The regulation of bidirectional mitochondrial transport is coordinated with axonal outgrowth

The functional asymmetry of eucaryotic cells gives rise to different requirements for macromolecules and energy in different regions of the cell. To address these needs, cells have evolved complex and tightly controlled mechanisms to transport and position different organelles in specific regions of the cytoplasm (e.g. see Ho et al., 1989; Hollenbeck and Swanson, 1990; Lee et al., 1989; Corthésy-Theulaz et al., 1992). The elaboration of organelle transport to support cellular asymmetry reaches its zenith in neurons, whose long axonal processes lack synthetic capacity and are supported by the copious bidirectional traffic of organelles between the cell body and terminal region (reviewed by Grafstein and Forman, 1980). Axonal organelle transport seems likely to be driven entirely or in part, by the microtubule-based motor proteins kinesin and cytoplasmic dynein (Vale et al., 1985a; Paschal et al., 1987), which generate movement in opposite directions along the axon’s polarized microtubule array (Burton and Paige, 1981; Heidemann et al., 1981). However, despite extensive studies of the biochemical and biophysical characteristics of the proteins that drive movement, the mechanisms by which transport is regulated to appropriately position specific classes of organelles are essentially unknown. In fact, with one exception (Hollenbeck and Bray, 1987), we lack information about how the movement of any particular class of neuronal organelle is modulated in response to normal physiological changes.

In this study, we have examined the regulation of mitochondrial movement in the axons of peripheral neurons grown in culture. Since ATP is a small molecule and therefore freely diffusible in the cell, the specific location of mitochondria within the cell might in principle seem to be unimportant, and one might wonder why their transport should be regulated at all. However, it has been clear for some time that many cell types, such as epithelia, do position their mitochondria in regions of intense ATP consumption (Fawcett, 1981; Zielinski et al., 1988; Van Blerkom, 1991), indicating that even cells of relatively modest dimensions do not rely entirely on diffusion to ensure adequate delivery of ATP to ion pumps and other ATPases. In the highly elongated axonal processes of growing peripheral neurons, the active growth cone provides a region of intense ATP consumption so distant from the cell body that it could require the positioning of mitochondria nearby via regulated axonal transport. To test this hypothesis, we observed and quantified the distribution and transport behavior of mitochondria in both growing and non-growing axons, under aerobic and anaerobic conditions. Because axonal mitochondria are both easily recognizable and constrained to move in simple linear paths by the small diameter of the axon, they provide an ideal subject for quantitative studies of the regulation of motility. Our data reveal that mitochondrial transport in the distal axon is regulated in concert with the growing state of the axon but independent of mitochondrial ATP-generating capacity.

All tissue culture media and supplements, and cytochalasin E, were obtained from Sigma Chemical Co. (St. Louis, MO). 4-(4-diethylaminostyryl)-N-methylpyridinium (4-di-2-ASP) and rhodamine 123 were obtained from Molecular Probes, Inc. (Eugene, OR). Texas Red-conjugated donkey anti-rabbit IgG was obtained from Vector Labs (Burlingame, CA). Anti-cytochrome c oxidase polyclonal antibody was a gift from A. Miranda. Ruled coverslip masks were prepared by Agar Scientific (Cambridge, UK).

Sympathetic chain ganglia were dissected from 9- to 11-day-old chicken embryos and cultured as whole ganglia or dissociated and grown as single cells on coverslips as previously described (Hollenbeck et al., 1985). Cells were grown for 18-26 h in a humidified 37°C incubator in Liebovitz L-15 medium supplemented with 10% fetal bovine serum, 0.6% glucose, 2 mM L-glutamine, 100 i.u./ml penicillin, 100 μg/ml streptomycin, 50 ng/ml nerve growth factor, and 0.5% methyl cellulose. Coverslips were treated with 1 mg/ml polylysine followed by a laminin-enriched fraction of conditioned medium (prepared as per Lander et al., 1982) for 20-60 min each before plating out cells. Live cells were observed using a Zeiss IM35 or Axiophot microscope, and cultures were maintained at 37°C with an air curtain stage-warmer.

