Neurofilament cross-bridging competes with kinesin-dependent association of neurofilaments with microtubules

Neurofilaments (NFs) are among the most abundant constituents of the axonal cytoskeleton. NFs consist of subunits, termed NF-H, NF-M, NF-L (corresponding to heavy, medium and light, respectively, in reference to their molecular mass) and α-internexin (Yuan et al., 2006b). The C-terminal regions (`sidearms') of NF-H and NF-M contain multiple phosphorylation sites (Julien and Mushynski, 1998), and protrude laterally from the filament when phosphorylated (Sihag et al., 2007). Phosphorylation regulates the interaction of NFs with each other and with other cytoskeletal structures, and, in doing so, mediates the formation of a cytoskeletal lattice that supports the mature axon (Nixon, 1998; Pant and Veeranna, 1995; Sihag et al., 2007). A considerable body of evidence supports the notion that sidearm phosphorylation regulates NF axonal transport, although the extent of this regulation remains controversial (Barry et al., 2007; Rao et al., 2003; Rao et al., 2002; Shea and Chan, 2008; Shea et al., 2003; Sihag et al., 2007; Yuan et al., 2006a).

NF-H sidearm phosphorylation is regionally and/or temporally associated with retardation of NF transport (Archer et al., 1994; Hoffman et al., 1983; Lewis and Nixon, 1988; Nixon and Logvinenko, 1986). NF-H subunits can be phosphorylated to the extent that they migrate on SDS gels to the 200-kDa band and transport 50% slower than do hypophosphorylated (160 kDa) NF-H subunits (Jung et al., 2000a; Jung et al., 2000b). Moreover, the subset of 200-kDa NFs displaying a developmentally delayed phospho-epitope (RT97) transports 50% slower than total 200-kDa NF-H (Jung et al., 2000a; Jung et al., 2000b). Phosphatase inhibition in situ increases phospho-immunoreactivity and decreases NF transport by 44%, without decreasing transport of other cargo (Jung and Shea, 1999). NF-H overexpression slows NF axonal transport in situ (Collard et al., 1995; Marszalek et al., 1996). Finally, mutation of NF-H sidearm consensus sites for Cdk5 (which phosphorylates NFs and inhibits transport) to alanine increases NF transport, whereas mutation to aspartic acid slows NF transport (Ackerley et al., 2003).

One mechanism by which sidearm phosphorylation could impact transport is by dissociation of NFs from their anterograde motor, kinesin (He et al., 2005; Motil et al., 2006; Prahlad et al., 2000; Shah et al., 2000; Shea and Yabe, 2000; Theiss et al., 2005; Xia et al., 2003; Yabe et al., 1999). Hypophosphorylated NFs preferentially associate with kinesin, whereas sidearm phosphorylation progressively restricts NFs from association with kinesin, to the point where they do not associate with kinesin at all but instead demonstrate selective affinity for dynein (Motil et al., 2006). The ability of kinesin and dynein (motors that mediate `fast' axonal transport) to translocate NFs (which undergo `slow' transport) has been validated by the demonstration that NFs undergo a series of rapid bursts at a fast rate, interspersed with prolonged pauses, which averages out slow transport (Roy et al., 2000; Wang et al., 2000).

A second mechanism by which sidearm phosphorylation could impact transport is by promotion of NF-NF associations that compete with transport. Sidearm phosphorylation promotes the formation of NF `bundles' that are centrally situated along axons (Yabe et al., 2001a; Yabe et al., 2001b). So-called `bundled' NFs undergo transport ≥200% slower than surrounding `individual' NFs (Yabe et al., 2001a). Bundles are selectively comprised of phospho-NFs; phosphatase inhibition increases bundle size within axonal neurites (Yabe et al., 2001a). The above-described pseudo-phosphorylated mutants are concentrated within bundles, whereas the pseudo-nonphosphorylated forms are excluded (Chan et al., 2005). Notably, this is the opposite of the affinity of these mutant forms with kinesin (Shea and Chan, 2008), consistent with the reciprocal influence of NF phosphorylation on transport and NF-NF associations. Phospho-mediated NF-NF associations are likely to generate a `macro-structure' that precludes effective transport (Shea and Flanagan, 2001; Shea and Yabe, 2000).

