14-3-3 proteins interact with neurofilament protein-L and regulate dynamic assembly of neurofilaments

Neurofilaments are type IV intermediate filaments (IFs), and are the main IFs of neurons and an important component of axons. Neurofilaments are obligate heteropolymers of neurofilament proteins: NF-L, NF-M and NF-H (Lee et al., 1993). In common with other IF proteins, neurofilament triplet proteins are composed of an α-helical rod domain, a globular N-terminal head domain and a non-α-helical C-terminal tail domain. The central core of neurofilaments is made up of the rod domains of NF-L, NF-M and NF-H, and the tail domains of NF-M and NF-H are structural components of cross-bridges between adjacent neurofilaments and are involved in the parallel bundling of neurofilaments (Hirokawa et al., 1984; Nakagawa et al., 1995; Chen et al., 2000). The expression of NF-L alone in Sf9 cells lacking endogenous cytoplasmic IFs results in the formation of 10-nm short filaments, while NF-M or NF-H alone cannot self-assemble into 10-nm filaments in vivo (Hirokawa et al., 1984; Nakagawa et al., 1995; Chen et al., 2000). Co-assembly experiments with truncated NF-L showed that the head domain of NF-L is more important for assembly than the tail domain (Chin et al., 1991; Ching and Liem, 1993; Kim et al., 2007). And phosphorylation of the NF-L head domain is important for the regulation of neurofilament assembly (Hisanaga et al., 1990; Sihag and Nixon, 1991; Hisanaga et al., 1994; Gibb et al., 1998; Hashimoto et al., 1998). A series of NF-L head domain phosphorylation sites has been reported: Ser2, 12, 26, 27, 33, 37, 41, 43, 49, 51, 55, 57, 62, 66 (Gonda et al., 1990; Sihag and Nixon, 1991; Cleverley et al., 1998; Hashimoto et al., 1998; Nakamura et al., 2000; Trimpin et al., 2004). However, only Ser2, 55, 57 and 66 are reported to be endogenous phosphorylation sites (Giasson et al., 1996; Hashimoto et al., 2000; Trimpin et al., 2004). Phosphorylation of the NF-L head domain is thought to regulate neurofilament assembly (Sihag and Nixon, 1991; Hisanaga et al., 1994; Giasson et al., 1996; Cleverley et al., 1998; Hashimoto et al., 1998).

Charcot–Marie–Tooth disease (CMT) is the most common inherited degenerative disease of the peripheral nervous system. Its clinical manifestations are limb weakness, atrophy and sensory loss. The first CMT NF-L mutation, Gln333Pro, was linked with CMT disease type 2 in 2000 (Mersiyanova et al., 2000), then, more and more CMT NF-L mutations were revealed, many of which are located in the NF-L head domain: NF-L E7K, P8Q, P8R, P8L, P22S, P22T, and E89K (Jordanova et al., 2003; Sasaki et al., 2006). These mutations affect neurofilament assembly and NF-L head domain phosphorylation (Brownlees et al., 2002; Perez-Olle et al., 2002; Perez-Olle et al., 2004; Sasaki et al., 2006). However, the mechanism of how CMT NF-L mutations lead to abnormal neurofilament assembly and finally cause the disease is unclear.

The 14-3-3 proteins, abundant in brain (Moore et al., 1968), are important signal transduction adaptors. In mammals, there are seven isoforms of these proteins: β, γ, ε, ζ, η, θ, and σ. The interactions of 14-3-3 proteins are dependent on specific phospho-serine-containing motifs of the target proteins (Muslin et al., 1996). There are two kinds of 14-3-3 binding motifs: RSXpSXP and RXXXpSXP (Muslin et al., 1996; Yaffe et al., 1997). 14-3-3 proteins interact with many IFs and these interactions depend on head domain Ser/Thr phosphorylation of the IF proteins. 14-3-3 proteins associate with keratin 18 via Ser33 phosphorylation (Liao and Omary, 1996; Ku et al., 1998); they bind to vimentin through its head domain phosphorylation (Tzivion et al., 2000); and they interact with GFAP via Ser8 phosphorylation (Li et al., 2006). It is reported that 14-3-3 proteins play important roles in the regulation of IF dynamics (Liao and Omary, 1996; Li et al., 2006; Sivaramakrishnan et al., 2009). Since NF-L assembly is regulated by head domain phosphorylation and the mechanism is unclear, we focused on elucidating whether 14-3-3 proteins interact with NF-L and regulate NF-L assembly, and whether these proteins play a role in the abnormal neurofilament assembly caused by CMT NF-L mutations. Our results showed that 14-3-3 proteins interacted with NF-L and affected NF-L dynamic assembly; furthermore, 14-3-3 proteins also alleviated the aggregate formation of CMT NF-L mutants, which reveals 14-3-3 proteins as possible molecular targets for human CMT disease therapy.

