Neurofilament heavy chain side arm phosphorylation regulates axonal transport of neurofilaments

Neurofilaments are transported through axons by a process termed slow axonal transport. However, recent studies have shown that they move at conventional fast rates of ∼1 μm/s but that this movement is interrupted by prolonged pauses (Yabe et al., 1999; Prahlad et al., 2000; Roy et al., 2000; Shah et al., 2000; Wang et al., 2000; Wang and Brown, 2001). At any one time, most neurofilaments are therefore stationary, giving rise to an overall slow rate of transport.

The molecular mechanisms that regulate neurofilament transport are not properly understood, but a body of evidence associates increased phosphorylation of neurofilament side arms with slower transport rates (Ackerley et al., 2000; Sanchez et al., 2000; Yabe et al., 2001). In most mature neurons, neurofilaments comprise three subunit proteins, neurofilament light, middle, and heavy chains (NFL, NFM, and NFH),^* and the carboxy-terminal domains of NFM and NFH form side arms that extend from the filament. These NFM/NFH side arms are phosphorylated in axons, with that of NFH being particularly heavily phosphorylated (Pant et al., 2000). Much of this phosphate is located in a domain that contains repeats of the motif lys-ser-pro (KSP). Kinases that phosphorylate the serines in these KSPs include Cdk5/p35, GSK-3α/β, and members of the MAPK/SAPK family (Pant et al., 2000). Here, we have investigated the role of side arm phosphorylation in neurofilament transport by analyzing movement of EGFP-tagged phosphorylation mutants of NFH in neurons. Our results provide direct experimental evidence to support a role for NFH side arm phosphorylation as a regulator of neurofilament transport.

We confirmed that amino-terminal tagging of NFH with GFP does not influence its ability to form neurofilaments by studying its assembly properties in SW13− cells that do not contain intermediate filaments. Transfection of GFP-tagged wild-type NFH (GFP–NFHwt) or phosphorylation mutants of NFH (GFP–NFHala and GFP–NFHasp) with NFL, NFL + NFM, and NFL + NFM + NFH into SW13− cells all led to the formation of neurofilament networks that were not noticeably different from those formed by wild-type, untagged NFH (unpublished data; Fig. 1, A–F). Transfected GFP–NFHwt, GFP–NFHala, and GFP–NFHasp also colocalized with endogenous neurofilaments in rat cortical neurons (Fig. 1, G–L). These results are in agreement with earlier studies of GFP–NFH/M (Ackerley et al., 2000; Roy et al., 2000; Wang et al., 2000).

NFH and NFM side arms are heavily phosphorylated in axons but not cell bodies. Thus, antibodies 8D8 and RT97, which detect phosphorylated NFH/NFM side arms, and antibody RMO45, which detects phosphorylated NFM side arms, all labeled axons but not cell bodies in cortical neurons (Fig. 1, M and N, 8D8 and RT97). Labeling of axons with RT97 commenced ∼20 μm from cell bodies, which was determined by analyzing RT97 and NA1211 dual-labeled neurons; NA1211 detects NFH irrespective of its phosphorylation status. These results are consistent with earlier analyses of NFM/H phosphorylation in cortical neurons (Ackerley et al., 2000; Brownlees et al., 2000).

Transfection of GFP–NFHwt, GFP–NFHala, or GFP–NFHasp did not alter this RT97/8D8 labeling pattern over the time periods in which GFP–NFH transport was investigated (140–260 min and 48 h after transfection, see below); cell bodies remained negative and axons positive for RT97/8D8. Furthermore, regions of axons that contained GFP–NFHwt showed increased RT97/8D8 staining, which shows that phosphorylation of the GFP–NFH species on these epitopes is mimicking that of endogenous NFH (unpublished data; Fig. 1, O–R).

Most phosphorylation sites in NFH side arm are within a domain that contains repeats of the motif KSP (Fig. 2 A), but only 12 of these have been formally identified as in vivo phosphorylation sites in the rat (Elhanany et al., 1994). KSPs with the motif KSPXK are believed to be targeted by Cdk5/p35, and several lines of evidence indicate that these are functionally significant (Shetty et al., 1993; Guidato et al., 1996; Sun et al., 1996; Bajaj and Miller, 1997). We therefore chose to analyze the effect of Cdk5/p35 phosphorylation on NFH transport. The known in vivo phosphoserines within the consensus Cdk5/p35 motif are those in repeats 44–48, 50, and 51 (Elhanany et al., 1994), and these were mutated to alanine to preclude phosphorylation or aspartate to mimic permanent phosphorylation (Fig. 2 A). There are many examples where the substitution of such a charged residue accurately mimics the functional effect of phosphorylation (Eidenmuller et al., 2000).