Either (or both) of these mechanisms would result in relatively slower axonal transport of phosphorylated NFs. Discerning between the two has been difficult in situ or cultured cells, because manipulation of NF phosphorylation state impacts both NF-NF associations and NF transport. We set out herein to determine which of these phenomena represented the major mechanism by which C-terminal phosphorylation modulates NF axonal transport by a series of cell-free analyses.

NFs in axons in situ (Leterrier et al., 1996; Lewis and Nixon, 1988) and axonal neurites of NB2a/d1 cells (Yabe et al., 2001a; Yabe et al., 2001b) consist of a bundle of highly phosphorylated NFs exhibiting a high degree of NF-NF associations, with a surrounding population of `individual' less and/or nonphosphorylated NFs that do not exhibit NF-NF associations (Fig. 1A). Sedimentation of standard cytoskeletal preparations derived from murine sciatic nerve, murine spinal cord and mouse NB2a/d1 neuroblastoma cell sources over a 1 M sucrose cushion as described (Yabe et al., 2001a; Yabe et al., 2001b) generated fractions enriched in bundled or individual NFs; the sucrose interface was enriched in individual NFs, whereas bundled NFs sedimented through the sucrose cushion (Fig. 1B-D,F). Bundles from these three sources were four- to ninefold longer than individual NFs (Table 1). Immunoblot analyses demonstrated that the resulting bundled and individual NF preparations were rich in phospho- and nonphospho-epitopes, respectively (Fig. 1E). Prior to incubation with Triton X-100, cytoskeletal preparations contained microtubules (MTs) aligned with individual NFs and surrounding NF bundles (Fig. 1G); these microtubules were lost during incubation with Triton X-100 and subsequent centrifugation at 4°C (Yabe et al., 2001a).

Rhodamine-tagged MTs (Rhod-MTs) were mixed with the heavy chain of kinesin, aliquoted into observation chambers, and rinsed free of excess kinesin and unattached MTs (Fig. 2A,B). As noted in the Materials and Methods, we used sparse MT preparations consisting of relatively short MTs (e.g. compare Fig. 2A,B) to minimize the likelihood of artifactual entrapment of NFs within a MT array. GFP-NF preparations were then aliquoted into chambers, and the number of bundled and individual NFs was quantified prior to and after rinsing. Both bundled and individual NFs were observed prior to rinsing, but only individual NFs were retained following rinsing (Fig. 2C-I). The number of NFs per 100× microscopic field was reduced by approximately 60% following rinsing (n=17 fields). However, a 78% increase was observed in NFs adjacent to MTs following rinsing (Fig. 2I), indicating selective adhesion of those NFs adjacent to MTs.

We considered that the relative size of bundles might interfere with their binding to MTs in this assay. To examine whether indeed this was the case, aliquots of material rinsed from the chambers were examined, which revealed that bundles were coated with MTs (Fig. 2J); the through-focus series of images indicated that MTs surrounded the bundle and were washed away with it. This was similar to the observation of endogenous MTs surrounding the bundle in cytoskeletal preparations prior to Triton X-100 extraction (Fig. 1G). This observation suggested that MTs were associated with those bundled NFs to which they had access (i.e. the outer layer). This phenomenon might be analogous to bundles being too large to undergo MT-mediated axonal transport in situ (Yabe et al., 2001a; Yabe et al., 2001b).