To investigate whether 14-3-3 proteins associate with neurofilament proteins in mouse brain, pull-down assays of the seven isoforms of GST–14-3-3 fusion proteins were carried out. The results showed that all seven isoforms of 14-3-3 exclusively associated with NF-L, but no interaction between 14-3-3 and either NF-M or NF-H was detected (Fig. 1A). The results also showed stronger interactions of 14-3-3β and 14-3-3γ with NF-L than the other 14-3-3 isoforms (Fig. 1A). Mouse brain lysate also contained other IF proteins such as GFAP and vimentin, which were capable of co-assembly with NF-L to form 10-nm IFs (Chin et al., 1991; Ching and Liem, 1993; Li et al., 2006). To determine the interaction between 14-3-3 and NF-L is independent of other IF proteins, SW13 Cl2 Vim⁻ cells that lack endogenous cytoplasmic IF proteins were used to perform co-immunoprecipitation assays. Analysis of co-immunoprecipitation of green fluorescent protein (GFP)-tagged 14-3-3γ with anti-GFP antibody revealed that 14-3-3γ still associated with NF-L in the absence of endogenous IF proteins (Fig. 1B). Further analysis revealed that 14-3-3 not only associated with soluble NF-L subunits, but also interacted with pellets of filaments (supplementary material Fig. S4). Furthermore, co-immunoprecipitation assay with anti-14-3-3γ antibody showed that endogenous 14-3-3γ of SW13 Cl2 Vim⁻ cells also interacted with NF-L (Fig. 1C). To determine the capacity of the seven isoforms to associate with NF-L in the absence of endogenous IFs in vivo, NF-L and each GFP-tagged 14-3-3 isoform were co-transfected into SW13 Cl2 Vim⁻ cells, and the co-immunoprecipitation results showed that all seven isoforms associated with NF-L, with 14-3-3γ displaying the strongest association (Fig. 1D), which confirmed the results of the 14-3-3 pull-down assays (Fig. 1A).

To investigate the co-localization of 14-3-3 and NF-L, indirect immunofluorescence assays were carried out in SW13 Cl2 Vim⁻ cells and dorsal root ganglion (DRG) neurons. In SW13 Cl2 Vim⁻ cells, transfected NF-L displayed as morphology of small granules dispersed throughout cytoplasm (Fig. 2A), and the granule morphology is common to either human NF-L (hNF-L) or mouse NF-L (mNF-L) or mouse NF-L (mNF-L) with a FLAG tag (mNF-L–FLAG) (supplementary material Fig. S1A,B). Interestingly, these granules were revealed as a mass of short 10-nm filaments under transmission electron microscopy (TEM) (supplementary material Fig. S1B). To confirm this phenomenon, Sf9 cells that are also lack of endogenous IF proteins were infected with baculoviruses expressing hNF-L, mNF-L or mNF-L–FLAG, clusters of short 10-nm filaments were observed in the cytoplasm (data not shown), and 0.1% Triton X-100 treatment of the infected Sf9 cells before fixation got a clear view of 10-nm short filaments assemble of mNF-L–FLAG (supplementary material Fig. S1C). In SW13 Cl2 Vim⁻ cells, co-localizations between endogenous 14-3-3γ dots and NF-L granules were observed (Fig. 2A). When cells were co-transfected with NF-L and NF-M, they co-assembled into neurofilaments (Fig. 2B; supplementary material Fig. S1D). Some of endogenous 14-3-3γ stained as fibrous distribution and co-localized with neurofilaments (Fig. 2B, top). When cells were co-transfected with NF-L, NF-M and 14-3-3γ–red fluorescent protein (RFP), co-localization between fibrously distributed 14-3-3γ–RFP and neurofilament bundles was also observed (Fig. 2B, bottom). And there are also some other neurofilament bundles showed no co-localization with 14-3-3γ–RFP (Fig. 2B, bottom). Immunofluorescence staining of DRG neurons revealed that, some fibrously distributed 14-3-3 co-localized with neurofilaments in axons (Fig. 2C, top) and growth cone (Fig. 2C, bottom). In order to determine the direct interaction between 14-3-3 and NF-L, we performed fluorescence resonance energy transfer (FRET) analyses between cyan fluorescent protein (CFP)-tagged 14-3-3 (donor) and yellow fluorescent protein (YFP)-tagged NF-L (acceptor) in SW13 Cl2 Vim⁻ cells. When the selected co-localization granules were bleached of acceptor fluorescence (NF-L–YFP), the intensity of donor fluorescence (CFP–14-3-3) increased over the value before bleaching (Fig. 2D,E, labels A1, D1). However, the control granules without bleaching showed no significant change in the fluorescence intensity of donor and acceptor (Fig. 2D,E, labels A2, D2). FRET analysis of CFP–14-3-3 and NF-L–YFP showed about 17% FRET efficiency (Ef) of the bleached zone compared with the control efficiency (Cf) of about 3% in the unbleached zone (Fig. 2F).

