Neuronal survival and plasticity critically depend on constitutive activity of the transcription factor nuclear factor-κB (NF-κB). We here describe a role for a small intracellular fibroblast growth factor homologue, the fibroblast growth factor homologous factor 1 (FHF1/FGF12), in the regulation of NF-κB activity in mature neurons. FHFs have previously been described to control neuronal excitability, and mutations in FHF isoforms give rise to a form of progressive spinocerebellar ataxia. Using a protein-array approach, we identified FHF1b as a novel interactor of the canonical NF-κB modulator IKKγ/NEMO. Co-immunoprecipitation, pull-down and GAL4-reporter experiments, as well as proximity ligation assays, confirmed the interaction of FHF1 and NEMO and demonstrated that a major site of interaction occurred within the axon initial segment. Fhf1 gene silencing strongly activated neuronal NF-κB activity and increased neurite lengths, branching patterns and spine counts in mature cortical neurons. The effects of FHF1 on neuronal NF-κB activity and morphology required the presence of NEMO. Our results imply that FHF1 negatively regulates the constitutive NF-κB activity in neurons.
In the nervous system, the transcription factor NF-κB consists of hetero- or homodimers made up of p50, p65 or c-Rel (Schmidt-Ullrich et al., 1996; Kaltschmidt and Kaltschmidt, 2009). The N-terminal Rel homology domain (RHD) contains sequences responsible for dimerization, nuclear translocation and DNA-binding. In contrast to p50, p65/RelA and c-Rel contain a strong transactivation domain. In the so-called canonical pathway, NF-κB complexes are retained in the cytoplasm by sequestration to the ankyrin-repeat domains of the nuclear factor κB inhibitor-α (IκBα), while signal-induced N-terminal phosphorylation results in proteasomal degradation of the latter and translocation of the transcription factor to the nucleus (Perkins, 2006). The IKK-complex, a heteromeric holoenzyme that phosphorylates IκBα on two highly conserved serine residues (Ser32, Ser36), is composed of three specific subunits IKKα, IKKβ and a varying number of the scaffold protein IKKγ, also known as NEMO (NF-κB essential modulator) (Yamaoka et al., 1998). NEMO is considered to be a key regulator of NF-κB signaling, as genetic ablation results in diminished activation of NF-κB in response to a number of inflammatory stimuli (Rothwarf et al., 1998; Makris et al., 2002; van Loo et al., 2006).
NF-κB controls neurite outgrowth in the developing CNS and PNS (Gutierrez et al., 2005; Li et al., 2010), and is thought to be involved in the regulation of learning and memory (Albensi and Mattson, 2000; Meffert et al., 2003). Additionally, calcium influx induced by synaptic transmission is considered to be the driving force for constitutive neuronal NF-κB activity in the adult nervous system (Lilienbaum and Israël, 2003; Meffert et al., 2003; O’Sullivan et al., 2010). In contrast, little is known regarding neuron-specific proteins in control of NF-κB signaling. Considering the important role of NEMO in the canonical NF-κB signaling pathway, we considered NEMO a valid bait to identify new upstream regulators of neuronal NF-κB signaling. Using a protein microarray approach, we identified the intracellular protein fibroblast growth factor homologous factor 1b (FHF1b) as a novel interactor of NEMO, and investigated the biological role of FHF1b in the regulation of neuronal NF-κB signaling, and in the maintenance of neuronal morphology in adult neurons.
Protein microarrays identify FHF1b as a novel interaction partner of NEMO
Affinity-purified and biotinylated full-length human NEMO was used to probe a human protein microarray containing ∼8200 full-length proteins to identify novel factors controlling NF-κB signaling (Fenner et al., 2010) (Fig. 1A). Two canonical NEMO interactors, IKKα and IKKβ, were among the top hits based on Z score analysis, demonstrating the validity and robustness of this approach (Fig. 1B). Intriguingly, a new neuron-specific binding partner, fibroblast growth factor homologous factor 1b (FHF1b/FGF12b), was among the highest affinity hits (Fig. 1A,B; supplementary material Fig. S1). Initial experiments confirmed positive interaction between overexpressed FHF1b and NEMO in HEK293 cells, as determined by co-immunoprecipitation (Fig. 1C) and GST pulldown assays (Fig. 1D). Additionally, mammalian two-hybrid assays in neuronal PC12 cells and primary neurons (Fig. 1E,F) demonstrated that co-expression of GAL4-binding domain tagged NEMO and transactivation domain tagged FHF1b significantly enhanced GAL4-driven normalized luciferase activity (Fig. 1F) (control 0.04±0.02, MyoD/Id 0.47±0.18, NEMO/FHF1b 2.70±1.40).