The mixing of NFs with purified murine brain kinesin (Yabe et al., 2000) increased the percentage of NFs adjacent to MTs from 40 to 55%, and nearly doubled the average length of those NFs adjacent to MTs. Conversely, mixing NF preparations with an anti-kinesin antibody known to inhibit MT-based NF transport (Yabe et al., 1999) reduced both the number and length of NFs adjacent to MTs (Fig. 2K). Although addition of ATP has been shown to foster motor-driven translocation of NFs along MTs in this assay system, it was not required for initial kinesin-mediated association of NFs with MTs (Shah et al., 2000), nor is it required for initial association of kinesin to MTs (Hackney, 1994; Vugmeyster et al., 1998). Consistent with these prior studies, aliquoting of ATP or its nonhydrolyzable analogue AMP-PNP in the observation chamber did not significantly increase the percentage of NFs adjacent to MTs: 55±13% NFs were adjacent to MTs in the presence of kinesin without ATP or AMP-PNP, 66±13% were adjacent following aliquoting of ATP and 65±19% were adjacent following aliquoting of AMP-PNP. Although NFs can bind directly to MTs in the apparent absence of an intervening motor protein (Hisanaga et al., 1993), the reduction of NF association with MTs by anti-kinesin antibodies confirms that kinesin within NF preparations (Fig. 1) contributes to the association of NFs with MTs in this cell-free assay.

Isolated bundles remain intact for >7 days and retain their phospho-epitopes; by contrast, during 4-7 days incubation with 1 mM EGTA, bundles progressively increased in diameter and ultimately completely dissociated into individual NFs of 6.7±2.5 μm in length (mean ± s.e.m.; n=5). During this time, bundles underwent visible, progressive depletion of NFs (Fig. 3A,C). Approximately 15±5% of the total NF profiles within the bundled fraction appeared as individual NFs (e.g. Fig. 1). Incubation with EGTA as above increased the percentage of total NFs appearing as individual NFs to 97±6%. By contrast, incubation with 1 mM EDTA (to chelate Mg²⁺ within the buffer; see Materials and Methods) instead of an additional 1 mM EGTA increased the percentage of individual NF profiles to 63±34%, whereas incubation with 1 mM EDTA along with 1 mM EGTA increased this percentage to 90±17% (i.e. the level achieved with EGTA alone). Bundle dissociation was probably due to depletion of Ca²⁺-mediated `cross-bridging' of phosphorylated C-terminal sidearms (Gou et al., 1998), because EGTA has a selective affinity for Ca²⁺, whereas EDTA chelates both Ca²⁺ and Mg²⁺ (Sanui and Pace, 2005). These data do not discount a role for magnesium ions in the formation NF-NF associations leading to bundling (Leterrier and Eyer, 1987), but indicate an essential role of Ca²⁺ in this process.

Incubation of bundles with 0.5 U/ml alkaline phosphatase (without additional EGTA) depleted NF phospho-epitopes and fostered the dissociation of bundles, confirming a role for NF phosphorylation itself in bundle formation (Fig. 3). NFs that were dissociated from bundles by alkaline-phosphatase treatment were of a length of 5.6±3.5 μm (mean ± s.e.m.; n=38); this length was identical (P<0.79) to the length of individual NFs isolated above the sucrose cushion (Table 1) as well as those dissociated by EGTA (above). Notably, the length of bundled and individual NFs as observed in initial cytoskeletal homogenates was not altered following sedimentation over sucrose. Bundle width increased, bundle length decreased and depletion of NF content was readily apparent during incubation with alkaline phosphatase or EGTA (Fig. 3A,C). Notably, individual NFs that dissociated from bundles were the same length as NFs originally isolated as individual NFs or observed within cytoskeletons prior to sedimentation over sucrose (Fig. 3C). Whereas incubation with alkaline phosphatase depleted NF phospho-epitopes and increased nonphospho-epitopes, NFs derived from bundles that had been dissociated via EGTA retained their phospho-epitopes (Fig. 3B).

We next aliquoted an excess of Ca²⁺ (2 mM) to spinal cord NFs that had been dissociated from bundles following the addition of 1 mM EGTA. We observed a significant increase in longitudinally oriented NF doublets and triplets (Fig. 3D). This was not observed following Ca²⁺ addition to NFs originally isolated as individual NFs.

We next examined whether or not spinal cord NFs that were dissociated from bundles would align with MTs in our cell-free assay. To accomplish this, individual NFs, previously bundled NFs dissociated by incubation with alkaline phosphatase and previously bundled NFs dissociated by incubation with EGTA were mixed with bovine brain kinesin and dispensed into observation chambers containing MTs. We observed that, despite differences in C-terminal phosphorylation, statistically identical percentages of each of these NF populations aligned with MTs: 63±15% (mean ± s.e.m.) of NFs dissociated via EGTA and 62±12% of NFs dissociated via alkaline phosphatase associated with MTs; these values were statistically indistinguishable (P<0.21) from the 46±13% of individual NFs that associated with MTs (Fig. 4).

