|Home | About | Journals | Submit | Contact Us | Français|
Axonogenesis involves a shift from uniform delivery of materials to all neurites to preferential delivery to the putative axon, supporting its more rapid extension. Waves, growth cone-like structures that propagate down the length of neurites, were shown previously to correlate with neurite growth in dissociated cultured hippocampal neurons. Waves are similar to growth cones in their structure, composition and dynamics. Here, we report that waves form in all undifferentiated neurites, but occur more frequently in the future axon during initial neuronal polarization. Moreover, wave frequency and their impact on neurite growth are altered in neurons treated with stimuli that enhance axonogenesis. Coincident with wave arrival, growth cones enlarge and undergo a marked increase in dynamics. Through their engorgement of filopodia along the neurite shaft, waves can induce de novo neurite branching. Actin in waves maintains much of its cohesiveness during transport whereas actin in non-wave regions of the neurite rapidly diffuses as measured by live cell imaging of photoactivated GFP-actin and photoconversion of Dendra-actin. Thus, waves represent an alternative axonal transport mechanism for actin. Waves also occur in neurons in organotypic hippocampal slices where they propagate along neurites in the dentate gyrus and the CA regions and induce branching. Taken together, our results indicate that waves are physiologically relevant and contribute to axon growth and branching via the transport of actin and by increasing growth cone dynamics.
After terminal differentiation, neurons form neurites which then grow, arborize, and respond to environmental cues to form complex synaptic connections. The extending neurites are initially similar, but then acquire different identities becoming either axons or dendrites during polarization, which is necessary to neuronal function. Cultured hippocampal neurons have served as a model for many aspects of development including axon specification and branching (Craig and Banker, 1994; da Silva and Dotti, 2002). Axon specification involves cytoskeletal reorganization (Witte and Bradke, 2008), as well as the regulation of intracellular transport (Horton and Ehlers, 2003). Prior to axon formation, in what are known as stage 2 hippocampal neurons, transport is relatively uniform in all processes. Later, during the stage 2–3 transition in which axon specification and growth occur, there is an asymmetric increase in anterograde transport into the presumptive axon (Bradke and Dotti, 1997). This increase in transport is reflected by an engorgement and increased dynamics of the growth cone of the emergent axon (Bradke and Dotti, 1997, 1999; Kunda et al., 2001). Accumulating in the presumptive axon are proteins that impact the development of neuronal polarity including many cytoskeletal regulators (Shi et al., 2003; Schwamborn and Puschel, 2004; Inagaki et al., 2001; Garvalov et al., 2007). A general increase in transport could contribute to the increased levels of these proteins in the axon; however, other mechanisms such as selective degradation of proteins in minor processes (Schwamborn et al., 2007) and selective association with axon-specific kinesin motors (Jacobson et al., 2006) are also at work.
In hippocampal neurons, axons as well as dendrites undergo periodic extension and retraction. Ruthel and Banker (1998, 1999) initially identified and characterized another periodic phenomenon that they called “waves” which could underlie, at least partially, the pulsatile outgrowth. Waves appear similar to growth cones, often originate at the soma, travel down the length of axons and dendrites, and are associated with spurts of outgrowth. Waves travel at a speed consistent with slow component b (SCb) of axonal transport (~3 μm/min), which carries actin and actin associated proteins and which may be rate limiting for neurite growth (Wujek and Lasek, 1983; McQuarrie and Jacob, 1991). Thus waves may represent one mechanism to supply growth cones with actin as well as other growth promoting proteins.
Here we use live-cell imaging to characterize wave behavior and effects on neuronal development. We confirm that waves are indeed similar to growth cones. Waves occur in all processes in stage 2 neurons but become more frequent in presumptive axons during the stage 2–3 transition and their prevalence increases with treatments that enhance axonogenesis. Waves deliver actin and associated proteins to growth cones or to filopodia along the neurite shaft, which are often associated with neurite growth and filopodial engorgement to create neurite branches, respectively. Waves occur in organotypic hippocampal slices, demonstrating that they can contribute to growth in a relevant physiological setting. The implications from these results are that waves contribute to neuronal branching and axon growth.