For studies of mitochondrial transport in non-growing axons, elongation was blocked by either a substratum-associated barrier or mild drug treatment as previously described (Hollenbeck and Bray, 1987). Briefly, substratum-associated barriers were produced by shadowing silicon monoxide through a metallic electroformed mask to create 7 μm wide × 50-800 μm long adherent tracks on silane-treated coverslips. After a brief treatment of the coverslips with laminin, sympathetic neurons were plated onto them at low density. During subsequent growth, axons extended only to the ends of the tracks where their elongation was blocked by the surrounding siliconized surface. In other experiments, a low dose of cytochalasin E (1 μg/ml) was administered to cells to halt outgrowth rapidly.

To grow cells under anaerobic conditions, cultures were maintained as usual for 16 h, then the culture medium was withdrawn and replaced with a 1 mm-deep layer of medium containing 10 mM HEPES but lacking methyl cellulose. Culture dishes were sealed in a BBL GasPak Pouch (Becton Dickinson Microbiology Systems, Cockeysville, MD) for 4-24 h. Cells were fixed immediately after removal from the GasPak.

For analyses of mitochondrial distribution and motility, unbranched axons of uniform caliber with distinct growth cones were selected, and measurements were made in the most distal 100 μm of each axon. Morphometry was performed using either BioRad MRC 600 software or an Image-1 Image Analysis System (Universal Imaging Corp., West Chester, PA). In live aerobic cells, mitochondria were stained with either 4-di-2-ASP (20 μM for 10 min) or rhodamine 123 (0.1 μg/ml for 30 min) and their distributions were measured by one of three techniques: (1) positions of individual mitochondria were revealed by epifluorescent illumination and quantified by measuring the distance from the base of the growth cone to the center point of each axonal mitochondrion. (2) To determine the distribution of mitochondrial length in the axons, epifluorescent images of neurons were gathered using a BioRad MRC 600 confocal laser scanner mounted on a Zeiss Axiophot microscope. The confocal apertures were opened wide enough to allow the fluorescence from the entire thickness of the axon to be collected. The distance from the base of the growth cone to the proximal and distal tips of each mitochondrion were measured, and the data were compiled as the total mitochondrial length in each of ten 10 μm bins comprising the distal 100 μm of axon. (3) The distribution of mitochondrial electrochemical potential was determined in rhodamine 123-labeled axons by using the confocal microscope and image analysis software to quantify the fluorescence intensity in consecutive 5 μm segments of axon and in equivalently sized areas of laterally adjacent background for the 100 μm of axon proximal to the growth cone. After subtracting corresponding background values for each segment, adjacent 5 μm segments were paired and summed to produce total mitochondrial fluorescence intensities for ten 10 μm bins. Mitochondria within the growth cones themselves could not be accurately quantified due to the wide variation in growth cone dimensions, which makes measurements of mitochondrial density per unit length of axon meaningless there.

To determine whether mitochondrial distributions departed significantly from uniform, data were analyzed statistically by linear regression with analysis of variance, chi-square and Kolmogorov-Smirnov tests for grouped data. Distributions are refered to as “skewed” if they differed significantly from a uniform distribution by all three tests at a significance level of P < 0.01. Distributions are refered to as “flat”, “even” or “level” if they were not significantly different from a uniform distribution by at least two of these three tests at P < 0.05.

The mitochondrial distribution in anaerobic cells was determined by immunofluorescent staining of cytochrome c oxidase. Cells were removed from an anaerobic GasPak, fixed immediately for 15 min at 37°C, with 2.67% paraformaldehyde, 1.46% lysine, 0.23% sodium periodate, 0.73 mM CaCl₂ and 0.37 mM MgCl₂, in PBS, pH 7.4 (adapted from McLean and Nakane, 1974), permeabilized with 0.1% Triton X-100 in PBS, pH 7.4, and blocked with 3% BSA in PBS, pH 7.4. They were then incubated for 1 h in polyclonal anti-cytochrome c oxidase antibody diluted 1:300, washed, incubated for 1 h in Texas Red-conjugated goat anti-rabbit IgG diluted 1:1000, washed and mounted with 1 mg/ml phenylenediamine in 90% glycerol/10% 10 × PBS (Johnson and Nogueira Araujo, 1981). Labeled cells were viewed on the scanning confocal microscope and the distribution of mitochondrial length was quantified as described in technique (2), above.