To determine which domain of NF-L is required for binding of 14-3-3, mutants of NF-L with the head or tail domain truncated were constructed (Fig. 3A). After co-transfection of NF-M and either wild-type or truncated NF-L into SW13 Cl2 Vim⁻ cells, immunofluorescence analyses revealed the disrupted co-assembly of neurofilaments in head domain-truncated NF-L and NF-M, while the tail domain-truncated mutant still co-assembled with NF-M into filaments (Fig. 3B). These filament networks formed by tail domain-truncated NF-L and NF-M were irregular and distributed randomly in the cytosol, unlike the classical filament array extending from the nucleus to the periphery of the cell formed by wild-type NF-L and NF-M. These immunofluorescence data suggested the importance of the head domain for the assembly of NF-L (Fig. 3B). Co-immunoprecipitation assays demonstrated that deletion of the head domain of NF-L prevented the association of 14-3-3, whereas NF-L with the tail domain deleted retained association, further suggesting the necessity of the NF-L head domain for 14-3-3 association (Fig. 3C). Commonly, phosphorylation of binding targets is required for 14-3-3 interaction. To determine whether interaction between 14-3-3 and NF-L depends on phosphorylation, two 14-3-3 binding motifs at the head domain of NF-L were predicted at website: http://scansite.mit.edu/motifscan_seq.phtml (Fig. 3D). However, mutation of both predicted phosphorylation sites only diminished 14-3-3 binding to some extent, but not completely (Fig. 3E), indicating that some other phosphorylation sites were also required for 14-3-3 binding. It is known that Ser2, 55, 57 and 66 are phosphorylation sites of the NF-L head domain in vivo (Giasson et al., 1996; Hashimoto et al., 2000; Trimpin et al., 2004). Mutation of these sites and the predicted 14-3-3 binding sites simultaneously further diminished the 14-3-3 binding. Only the NF-L with mutation of all serines of the head domain showed completely inhibited 14-3-3 binding (Fig. 3E,F), indicating multiple phosphorylation sites involved in the interaction of 14-3-3 with NF-L. Treatment with the phosphatase-2A inhibitor, okadaic acid (OA), increases the head domain phosphorylation of NF-L (Sacher et al., 1994). After OA treatment, we found that the association between wild-type NF-L and 14-3-3 was increased, whereas the interaction with NF-L having all serines of the head domain mutated remained unchanged (Fig. 3G,H), which confirmed the necessity of phosphorylation of the NF-L head domain for binding of 14-3-3.

To investigate which phosphorylation sites of the NF-L head domain mediated 14-3-3 binding, a series of single mutations of reported serine phosphorylation sites of this domain were constructed. Co-immunoprecipitation results demonstrated that single site mutations of Ser41, 43, 49 and 55 significantly reduced 14-3-3 binding, and mutations of Ser26 and 62 also decreased it, although not so markedly (Fig. 4A). Immunofluorescence assays of SW13 Cl2 Vim⁻ cells transfected with these NF-L mutants revealed that those mutants affecting 14-3-3 binding, including S26A, S49A, S55A and S62A, resulted in the formation of sheet-like aggregates. In contrast, wild-type NF-L displayed as granules dispersed in the cytosol (Fig. 4B,C), indicating the important role of 14-3-3 in regulating NF-L assembly. However, some mutants, such as S41A and S43A, which significantly reduced 14-3-3 binding, displayed no formation of aggregates, suggesting these sites affected 14-3-3 binding only (Fig. 4B,C). To investigate whether these NF-L mutants affect the co-assembly with NF-M, NF-L mutants and NF-M were co-transfected into SW13 Cl2 Vim⁻ cells. Results showed that all of these mutants could co-assembly with NF-M into neurofilament networks (supplementary material Fig. S2). Therefore, co-assembly with NF-M into neurofilament networks was not affected by mutation of single phosphorylation site of NF-L. It is reported that Ser41, 43, 49, 55 and 62 are phosphorylated by protein kinase A (PKA) (Cleverley et al., 1998; Hashimoto et al., 2000; Nakamura et al., 2000), and our study demonstrated that mutation of these phosphorylation sites inhibited the interaction between 14-3-3 and NF-L (Fig. 4A), which together suggested the role of PKA in the regulation of 14-3-3 binding to NF-L. Assays with the PKA inhibitor KT5720 and the activator forskolin showed that the inhibitor decreased the interaction between 14-3-3 and NF-L, while the activator increased the interaction (Fig. 4D). Furthermore, the interaction between the NF-L S55A mutant and 14-3-3 was still increased by treatment with the PKA activator, indicating that several PKA phosphorylation sites of NF-L were involved in 14-3-3 association besides Ser55.