The C-terminal NF kinase Cdk5 increases the formation of RT97-immunoreactive NF bundles within intact cells (Shea et al., 2004). We therefore examined whether or not manipulation of Cdk5 activity would impact NF-NF and/or NF-MT association in cell-free analyses. Incubation of NFs under cell-free conditions with Cdk5 and its activator p35 increased RT97 immunoreactivity (Fig. 5A; compare with Fig. 1E). NFs incubated with Cdk5+p35 displayed an approximate 33% increase (mean ± s.e.m.; P<0.05) in NF-NF associations and a corresponding 36% reduction (P<0.05) in alignment with MTs. In addition, we treated NB2a/d1 cells expressing GFP-tagged NF-M with roscovitine, which reduces NF C-terminal phosphorylation and increases transport into and along axonal neurites (Moran et al., 2005; Shea et al., 2004); GFP-tagged NFs isolated from roscovitine-treated cultures displayed an increase in lower-molecular-weight isoforms and associated with MTs in our cell-free assay approximately 30% more (P<0.05) than did those from cultures not treated with roscovitine (Fig. 5B).

The organization of axonal NFs has been the subject of long-standing controversy. Models that describe the manner in which NFs transport range from the so-called `relentless' translocation of a homogeneous population of NFs (Lasek, 1986; Lasek et al., 1992; Lasek et al., 1993) to translocation of only some NFs while others incorporate into a distinct cytoskeletal macrostructure that is essentially stationary (Nixon, 1993; Nixon, 1998; Nixon and Logvinenko, 1986). These models display considerable inherent similarity; proponents of each agree that NFs transport over a broad range of rates and that these rates are derived at least in part from the length of time that NFs are associated with their transport vector (Lasek et al., 1993; Nixon, 1993; Nixon, 1998). The essential difference between these basic models is whether axonal NFs constitute a single population with a continuum of transport rates, or whether the differences in transport rates actually characterize distinct NF populations. Characterization of the so-called stationary cytoskeleton encompasses the concept that it consists of cross-linked, extensively phosphorylated NFs that as a whole do not exhibit net transport, whereas individual NFs may enter and depart from this macrostructure from the so-called `moving wave' of less phosphorylated NFs (Lewis and Nixon, 1988; Nixon, 1998; Nixon and Logvinenko, 1986). Studies using GFP-tagged NF subunits have provided considerable insight into NF dynamics, including demonstrating that NFs undergo bursts of rapid transport interspersed with protracted pauses (Roy et al., 2000; Wang and Brown, 2001), and, furthermore, that NFs could undergo kinesin-mediated anterograde transport (Yabe et al., 1999). In these studies, the vast majority of GFP-NFs did not undergo translocation during the observation period. Taken together, these findings led us to hypothesize that the stationary cytoskeleton consists of NFs that have dissociated from their transport complex owing to one or more crucial phosphorylation events that promote NF-NF associations, leading to bundling, and, as a consequence, these NFs are sterically hindered from undergoing continued transport owing to their inability to re-associate with motors (Shea and Flanagan, 2001; Shea and Yabe, 2000). The findings of the present study are consistent with these prior hypotheses. An underlying assumption in all of the above models was that stationary NFs had the inherent ability to undergo transport. More recently, Trivedi et al. provided what could be considered the `missing link' for the above hypotheses by demonstrating that non-moving NFs could resume rapid transport (Trivedi et al., 2007). Our data demonstrating the ability of previously bundled NFs to undergo kinesin-mediated MT association at a level indistinguishable from that of individually isolated NFs extends these findings of Trivedi et al. by suggesting that all populations of NFs within the stationary cytoskeleton are capable of resuming kinesin-mediated transport.