The movements of mitochondria were quantified in two discrete regions of growing and non-growing axons: (1) between 0 and 40 μm from the base of the growth cone and (2) between 60 and 100 μm from the base of the growth cone. This separation of regions eliminates possible confusion from behavior occurring on the border between observed zones. Axons were viewed by videoenhanced phase-contrast microscopy, using a Hamamatsu Newvicon camera, an Image-1 image analysis system to provide background subtraction and contrast enhancement, and a Panasonic AG-7300 Super-VHS video cassette recorder to record images. Videotaped movements of individual mitochondria were quantified with a Mark-V video measurement system (M. Walsh Electronics, San Dimas, CA) driven by Measure software (gift of Steven M. Block, Rowland Institute, Cambridge, MA). Based upon comparisons between phase-contrast and rhodamine 123 epifluorescence images, phase-dark organelles which were 0.5 μm across and at least 1 μm long were considered to be mitochondria. To prevent bias in selection of mitochondria for measurement, all mitochondria visible in the regions 0-40 μm and 60-100 μm from the growth cone over a 5 min period were tracked for the entire time they remained in focus. All mitochondrial displacement data for one region were then pooled. To correct for apparent movement of mitochondria due to drift of the stage, noise in the tracking system, etc., a threshold velocity of 0.05 μm/s was selected for two reasons: this was the average velocity of mitochondrial movements identified qualitatively as drift during recording, and was also equal to the value of the distance error in mouse positioning divided by the average duration of one measurement. Velocities below this “drift” velocity were considered to be zero. “Maximum” velocities were calculated from any movements which were sudden, clearly visible and of uniform velocity for their entire duration.

To ascertain whether the growth state of the axon had any effect on the transport and positioning of its mitochondria, we compared the mitochondrial distributions of growing and non-growing axons from the same cell body. Plating chick embryonic sympathetic neurons onto coverslips containing defined adherent tracks (as described in Materials and methods) constrained the pattern of their outgrowth to the production of either a single axon or two axons growing at 180° from each other, from opposite sides of the cell body. When a growth cone arrived at the end of a track, its continued outgrowth was blocked by the non-adhesive siliconized surface of the coverslip; in cases where one axon of a bipolar neuron was halted while the other continued to elongate, the distribution of mitochondria in a growing and a non-growing axon from the same cell could be compared. Mitochondrial 4-di-2-ASP staining in these neurons revealed a striking difference between actively elongating and blocked axons: in elongating axons, the mitochondrial distribution appeared to form a gradient with the highest density near the growth cone, while blocked axons had lost this gradient and also had fewer mitochondria overall (Fig. 1).

To quantify these observations, we used the lipophilic cationic dyes 4-di-2-ASP (Magrassi et al., 1987) and rhodamine 123 (Johnson et al., 1981) to reveal the positions of mitochondria in live cells, and we halted outgrowth with a low dose of cytochalasin E (Cooper, 1987; Hollenbeck and Bray, 1987), so that the time at which growth cone activity ceased could be known with some accuracy for a large population of axons. 4-di-2-ASP reveals the position and dimensions of the mitochondria, while the fluorescence intensity of rhodamine 123 reports the magnitude of mito-chondrial transmembrane potential as well (Johnson et al., 1981), and thus indicates not just the physical distribution of mitochondria, but the distribution of oxidative phosphorylation capacity. When we quantified the rhodamine 123 fluorescence intensity in the 100 μm of axon adjacent to active growth cones, we found a significantly skewed distribution, with a sevenfold higher density of mitochondrial fluorescence near the growth cone than at a distance of 100 μm away (Fig. 2A). When axonal elongation was halted, this distribution gradient disappeared as the mitochondria became redistributed away from the growth cone, assuming a statistically flat, uniform distribution within 1 h (Fig. 2B). Furthermore, the total mitochondrial length in the entire distal 100 μm of the axon, as represented by the area under the distribution histograms, became substantially reduced, indicating that mitochondria were leaving the distal axon entirely. This growth-dependent gradient of mitochondrial distribution, and its dissipation when growth was halted, were also seen when the distribution was quantified with 4-di-2-ASP in terms of physical length of mitochondria per unit length of axon, or by simply plotting the positions of the midpoint of each mitochondrion regardless of its length (data not shown). In addition, similar quantitative results were obtained when outgrowth was halted by culturing cells on patterned substrata as described above (Fig. 1) rather than by drug treatment. All distributions were analyzed statistically for significant deviation from an even distribution as described in Materials and Methods. We saw no evidence of mitochondrial division in the axon.