Rat and mouse hippocampal neurons were dissected at day E18 and E16.5, respectively, and either used directly in experiments or frozen as previously described (Mattson and Kater, 1988; Garvalov et al., 2007). Neurons were cultured on poly-d-lysine (Sigma) coated glass coverslips in either neurobasal (Gibco) supplemented with B-27 (Invitrogen, San Diego, CA) and glutamine (Minamide et al., 2000) or in glial conditioned N2 medium as previously described (de Hoop, 1998). The myosin II inhibitor, blebbistatin (Sigma Aldrich) was dissolved in DMSO at 100 mM and applied to neuronal cultures at a final concentration of 2.5 μM. For long-term cultures, blebbistatin was added to growth medium 16 h after plating through the termination of the experiment. In some experiments neurons were plated on mixed poly-d-lysine-laminin (10mg/ml) substrate.
Replication deficient adenoviruses were produced (Minamide et al., 2003) for transgene expression in neuronal cultures essentially as described (Garvalov et al., 2007). The viruses used in these studies include wild-type and nonphosphorylatable (active, S3A) human cofilin constructs fused to monomeric Red Fluorescent protein (mRFP; Campbell et al., 2002). An adenovirus containing the neuron-specific enolase promoter (NSE; Forss-Petter et al., 1990) driving cofilin expression was also used for live-cell imaging experiments. Beta-actin fused at its C-terminus to mRFP, GFP, a photo-activable (pa) GFP (kind gift of Jennifer Lippincott-Schwartz, National Institute of Health) or the green-to-red photo-convertible(pc)-Dendra protein (kind gift of Michael Tamkum) were also used to visualize actin in live cells. Adenoviral mediated expression of hippocampal slice preparations were performed as previously. Slices were prepared for time-lapse imaging 2–4 days post-infection (see below).
The fixation, blocking and immunostaining were performed essentially as previously described (Garvalov et al., 2007). Neurons were fixed 30 min at 37°C in 4% paraformaldehyde in cytoskeletal preservation buffer (10 mM MES pH 6.1, 138 mM KCl, 3 mM MgCl2, 10 mM EGTA, and 0.32 M sucrose). After blocking, immunostaining was performed with the following primary antibodies: Tau1 (1:1000; Chemicon), β-tubulin (1:300; Sigma), Rap1 (1:250; BD Biosciences), Gap43 (1:300; De la Monte et al., 1989), cdc42 (1:250; Santa Cruz Biotech), Arp3 (mouse monoclonal, clone FMS338, Sigma-Aldrich), acetylated tubulin (clone6-11B-1, 1:50,000; Sigma-Aldrich), tyrosinated tubulin (YL1/2, 1:40,000; Abcam) total cofilin (MAb22;14.8 μg/ml, Abe et al., 1989) and for phospho-cofilin (1.1 μg/ml, Meberg et al., 1998). IgG secondary antibodies (Molecular Probes, Eugene, Oregon) used were Fluorescein goat anti-mouse IgG, Texas Red goat anti-rabbit, Alexa 650 goat anti-mouse and Alexa 350 goat anti-mouse (all at 1:400). Texas Red- or Alexa 488-conjugated phallodin (1:200) was used to visualize actin filaments. Coverslips were mounted with Prolong Anti-fade (Molecular Probes). Images were acquired with a Nikon (Tokyo, Japan) Diaphot inverted microscope using 20X air (0.75 NA), 40x oil (1.3 NA) and 60X oil (1.4 NA) objectives and a Coolsnap ES CCD camera (Roper Scientific, Tucson, AZ). Metamorph software (Universal Imaging, Westchester, PA) was used for image acquisition and analysis.
Organotypic hippocampal slice cultures were performed on postnatal day 0 to 5 Sprague-Dawley rat pups or Thy1-YFP mouse pups (line H; (Feng et al., 2000); Jackson Laboratory) as previously described (Davis et al., 2009). In some cases, small DiI crystals (Invitrogen-Molecular Probes) were placed in the hilus region of the dentate gyrus as described (Dailey et al., 1994), resulting in labeling of cells in the dentate gyrus as well as mossy fiber tracts (Supplementary Fig. 1). Slices were incubated for 3–48 h at 35° C prior to fixation or live-cell microscopy.