Endogenous FHF1 protein is expressed at the axon initial segment of mature hippocampal neurons in culture and in the adult mouse cortex
Previous studies using green fluorescent protein (GFP)-tagged FHF1a/b indicated that these proteins localize to the axon initial segment (AIS) (Goldfarb et al., 2007; Wang et al., 2011). The AIS is characterized by a high accumulation of the cytoskeletal scaffolding protein ankyrinG, which is involved in the attachment of further AIS-specific protein complexes (Rasband, 2010). We found that cultured neurons expressed Fhf1 mRNA (supplementary material Fig. S2E) and displayed a marked FHF1 immunofluorescence in the AIS of mature neurons (Fig. 2A). Specificity of the FHF1 antibody was confirmed by plasmid-mediated overexpression of FHF1 and Fhf1-RNA interference in cultured cortical neurons (Fig. 2B,C; supplementary material Fig. S3A). We also found endogenous FHF1-immunoreactivity in extracts obtained from the neocortex of adult mice (Fig. 2D). Furthermore, we used a pepsin-based antigen-retrieval method on floating sections (Lorincz and Nusser, 2008), obtained from the cortex of adult mice, to detect endogenous FHF1-immunoreactivity in situ at the AIS (Fig. 2E; Fig. 2F as negative control). We observed that this was closely associated with the ankyrinG-positive AIS (Fig. 2G).
FHF1 interacts with NEMO and the IKK complex at the AIS in situ
Having established the expression of FHF1 in the AIS, we next examined whether FHF1 interacted with NEMO and the IKK signaling complex at this site. Detergent extraction of living neurons isolates cytoskeleton-associated structures, while cytosolic proteins are eliminated prior to immunofluorescence analyses (Winckler et al., 1999). FHF1 immunoreactivity clearly decorated the axon-initial segment in mature cortical neurons following detergent extraction of somatic proteins prior to fixation in vitro (Fig. 3A; supplementary material Fig. S3B). We then investigated whether NEMO was localized in the AIS. Similar to FHF1, NEMO immunoreactivity decorated the AIS in mature cortical neurons following detergent-extraction (Fig. 3B; supplementary material Fig. S3C). Depletion of Nemo mRNA using siRNA in mature cortical neurons significantly reduced NEMO immunoreactivity and phosphorylated-IKKα/β immunoreactivity in the AIS (Fig. 3C,C′, decrease of 45.3%±8.01%; supplementary material Fig. S2A–D; specificity of the phospho-IKKα/β antibody was examined in supplementary material Fig. S3D). Studies in cortical neurons using confocal microscopy revealed FHF1 immunoreactivity in speckled clusters along the phospho-IKKα/β-positive AIS (Fig. 3D,E). Quantitative co-localization analyses confirmed that the highest immunofluorescence intensities of FHF1 and pIKKα/β occurred within the cell membrane along the AIS (Fig. 3D,E) [Pearson’s correlation coefficient Rr = 0.74 (AIS only, D) and Rr = 0.52 (non-AIS, D); Rr = 0.68 (E, AIS)]. Subcellular fractionation of adult brain extracts revealed a high expression of FHF1 protein in the cytosolic and lipid fractions, while NEMO concentrated in the nuclei, mitochondria and likewise in the lipid fraction (Fig. 3F).
To show interaction of FHF1 with the proteins of the IKK-complex in situ, we employed proximity ligation assays (PLA). These assays combine antibody-labelling with signal detection by rolling-circle amplification of secondary antibody-linked oligonucleotides and subsequent fluorescent staining, thus indicating protein interactions by detection of molecular proximity in situ, similar in sensitivity to fluorescence resonance energy transfer (Söderberg et al., 2006) (Fig. 4A). Firstly, we overexpressed NEMO and FHF1b together with EGFP in primary cortical neurons. Transfected cells displayed a punctate PLA fluorescence across the neuronal cell body, and enrichment in the proximal area of some neurites (Fig. 4B,C). We also detected a marked accumulation of PLA spots within the AIS in cells transfected with either, FHF1b (Fig. 4D) or NEMO (Fig. 4E) alone and probed for NEMO-FHF1b interactions (Fig. 4D,E). Cells were additionally immunostained for ankyrinG to confirm the localization of PLA spots to the AIS (Fig. 4E). As a negative control, we performed PLA assays with EB1 (Piehl and Cassimeris, 2003), a microtubule-associated protein that is significantly enriched in the AIS (Nakata and Hirokawa, 2003). PLA assays using EB1-GFP did not reveal a significant association with the IKK-complex (Fig. 4F). Quantification of PLA spots showed that FHF1-NEMO interactions were significantly more frequently observed in the AIS than interactions in the negative control experiment (Fig. 4G) (28±4 versus 3±1 spots). Next, we examined interactions between endogenous NEMO and FHF1 proteins by PLA assay. We detected PLA fluorescent aggregates – suggesting NEMO and FHF1 interactions – in the neuronal soma and proximal neurites of non-transfected cells, resembling the localization of the AIS (Fig. 4H), but no apparent fluorescence was found in the nuclei. Similar results were obtained using FHF1 and IKKβ primary antibodies, suggesting binding of FHF1 to the IKK complex in the AIS and soma (Fig. 4I). Additional post-PLA ankyrinG-labelling and confocal microscopy confirmed the interaction of endogenous FHF1 and NEMO occurred along the ankyrinG-positive AIS membrane (Fig. 4J). Furthermore, endogenous FHF1-NEMO molecular proximity was observed more frequently in the AIS, as in post-synaptic densities (PSD-95), MAP-2 positive dendrites, or as compared to the AIS of the negative control reaction (Fig. 4K; supplementary material Fig. S4). Proximity-ligation between overexpressed NEMO and overexpressed IKKβ protein served as a positive control, while no proximity-ligation was observed using FLAG and XPress antibodies (Fig. 4L,M).
