extraction of DNA using 1% SDS, 0.1% NaHCO 3 . The samples were incubated at 65 °C for 2-4 hours and purified using a PCR purification kit. PCR was conducted using standard procedures at 45 cycles using the following primer pairs (5′ -> 3′): Statistical analysis. We employed GraphPad Prism for statistical analyses (GraphPad Software Inc., La Jolla, USA). Significance was determined following parametric testing and using the test as detailed in the figure legends. All data are represented as mean ± SEM, data values are represented by a circle, p values ≤ 0.05 were considered to be significantly different and marked by an asterisk.