Live-cell microscopy was performed on an Olympus IX81 microscope equipped with a Yokagawa CSU22 spinning disk confocal head, AOTF and an EM-CCD cascade II camera (Photometrics). Diode lasers with 473 nm and 561 nm peak wavelengths were used for GFP and RFP excitation, respectively. DIC images were captured with an HQII camera. The objectives used include: 10x (0.3 NA), 20x (0.17 NA) air objectives, 40X DIC (1.35 NA), and 60x (1.42 NA) oil objectives. Computer controlled acquisition was performed with Slidebook Software (Intelligent Imaging Innovations, Denver, CO). For some experiments a phase contrast objective on the heated stage of a Nikon Diaphot microscope equipped with a Metamorph controlled CCD camera was used.
Time-lapse imaging for analysis of wave effects on neuronal development was performed either in glass bottom T-25 flasks or 35 mm dishes (German glass coverslips sealed with aquarium sealant). E16.5 mouse hippocampal neurons were cultured in neurobasal medium and pH equilibrated in a 5% CO2 incubator prior to imaging. The 35 mm dishes were used in a humidified, CO2 regulated system at 37° C. For long-term experiments, images were acquired every 2.5–5 min. For short-term experiments, images were acquired every 10s. Fluorescence image acquisition parameters varied depending on the particular experiment and were empirically determined to balance signal intensity and minimize cellular toxicity. In some cases neutral density filters were used to decrease excitation energy.
Adenovirally expressed photo-activatable(pa)-GFP or photo-convertible(pc)-Dendra (Gurskaya et al., 2006) fused to β-actin was activated 3 days postinfection with a computer-controlled Hg light source (405 nm) under control of the Mosaic Digital Diaphragm system (Photonics instruments). For pa-GFP-β-actin, waves were identified using bright field optics and a selected region encompassing the wave was photo-activated for 1s. Confocal images at 473 nm excitation were then acquired at 30–60 s intervals. In neurons expressing Dendra-β-actin, waves were identified by green fluorescence (473 nm). A region of the neurite with or without a wave was targeted with the Mosaic FRAP-system and photo-converted (400 ms). Following photo-conversion, images were captured using 561 nm excitation (500 ms) at 5-, 20-, or 60-second intervals. Before photo-conversion, an image using 473 nm excitation was captured to show the region of the neuron targeted. Though the first frames of the 561 nm time-series are marked “0-sec post-conversion”, it is actually delayed by ~10 s because of the operational delay in switching between imaging modes. Line-scans were made in Metamorph by manually drawing a one pixel-wide line (approximately 200 pixels long) through the targeted neurite and the intensity profiles were exported to Excel. Peak of fluorescence decay curves were made by manually drawing a one pixel-wide line that spanned the length of the original targeted boxed region and the average intensity values were exported to Excel. Curve-fitting using an exponential equation was done in Excel.
For live-cell imaging of hippocampal slices, an area of the membrane with one to three slices adhered was excised and placed slice-side down on a glass bottomed 35 mm dish. The slice was stabilized with a plasma clot made by mixing 20 μl chicken plasma with 6 μl thrombin. After clot formation, 400–500 μl of slice medium was added, completely submerging the slices. Diode lasers (561 nm and 473 nm) were used for fluorescence imaging of DiI and Thy1-YFP labeled neurons, respectively. Growing neurites and dynamic growth cones were maintained in slices for >24 h. In some cases images were analyzed using Metamorph software.
Growth cone and wave dynamics were measured from time-lapse DIC images from which was calculated the change in area every ten seconds (ΔA/10s) (Endo et al., 2003) by superimposing adjacent images from an image stack using Metamorph software. The average change in area from a sequence of 10 images was used for each growth cone and wave. Growth cones were characterized as inactive (<12% ΔA/10s) or active (≥ 12% ΔA/10s).