The interaction between FHF1 and NEMO limits NEMO’s wild-type-, K48- and K63-linked polyubiquitylation and IKK-kinase activity
K63-linked polyubiquitylation of NEMO has been identified as essential for IKK-signalosome activity via recruitment of upstream activators TAB2/TAB3 to the signalosome and thus facilitates the activation of the IKKβ-Kinase TAK1 (Gautheron and Courtois, 2010; reviewed by Schmukle and Walczak, 2012). To assess the effects of FHF1 on NEMO and IKK signalosome activity, we determined the phosphorylation status of the inhibitor of κB (IκBα) following a canonical NF-κB stimulus (lipopolysaccharide). Intriguingly, we found that FHF1b-transfection induced an ablation of LPS-induced phosphorylation of the IκBα protein and thus might be seen as a suppressor of the activity of the NF-κB transcription factor (Fig. 5A). Having demonstrated that FHF1b overexpression resulted in reduced IKK-signalosome activity, we next evaluated whether FHF1b affected polyubiquitylation of NEMO. Co-transfection of FHF1b reduced the basal level of ubiquitylation of NEMO (Fig. 5B), indicating that FHF1 negatively regulated polyubiquitylation of NEMO. To determine if FHF1b was affecting either K48-linked (thus implicating FHF1b in the regulation of NEMO stability) or K63-linked polyubiquitylation (and hence assembly of the IKK-signalosome), we employed ubiquitin-mutant proteins in this assay, where all five lysine-residues are mutated to arginine, except residues K48 or K63 (R48K, R63K) (Ramakrishna et al., 2011), respectively. Interestingly, FHF1b co-transfection resulted in reduced R48K and R63K polyubiquitylated NEMO, indicating that, in addition to regulating NEMO activity, FHF1 may also affect NEMO stability (Fig. 5C). This finding is supported by analysis of NEMO protein levels following overexpression of FHF1b in primary neurons, in which enhanced NEMO levels were observed, with no alteration to p65/NF-κB levels (Fig. 5D).
Constitutive transcriptional activation by NF-κB is repressed by FHF1b overexpression and potentiated by FHF1 silencing
Having demonstrated that FHF1 reduced IKK-signalosome activity and K63-linked polyubiquitylation of NEMO, we next examined the impact of FHF1 on NF-κB activation. First, endogenous constitutive NF-κB activity during development was probed using antibodies specific to active p65/NF-κB. Nuclear localization of the activated transcription factor peaked around DIV7 and subsided thereafter (supplementary material Fig. S5A). FHF1 expression levels during in vitro development were obtained by immunofluorescence analyses in dissociated cortical neurons to determine optimized time points for functionally meaningful gene silencing experiments. FHF1 expression levels increased consistently during in vitro development up to DIV18, when a strong co-localization to the ankyrinG-positive axon-initial segment and less somatic or nuclear accumulation was observed (Fig. 6A). We then employed NF-κB-specific reporter-gene assays to examine the influence of FHF1 on p65 transactivation potential. Interestingly, overexpression of FHF1b in primary mature neurons resulted in a reproducible and significant decrease in NF-κB-dependent transcriptional activation (Fig. 6B). Expression of two FHF1 splice variants, FHF1a and FHF1b (truncated N-terminus), in mammalian cells prevented selective gene silencing of FHF1b. Instead, we used small-hairpin RNA directed against a common region of both splice variants and obtained a significant reduction of total FHF1 mRNA in transfected neurons (Fig. 2C; supplementary material Fig. S2E). Most notably, NF-κB activity increased significantly by approximately threefold compared to controls in FHF1-depleted mature neurons (DIV15-19) (Fig. 6D,E), in keeping with the inhibitory effects of FHF1 on polyubiquitylation of NEMO and phosphorylation of IκBα. No significant impact of Fhf1 silencing on NF-κB activity was measured in early and immature cortical neurons, exhibiting relatively low FHF1 protein expression levels in culture (DIV4-7) (Fig. 6A,C). Sequence-intrinsic effects were additionally confirmed by transfection of single shRNA constructs into rat neurons. Single Fhf1-mRNA silencing constructs induced κB-dependent luciferase activity less effectively than co-transfection with two silencing vectors (supplementary material Fig. S5B). As expected by sequence disparities, the constructs failed however to induce κB-dependent luciferase activity in a human cell line (supplementary material Fig. S5C), thus excluding unspecific effects.
We next sought to determine the requirement for NEMO in the repressive effect of FHF1 on NF-κB transcriptional activity. Ablation of FHF1 and NEMO in cortical neurons resulted in reductions of the respective mRNA copy numbers (Fig. 6G,H). Co-silencing of NEMO and FHF1 in cortical neurons demonstrated that the suppressive effect of FHF1 on NF-κB activity indeed required NEMO (Fig. 6E) (Fhf1 shRNA increase by 3.30±1.14 fold; Nemo and Fhf1 shRNA decrease by 0.55±0.44 fold over control). Additionally, transfecting increasing amounts of NEMO effectively neutralized the suppressive effect of FHF1b on NF-κB activity, while a NEMO mutant defective in ubiquitin-binding failed to enable NF-κB activity, establishing FHF1b upstream of NEMO-induced signal transduction (Fig. 6F) (FHF1b-induced decrease by 0.25±0.10 fold, FHF1b and NEMO wt (2.5 µg) increase by 2.48±0.07 fold, FHF1b and NEMO-L329P (2.5 µg) decrease by 3.99±0.32 fold).