In order to understand how mitochondrial movements were modulated to form the distribution gradient in growing axons and to disperse it in non-growing axons, we analyzed the motile behavior of individual mitochondria by videoenhanced phase-contrast microscopy. Mitochondria were tracked and their motile behavior was analyzed in two separate regions of the distal axon: “near” the growth cone (0-40 μm from the base of the growth cone) and “far” from the growth cone (60-100 μm from the base of the growth cone). Separating the distal 100 μm of axon into two non-overlapping zones eliminated possible confusion of classifying types of behavior occurring in the border region. The duration and distance of all mitochondrial movements or stationary periods were then measured both near to and far from the growth cone in growing and non-growing axons. To ensure that measurements were taken during the time that the distribution gradient was being dissipated, data were gathered within 20 min of blocking growth.

Although individual mitochondria throughout the distal axon displayed movement in both directions along with periods of quiescence (Fig. 3), calculation of the rates of net directional transport and of the motile duty cycle (percentage of time spent moving) for the entire population revealed both growth-dependent and position-dependent characteristics of mitochondrial movement (Table 1). First, anterograde and retrograde movements combined to produce net anterograde transport in growing axons and net retrograde transport in non-growing axons. Second, the rate of net anterograde movement in growing axons was highest when mitochondria were still far from the growth cone and slowed to near zero in the region near the growth cone. In non-growing axons, however, the rate of net retrograde movement was highest near the growth cone and decreased somewhat when mitochondria were farther away. These modulations of net transport would clearly serve to deliver mitochondria to an active growth cone, and to reverse the process when growth cone advance ceased.

The duty cycle of the mitochondria revealed a striking pattern as well: the majority of the mitochondria’s time in the axons was spent stationary (Table 2). Because of this widely imbalanced duty cycle, we analyzed the moving mitochondria separately from the persistently stationary ones in order to reveal more about the underlying mecha-nisms regulating the net transport. A mitochondrion was classified as “moving” if it showed significant motion during the observation time as judged by one of two crite-ria: (1) achieving a velocity at any time during tracking that met or exceeded the slowest “maximum” velocity measured in that direction; (2) displaying persistent movement that achieved a net velocity over the entire time it was observed of 0.05 μm/s. “Maximum” velocities and the choice of this threshold velocity are explained in Materials and Methods.

When these criteria were applied to all mitochondria in growing and non-growing axons, three patterns emerged (Table 2). First, in growing axons, there was a large regional difference in duty cycle. Only 25% of the mitochondria near the growth cone moved significantly during observation, with the other 75% remaining stationary, while far from the growth cone the inverse was the case, with 75% of the mitochondria moving and only 25% remaining stationary. Second, blocking axonal growth significantly increased the percentage of moving mitochondria. Near the blocked growth cones almost 60% of the mitochondria moved, as opposed to 25% in growing cells, and far from the growth cone there was also a small rise in the percentage that were moving. Third, when axonal growth was blocked, the percentage of mitochondria that moved only anterogradely dropped both near and far from the growth cone, while the percentage moving exclusively in a retrograde direction remained virtually the same as in growing axons.

To address the regulation of mitochondrial movement separately from the balance between persistently moving and stationary states, we segregated and analyzed the data for moving mitochondria alone (Table 3). As with the total mitochondrial population, the moving mitochondria underwent net anterograde transport in growing axons and net retrograde transport in blocked axons. Also, even these mitochondria, which moved significantly at least some time during observation, spent the majority of their time stationary. But the fraction of time that they spent in motion changed in one significant way when axonal outgrowth was blocked: the fraction of time mitochondria spent moving in the anterograde direction dropped dramatically when growth stopped, from 33% to only 6% of their time near the growth cone and from 17% to 9% far from the growth cone. Meanwhile, the fraction of time mitochondria moved in the retrograde direction changed very little when growth stopped either near to or far from the growth cone.