Transfection of NFHwt alone into COS cells revealed that it had an accelerated rate of migration on SDS-PAGE compared with hyperphosphorylated NFH from brain when blots were probed with antibody NA1211 (Fig. 2 B). However, cotransfection with Cdk5/p35 induced a shift in a proportion of transfected NFHwt so that it migrated close to NFH from brain; this shift was abolished in cells treated with the Cdk5/p35 inhibitor roscovitine (Fig. 2 B). Similarly, cotransfection of NFHwt with Cdk5/p35 increased its reactivity with antibodies, 8D8 and RT97, that detect phosphorylated NFH side arms (Fig. 2 B), which is in agreement with several previous studies (Guidato et al., 1996; Sun et al., 1996; Bajaj and Miller, 1997).

Mutant NFH in which the Cdk5/p35 sites were mutated to alanine (NFHala) comigrated with NFHwt when transfected alone into COS cells, but mutation of the sites to aspartate (NFHasp) induced a minor upward shift in a proportion of the transfected protein. This may be due to the aspartate residues mimicking phosphorylation that primes phosphorylation on other residues to induce a shift. NFHala and NFHasp did not react or reacted only weakly with antibodies 8D8 and RT97. However, after cotransfection of the mutants with Cdk5/p35, increased labeling with antibodies 8D8 and RT97 was observed; although this was significantly weaker than NFHwt in Cdk5/p35-cotransfected cells (Fig. 2 B). Cotransfection of NFHala or NFHasp with Cdk5/p35 also induced upward mobility shifts in a proportion of the transfected protein. However, the shift for NFHala was minor, whereas that for NFHasp was more pronounced so that it migrated close to NFH from brain. These results are consistent with the notion that some, if not all, of the sites mutated in NFHala are phosphorylated by Cdk5/p35 but that Cdk5/p35 targets other, as yet unidentified, sites.

The failure of NFHasp to react strongly with 8D8 and RT97 is as expected. Substituting charged residues in tau mimics the functional effects of phosphorylation but does not generate the epitopes for phosphorylation-dependent antibodies (Eidenmuller et al., 2000).

We calculated the transport rate of NFHwt, NFHala, and NFHasp in axons using a previously described assay (Ackerley et al., 2000). GFP–NFHwt was transported at a rate of ∼80 μm/h, and this was not significantly different from that of GFP–NFM (Fig. 3 A). These rates were linear over the course of the experiment, which suggests that very early movement of nonphosphorylated neurofilaments close to cell bodies (Sanchez et al., 2000) is not being studied. Indeed, dual labeling of GFP–NFHwt-transfected cells with RT97 revealed that phosphorylation on NFH to generate this epitope commences ∼20 μm from the cell body (PFig. 1, O and P). This is less than the distance travelled by GFP–NFHwt/ala/asp at the first time point analyzed (Ackerley et al., 2000; Fig. 3, legend).

We next examined how the mutation of phosphorylation sites affects NFH transport. GFP–NFHasp was transported significantly slower (52 μm/h) and GFP–NFHala significantly faster (108 μm/h) than GFP–NFHwt (Fig. 3 B). GFP–NFHala transport was determined for the first four time points, because at later times, its signal became progressively weaker toward the front, which precluded accurate measurements. The weaker signal obtained from GFP–NFHala at these later time points is consistent with the more rapid transport of a finite amount of GFP–NFHala with a concomitant dilution of fluorescent signal.

We also calculated the average distance moved by GFP–NFHwt and GFP–NFHasp 240 min after transfection in the presence of the Cdk5/p35 inhibitor roscovitine. Cdk5/p35 is active in cortical neurons (Nikolic et al., 1996). 20 μM roscovitine or vehicle was applied to the cells 130 min after transfection and maintained throughout the course of the experiment. Roscovitine did not induce any noticeable changes in the cells, including alterations to nuclear morphology. However, roscovitine significantly increased the distance travelled by GFP–NFHwt at this time point but had little effect on GFP–NFHasp (Fig. 3 C). Thus, mutation of Cdk5/p35 sites in NFH to preclude or mimic permanent phosphorylation modulates NFH transport, and inhibition of Cdk5/p35 activity with roscovitine significantly accelerates NFH transport. For all of the above assays, three independent experiments were performed using different preparations of plasmid DNAs.