FHF1 modulates NF-κB-dependent control of structural homeostasis
Activation of NF-κB has been implicated in neurite outgrowth (Gutierrez et al., 2005). We next sought to determine whether FHF1 controls the neuronal morphology of mature cortical neurons with established neurite trees. Firstly, we derived Sholl profiles by manual evaluation of p65-depleted neurons to examine the implication of the transcription factor in the maintenance of neuronal morphology. Ablation of p65 in mature neurons significantly reduced neuronal arborisation and neurite length compared to control neurons (Fig. 7A,A′). We next investigated whether silencing of FHF1 expression resulted in the opposite effect. Interestingly, FHF1 ablation increased the number of intersections across the neurite tree (Fig. 7B,B′). These findings were further supported with a high-content imaging approach on mature cortical neurons (Buchser et al., 2010), which revealed a significant increase in relative neurite length and branching point counts in FHF1-depleted neurons (increase of 39.5±9.9% and increase of 42.1±11.6%), while p65 ablation resulted in the opposite (Fig. 7C–E) (decrease of 19.9±4.5% and 13.6±6.6%, DIV15-20). In accordance with the above, Sholl profiles obtained from NEMO-depleted neurons revealed a significant reduction of neurite lengths and branching patterns in mature neurons (Fig. 7F,G), with a comparable result using the high-content imaging approach (Fig. 7G′). We also investigated whether the increase in neurite growth induced by FHF1 depletion required NEMO by comparing the effects of NEMO, Fhf1, and combined NEMO and Fhf1 gene silencing over a period of 5 days. An increase in average neurite length, neurite counts and a trend towards increased branching was observed in FHF1-depleted neurons, while a decrease was noted in NEMO-depleted cells. As shown by double-silencing of FHF1 and NEMO, the effect on NF-κB achieved by co-silencing of NEMO was dominant over FHF1 ablation, suggesting that the regulation of neuronal morphology by FHF1 required NEMO (Fig. 7H–J).
FHF1 regulates neurite spine densities
Finally, we sought to determine whether FHF1 also modulates spine structural plasticity. Cortical neurons at stage 5 (DIV15-25) in culture exhibit elaborate spine patterns on their dendrites (Dotti et al., 1988; Barnes and Polleux, 2009). We determined the impact of FHF1 depletion on their spine patterns using shRNA-mediated mRNA silencing. Neuronal spines were detected using a voxel-clustering algorithm based on the skeletonization of 3D-stacks (Rodriguez et al., 2006; Rodriguez et al., 2008). Concurrent with our morphological findings, the FHF1-depleted neurons exhibited an increase in average neurite length per unit volume (Fig. 8A,B) (Scr shRNA 712.25±64.9% to Fhf1 shRNA 1015.42±181.7%). Likewise, image analysis revealed a significant increase in the average spine count per unit volume (Fig. 8C,D) (303.0±66.9% to 719.4±142.7%) and in average spine density per neurite shaft length (Fig. 8C,E) (39.9±5.9 to 70.1±5.1 spines per 100 µm neurite shaft).
In the present study, we identified an AIS-enriched protein, the fibroblast growth factor homologous factor 1 (FHF1/FGF12), as a novel interactor of the canonical NF-κB modulator IKKγ/NEMO within mature neurons. We furthermore demonstrate that a disruption of this interaction impacts on NF-κB signalling and maintenance of neuronal morphology. It is well established that cortical neurons exhibit a pronounced constitutive activity of the transcription factor NF-κB (Kaltschmidt et al., 1994) and its trans-activating capacity is considered essential for development and adult neuronal tissue maintenance (Schmidt-Ullrich et al., 1996; Bhakar et al., 2002; Chiarugi, 2002; Fridmacher et al., 2003). The continuous drive for the activation of NF-κB was suggested to be imparted by either post-synaptic activation mediated through the neurotransmitter glutamate or spontaneous calcium fluctuations (Kaltschmidt et al., 1995; Lilienbaum and Israël, 2003). This is mediated via post-synaptic activation and retrograde transport of p65/NF-κB from dendrites to the nucleus, using dynein-mediated transport mechanisms (Meffert et al., 2003; Mikenberg et al., 2007). We here demonstrate the potential existence of a second source of constitutive NF-κB activity in neurons that is controlled by FHF1 and requires NEMO, in line with the presumed essential role of NEMO in activation of the IKK-complex by a range of established stimuli (Yamaoka et al., 1998; Solt et al., 2009; Fenner et al., 2010).