Previous studies revealed the importance of 14-3-3 in the regulation of the dynamics of keratin and GFAP (Liao and Omary, 1996; Li et al., 2006; Sivaramakrishnan et al., 2009). To determine the role of 14-3-3 in the regulation of NF-L dynamics and neurofilament assembly, fluorescence recovery after photobleaching (FRAP) analyses were carried out (Fig. 5A). The results revealed that co-expression of 14-3-3 clearly increased the dynamic exchange rate of NF-L, with an average t1/2 of 16±1 s and a rate constant of 0.049±0.004, compared to an average t1/2 of 23±3 s and rate constant of 0.035±0.003 in control cells (Fig. 5B,C). Since 14-3-3 facilitated the dynamic exchange rate of NF-L, we set out to assess the role of 14-3-3 overexpression in the regulation of the neurofilament network. Normally, NF-L and NF-M could co-assemble into extensive neurofilament networks; however, upon co-expression of NF-L and NF-M with 14-3-3, overexpression of 14-3-3 disassembled the neurofilament network and little extensive neurofilament networks were observed (Fig. 5D, lower panels). However, in cells with low amounts of 14-3-3 expression, the neurofilament assembly was unaffected (Fig. 5D, upper panels). Statistical analysis showed that the percentage of cells with neurofilament network was decreased from 93% to 49% owing to 14-3-3 overexpression (Fig. 5E).

Reported CMT NF-L mutants show defects in neurofilament assembly and transport (Perez-Olle et al., 2004), while P22S and P22T mutants abolish phosphorylation of Thr21 in the NF-L head domain (Sasaki et al., 2006). P8 and P22 residues are mutational hot-spots of NF-L in CMT disease, and so far, five mutations of these two residues have been reported in CMT disease: P8Q, P8R, P8L, P22S and P22T (Jordanova et al., 2003; Perez-Olle et al., 2004; Sasaki et al., 2006; Dequen et al., 2010). To investigate the effects of these mutations on the association of 14-3-3 with NF-L, all five mutants were constructed. The results of co-immunoprecipitation showed that mutants P8Q, P8L and P22S clearly inhibited 14-3-3 binding (Fig. 6A). The fluorescence result demonstrated that these CMT mutants formed huge aggregates in the cell body, in contrast to the small granules formed by wild-type NF-L (Fig. 6B). Since these CMT mutants affected 14-3-3 association and NF-L self-assembly, we then investigated their effect on the co-assembly with NF-M into neurofilaments. Fluorescence analysis of cells co-transfected with NF-M and CMT mutants revealed that the P8Q and P8R mutants retained the ability to co-assemble with NF-M into neurofilament networks, whereas the P8L, P22S and P22T mutants significantly inhibited this co-assembly (Fig. 6C). Statistical analysis showed that 91.0% of cells co-transfected with wild-type NF-L and NF-M had networks of co-assembly neurofilaments, and this value was decreased to 50.3% in the P8Q mutant and 63.7% in the P8R mutant; and the percentage of cells with neurofilament networks was significantly decreased to 19.7% in the P8L mutant, 16.9% in P22S, and P22T mutant completely inhibited formation of neurofilament network (Fig. 6D), which is consistent with the report that among the three mutations of the P8 residue in patients, a more serious phenotype is caused by P8L mutation (Jordanova et al., 2003). Interestingly, statistical analysis demonstrated that co-expression of 14-3-3 with CMT mutants reduced the formation of huge aggregates by CMT mutants (Fig. 6E,F). However, co-expression of 14-3-3 with CMT mutants and NF-M, failed to rescue the deficiency of CMT mutants to co-assembly with NF-M into neurofilament networks (supplementary material Fig. S3).