The mitochondria in the moving pool occasionally under-went periods of rapid, persistent movement, which were measured and averaged separately (“maximum velocities”, Table 3). There was some variation in the maximum velocities of mitochondria, but the retrograde velocities were higher than the anterograde velocities. This is consistent with in vitro studies on the putative motor proteins for mito-chondrial transport: kinesin driving anterograde transport (Vale et al., 1985a,b) and dynein driving retrograde transport (Paschal et al., 1987; Schnapp and Reese, 1989).

Finally, we asked the question of whether the regulation of mitochondrial distribution and motility was coupled to their normal function - the production of ATP by oxidative phosphorylation. To do this, we grew neurons under anaerobic conditions for 8-24 h in order to inhibit oxidative phosphorylation, then fixed them and determined the distribution of mitochondria in the distal axons by anti-cytochrome c immunofluorescence (Kuhn-Nentwig and Kadenbach, 1985). In cultures grown anaerobically for 8 h, the mitochondrial distribution in actively growing axons was skewed toward the growth cone (Fig. 4A), in a manner similar to that seen in aerobically grown cultures, with a twofold gradient between the region immediately adjacent to the growth cone and the region 100 μm away. When axonal outgrowth was blocked in these cultures, the mitochondria underwent a retrograde redistribution (Fig. 4B). In cultures maintained under anaerobic conditions for 24 h, growing axons also showed a mitochondrial distribution that was skewed toward the growth cone, but this gradient had a smaller magnitude (Fig. 4C). In addition, the total mitochondrial density in the distal axons of these cells was 60% lower than in aerobically growing, and 50% lower than in 8 h anaerobically growing neurons.

In this study, we have shown that the bidirectional axonal transport of mitochondria is regulated in concert with the growth state of the axon. In axons which were elongating, mitochondria were clustered into a steep gradient near the active growth cone, while in axons whose elongation was blocked, the mitochondria became evenly distributed. Because this behavior was observed whether we measured the distribution of mitochondrial number, mitochondrial length or mitochondrial rhodamine 123 fluorescence, it was clear that it represented regulated movement of mitochondria, and not just changes in their size or membrane potential in response to physiological changes in the axon.

While individual mitochondria moved bidirectionally (Fig. 3), tracking the average transport behavior of a large population of mitochondria confirmed that regional and growth-dependent regulation of their net transport created and dissipated the gradient (Table 1). In growing axons, the mitochondria moved in a net anterograde direction. When growth was blocked, mitochondrial transport rapidly shifted to net retrograde and the mitochondrial distribution gradient was quickly eliminated. Furthermore, the rate of net transport varied with their distance from the growth cone: in growing axons, net anterograde velocity was 4.4 times greater far from the growth cone than it was near the growth cone, while in halted axons the net retrograde velocity was 1.8 times greater near the growth cone than far away. Thus, the mitochondrial distribution gradient is established in growing axons by the relatively rapid approach of mitochondria to the distal region, followed by a reduction of their net anterograde velocity to roughly that of growth cone advance (Hughes, 1953). The dissipation of the gradient occurs by the relatively rapid exit of mitochondria from the growth cone region, followed by a slowing of the net retrograde velocity as mitochondria progress farther from the axon terminus. The result is a reduced density and a uniform distribution of mitochondria along the axon length rather than a complete evacuation of the distal region.

A detailed analysis of individual mitochondrial behavior (Tables 2 and 3) revealed that the net transport described above was generated by regulation at two levels. First, the transition between moving and stationary states was regulated along the length of growing axons to control the number of mitochondria which remained clustered near the growth cone in growing vs non-growing axons. Second, anterograde transport, but not retrograde transport, was selectively regulated between growth states to control the level and direction of net mitochondrial transport. By coordination of these two regulatory mechanisms, the skewed mitochondrial distributions were created and maintained in growing cells and eliminated when growth ceased.

The number of mitochondria which exhibited significant movement varied considerably with distance from the growth cone and with the growth state of the axon. In growing axons, 75% of the mitochondria far from the growth cone were moving, but this fraction dropped by two-thirds as the mitochondria arrived near the growth cone. When growth was blocked, the moving mitochondrial fraction rose all along the axon, more than doubling near the growth cone and also rising far from the growth cone. These data are consistent with a conversion of mitochondria from a moving to a persistently stationary state in the distal region of growing axons, and a partial reversal of that process when axonal growth is halted.