Using an unbiased protein microarray approach, we identified protein interactions between NEMO and FHF1b, a family member of fibroblast growth factor homologous factors (FHF), which have repeatedly been described to be highly concentrated at the AIS in earlier studies (Lou et al., 2005; Goldfarb et al., 2007; Goetz et al., 2009; Laezza et al., 2009). FHFs are a subfamily of the fibroblast growth factor (FGF) superfamily, expressed throughout the developing nervous system (Smallwood et al., 1996). However these factors lack N-terminal secretion sequences (Goldfarb, 2005) and are considered bona fide intracellular proteins without the capability to bind to FGF-receptor complexes (Olsen et al., 2003; Goldfarb, 2005). FHFs share considerable mutual protein sequence homology (Goldfarb, 2005). Mutations in FHF4 cause ataxia (Wang et al., 2002), dystonia and dyskinesia in mice and the autosomal dominant missense mutation FHF4F145S gives rise to a form of progressive spinocerebellar ataxia (SCA27) (van Swieten et al., 2003; Xiao et al., 2007). Complete genetic ablation of FGF14 affects synaptic transmission and likewise results in ataxia and paroxysmal dyskinesia (Wang et al., 2002; Laezza et al., 2007; Xiao et al., 2007), with the phenotype of FHF1/FHF4 double-knockout animals being more pronounced (Goldfarb et al., 2007).
Even though FHF1 expression in the AIS, similar to its relatives FHF2-4, has been reported for overexpressed FHF1 fusion proteins (Goldfarb et al., 2007; Wang et al., 2011), its endogenous accumulation at the AIS in situ had not yet been documented. We found that upon antigen unmasking, FHF1 antibodies decorated the AIS in the neocortex and hippocampus, regions with considerably high constitutive NF-κB activity (Kaltschmidt et al., 1994; Bhakar et al., 2002; O’Sullivan et al., 2010). This pattern is also in line with FHF-transcript expression in the developing nervous system (Smallwood et al., 1996). We also demonstrated a clustering of the IKK-complex (i.e. activation-loop phosphorylated IκB-kinase and NEMO), and an interaction of FHF1 and NEMO in the AIS. In this context, it is interesting to note that other studies recently reported a clustering of constitutively-phosphorylated proteins of the NF-κB signal transduction cascade in the AIS (activation-loop phosphorylated IKKα/β, Ser32, 36-phosphorylated IκBα and transactivation-domain Ser536-phosphorylated NF-κB/p65) (Schultz et al., 2006; Sanchez-Ponce et al., 2008). We found that silencing of FHF1 protein levels resulted in a prominent, three to fourfold upregulation of κB-dependent reporter-gene transactivation, whereas overexpression of FHF1 resulted in a suppression of NF-κB, which could be alleviated by increasing levels of NEMO. Normally a membrane-receptor associated scaffolding protein (Windheim et al., 2008), nuclear NEMO is exported to the cytosol in resonse to genotoxic stimuli and calcium-signaling (Berchtold et al., 2007). NEMO has also been shown to link the IKK-complex to the actin cytoskeleton and transport via Myo1c (Nakamori et al., 2006). NEMO binds to K63-linked polyubiquitylated upstream complexes, a process which is thought to facilitate IKK-complex formation and phosphorylation (Schmukle and Walczak, 2012). We observed elevated NEMO levels in neurons following FHF1b expression, possibly the consequence of decreased proteasomal turnover due to decreased K48-linked polyubiquitylation. Likewise, we observed a decrease in K63-linked polyubiquitylated NEMO. As this process is believed to favour NEMO-dependent signal transduction towards IKK, ‘activated NEMO’ may decrease in the presence of FHF1b. As we found FHF1b itself was polyubiquitylated, it is reasonable to hypothesize that NEMO binds directly to the attached K63-linked chains on FHF1b. We hope that future work will elucidate the interdependencies or cross-talk between the polyubiquitylation patterns of the two proteins. In summary, our results suggest FHF1 interaction with NEMO may sequestrate the protein from the IKK-complex and limit NEMO’s Lys 63-linked polyubiquitylation (Tang et al., 2003; Henn et al., 2007) or nuclear translocation. We cannot fully exclude that FHF1 regulates neuronal NF-κB at sites other than the AIS, particularly the soma, nevertheless our data provide important evidence for a NEMO-dependent regulation of constitutive NF-κB signaling by FHF1. Notably, we found developmental regulation of constitutive activation of NF-κB peaked around DIV7 and decreased thereafter. Although we do not presume FHF1 to be the only regulator of constitutive NF-κB, this decline may partly be explained by the rise in FHF1 expression levels.
We also found a marked increase in the number of neurite branching points upon silencing of FHF1 in fully-differentiated neurons, along with a significant increase in neurite numbers and lengths. Our data suggest that this effect is mainly achieved through FHF1 interaction with NEMO. Consistently, the opposite effects were observed after NEMO or p65 silencing. Furthermore, the FHF1-silencing phenotype was regulated in a NEMO-dependent manner. NF-κB activation has previously been linked to enhanced neurite outgrowth (Gutierrez et al., 2005; Li et al., 2010). Recent evidence implicates NF-κB-dependent transcription of PKA and FOXO1 in axogenesis (Imielski et al., 2012), while pharmacologic interference with IKK activation has been shown to slow down axon outgrowth and AIS development (Sanchez-Ponce et al., 2008). We here show that constitutive NF-κB activity, mediated through a NEMO-dependent pathway, is necessary to maintain the morphology of established dendritic trees in differentiated neurons. Our results further suggest that the interaction of FHF1 with NEMO is required to maintain this state, and that any alteration in FHF1-NEMO interaction may lead to significant changes in neuronal morphology and dendritic structural plasticity. In this context, it is interesting to note that disruption of the AIS cytoskeleton during ischemia has been suggested to contribute to neuronal injury, where proteolytic cleavage of ankyrinG is irreversible (Schafer et al., 2009). Loss of this important upstream anchor caused loss of neuronal polarity, a transition from axonal to dendritic identity and the formation of dendritic spines on the former axon (Hedstrom et al., 2008; Sobotzik et al., 2009). Additionally, recent evidence suggests that the cell-adhesion molecule Neurofascin186 is involved in the stabilization of the axon initial segment (Zonta et al., 2011). Finally, it has also been shown that the expression of a constitutively-active IKK increases spine density in neurons (Russo et al., 2009). We observed significantly enhanced dendritic spine counts on FHF1-ablated cells, suggesting that FHF1 is involved in spine plasticity (Segal, 2010). Our study therefore establishes FHF1 as an important regulator of structural plasticity and neuronal morphology.