Previous studies demonstrated that 14-3-3 interacts with several IF proteins such as keratin, vimentin and GFAP (Ku et al., 1998; Tzivion et al., 2000; Kim et al., 2006; Li et al., 2006). However, the keratins are Type I/II IFs, while vimentin and GFAP are Type III. Here, we report that 14-3-3 proteins interact with a Type IV IF, NF-L, which is the core component of neurofilaments.

Seven isoforms of 14-3-3 are expressed in brain (Moore et al., 1968), named 14-3-3 β, γ, ε, ζ, η, θ and σ. GST–14-3-3 binding assays showed that all seven isoforms specifically interacted with NF-L, but not NF-M or NF-H (Fig. 1A). NF-L associated with GST–14-3-3 was in the soluble state, and it is thought that soluble IFs take the form of tetramers (Herrmann and Aebi, 1998). Although NF-M and NF-H are known to co-assemble with NF-L into neurofilaments, GST–14-3-3 binding assays suggested that these soluble NF-L tetramers associated with 14-3-3 were homopolymeric, explaining the specific binding of 14-3-3 to NF-L, but not NF-M or NF-H. There was another possibility that NF-L co-assembled into heteropolymers with vimentin or GFAP in brain, and thus the interaction with 14-3-3 was dependent on vimentin or GFAP, which were shown to associate with 14-3-3 in previous studies. To rule out this possibility, further co-immunoprecipitation assays were carried out in the SW13 Cl2 Vim⁻ cell line with no endogenous cytoplasmic IFs. Results showed that all seven isoforms of 14-3-3 still interacted with NF-L, independent of vimentin or GFAP (Fig. 1D). Although the seven isoforms are highly homologous, GST–14-3-3 binding assays and co-immunoprecipitation assays showed that different isoforms showed differential associations with NF-L. These assays demonstrated that 14-3-3γ had the strongest interaction. The differences in association may be due to competition between NF-L and other favored binding targets of each 14-3-3 isoform, since they are important signal transduction adaptor proteins with many binding targets. Previous studies also reported competitive interactions between 14-3-3 and IF proteins as well as other signal transduction proteins (Tzivion et al., 2000; Ku et al., 2002; Kim et al., 2006; Margolis et al., 2006).

14-3-3 proteins usually interact with phosphorylated targets. Two common 14-3-3 binding motifs are reported: RSXpSXP and RXXXpSXP (Muslin et al., 1996; Yaffe et al., 1997); pS represents the phosphorylation site of target proteins. So far, reported interactions between 14-3-3 and IF proteins are all dependent on phosphorylation of target the IFs, including keratin18 (Liao and Omary, 1996), vimentin (Tzivion et al., 2000), GFAP (Li et al., 2006) and keratin17 (Kim et al., 2006). Our study also demonstrated the requirement of head domain phosphorylation for interaction between 14-3-3 and NF-L (Fig. 3). However, among these binding targets, only Thr9 and Ser44 residues of keratin17 have been shown to affect 14-3-3 association, consistent with the predicted 14-3-3 binding motifs, although they do not exactly match the two classical 14-3-3 binding motifs (Kim et al., 2006). The reported Ser33 residue of keratin18 and the Ser8 residue of GFAP and vimentin do not match or contain classical 14-3-3 binding motifs. In our study, co-immunoprecipitation assays showed that several serine residues were necessary for 14-3-3 association (Fig. 4A). Among these residues, only Ser43 was the site of a predicted 14-3-3 binding motif of the NF-L head domain, which still partially matched the classical 14-3-3 binding motifs. None of the remaining residues matched 14-3-3 binding motifs. Notably, Ser57, another predicted 14-3-3 binding site and simultaneously a phosphorylation hot-spot of the NF-L head domain, had no effect on the interaction between 14-3-3 and NF-L (Fig. 4A). Together, the lack of classical 14-3-3 binding motifs in these 14-3-3-associated IF proteins suggested that the interaction between 14-3-3 and IF proteins is weak, unlike signal transduction proteins such as Raf-1 (Fu et al., 1994; Muslin et al., 1996), BAD (Zha et al., 1996) and Cdc25C (Peng et al., 1997), which contain classical 14-3-3 binding motifs and show strong interaction with 14-3-3.