The presence of a stationary pool of mitochondria has been suggested by evidence from both neuronal and non-neuronal cells, indicating apparent anchorage of mitochondria to stable elements in the axon. This includes morphological evidence showing attachment of mitochondria to the cell cortex (Forman et al., 1987; Lin et al., 1990) or the cytoskeleton, particularly the intermediate filaments (Hirokawa, 1982; Stromer and Bendayan, 1990), as well as some biochemical evidence suggesting stable associations with the cytoskeleton (Price and Gomer, 1989; Rendon et al., 1990; Leterrier et al., 1990). Such a stable attachment of the majority of mitochondria to the cytoskeleton upon arrival in the distal region of growing axons would explain both the maintenance of the distribution gradient seen there, and the “advance” of the stationary mitochondria at approximately the rate of growth cone advance (Hughes, 1953; Reinsch et al., 1991). It would also explain the reported codependence of mitochondrial distribution upon both the microtubule and intermediate filament systems (Summerhayes et al., 1983). The large increase in the moving pool of mitochondria when axonal outgrowth was halted could represent mitochondria freed from their anchorage to the “stationary” cytoskeleton, allowing modulation of their transport and subsequent redistribution.

The second level at which mitochondrial motility was controlled was the specific regulation of anterograde movement. This became apparent first when looking at the percentage of the whole mitochondrial population which moved exclusively anterogradely or retrogradely (Table 2). Here, when axonal outgrowth stopped, the percentage of mitochondria moving exclusively anterogradely dropped while the percentage moving exclusively retrogradely remained virtually the same. Separate analysis of the persistently moving mitochondrial population gave a complementary result: when axonal outgrowth ceased, the percentage of time that mitochondria spent moving anterogradely dropped sharply while the percentage of time spent moving retrogradely remained the same or dropped only slightly (Table 3). These differences held both near to and far from the growth cone, and were sufficient to explain the differences in the net direction of mitochondrial transport in growing and non-growing axons. This suggests that formation of the steep mitochondrial gradient in growing axons, and its dissipation in non-growing axons, is achieved almost entirely by regulation of the anterograde motor activity alone. When growth is blocked, the anterograde motor activity is reduced, allowing the steady activity of the retrograde motors to exert more influence. We find no evidence that mitochondrial retrograde motor activity is significantly regulated.

Three lines of evidence presented here argue that mitochondria bear motor activities for both the anterograde and retrograde directions of movement. First, mitochondria change direction frequently and rapidly (Fig. 3), seemingly too quickly for the exchange of one set of motors for another. Second, the duty cycle data in Tables 2 and 3 indicate that anterograde mitochondrial transport is regulated independently of retrograde transport. Third, mitochondria move with different maximum velocities in the anterograde and retrograde directions (Table 3). These maximum velocities seem more likely to be characteristic properties of the motors driving mitochondrial movement than are the slower average velocity values reported here or reported previously (Forman et al., 1987; Takenaka et al., 1990), since the slower average velocities reflect variations in resistance to organelle movement through the axoplasm. Ultrastructural studies (Heggeness et al., 1978; Tsukita and Ishikawa, 1980; Hirokawa, 1982; Schnapp and Reese, 1982; Miller and Lasek, 1985; Van Blerkom, 1991) and in vitro experiments (Brady et al., 1982; Martz et al., 1984) have strongly implicated microtubules as the transport filament for mitochondria, and current thinking favors the action of two distinct classes of motor proteins in the production of organelle movements (reviewed by Schroer and Sheetz, 1990; Vallee and Bloom, 1991): kinesins driving anterograde movement (Vale et al., 1985b; Porter et al., 1987; Saxton et al., 1988; Hirokawa et al., 1991), and dyneins driving retrograde movement (Schnapp and Reese, 1989; Schroer et al., 1989). The fact that the maximum retrograde velocities for mitochondria were higher than anterograde velocities is consistent with kinesin and dynein driving mitochondrial movements (Vale et al., 1985a,b; Paschal et al., 1987). If this is the case, then our data implicate the specific regulation of kinesin in the transition between net anterograde and retrograde mitochondrial transport. However, our data in no way rule out a role for novel classes of microtubule-based motor proteins (Cole et al., 1992; Sawin et al., 1992) or even actin-based motor proteins (Bradley and Satir, 1979; Brady et al., 1984; Kuznetsov et al., 1992) in mitochondrial motility. Although cytochalasin E can disrupt the actin cortex at high concentration, we administered a low dose to block axonal outgrowth by inhibiting actin polymerization in the growth cone, without compromising the actin cortex integrity.