Materials and Methods
Reagents and chemicals
Unless otherwise stated, chemicals were purchased from Sigma-Aldrich (Arklow, Ireland), Merck Chemicals (Nottingham, UK) or Tocris Bioscience (Bristol, UK). Cell culture media were purchased from Gibco-Invitrogen (Dun Laoghaire, Ireland). Pre-designed HuSH-29 small-hairpin RNA-vector constructs from Origene, Rockville, MD, USA.
Primary cultures of cortical neurons were prepared from neonatal Sprague-Dawley rats (Rattus norvegicus) as described previously (König et al., 2005). Neurons were cultured in in NMEM-B27 media (1×MEM, 1 mM sodium pyruvate, 26 mM NaHCO3, 2 mM Glutamax®, 1×B27 serum supplement and 33 mM β-D-Glucose (Sigma-Aldrich) at 37°C in a humidified atmosphere containing 5% (v/v) CO2. Astrocyte content of cultures was kept to a minimum by treatment with 1 µM cytosine β-D-arabino-furanoside. Mature neurons in vitro were defined as stage 5 cultures (older than DIV15) (Barnes and Polleux, 2009). Cell lines were cultured using standard techniques and media. Animal work was carried out under license from the Department of Health and Children of Ireland. Animal working procedures were additionally approved by the Research Ethics Committee of the Royal College of Surgeons in Ireland. Cell lines were grown according to standard techniques and transfected via Lipofectamine 2000® (Invitrogen) according to the manufacturer’s protocol.
Transfection of primary neurons
Freshly dissociated cortical neurons were transfected using the Amaxa nucleofector® II (Lonza, Cologne, Germany) using modified manufacturer’s protocols. An improved calcium-transfection protocol under ambient CO2 partial pressure was used to gently transfect mature neurons (Goetze et al., 2004). This technique preserved fine neuronal morphology and survival post-transfection in mature neurons up to DIV21, as observed by us and others (Karra and Dahm, 2010). Transfection efficiencies ranged typically between 5 and 20%, while triple-vector co-transfection efficiencies were determined using visibly-expressed fluorescent protein at 68±3.84% (n = 5 wells). The following transfection vectors were used: pCMV6-Entry-Myc-FLAG-FHF1b (PCR-amplified and subcloned into pCMV6-Entry from a human cDNA clone; NM_004113.3, Origene), pcDNA4/HisMaxA-NEMO (Fenner et al., 2009), p65-EGFP-N1 (a generous gift of E. Floettmann and M. Rowe, Cardiff, UK), EGFP-N1 (Clontech, Heidelberg, Germany) served to determine positively-transfected cells or as control, EB1-GFP (Piehl and Cassimeris, 2003), wild-type and K44M-IKKβ [kinase-dead, Addgene plasmids 11103 and 11104 (Mercurio et al., 1997)], wild-type and NEMO-L329P [Addgene plasmids 11970 and 11971 (Wu et al., 2006)] were used.
Generation of mammalian two-hybrid vectors
Two-hybrid complementation plasmids were generated by ligation into the according vectors as directed by the manufacturer’s protocol (CheckMate®, Promega, Southampton, UK).
RNA silencing and quantitative PCR
HuSH-29mer small-hairpin (shRNA) vectors against rat Nemo/Ikbkg in pRFP-C-RS, Fhf1 in pRS and RelA/p65 in pGFP-V-RS were obtained from Origene (Rockville, MD, USA). Of four targeting vectors, the two most effective sequences were determined by western blot (RelA) or qPCR (Ikbkg, Fhf12). RNA was isolated and its levels were measured by RT-qPCR based on a LightCycler RNA Master SYBR Green I kit (Roche) with a LightCycler 2 instrument, normalized against 18S-rRNA. Measured mRNA amounts were normalized against 18S-rRNA. Primers used for RT-qPCR were MM-FHF1-FWD (5′-TTCAGCCAGCAGGGATATTT-3′) and MM-FHF1-REV (5′-TCTCCATTCATGGCCACATA-3′) for FHF1, Rn-NEMO-FWD1 (5′-ATGGATCCATGAATAGGCACCTCTGGAAGAGCC-3′) and Rn-NEMO-REV1 (5′-AGGAATTCCTACTCAATGCACTCCATGACATG-3′).
Protein microarray screening
Biotinylated full-length human recombinant NEMO and BSA were used to probe protein microarrays (Invitrogen Protoarray version 4.0) as described previously (Fenner et al., 2010). Significant interactions were identified based on a Z-score cutoff value of 3.0.
