Introduction

Phosphatidylinositol phosphates (PtdInsPs) play multiple roles in cell growth and differentiation (Berridge, 1993; Nishizuka, 1995; Lemmon et al., 1996; Carlton and Cullen, 2005), acting as: (i) precursors of second messenger molecules; (ii) effectors for the localization and assembly of protein molecules; and (iii) components of the cellular plasma membrane.

Neurofilaments (NFs), the most abundant cytoskeletal components in large myelinated axons, are obligate heteropolymers composed of NF-L, NF-M and NF-H (Ching and Liem, 1993; Lee et al., 1993; Nakagawa et al., 1995). Each NF is composed of three domains: an α-helix-rich rod domain in the middle, flanked by an N-terminal head domain and a C-terminal non-α-helical tail domain (Fuchs and Weber, 1994; Grant and Pant, 2000; Al-Chalabi and Miller, 2003). NF-L is known to self-assemble into core filaments that then form cross-bridges with NF-M or NF-H (Fuchs and Weber, 1994; Nakagawa et al., 1995). Recent studies have revealed that overexpression or point mutation of NF-L can result in abnormal NF assembly and aggregation in neuronal cell bodies (Cote et al., 1993; Xu et al., 1993; Mathieu et al., 1995; Carter et al., 1996; Lee et al., 2008). NF aggregations due to overexpression or point mutations of NF-L have been associated with neurodegenerative diseases such as Charcot-Marie-Tooth disease (Watson et al., 1994; Fabrizi et al., 2004), Alzheimer's disease (Shepherd et al., 2002; Norgren et al., 2003) and amyotrophic lateral sclerosis (Hirano et al., 1984; Collard et al., 1995; Al-Chalabi and Miller, 2003). Notably, phosphorylation of the NF-L head domain can inhibit self assembly or induce disassembly of polymerized NF-L (Sihag and Nixon, 1991; Guan et al., 1992; Hisanaga et al., 1994; Gibb et al., 1996; Mukai et al., 1996; Yates et al., 2009). However, the factors and mechanisms that regulate NF assembly are not yet fully understood. Thus, we need to identify and characterize additional factors capable of regulating the cytoplasmic assembly, axonal transport and organization of NFs. Phospholipase C-γ1 (PLC-γ1) plays a pivotal role in cell growth and differentiation by hydrolyzing phosphatidylinositol 4,5-bisphosphate (PIP2) to inositol 1,4,5-trisphosphate (IP3) and diacylglycerol (DG) (Suh et al., 1988; Berridge, 1993; Rhee, 2001; Shin et al., 2002). PLC-γ1 has two pleckstrin homology (PH) domains for protein-protein and protein-lipid interactions (Gibson et al., 1994; Falasca et al., 1998; Chung et al., 2010): PH1 is located in the 150 amino acid residues closest to the N-terminus, while PH2 is located near the center of the molecule and is split by an SH2-SH2-SH3 domain (Gibson et al., 1994; Chang et al., 2002). PH domains bind with high specificity and affinity to phosphatidylinositol phosphates (PtdInsPs) such as PI(3)P, PI(4)P, PI(5)P, PI(3,4)P2, PI(3,5)P2, PI(4,5)P2 and PI(3,4,5)P3 (Lemmon and Ferguson, 2000; Cozier et al., 2004; Carlton and Cullen, 2005; Balla, 2005). During our study of the interaction between the PH domains of PLC-γ1 and NF-L (Kim et al., 2006), we found that NF-L also associates with PtdInsPs.

Here, we show for the first time that NF-L directly binds PtdInsPs to regulate NF-L self assembly. These findings provide important new insights into the roles of PtdInsPs in neuronal differentiation.

Results

PtdInsPs promote the association between NF-L and the PH1 domain of PLC-γ1

In our studies aimed at understanding the relationship between NF-L and PLC-γ1-dependent signaling, we found that the PH1 domain of PLC-γ1 directly binds to NF-L (Kim et al., 2006). Because the PH domain is known to be a PtdInsP-binding motif in numerous proteins (Lemmon and Ferguson, 1996; Klopfenstein et al., 2002; Klopfenstein and Vale, 2004), we examined whether PtdInsPs could be involved in the molecular interaction between the PLC-γ1 PH1 domain and NF-L. We performed a GST-PLC-γ1-PH1 pull-down assay using nerve growth factor (NGF)-treated PC12 cell lysates containing different PtdInsPs in vesicle form including PI(3)P, PI(4)P, PI(5)P, PI(3,4)P2, PI(3,5)P2, PI(4,5)P2 and PI(3,4,5)P3. As shown in Figure 1A, substantially more NF-L proteins bound to the GST-PLC-γ1-PH1 domain in PtdInsP-harboring cells compared to untreated controls. This result led us to examine whether NF-L itself can bind PtdInsPs.