Although neurofilaments move at an overall slow rate of transport, this is now known to be due to rapid movement that is interrupted by prolonged pauses. This has been revealed by analyses of GFP–NFM and GFP–NFH movement in transfected living superior cervical ganglion (scg) neurons (Roy et al., 2000; Wang et al., 2000). Scg neurons possess gaps in their axonal neurofilament network through which moving GFP–NFM/H can be visualized several days after transfection. We found that similar gaps in GFP–NFH fluorescence also appeared in the transfected cortical neurons 48 h after transfection (Fig. 4, A and B). These weaker areas of fluorescence enabled us to monitor neurofilament movement in a manner similar to that described for scg neurons.

To characterize GFP–NFHwt movement in cortical neurons, we acquired time-lapse movies from 20 transfected cells with the duration of observation ranging from 1.08 to 3.75 min. Moving fluorescent structures were defined as those that exhibited movement ≥0.3 μm in 5 s; pauses were thus any movement less than this. The length of GFP–NFHwt moving structures varied from 1.3 to 4.4 μm, and movement was bidirectional. The velocity of GFP–NFHwt movement, excluding any pauses, varied between 0.08 and 0.96 μm/s in the anterograde direction and between 0.08 and 0.93 μm/s in the retrograde direction, but pauses in movement were observed. These speeds and pausing characteristics are similar to those described for GFP–NFM and GFP–NFH in scg neurons (Roy et al., 2000; Wang et al., 2000). Examples of GFP–NFHwt movement are shown in Fig. 4, A–C (see Video 1).

NFH side arm phosphorylation is dynamic, but the phosphorylation states of GFP–NFHala and GFP–NFHasp on the seven sites investigated are defined irrespective of their subcellular location. We therefore compared the movement characteristics of these two mutants.

The lengths of GFP–NFHala and GFP–NFHasp moving fluorescent structures were not noticeably different from those of GFP–NFHwt, and their velocities, excluding pauses, were similar to GFP–NFHwt. The two mutants also exhibited bidirectional movement, but analyses of this revealed no significant differences. Because images of moving GFP–NFH and GFP–NFM reveal minipauses (Roy et al., 2000; Wang et al., 2000), we next considered whether GFP–NFHala and GFP–NFHasp displayed differences in pausing. Moving GFP–NFHala and GFP–NFHasp spent an average of 16 and 37% pausing, respectively, and statistical analyses revealed a significant difference. In a separate experiment, we also measured the total distances moved by GFP–NFHala and GFP–NFHasp per minute of microscope observation time. As the majority of neurofilaments are stationary at any one time, many movies do not show any movement. In this assay, GFP–NFHala moved significantly more than GFP–NFHasp. These results are summarized in Table I. Examples of GFP–NFHala and GFP–NFHasp movement and pausing are shown in Fig. 4 C.

The role of NFH side arm phosphorylation is not properly understood, but a body of evidence suggests that it is a mechanism for regulating neurofilament transport (de Waegh et al., 1992; Ackerley et al., 2000; Sanchez et al., 2000; Yabe et al., 2001). Here, we have used mutants of NFH to investigate the role of phosphorylation in transport. Using a bulk transport assay, we demonstrate that mutants mimicking permanent phosphorylation are transported slower, and that mutants that cannot be phosphorylated move faster than wild-type NFH. In these assays, inhibition of Cdk5/p35 activity with roscovitine also accelerates neurofilament transport. In a second series of experiments, we analyzed neurofilament movement in living neurons and demonstrate that mutation to mimic permanent phosphorylation significantly increases the proportion of time that neurofilaments spend pausing. Together, these data provide strong experimental evidence that NFH side arm phosphorylation is a regulator of neurofilament transport.

Very recently, Rao et al. (2002) created mice in which the side arm domain of NFH had been deleted by gene replacement. These animals show no alteration in neurofilament transport, which suggests that the side arm is not involved in this process. However, other transgenic studies have supported a role for NFH in neurofilament transport; ablation of the NFH gene accelerates neurofilament transport, and increased NFH expression inhibits transport (Collard et al., 1995; Marszalek et al., 1996; Zhu et al., 1998). Also, NFM and NFH phosphorylation states are intimately related; NFM phosphorylation is increased (including the RT97 epitope) in both NFH knockout and NFH side arm–deleted mice (Tu et al., 1995; Zhu et al., 1998; Sanchez et al., 2000; Rao et al., 2002). Thus, there are compensatory changes in NFM after modulation of NFH expression in mice, and these changes may well effect neurofilament transport.