Trivedi and colleagues hypothesized the existence of three NF `populations': those on a MT motor, those off the motor but able to re-associate with it, and a third population that was in some manner restricted from associating with motors. The findings presented herein are consistent with this model, and suggest that: (1) NF phosphorylation is in fact the mechanism that gives rise to this `restricted' NF population, and (2) that the `restricted' population is comprised of NF bundles. They further suggest that the surrounding individual NFs observed within axons represent the population that toggles on or off the motor, pending establishment of competing NF-NF associations (Shea and Yabe, 2000). Our cell-free analyses provide several lines of evidence supporting the notion that Ca²⁺-mediated, phospho-dependent NF-NF associations compete with kinesin-mediated NF-MT associations.

NFs that were dissociated from bundles by incubation with alkaline phosphatase (and were therefore depleted of phospho-epitopes), NFs dissociated by incubation with EGTA (which retained their phospho-epitopes) and NFs originally isolated as individual NFs all underwent kinesin-mediated MT association to an equivalent extent, suggesting further that kinesin-mediated association of NFs with MTs might be independent of NF C-terminal phosphorylation, and, therefore, that NF phosphorylation competes with NF axonal transport in an indirect manner – i.e. by promoting NF bundling and fostering the formation of a macro-structure too large to undergo transport. Our findings support a corollary of this hypothesis that the entire axonal NF content is capable of associating with MTs pending dissociation from the stationary cytoskeleton. This latter speculation remains provisional, because there might be a cascade of NF C-terminal phosphorylation events, the first of which might decrease NF-kinesin affinity, potentially followed by others that might promote initial NF-NF associations, and perhaps yet others that foster and/or maintain large-scale NF bundling; this possibility is underscored by the complex array of C-terminal phosphorylation sites as well as the existence of multiple NF C-terminal kinases (Shea and Chan, 2008; Sihag et al., 2007; Veeranna et al., 2008). We previously reported that cell-free phosphorylation of NFs to the extent that generated RT97 fostered their dissociation from kinesin (Chan et al., 2004); despite the above caveat, our findings herein prompt us to amend this prior conclusion to encompass that this extent of phosphorylation might have indirectly prevented this by inducing NF-NF associations (which were not assayed in this prior study) rather than directly inhibiting NF-kinesin association.

Bundles from the three separate sources used herein (sciatic nerve, spinal cord and NB2a/d1 cells) were several-fold longer than corresponding individual NFs. Following dephosphorylation or Ca²⁺ chelation, however, they dissociated into NFs of lengths identical to those originally isolated as individual NFs. Although our findings do not rule out the possibility that bundles contain some longer NFs, perhaps formed via subunit addition or end-to-end annealing, these findings suggest that bundles are formed by staggered association of shorter individual NFs. Our findings collectively suggest a mechanism by which a stable, elongated macro-structure can be assembled from shorter individual `building blocks'. In further support of this notion, individual and dissociated previously bundled NFs were statistically identical in length to NFs observed undergoing transport within cultured neurons [4.1 μm as reported by Wang and Brown (Wang and Brown, 2001); 9.8±8.9 μm as reported by Roy et al. (Roy et al., 2000)]. Whether or not individual and previously bundled NFs correspond to a `unit length' in situ remains to be determined, but we note that identical length of individual NFs and NFs dissociated from bundles affords the possibility for rapid replacement of a catabolized or dissociated NF with an individual, transporting NF.

The findings of the present study confirm that NF-NF associations compete with NF-MT association, and provide a mechanism by which C-terminal NF phosphorylation might indirectly contribute to the observed slowing in axonal transport of phospho-NFs. By demonstrating that disruption of previously bundled NFs allows them to undergo NF-MT association provides evidence that the so-called `moving' and `stationary' cytoskeletons apparently represent a continuum. A complete understanding of the interactions among NFs leading to the establishment and maintenance of the resident population referred to as the stationary cytoskeleton, including the potential role of site-specific phosphorylation by sequential kinase activities, remains to be determined.