Axons were defined as being ≥ 80μm in length and containing intermediate to high levels of Tau1 immunofluorescence (Schwamborn and Puschel, 2004; Jiang et al., 2005). Neurites with an absence of Tau1 staining were not counted as axons regardless of length. Neurons were counted with multiple axons when they displayed two or more neurites ≥ 80μm in length that immunostained for Tau1. Waves were defined in fixed specimens as F-actin rich growth cone-like structures with filopodial and/or lamellipodial features. From our live-cell imaging experiments we observed that the majority (>80%) of such structures were waves. Branches were defined as protrusions off of the primary neurite that were ≥ 20 μm.
Axon formation is defined as the moment when one neurite reaches ≥ 80 μm in length and is at least twice as long as the next longest neurite. Measurements of wave frequency were taken over three four hour intervals: stage 2, stage 2–3 and stage 3. For the stage 2–3 transition the 4 h flanking (2 h before and 2 h after) axon formation was used for analysis. Neurite growth measurements after wave arrival were performed using Metamorph imaging software. Bursts of outgrowth were measured from the position of growth cone tips immediately before to 10 min following wave arrival.
We confirmed the observations of waves reported by Ruthel and Banker (1998, 1999) including their growth cone-like morphology, velocity, and effects on outgrowth when they arrived at the distal neurite (Fig. 1, Supplementary Fig. 2; Movie 1). In cases where waves consolidate into the neurite shaft (Supplementary Movie 1, note the second wave) they were not associated with bursts of neurite growth. To further quantify the behavioral similarity between waves and growth cones, we compared their motility by live-cell DIC time-lapse imaging (Fig. 1C, D). Total area changes per 10 s intervals are similar between growth cones and waves. We also examined waves for microtubule organization (Fig. 1B) and for the presence of tyrosine-tubulin and acetylated-tubulin, which represent newly assembled microtubules and stable microtubules respectively (Fig. 1E.). The tips of the microtubules that splay out from the bundles in the neurite contain tyr-tubulin, demonstrating that they are actively growing, whereas the microtubules in the neurite shaft are predominantly acetylated (stable) microtubules.
In addition to actin, cortactin and GAP-43 previously shown to be in waves (Ruthel and Banker, 1998, 1999), we identified other normal constituents of growth cones including the Rho GTPases Rap1B, cdc42 and Rac1 (Kunda et al., 2001; Schwamborn and Puschel, 2004), ADF/cofilin (Garvalov et al., 2007), Arp3 (Strasser et al., 2004), slingshot (Abe et al., 2003), and LIM kinase 1 (Rosso et al., 2004) (Supplementary Fig. 3A,B). In addition, waves are enriched in tetanus insensitive vesicle associated protein (TI-VAMP)-containing vesicles (Supplementary Fig. 4), suggesting that waves are sites of membrane addition which may be crucial for neurite growth (Alberts et al., 2006). Taken together, these data confirm that waves are growth cone-like structure in their composition, cytoskeletal organization and dynamics.
Waves have not been studied in relation to axonogenesis during the stage 2 to 3 transition (axonogenesis--see definition in materials and methods). By time lapse imaging of individual neurons maturing from stage 2 to stage 3, the neurite that develops into the axon can be identified. Movies can be examined at earlier stages to identify waves in all neurites, quantify their frequency, and determine how they impacted axon development (see Supplementary Movie 2). Although waves occur in multiple neurites in stage 2, they occur more frequently in the neurite that later develops into the axon (Fig. 2B, Supplementary Movie 2). However, at this developmental stage the impact of waves on transient neurite elongation is essentially the same for all minor processes (Fig. 2B). During the stage 2–3 transition, wave frequency increases 2.8 fold in the developing axon over that occurring in the minor neurites (Fig. 2B). Furthermore, waves that arrive at the growth cones of the developing axon increase outgrowth by 1.5 fold that of the minor neurites. The frequency and impact of waves on axon outgrowth continues in young stage 3 neurons (Fig. 2B). Most neurons (>80%) undergoing axonogenesis generate waves. To conclude waves are more frequent in the developing axon and are associated with bursts of neurite outgrowth, perhaps providing an alternative mechanism for material delivery to the growth cone.