Figure 1
figure 1

PtdInsPs promote the interaction between the PH1 domain of PLC-γ1 and NF-L. (A) GST-PLC-γ1-PH1 domain fusion proteins were incubated with NGF-treated PC12 cell extracts (NF-L pool) in the presence of different PtdInsPs (50 µg/ml), and bound proteins were analyzed by Western blotting with anti-NF-L (upper) and anti-GST (middle) antibodies. The relative intensities of the NF-L bands were analyzed using image analysis software (Quantity One, Bio-Rad) (lower). Two independent experiments showed similar results. WCL represents whole cell lysate. (B) Dot-blot analysis shows that PtdInsPs directly bind NF-L. PIP strips were incubated with purified GST, GST-PH1, GST-NF-L (H), GST-NF-L (R) and GST-NF-L (T) fusion proteins (0.5 µg/ml) in TBT buffer containing 2% non-fat skim milk for 14 h at 4℃. After washes with TBT buffer, the membranes were incubated with a goat anti-GST antibody for 2 h at room temperature. Finally, the bound proteins were detected with an HRP-conjugated anti-goat antibody. Each dot contained 100 pM of PI. LPA, lysophosphatidic acid; LPC, lysophosphocholine; PI, phosphatidylinositol; PE, phosphatidylethanolamine; PC, phosphatidylcholine; S(1)P, sphingosine-1-phosphate; PA, phosphatidic acid; PS, phosphatidylserine. The asterisk (*) represents the position of PI(4,5)P2.

Full size image

PtdInsPs directly bind to NF-L

To examine the binding affinity and specificity of NF-L for phospholipids, we purified GST-fused NF-L proteins containing GST-NF-L(H), -NF-L(R) and -NF-L(T), which represent head (H), rod (R) and tail (T) domains of NF-L, from an E. coli expression system and applied the recombinant proteins to dot-blot analysis using Echelon strips (See 'Methods'). Both the NF-L(H) and NF-L(T) domains specifically bound all of the tested PtdInsPs with high affinity (Figure 1B), whereas the NF-L(R) domain showed a weak affinity for only PI(3,4,5)P3. Thus, NF-L directly binds to phospholipids containing the phosphatidylinositol rings of PtdInsPs but not to other tested lipids, such as phosphataidylinositol (PI), phosphatidylethanol (PE), phosphatidylcholine (PC) and sphingosine 1-phosphate (S(1)P) (Figure 1B). NF-L(H) and NF-L(T) showed the highest affinity for PI(3,5)P2, which strongly promoted the molecular interaction between NF-L and the PLC-γ1 PH1 domain (Figure 1A). NF-L(H) and NF-L(T) showed the lowest affinity to PI(4,5)P2, which only moderately promoted the interaction between NF-L and the PLC-γ1 PH1 domain (Figure 1A). The PLC-γ1 PH1 domain also showed high affinities for PI(3,5)P2 and PI(3)P, and had a higher affinity to PI(4,5)P2 than did the NF-L protein (Figure 1B).

Since our results revealed that the NF-L molecule binds to PtdInsPs but not to other phospholipids (e.g., PI, PC or PE), we speculated that the negatively charged phosphate groups in 3rd, 4th and 5th carbon positions of the inositol ring may supply the binding affinity to NF-L. If this is true, then the positively charged basic amino acid residues of NF-L may be involved in binding the phosphate group of the inositol ring. To determine the amino acid residue(s) in NF-L responsible for binding PtdInsPs, we constructed point mutants in which Arg15/Arg16, Arg53, Arg54 and Arg53/Arg54, one or two basic amino acids in the head domain of NF-L, were replaced with alanine (Ala; a neutral amino acid) (Figure 2A). As shown in Figure 2B, the R54A and R53A/R54A mutants showed almost no affinity for the tested PtdInsPs (Figure 2B), while the R53A and R15A/R16A mutants showed only slight decreases in their binding affinities for the tested PtdInsPs (Figure 2B). These results suggest that the positively charged Arg54 is responsible for the binding of NF-L with PtdInsPs. The Ser55 residue (S; boxed in Figure 2A) may be phosphorylated by some protein kinases (Guan et al., 1992; Gibb et al., 1996; Grant and Pant, 2000).