Cytosolic extraction of neurons prior to fixation was performed by incubation of live cells for 3–5 minutes at 4°C with 1% (m/v) Triton X-100 in a microtubule-destabilising buffer as described previously (Robinson et al., 1991) (10 mM Na3PO4 pH 7.4, 1 mM MgCl2, 3 mM CaCl2, 150 mM NaCl). Cells were subsequently fixed and immunostained.
Immunofluorescence and microscopy
Primary cultured neurons were fixed in 2 or 4% paraformaldehyde (PFA). Cells were permeabilized by incubation with ice-cold PBS containing 0.1% (w/v) Triton X-100 and blocked with 0.3% (w/v) Triton X-100, 5% (v/v) horse serum in PBS. They were incubated overnight or weekend at 4°C in primary antibody. The following primary antibodies were used: anti-FHF1 (1∶100; Abcam, Cambridge, UK); anti-pIKKα/β (1∶500; clone 16A6, Cell Signaling Technology, Hitchin, UK); mouse and rabbit anti-ankyrinG antibodies (1∶500; both SCBT, Heidelberg, Germany); mouse monoclonal and rabbit polyclonal anti-NEMO (1∶100; FL-419, SCBT, and 1∶100; Abcam), anti-EB1 (BD Biosciences, Oxford, UK), anti-PSD-95 (1∶250), anti-MAP2 (1∶500; both SCBT) and anti-p65 (1∶500; active subunit, Millipore). Cultures were incubated for 1 hour at RT with secondary antibodies tagged with either Alexa-Fluor 488 nm or Alexa-Fluor 568 nm (1∶500, Invitrogen). Photomicrographs were taken using a SPOT RT SE 6 Camera (Diagnostic Instruments, Sterling Heights, MI, USA) on an Eclipse TE 300 inverted microscope (Nikon, Kingston upon Thames, UK) with Mercury-arc excitation. Confocal images were acquired on a LSM 710 (Zeiss, Jena, Germany) with a DPSS laser diode, an argon, and helium/neon laser and 40×and 63×oil immersion objectives (NA 1.3 and 1.40, Zeiss) at RT. Brain sections: Adult C57BL/6 mice (Mus musculus) were euthanized by anaesthesia overdose. Brains were removed, post-fixed and cut into 20–50 µm-thick coronal sections. For FHF1-epitope unmasking the free-floating slices were incubated with 0.2 mg/ml pepsin (Dako, Glostrup, Denmark) as described previously (Lorincz and Nusser, 2008). Co-localization analyses were performed using CoLocalizer Express® software (Japan, Switzerland) (Zinchuk and Grossenbacher-Zinchuk, 2011).
Proximity ligation assays
Interactions between neuronal signaling proteins were visualized using pairs of antibodies raised in different species as indicated by Olink Bioscience (Uppsala, Sweden) PLAs using the antibodies as indicated were performed as described in the Duolink manual.
Glutathione-S-transferase (GST) pulldown, co-immunoprecipitation and western blotting
GST-NEMO pulldown and co-immunoprecipitation from lysates of transfected HEK-293T cells expressing FHF1b-FLAG as well as electrophoresis and western blotting and immunodetection were performed using standard techniques and as described previously (Fenner et al., 2010). Brain extracts for western-blotting experiments were prepared using non-ionic-detergent extraction (1% IGEPAL, Sigma Aldrich) in a sodium/sucrose-buffer. Following brain homogenization the lysate was centrifuged at 27,000×g at 4°C for 20 minutes. Electrophoresis was performed following volume reduction (Microncon YM-10, Millipore). For determination of IKK-kinase activity, HEK293T cells were transfected with 1 µg of plasmid per dish expressing either empty pEntry vector or pEntry-FHF1-Myc for 24 hours prior to stimulation as indicated. Cells were lysed on ice in 1 ml Tris-based lysis buffer (50 mM Tris-HCl, 150 mM NaCl, 2 mM EDTA, 1% Triton X-100, 1 mM DTT, 1 mM PMSF, 1 µg/ml aprotinin, leupeptin, pepstatin, 1 mM Na3VO4, 1 mM NaF). Samples were immunoblotted and exposed to anti-phosphorylated IκB (Ser32/36, Cell Signaling Technology) antibodies.
Subcellular fractionation of mouse brain extracts
Adult mice were sacrificed and brains immediately removed, homogenized by dounce homogenization in STE-buffer (0.32 M Sucrose, 10 mM Tris-HCl, 1 mM EDTA, pH 7.2), nuclei pelleted by centrifugation (5 minutes, 1200×g), anti-NeuN (1∶1000, Millipore), the cytosol (anti-GAPDH, 1∶1000, Millipore) was separated from membrane content by 10 min centrifugation (1200×g), the latter was layered on top of a discontinuous Ficoll gradient (7.5% w/v and 10% w/v in STE-buffer) to separate the mitochondrial pellet (anti-Porin, 1∶500, Abcam) from the lipid layer (anti-ankyrinG) recovered from the gradient by centrifugation at 28,000 rpm (45 minutes, Sorvall wx Ultra 100, Fisher Scientific, Dublin, Ireland). Proportionate volumes from the fractions were separated by 6 to 12% gradient gel electrophoresis. Each fraction marker set was used from a single membrane with two identically loaded membranes used for the different marker proteins.