The precise mechanisms by which phosphorylation inhibits neurofilament transport are not clear. One possibility is that phosphorylation leads to a detachment of neurofilaments from their motors. Alternatively, phosphorylation may influence neurofilament interactions with other axoplasmic components that somehow alter movement. The identification of neurofilament motors will enable a better understanding of this process.

Expression plasmids for rat NFL, NFM, NFH, Cdk5, and p35 were previously described (Guidato et al., 1996; Ackerley et al., 2000). Amino-terminal GFP-tagged rat NFH was prepared by cloning the cDNA into pEGFP-C3 (CLONTECH Laboratories, Inc.).

Serines within KSP repeats 44–48, 50, and 51 of NFH side arm were mutated to either alanine or aspartate using a Chameleon mutagenesis kit (Stratagene) in three consecutive rounds. Serines in repeats 44 and 45 were mutated using oligonucleotides ATCTCTTCCTTCACAGGGGCCTTGGCCTTCTCAGGGGCCTTGGCCTCAGCTAGGGATTTTGCAC (ala) or ATCTCTTCCTTCACAGGGTCCTTGGCCTTCTCAGGGTCCTTGGCCTCAGCTAGGGATTTTGCAC (asp). Serines in repeats 50 and 51 were mutated using oligonucleotides CTGGTCTCTTCCTTCTCGGGGGCCTTGGCCTCCTCCTTGGCAGGAGCTTTGACCTGCTCAGGGGATCTGATGT (ala) and CTGGTCTCTTCCTTCTCGGGGTCCTTGGCCTCCTCCTTGGCAGGATCTTTGACCTGCTCAGGGGATCTGATGT (asp). Serines in repeats 46, 47, and 48 were mutated using oligonucleotides ATCCAGAGTCTTGGCCTTCTCAGGAGCCTTGGCCTCCTCCTTCATGGGGGCCTTGGCCT-TCTCGGGGGCTTTCACCTCAGCTGGAGGCTTGAT (ala) and ATCCAGAGTCTTGGCCTTCTCAGGGTCCTTGGCCTCCTCCTTCATGGGGTCC-TTGGCCTTCTCGGGGTCTTTCACCTCAGCTGGAGGCTTGAT (asp).

SW13− cells, COS cells, and cortical neurons were grown and transfected as previously described (Guidato et al., 1996; Ackerley et al., 2000). Roscovitine was obtained from Calbiochem and prepared as a 20 mM solution in DMSO.

COS cells were processed for SDS-PAGE and immunoblotting, and SW13− cells and neurons were fixed and immunostained as previously described (Ackerley et al., 2000). The primary antibodies used were NA1211 (Affiniti), NR4 (Sigma-Aldrich), 8D8, and RT97.

Bulk axonal transport of GFP–NFM and GFP–NFH was analyzed in cortical neurons using a previously published method (Ackerley et al., 2000; Brownlees et al., 2002; Utton et al., 2002). Statistical analyses were performed using One-way ANOVA tests, and the rates of transport were calculated using linear regression analyses.

To study GFP–NFH movement in living neurons, cortical neurons were transfected as for the bulk transport assays and observed 48 h later using a Carl Zeiss MicroImaging, Inc. Axiovert S100 microscope and a 100× Fluar lens. Cells grown on coverslips were viewed in normal medium using an Open Perfusion Micro-Incubator equipped with bipolar temperature controller (models PDM1–2 and TC-202; Medical Systems Corp.) that was attached to the microscope stage. Cells were gassed with 5% CO₂. Images were captured from 1-s exposures taken every 5 s using a CCD camera (model RTE/CCD-1300-Y/HS; Princeton Instruments). The shutter was controlled using a motorized shutter controller (Lambda 10-2; Sutter Instruments). Filter sets for EGFP were from Chroma Technology Corp. Images were analyzed using Metamorph Software, and Quicktime movies were prepared using Adobe Premiere.

Only cells that exhibited a healthy appearance were monitored, and we did not acquire data from cells expressing high levels of GFP–NFH (as judged by brightness of fluorescence) because this can induce pathological changes (Roy et al., 2000). Statistical analyses of GFP–NFH movement in living transfected neurons were performed using chi-square, One-way ANOVA, and Mann-Whitney tests as indicated in the text.

The supplemental material (Video 1) is available. Video 1 is a Quicktime movie of GFP–NFHwt anterograde movement in a transfected cortical neuron. 1-s exposures were taken every 5 s.

We thank Ron Liem and Li-Huei Tsai for plasmids.

This work was supported by grants from the Medical Research Council, Wellcome Trust, UK Motor Neurone Disease Association, and Kings Healthcare Trust.