Cytoskeletons were prepared from sciatic nerve of adult FVB mice of both genders constitutively expressing green fluorescent protein (GFP)-tagged NF-H (Letournel et al., 2006), NB2a/d1 cells constitutively expressing GFP-tagged NF-M (Yabe et al., 2003; Yabe et al., 2001a) or murine spinal cord (which were labelled with rhodamine after harvesting as described below). Cytoskeletons were prepared by sedimentation of homogenates of the above tissues and cells at 15,000 g for 15 minutes in 50 mM Tris (pH 6.8) containing 5 mM EDTA, 1 mM phenylmethanesulphonylfluoride (PMSF) and 50 μg/ml leupeptin as described (Yabe et al., 2001a). To enrich for NFs, cytoskeletal preparations were resuspended in 1% Triton X-100 in 50 mM Tris (pH 6.8) containing 5 mM EDTA, 1 mM PMSF and 50 μg/ml leupeptin as described (Yabe et al., 2001a) then centrifuged again at 15,000 g for 15 minutes. NFs for experiments carried out herein were isolated and processed on at least two separate occasions from separate mice and NB2a cultures. Additional NB2a/d1 cultures were treated with 100 μM roscovitine for the final 2 hours prior to harvest (Moran et al., 2005; Shea et al., 2004).

NFs from sciatic nerve and NB2a/d1 cells were already labelled with GFP. Spinal-cord NFs (isolated from mice not expressing GFP-tagged NFs) were labelled with rhodamine as described (Leterrier et al., 1996; Wagner et al., 2003). Briefly, spinal-cord NFs from adult C57BL6 mice of both genders were resuspended in 0.1 M 2-(N-morpholino)ethanesulfonic acid (MES; pH 6.8) containing 1 mM MgCl₂, 1 mM EGTA, 1 mM PMSF and 50 μg/ml leupeptin. For simplicity of writing, GFP-tagged and rhodamine-tagged NFs are referred to as `GFP-NFs' and `Rhod-NFs', respectively.

To obtain fractions enriched in bundled or individual NFs, NFs that had been enriched from cytoskeletons by sedimentation in 1% Triton X-100 (above) were resuspended in 50 mM Tris (pH 6.8) containing 5 mM EDTA, 1 mM PMSF and 50 μg/ml leupeptin, layered over this the same buffer containing 1 M sucrose and centrifuged again at 15,000 g for 15 minutes; a fraction enriched in individual NFs was recovered at the sucrose interface, while a fraction enriched in bundled NFs sedimented through the sucrose cushion (Yabe et al., 2001a). We refer herein to NFs not contained within central bundles and isolated at the sucrose interface as `individual' in accordance with our previous studies (e.g. Yabe et al., 2001a); it should be noted, however, that we cannot readily distinguish single NFs from, for example, two or three closely aligned NFs, owing the limit of resolution of fluorescence microscopy (e.g. Wang and Brown, 2001). Although electron microscopic analyses confirm the presence of individual NFs, use of the term `individual' NFs with reference to fluorescent analyses is for simplicity of writing only. Aliquots of material recovered from the pellet and interface following sedimentation over sucrose were subjected to immunoblot analyses using antibodies directed against phospho- (RT97, SMI-31) and nonphospho- (SMI-32) epitopes, a polyclonal anti-tubulin antibody and a pan-kinesin antibody previously used to visualize kinesin in association with NFs (Yabe et al., 2000; Yabe et al., 1999). Immunoreactive species were densitometrically quantified in digitized images using NIH Image as described previously; the background signal from an adjacent, identically sized area in the identical lane was subtracted from each reactive species to be quantified (Yabe et al., 1999; Yabe et al., 2001a). All samples to be compared densitometrically were electrophoresed on the same gel, transferred to nitrocellulose and visualized simultaneously. RT97 was a generous gift of Brian Anderton (Inst Psych, UK); SMI antibodies were obtained from Sternberger Monoclonals (Baltimore, MD). Aliquots of various NF preparations were viewed by fluorescent microscopy as described (Yabe et al., 2001a). NFs from all three sources exhibited identical behaviour in all cell-free analyses used herein; in most cases, only one representative example is shown. All animal procedures were approved by our Institutional Animal Care and Use Committee.