It is reported that the head domain phosphorylation of NF-L is regulated by phosphatase-2A (Sacher et al., 1994; Giasson et al., 1996), and okadaic acid (OA) treatment, which specifically inhibits phosphatase-2A activity, increases the head domain phosphorylation of NF-L (Sacher et al., 1994). Since interaction between 14-3-3 and NF-L was dependent on phosphorylation, OA treatment markedly increased the association between 14-3-3 and NF-L (Fig. 3G), which was probably due to the increase phosphorylation level of the NF-L head domain by the inhibitory activity of phosphatase-2A. There are seven PKA phosphorylation sites at the head domain of NF-L: Ser2, 12, 41, 43, 49, 55 and 62 (Giasson et al., 1996; Cleverley et al., 1998; Nakamura et al., 2000). Our study revealed that mutation of Ser41, 43, 49, 55 and 62 had an evident effect on reducing the association between 14-3-3 and NF-L (Fig. 4A), indicating the role of PKA in the regulation of 14-3-3 binding. Further assays of PKA inhibitor and activator treatment confirmed this hypothesis (Fig. 4D). Mutation of Ser26, a phosphorylation site of Rho kinase (Hashimoto et al., 1998), also decreased 14-3-3 association (Fig. 4A), indicating that other kinases are also involved in the regulation of 14-3-3 association. The phosphorylation level of the NF-L head domain is regulated by a series of kinases and phosphatases, which results in changeable 14-3-3 associations. To some extent, binding of 14-3-3 to specific phosphorylation sites may prevent dephosphorylation of NF-L by phosphatase. Since many serine phosphorylation sites at the head domain of NF-L were involved in 14-3-3 association, mutation of only one or even several serine phosphorylation sites resulted only in partial reduction of 14-3-3 association (Fig. 3E; Fig. 4A). In fact, only mutation of all serines of the NF-L head domain completely disrupted 14-3-3 binding (Fig. 3E,G). Therefore, in terms of each functional phosphorylation site alone, the interaction with 14-3-3 is not strong, but the many phosphorylation sites of the NF-L head domain substantially enhance the association of 14-3-3, which also complicates the regulation of interaction between 14-3-3 and NF-L.

Transmission electron microscopy (TEM) shows that NF-L self-assembles into 10-nm filaments (Nakagawa et al., 1995; Chen et al., 2000), however, in this study, no filament structure was observed under fluorescence microscopy when expressed NF-L alone in SW13 Cl2 Vim⁻ cells. NF-L alone showed the morphology of small granules diffused throughout the cytoplasm in SW13 Cl2 Vim⁻ cells (Fig. 2A; supplementary material Fig. S1B). According to our TEM result, these granules are tangles of short 10-nm filaments (supplementary material Fig. S1B). However, co-transfected NF-L and NF-M can co-assembled into clear neurofilament networks (Fig. 2B; Fig. 3B; supplementary material Fig. S1D). It is suggested that these co-assembled neurofilaments seen under fluorescence microscopy are parallel bundle of 10-nm filaments, just like the phenomenon in Sf9 cells reported in previous studies (Nakagawa et al., 1995; Chen et al., 2000).

The disassembly of neurofilament network due to 14-3-3 overexpression (Fig. 5D) and the association of 14-3-3 with soluble NF-L subunits (supplementary material Fig. S4) together indicate the disassembly role of 14-3-3. Therefore, on one hand, 14-3-3 associates with the soluble NF-L subunits to inhibit assembly; on the other hand, 14-3-3 binds to phosphorylated neurofilaments and leads to neurofilament network disassembly. Since the regulation of IF disassembly is dependent on phosphorylation state, hyperphosphorylation of NF-L leads to the disassembly of neurofilaments (Giasson and Mushynski, 1998). Here, specific interaction of 14-3-3 proteins with phosphorylated NF-L subunits also indicated the role of 14-3-3 and NF-L phosphorylation in the disassembly of neurofilaments. What is more, binding of 14-3-3 to phosphorylated NF-L subunits may prevent the dephosphorylation of these subunits by phosphatases, maintaining the hyperphosphorylation state of the subunits, which facilitates the disassembly of neurofilaments. 14-3-3 specifically interacts with phosphorylated IF proteins, and many studies have revealed the role of 14-3-3 in IF disassembly. For example, during astrocyte mitosis, phosphorylation of GFAP and binding of 14-3-3 are followed by disassembly of glial filaments (Li et al., 2006); phosphorylation of the Ser33 site of keratin18 together with 14-3-3 binding during mitosis results in sequestration of keratin18 into the soluble pool, enhancing its solubility (Liao and Omary, 1996). It is reported that disruption of association between 14-3-3 and keratin18 results in the inhibition of keratin network disassembly induced by shear-stress and significantly retards FRAP recovery of keratin18 (Sivaramakrishnan et al., 2009). In our study, 14-3-3 overexpression not only resulted in neurofilament disassembly, but also increased the dynamic exchange rate of NF-L subunits (Fig. 5), indicating the important role of 14-3-3 in the regulation of NF-L assembly. And the co-localization data suggests that NF-L granules co-localized with 14-3-3 maybe the soluble portion of NF-L, whereas the independent NF-L granules without 14-3-3 association maybe the assemble 10-nm filaments (Fig. 2A). There are also fibrously distributed 14-3-3 co-localized with neurofilament bundles in SW13 Cl2 Vim⁻ cells co-transfected with NF-L and NF-M, and with neurofilaments in DRG neurons (Fig. 2B,C). These neurofilaments associated with 14-3-3 are possibly going to disassembly.