Figure 2
figure 2

PtdInsPs directly binds to Arg54 of NF-L. (A) Schematic representation of the NF-L head domain containing point mutations in arginine residues, which were replaced with the neutral amino acid, alanine. The boxed "S" represents a possible phospho-Ser. (B) Two site-directed point mutants of NF-L, R54A and R53A/R54A, had no binding affinity for PtdInsPs. Point-mutant GST-fusion proteins were used for dot-blotting as described in Figure 1 (B). The asterisk (*) represents the position of PI(4,5)P2.

Full size image

We also assessed the PIP2 binding affinity for R53A/R54A mutant and wild-type NF-L using PIP-bound agarose beads (Figure 3). The beads were incubated with GST-R53A, -R53A/R54A or -wild-type NF-L for 1 h at 4℃, and then extensively washed with buffer. As shown in Figure 3, R53A/R54A mutant NF-L showed a substantially reduced affinity for PIP2 compared with that of wild-type NF-L(H), suggesting that the head domain of NF-L lost its affinity for PIP2 upon mutation of Arg54 residue. In addition to NF-L head domain, there are so many basic amino acid residues in the tail domain of NF-L that shows a high affinity for PtdInsPs (Figure 1B). We then tried to identify the binding partner residues of PtdInsPs within the tail domain, none of 10 point mutants (substitutes alanine for arginine or lysine residue) showed decreased binding affinity for PtdInsPs. So we used only head domain mutant, NF-L(H)-R53A/R54A, for further experiments.

Figure 3
figure 3

The R53A/R54A mutant shows reduced binding affinity for PIP2. Bacterially expressed GST-NF-L proteins were incubated with PIP2-bound agarose beads for 1 h at 4℃, and the bound proteins were quantified by Western blotting.

Full size image

PIP2 accelerates NF-L self assembly

Since purified NF-L can self assemble in vitro (Hisanaga and Hirokawa, 1990; Lee et al., 1993), we next compared NF-L self assembly between R53A/R54A mutant and wild-type NF-L using the GST-fusion proteins. As shown in Figure 4A, R53A/R54A mutant NF-L showed higher self-assembly activity than wild-type NF-L, whereas the control GST protein showed basal activity (Figure 4A). Next, we compared the levels of NF-L self assembly between wild-type GST-NF-L and GST- NF-L-R53A/R54A mutant in the presence of PtdInsPs. The self-assembly reaction of the GST- NF-L-R53A/R54A mutant was unchanged in the presence of PIP2, whereas that of wild-type NF-L was greatly increased by the presence of PIP2 (Figure 4B), suggesting that binding of PIP2 accelerates NF-L self assembly in vitro. We tested other PtdInsPs that showed binding to the NF-L head domain (Figure 1B), and found that their self-assembly-promoting effects on NF-L were similar to those of PIP2 (data not shown).

Figure 4
figure 4

PtdInsP promotes NF-L assembly by direct binding to the Arg54 of NF-L. (A) Comparison of NF-L self assembly between wild-type and R53A/R54A mutant NF-L in vitro. Bacterially expressed GST-NF-L proteins were incubated in NF-L polymerizing buffer and the optical density (OD) was measured over time. (B) PIP2 affects NF-L self assembly. Wild-type and R53A/R54A mutant NF-L were incubated in the presence of PIP2, and the OD was measured over time. Wild-type NF-L self assembly was promoted by PIP2, whereas that of the mutant (which did not bind PIP2) was not.

Full size image

The R53A/R54A mutant shows accelerated NF-L assembly in vivo

Finally, we compared the level of filament formation between the mutant and wild-type NF-L proteins in intact cells. The vimentin-deficient human adrenal carcinoma-derived SW13 cell line has been used for the examination of intermediate- filament formation (Lee et al., 1993). Here, we examined the filament formation levels in SW13 cells co-transfected with vectors encoding rat NF-M with either wild-type NF-L (wild set) or NF-L-R53A/R54A (mutant set). The transfected cells were fixed and double stained with anti-NF-M and anti-NF-L antibodies at various time points. As shown in Figure 5, SW13 cells transfected with the mutant formed intermediate filaments within 15 h after transfection, whereas cells transfected with the wild-type construct needed at least 18 h to form filaments. The mutant set also developed rod-like thick filaments over time, while wild set showed thread-like filaments (Figure 5). This result suggests that in vivo intermediate-filament formation by NF-L is regulated by binding between PtdInsPs and the Arg54 residue.

Figure 5
figure 5

The R53A/R54A mutant shows accelerated NF assembly in intact cells. The cDNAs encoding rat NF-M/NF-L (wild set, left) or NF-M/NF-L-R53A/R54A (mutant set, right) were transiently co-transfected into SW13 cells (Vim-) grown on glass coverslips. The cells were fixed in 4% paraformaldehyde at the indicated times, and subjected to immunostaining with anti-NF-L and anti-NF-M antibodies, followed by detection using FITC- and rhodamine-conjugated secondary antibodies, respectively.