HEK293T cells were seeded at 5×105 ml on 10 cm dishes. Dishes were transfected with 1 µg of plasmid encoding NEMO-His, FHF1-Myc and 3 µg of plasmid encoding wild-type HA-Ubiquitin (HA-Ubiq. wt) or plasmids encoding mutant ubiquitin constructs wherein all lysine residues are replaced by arginine residues except at Lys-48 and Lys-63 (HA-Ubiq. R48K and HA-Ubiq. R63K, respectively, kindly supplied by Andrew Bowie, Trinity College Dublin) for 24 hours. Ubiquitylation of NEMO was assessed using anti-His antibody (1∶1000, SCBT). Confirmation of successful transfection of plasmids was determined through immunoblotting with anti-Myc (1∶2000, Abcam) and anti-HA (1∶1000, SCBT).
Luciferase reporter-gene assays for NF-κB and mammalian two-hybrid luciferase assays
Transcriptional activation of p65/NF-κB was monitored using a vector with six tandem repeats of the κB enhancer element upstream of the coding sequence of the firefly-luciferase (NF-κB-luc; P. Baeuerle, Freiburg, Germany) along with a plasmid encoding Renilla-luciferase under constitutive thymidine-kinase promoter (phRL-TK-luc; Promega) as a transfection control for normalisation in DUAL-Luciferase® assays. Protein-protein interactions in PC12 cells and neurons were confirmed using the CheckMate®/Flexi mammalian two-hybrid system (Promega, Southampton, UK). Proteins of interest were co-expressed with a GAL4-DNA-binding domain (FN11A) and a herpes simplex-derived VP16 transcriptional activation domain (FN10A), respectively. Interactions were analysed by measuring firefly-luciferase transcriptional activation downstream of five GAL4-binding sites and a minimal TATA-box as promoter. Expression of Renilla-luciferase, encoded under control of a SV40 promoter on the pFN11A vector, was measured as transfection control. Primary cortical neurons were transfected with the indicated plasmids at DIV5. Cells were lysed 48 hours post transfection and DUAL-luciferase assays® were performed as described above.
Sholl plots and high-content analyses of neuronal morphology parameters
Photomicrographs of EGFP or RFP-expressing mature rat hippocampal neurons were taken at 20×magnification (NA 0.45) using an Eclipse TE 300 inverted microscope (Nikon) and the appropriate filter settings as indicated. Coordinates of the branching points plus neurite end-points in relation to the soma were estimated manually. We used a MATLAB-based algorithm (2007b, MathWorks, Cambridge, UK) to generate the corresponding Sholl profile as described previously (Gutierrez and Davies, 2007). A LOWESS-algorithm with a smoothing window of 10 points was applied to derive a regression line for the Scholl profile using GraphPad Prism. The region of difference between the profile Sholl sections (points binned within multiplies 20 µm distance) was established and evaluated using Pearson’s chi-square test with P≤0.05 corresponding to the significant difference between the sections. For high content analysis, images from several fields per well were acquired. Neurites were segmented at fixed intensity threshold and delineated using skeletonization of their binary masks. The total neurite length and branching points were normalized to the respective control cells. All image processing operations were executed using Neuronal Profiling BioApplication of the Discovery Toolbox Software (ThermoFisher, Horsham, UK). The cells were imaged in 3D (Z-stacks) using the Zeiss LSM 710. Fluorescence of co-transfected EGFP was excited using 488 nm light and registered in 493–600 nm emission range. The z-stacks were collected with confocal pinhole set to 1.0 Airy unit at the maximum of EGFP emission, which corresponded to ∼600 nm of optical section thickness. Each dataset contained one neuronal cell, which was segmented based on statistical thresholding (1st percentile of dataset intensity distribution). Neuronal spines were detected using a voxel clustering approach (Rodriguez et al., 2008). The image processing and analysis procedures were executed using Neuron Studio version 0.9.92 (Rodriguez et al., 2006).
Statistical analyses and algorithmic procedures were performed using Matlab (MathWorks Inc., Massachusetts, USA), GraphPad Prism (GraphPad Software, Inc., San Diego, USA) or SPSS (IBM, Dublin, Ireland) software suites. Normal distribution of data was tested using Shapiro–Wilk, Kolmogorov–Smirnov or D’Agostino–Pearson omnibus test. Statistical significance was determined using two-tailed Student’s t-tests, one-tailed t-test where applicable on one-tailed hypotheses, or one-way ANOVA followed by Tukey’s post-hoc test for normal distribution of data. For non-parametric data, Mann–Whitney or Kruskal–Wallis test was used. Statistical significance defined at the level of P≤0.05 was marked by asterisks. Bars represent the arithmetic mean ± standard error of the mean (s.e.m.), unless otherwise stated.
The authors thank Heiko Düssmann, Tobias Engel and Caoimhín Concannon for technical discussions and advice. We would further like to thank Sarah Cannon for technical help and Orla Watters for critical reading of the manuscript.
This work was supported by Science Foundation Ireland [grant number 08/IN.1/B1949]; the Higher Education Authority, Ireland (PRTLI Cycle 4); and the European Union (EU FP6-Mobility, Marie Curie Transfer of Knowledge Fellowships and EU FP7-CEMP programme, jointly funded by the National Biophotonics and Imaging Platform, HEA).