Mutations in NF-L cause CMT disease in humans. CMT patients with NF-L mutations are classified as CMT2E (Mersiyanova et al., 2000; De Jonghe et al., 2001). In our study, proline mutations of the NF-L head domain in CMT disease disrupted 14-3-3 association (Fig. 6A). It is possible that mutations of proline residues change the secondary and tertiary structure of the NF-L head domain, and as a result, inhibit 14-3-3 association, since this is dependent on the head domain phosphorylation of NF-L. There is another possibility, that proline mutations affect Ser/Thr phosphorylation of the NF-L head domain, therefore inhibiting phosphorylation-dependent 14-3-3 association; there is a report that Pro22 mutation of the NF-L head domain inhibits nearby Thr21 phosphorylation (Sasaki et al., 2006). Among five head domain proline mutations of NF-L (P8Q, P8R, P8L, P22S and P22T), mutations P8Q, P8L and P22S significantly inhibited 14-3-3 interaction (Fig. 6A). Interestingly, there is a correlation between the ability of CMT mutants to interact with 14-3-3 and the motor nerve conduction velocity (NCV). With regard to the mutants P8Q, P8R and P8L, a lower interaction with 14-3-3 (Fig. 6A), was correlate with a lower motor NCV in the CMT patients (Jordanova et al., 2003). This correlation suggests a role of 14-3-3 in the pathogenesis of CMT disease, due to abnormal regulation of NF-L assembly. These reported proline mutations all induced the formation of huge aggregates of NF-L (Fig. 6B). Furthermore, the P8L, P22S and P22T mutations even disrupted co-assembly with NF-M into neurofilament arrays (Fig. 6C). It is possible that disruption of 14-3-3 binding leads to the formation of aggregates. One of CMT patients' clinical symptoms is distal predominance of motor and sensory neurons. Since NF-L is an important component of axons, formation of NF-L aggregates has a negative effect on axon development. Therefore, prevention of NF-L aggregate formation may be an aspect of therapy for CMT disease.

In summary, this study reports the interactions between 14-3-3 proteins and NF-L, and the regulation of NF-L assembly and dynamic exchange by 14-3-3. The formation of NF-L aggregates in CMT patients with NF-L mutations may be due to inhibition of 14-3-3 binding to these mutants, result in abnormal neurofilament assembly. Our results suggest 14-3-3 proteins as potential molecular targets for human CMT disease therapy.

All experiments on animals were performed according to the standards of Animal Facility of Peking University on Animal Care and National Institutes of Health guidelines. Seven isoforms of 14-3-3 cDNA were synthesized by RT-PCR using mRNA samples from mouse brain. The cDNA of NF-L and NF-M were kindly provided by Dr Nobutaka Hirokawa (University of Tokyo, Tokyo, Japan) (Nakagawa et al., 1995). NF-L deletion mutants (NF-L-ΔH (Δ1-93aa), NF-L-ΔT (Δ403-543aa), NF-LΔ1-40aa, NF-LΔ1-50aa, NF-LΔ1-60aa, NF-LΔ1-70aa and NF-LΔ42-98aa) and phosphorylation site mutants (S2A, S12A, S26A, S27A, S33A, S37A, S41A, S43A, S49A, S51A, S55A, S57A, S62A and S66A) were constructed by PCR amplification of NF-L cDNA. The NF-L cDNA, NF-L mutants and NF-M cDNA were subcloned into p3XFLAG-CMV-14 vector (Sigma-Aldrich, Inc.). NF-L cDNA was also subcloned into pEGFP-N3 vector (Clontech Laboratories, Inc.). The seven isoforms of 14-3-3 cDNA were inserted into pEGFP-N3 vector (Clontech Laboratories, Inc.). 14-3-3γ was also inserted into pHM6 vector (Roche Molecular Biochemicals) and pmRFP-N3 vector [enhanced GFP (EGFP) tag of pEGFP-N3 vector replaced with monomeric RFP (mRFP) tag].