In some experiments, fractions enriched for bundled NFs were incubated with 0.5 U/ml Escherichia coli alkaline phosphatase for 4 hours at 37°C (Shea et al., 1997), or 1 mM EGTA for 2 hours, each of which dissociated bundles into individual NFs; NFs dissociated in this manner were referred to as `previously bundled NFs'. Aliquots of previously bundled NFs or individual NFs were incubated with 2 mM Ca²⁺. Additional aliquots of NF preparations were incubated for 2 hours with 0.1 μg/μl purified Cdk5 and its activator p35 (1225 U/mg activity; Upstate Biochemicals, Lake Placid, NY) with excess ATP and a cocktail of protease inhibitors as described previously (Shea et al., 2004). Aliquots of various NF preparations were viewed by fluorescent microscopy. All chemicals were obtained from Sigma-Aldrich (St Louis, MO) unless otherwise indicated.

Rhodamine-conjugated or fluorescein-conjugated tubulin was polymerized into taxol-stabilized MTs using a kit from Cytoskeleton (Denver, CO) according to the manufacturer's instructions as described (Shah et al., 2000). Resulting `Rhod-MTs' or `Fluor-MTs', respectively, were mixed with the heavy chain of kinesin and dispensed (total volume 26 μl) into `observation chambers' consisting of a coverslip adhered to a slide via two pieces of double-stick tape along the edges; the added kinesin allows MTs to adhere to the chamber. After 5 minutes, chambers were then rinsed with 15 μl buffer to deplete non-adherent MTs and any excess kinesin. Bundled, individual or previously bundled GFP-NFs or Rhod-NFs (6 μl total volume) were aliquoted into the chamber, the chamber was then rinsed with 15 μl buffer to remove non-adherent NFs and NFs were monitored at 100× under fluorescein and rhodamine optics (Shah et al., 2000; Yabe et al., 1999). In some experiments, 100 μM ATP or the nonhydrolyzable ATP analogue AMP-PNP at the same concentration were aliquoted into the chambers (Shah et al., 2000; Yabe et al., 1999). In all cases, multiple images encompassing areas of the observation chamber containing NFs and MTs were captured rapidly to avoid photobleaching at 100× under identical conditions using a Zeiss Axiovert microscope operated by OpenLab software (Improvision, Waltham, MA). Calculations were then carried out on stored images, which were processed only for brightness and contrast to ensure visualization of all MTs and NFs, using OpenLab software. Images of Rhod-MTs (captured under rhodamine optics) and GFP-NFs (captured under fluorescein optics), or Fluor-MTs (captured under fluorescein optics) and Rhod-MTs (captured under rhodamine optics), were merged via OpenLab software. The percentage of NFs aligned with MTs was then calculated versus the total number of NFs remaining with the chamber after rinsing by examination of at least ten images obtained from each chamber from at least two independent experiments. The number and length of MTs formed under these conditions varied in accordance with incubation time; we routinely incubated tubulin for 2-3 hours as described by the manufacturer, which yielded relatively sparse preparations of short MTs (see Results). This was carried out to minimize the possibility that NFs could become non-specifically entangled within the resultant MT preparation.

NFs associate with MTs and undergo kinesin-mediated transport both within cells and in this cell-free assay (Shah et al., 2000; Yabe et al., 1999). Rather than quantify translocation along MTs, we instead quantified the extent of association with MTs. Translocation of all but extremely short NFs is likely to require the coordinated activity of multiple kinesin molecules. More importantly, some NFs might undergo axonal transport by kinesin-mediated association with MTs that themselves actively translocate against other MTs and/or the actin cortex (Hasaka et al., 2004; Jung et al., 2004; Myers et al., 2006; Susalka and Pfister, 2000). Finally, kinesin-mediated translocation of NFs can be rapid, and many such events are likely to have transpired prior to image capture. This latter possibility was further accentuated by our use of sparse NF and MT preparations as described above. Quantification of kinesin-mediated association of NFs with MTs is the only method that can encompass all of these possibilities.

Values presented represent the mean ± s.e.m.; statistical comparisons were carried out using Student's t-test. Significant differences among values are indicated on graphs by asterisks.