The reversal of net mitochondrial transport from anterograde to retrograde when axonal outgrowth is blocked raises the question of which intracellular signals control that switch. Studies performed on regenerating or ligated nerves have demonstrated the reversal of axonal transport direction (Bray et al., 1971; Frizzel et al., 1976; Bulger and Bisby, 1978; Smith, 1988), and some have implicated proteases (Sahenk and Lasek, 1988; Smith and Snyder, 1991) or acidification (Edmonds and Koenig, 1987; Sahenk and Brown, 1991) in the reversal process. But these experiments involved severe insult to the cells and reversal of organelles that may not otherwise move retrogradely, and therefore may not be relevant to mitochondria which routinely change direction. One report suggests that ADP could be the molecule which signals mitochondria to cease movements (Bereiter-Hahn and Vöth, 1983), which could explain the clustering of mitochondria in regions of high ATP demand. The unique ionic environment of the growth cone, which is thought to control its motility (Cohan et al., 1987; Lankford et al, 1988; Mattson et al., 1988; Lankford and Letourneau, 1991; Rehder and Kater, 1992), may also play a role in controlling mitochondrial transport, which, as shown here, is dependent upon axonal outgrowth and is specifically regulated in the growth cone region. These intracellular signals may act through local modification of the phosphorylation state of the mitochondrial motor proteins or their putative receptors. Kinesin, the kinesin-associated protein kinectin and dynein are all phosphoproteins (Sato-Yoshitake et al., 1991; Farshori and Goode, 1991; Toyoshima et al., 1992; Chilcote and Johnson, 1990; Hamasaki et al., 1991; Hollenbeck, 1993), and may affect the motor activity (Chilcote and Johnson, 1990; Hamasaki et al., 1991) or the affinity of motor proteins for their receptors (Sato-Yoshitake et al., 1991; Farshori and Goode, 1991). Phosphorylation could serve to regulate the transition between the moving and stationary states in the growth cone region by acting at any of these levels. At least one kinase localizes to growth cones (Maness et al., 1988), and others may as well.

Does the regulation of mitochondrial transport depend upon whether or not the mitochondria are producing ATP? A guiding assumption in this study is that the reason mitochondrial distribution and transport are regulated at all in many cell types is to ensure adequate ATP levels at foci of intense energy consumption. An interesting question then was whether the regulated transport of mitochondria in axons was a “default” process, or whether it depended upon their capacity to supply ATP to the cytosol. Experiments performed under anaerobic conditions showed that after 8 h of anaerobic culture, growing axons still displayed a mitochondrial distribution gradient although a shallower one than in aerobic cultures (Fig. 4A), and that blocking outgrowth in these cultures caused mitochondrial redistribution as it had in aerobic cultures (Fig. 4B). Since the mitochondria were redistributed here, this argues that the skewed distribution in the growing anaerobic culture was not simply due to the lack of sufficient ATP to power a redistribution. Growing cultures maintained under anaerobic conditions for 24 h (Fig. 4C) revealed that even though the overall density of mitochondria in the distal axon had dropped to one half of the aerobic value, the few mitochondria which remained were still maintained in a skewed distribution. The reduction in mitochondrial density was most likely due to breakdown of mitochondrial inner membranes (Luzikov et al., 1971, 1973) and subsequent loss of cytochrome c oxidase for oxidative phosphorylation and for immunofluorescence detection. Thus, the regulatory mechanisms which retain mitochondria near the growth cone of growing axons, and free them to move retrogradely in non-growing axons, are not strictly dependent on aerobic mitochondrial metabolism.

The authors thank Dr A. Miranda for the generous gift of antibodies against cytochrome c oxidase, and Drs P. C. Bridgman and L. B. Chen for critically reading the manuscript. This work was supported by a grant from the National Institutes of Health (NS27073) and a Basil O’Connor Starter Scholar Award from the March of Dimes Birth Defects Foundation.