Full size image

Discussion

We herein show that the PtdInsPs directly interact with NF-L and regulate neuronal differentiation by binding with the NF-L head domain along with the PH1 domain of PLC-γ1. We also examined whether PtdInsPs affect NF polymerization, and show for the first time that a number of PtdInsPs, including PI(3)P, PI(4)P, PI(5)P, PI(3,4)P2, PI(3,5)P2 and PI(3,4,5)P3, directly bind NF-L molecules (Figure 1B). All of the tested PtdInsPs bound the NF-L head and tail domains with different affinities, and were found to promote NF-L self assembly in vitro. Interestingly, among all tested PtdInsPs, PI(4,5)P2 showed the poorest binding to NF-L.

Previously, researchers had hypothesized that charge-dependent polymerization (phosphorylation/dephosphorylation) of the head domain of NF-L is responsible for regulating NF-L self assembly (Sihag and Nixon, 1991; Hisanaga et al., 1994; Nakamura et al., 2000; Yates, 2009). The head domain of mammalian NF-L has an Arg-Arg-Ser (RRS) sequence (amino acid residues 53-55), where S55 can be phosphorylated by protein kinases to preclude NF-L self assembly (Sihag and Nixon, 1991; Nakamura et al., 2000; Yates et al., 2009). These previous findings seem to indicate that binding of the negatively charged phosphate group from residue Ser55 of the NF-L head domain precludes self assembly (Gibb et al., 1996; Grant and Pant, 2000; Yates et al., 2009).

Interestingly, our present findings suggests a new regulation mechanism wherein the positively charged Arg54 acts as an adaptor for PtdInsPs, which function as positive effectors for NF-L assembly. Consistent with this hypothesis, we found that the R53A/R54A double mutant was resistant to PtdInsP-induced NF-L self assembly. However, in the absence of any PtdInsP, the R53A/R54A mutant was more efficient than the wild-type protein for self assembly both in vitro (Figure 4) and in vivo, as assessed by intermediate-filament formation in SW13 cells (Figure 5). The mutant-transfected SW13 cells showed a characteristic pattern of filament assembly, characterized by rapid self assembly and the formation of thick, rod-like filaments (Figure 5). Our observations that replacement of the positively charged arginine with a neutral amino acid increased the efficiency of NF-L self assembly may suggest that the positive charge of Arg54 attenuates the self assembly of wild-type NF-L.

In sum, our present findings suggest that Arg54 plays a key role in NF-L self assembly and neurofilament formation, likely via an association with PtdInsPs and/or the direct attenuation of NF-L self assembly. These observations provide important new insights into the roles of PtdInsPs in neuronal differentiation.

Methods

Antibodies and cell culture

Monoclonal anti-NF-L (mAb 1615) and polyclonal anti-NF-L were purchased from Chemicon (Temecula, CA). Monoclonal anti-NF-M (mAb RNF403) was purchased from MP Biochemicals (Aurora, OH). The horseradish peroxidase (HRP)-conjugated goat anti-mouse and goat anti-rabbit antibodies were purchased from Upstate Biotechology Inc. (Lake Placid, NY). Fluorescein-conjugated Affinipure goat anti-rabbit IgG and rhodamine-conjugated Affinipure goat anti-mouse IgG were from Jackson ImmunoResearch Laboratories (West Grove, PA). All PtdInsPs were purchased from Sigma-Aldrich (St. Louis, MO), and the Echelon Strips (P-6001) and PIP-bound agarose beads (P-B045a) were purchased from Echelon Bioscience Inc. (Salt Lake City, UT).

PC12 cells were cultured at 37℃ in Dulbecco's modified Eagle's medium (DMEM) containing 10% fetal bovine serum (FBS), 5% horse serum and 100 U/ml penicillin-streptomycin. The human adrenal carcinoma SW13 cell line (Vim-) was grown at 37℃ in DMEM supplemented with 10% FBS.