GFP cDNA was subcloned into vector pET-28a (Novagen). His–GFP expressed in E. coli BL21 was purified and used as an antigen to immunize rabbits.

Antibodies were sourced as follows: mouse anti-NF-L (clone DA2; Abcam), rabbit anti-NF-L (a gift from Dr Nobutaka Hirokawa), mouse anti-NF-M (clone NN18; Sigma), rabbit anti-NF-H (a gift from Hirokawa), mouse anti-14-3-3γ (clone CG31; Millipore), rabbit anti-14-3-3γ (Santa Cruz Biotechnology), mouse anti-GFP (clone 1E4; Abcam), mouse anti-FLAG (clone M2; Sigma).

Cultures of DRG neurons were prepared as described previously (Okabe and Hirokawa, 1990). SW13 Cl2 Vim⁻ cell lines (a kind gift from Dr Robert M. Evans, University of Colorado Health Sciences Center, Denver, CO) were grown in Dulbecco's modified Eagle's medium (DMEM) (Gibco) with 10% fetal bovine serum (FBS) at 37°C in 5% CO₂, and transfected with jetPEI (Polyplus transfection) for immunofluorescence experiments, or Lipofectamine 2000 reagent (Invitrogen) for immunoprecipitation experiments.

The seven isoforms of GST–14-3-3 fusion proteins were immobilized on glutathione–Sepharose 4B beads (Amersham) for binding assays. Adult mouse brains were homogenized in lysis buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1% Triton X-100, protease inhibitor cocktail; Roche). The homogenate was centrifuged twice at 12,000 g for 30 min and the supernatant was added to the affinity column. After washing with lysis buffer several times, the eluted samples were analyzed by western blotting.

Cells were washed twice with PBS and lysed in lysis buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1% Triton X-100, 2 mM EDTA, protease inhibitor cocktail). The lysate was centrifuged at 16,000 g for 30 min. The supernatant was mixed with antibody at 4°C overnight. Then 20 µl Protein A–Sepharose beads (Pharmacia) were added and the mixture was rotated at 4°C for 2 h. After washing three times with lysis buffer, the beads were eluted and boiled in 2×SDS sample buffer and subjected to western blot analysis.

Cells grown on coverslips were washed with PBS, fixed in 4% paraformaldehyde for 15 min and permeabilized with 0.2% Triton X-100 for 10 min at room temperature. After blocking with 1% BSA for 30 min, cells were incubated with primary antibodies for 1 h at room temperature. The secondary antibodies were FITC, TRITC, Alexa-488- or Alexa-568-conjugated goat anti-mouse or -rabbit IgG. Cell samples were observed under an LSM-710 confocal laser-scanning microscope (Zeiss, Germany), a TCS SP2 confocal microscope (Leica, Germany), or a TH4-200 fluorescence microscope (Olympus, Japan).

The acceptor photobleaching method was used to measure FRET. FRET was carried out with the LSM-710 confocal microscope. The fluorescence of GFP-linked fusion protein (donor) and RFP-linked protein (acceptor) was recorded before and after bleaching of RFP. FRET efficiency was calculated by Zeiss FRET software. GFP-tagged NF-L and RFP-tagged 14-3-3γ were co-transfected into SW13 Cl2 Vim⁻ cells to detect direct interactions between 14-3-3γ and NF-L.

FRAP was carried out with the TCS SP2 confocal microscope. Phase-contrast images of cells were taken before and immediately after FRAP to ensure that there were no significant changes in cell shape or position. Bar-shaped regions were bleached using the line-scan function at 488 nm (100% power), and recovery of fluorescence was monitored (15% power) using the time-series function at 10 s for up to 100 s. SW13 Cl2 Vim⁻ cells were transfected with pEGFP-NF-L and pHM6-14-3-3γ simultaneously or with control pHM6 empty vector, to assess the effect of 14-3-3γ on NF-L dynamics.

Transfected cells were treated with 100 nM OA (Calbiochem), 1 µM KT5720 (Calbiochem) or 15 µM forskolin (Sigma) for 2 h, then harvested for further experiments.

Pixel density of the bands from western blots was measured by Image J (NIH). The unpaired Student's t test was used to perform comparisons. P<0.05 was considered significant.

We are grateful to Robert M. Evans (University of Colorado Health Sciences Center, USA) for SW13 Cl2 Vim⁻ cell lines.