Vector constructs and protein expression

For expression in E. coli, the cDNA sequence encoding the amino-terminal PH domain (PH1) of rat PLC-γ1 (amino acids 25-145) (Suh et al., 1988) was ligated into the pGEX-5X-1 vector (Amersham Pharmacia Biotech Inc. Piscataway, NJ) for expression of a GST-PH1 fusion protein, as previously described (Chang et al., 2002). PCR-amplified rat cDNAs encoding NF-L (purchased from the American Type Culture Collection, Manassas, VA) were ligated into the EcoRI/SalI restriction site of pGEX-5X-1. For the expression of domain-specific NF-L fusion proteins, PCR-amplified cDNAs for the NF-L head domain [NF-L(H), amino acid residues 1-93], rod domain [NF-L(R), amino acid residues 93-397], head/rod domain [NF-L(H/R), amino acid residues 1-397], and tail domain [NF-L(T), amino acid residues 398-542] were separately inserted into the EcoRI/SalI sites of pGEX-5X-1. The GST-NF-L-R54A and GST-NF-L-R53A/R54A point mutants were engineered by PCR-based point mutation (Takara Bio Inc. Kyoto, Japan). All constructs were prepared using a Qiagen Plasmid Maxi Kit (Qiagen Inc., Santa Clarita, CA) and confirmed by direct sequencing of the ligation and mutation sites.

The GST fusion proteins were expressed in E. coli, and lysates were incubated with glutathione sepharose (GSH) beads. The bound beads were washed extensively with Igepal buffer [20 mM Tris-Cl, pH 7.5, 1% Igepal CA-630, 300 mM NaCl, 2 mM MgCl2, 1 mM EDTA, 10 µg/ml aprotinin, 10 µg/ml leupeptin, 1 mM phenylmethylsulfonyl fluoride (PMSF) and 1 mM sodium orthovanadate], and the purified proteins were eluted with a buffer containing reduced glutathione.

To examine intermediate-filament formation in SW13 cells, cDNAs encoding rat NF-L (purchased from ATCC) and NF-M (kindly provided by Dr. S. Hisanaga, Tokyo Metropolitan University, with the permission of Dr. R. K. H. Liem, Columbia University) were ligated into the pCMV-FLAG (Clontech Laboratories, Mountain View, CA) and pREFA vectors (Chang et al., 1997), respectively. The resulting constructs, pFLAG-NF-L and pREFA-NF-M, produced proteins that were FLAG-tagged at their N-termini. SW13 cells were transiently co-transfected with pFLAG-NF-L and pREFA-NF-M using a ProFection kit (Promega, Madison, WI), which applies the calcium-phosphate coprecipitation method. The cells were fixed at 37℃ for 10 min in 4% paraformaldehyde for either immunostaining using anti-NF-L and anti-NF-M antibodies, or immunoblotting with an anti-FLAG antibody. For green fluorescent protein (GFP)-fused NF-L expression, PC12 cells were transiently transfected with constructs encoding GFP-NF-L or GFP-NF-L-R53A/R54A fusion proteins, using a Pro-Fection kit.

Immunofluorescent microscopy

SW13 cells were seeded on glass coverslips in 6-well plates, transfected using a ProFection kit, and grown for 2 days in DMEM. The cells were fixed at 37℃ for 10 min in 4% paraformaldehyde, and then incubated with affinity purified monoclonal anti-NF-M or polyclonal anti-NF-L antibodies for 1 h at room temperature in a humidified chamber. Following complete washing with PBS, the cells were incubated with fluorescein-conjugated Affinipure goat anti-rabbit IgG or rhodamine-conjugated Affinipure goat anti-mouse IgG. Immunostained cells were photographed under a confocal laser-beam fluorescent microscope (Perkin-Elmer Inc., Waltham, MA).

Dot-blot analysis

The ability of the PH1 domain and NF-L proteins to bind different phospholipids was examined by dot-blot analysis using Echelon P-6001 strips (Echelon Bioscience Inc.). Each membrane was blocked with TBT buffer for 1 h and then incubated with purified GST, GST-PH1, GST-NF-L or point-mutant GST-NF-L fusion proteins (0.5 µg/ml) in TBT buffer containing 2% non-fat skim milk for 14 h at 4℃. After washes with TBT buffer, the membranes were incubated with goat anti-GST antibodies for 2 h at room temperature. The membranes were then washed extensively with TBT buffer, and then the bound proteins were detected with HRP-conjugated anti-goat antibodies and visualized using an ECL detection system.

In vitro assay of NF-L self-assembly

To assess NF-L polymerization in vitro, purified GST-NF-L or mutant NF-L fusion proteins obtained from E. coli were dissolved in 200 µl of NF-L polymerizing buffer (20 mM Tris-Cl, pH 7.0, 1 mM dithiothreitol, 1 mM MgCl2, 10 µg/ml leupeptin, 10 µg/ml aprotinin, 2 µg GST-NF-L) and incubated for 0.5 h at 35℃, and then optical density was measured at 350 nm. To examine the effect of PtdInsPs on NF-L self assembly, appropriate amounts of phospholipid vesicles (Li et al., 2006) were added before incubation and optical density was measured at 350 nm over time.