A fundamental aspect of the wiring of the mammalian brain is neuronal branching, with individual neurons often making connections to multiple target cells. For example, in vivo, CA3 pyramidal neurons have axons that branch forming Schaffer collaterals that innervate the CA1 region as well as collaterals connecting with the dentate gyrus (Gomez-Di Cesare et al., 1997). In our live-cell imaging experiments, we not only observed waves that can traverse existing neurite branches (Ruthel and Banker, 1999), but also observed waves initiating nascent neurite branches. In these cases, a wave arrives at a region of the neurite shaft where a filopodium protrudes and causes elongation and engorgement of this filopodium, giving rise to a new branch point (Fig. 1C). These new branches are persistent and are maintained in culture for the time period of our observations (up to 24 hours). At other sites where waves terminate along the neurite shaft, we observed persistent increases of neurite caliber. The propagation of multiple waves in tandem also influence neurite branch formation and growth (Supplementary Fig. 5, Supplementary Movie 3).
We next examined the effects of cytoskeleton regulators on waves that are known to influence growth cone dynamics and axon growth. Cofilin positively regulates growth cone dynamics (Endo et al., 2003) and has recently been shown to positively impact axon formation (Garvalov et al, 2007). We found a high ratio of active cofilin in waves, similar to levels observed in the growth cones of developing axons (Fig. 3A). Furthermore, an increased percentage of neurons over-expressing either wild-type or the non-phosphorylatable (S3A) active cofilin have wave-like structures compared to control neurons or those expressing the inactive (pseudophosphorylated) S3E mutant (Fig. 3B,C).
Myosin II modulates growth cone motility and at least a portion of F-actin retrograde flow in neuronal growth cones (Bridgman et al., 2001) and is antagonistic to neurite growth in dorsal root ganglion neurons (Gallo et al., 2002) and axon formation in hippocampal neurons (Kim and Chang, 2004). Therefore, we examined the effects on wave formation of the specific myosin II inhibitor blebbistatin, which enhances neurite length and arborization in hippocampal neurons (Fig 4A). The addition of 2.5 μM blebbistatin induces the formation of supernumerary axons (Fig 4B) and greatly enhances axon branching (Fig. 4C). Neurons treated with blebbistatin showed a ~20% increase in the percent of neurons with wave-like structures (Fig. 4D). Live-cell imaging experiments (Fig. 5A) corroborated this finding, with wave frequency increasing 22% in the same neurons after blebbistatin addition (Fig 5B). Furthermore, the impact of waves on neurite growth is modulated with myosin inhibition (Supplementary Movies 4 and 5). Preceding wave arrival at the growth cone, the neurite retracts several microns (Fig. 5A,C). Treatment with blebbistatin reduces by over 2-fold the distance of retraction compared to untreated controls. The average magnitude of outgrowth following blebbistatin treatment also increased 1.5 fold relative to control neurons (Fig. 5D, Supplementary Movies 4 and 5). In the presence of blebbistatin, waves have increased filopodial dynamics and induce axonal branches more frequently than in control neurons.
Laminin increases axon growth in hippocampal neurons (Esch et al., 1999) and can rescue neurite growth of cortical neurons lacking Ena/Vasp proteins (Dent et al., 2007). Hippocampal neurons cultured on laminin (Supplementary Fig. 6) display a significant increase in the percent of neurons extending one or more axons and have increased neurite branch density and more waves. Several instances of waves inducing axonal branches in neurons growing on laminin were observed (Supplementary Movie 3, second neuron). Taken together these data suggest that myosin II inhibition or growth on laminin enhances axon growth and branching, at least in part via increased wave activity.
Preceding the rapid growth of the axon, there is an increase in the size and dynamics of the presumptive axon growth cone (Bradke and Dotti, 1997; Bradke and Dotti, 1999; Kunda et al., 2001). Therefore, we sought to determine if wave arrival at the distal neurite increased growth cone size and dynamics. Short interval DIC time-lapse imaging revealed that wave arrival at the neurite tip (Fig. 6A) increased growth cone size by 2.6 fold and growth cone dynamics by 2 fold (Fig. 6B, C, Supplementary Movies 5). Furthermore, we often observed collapsed, non-dynamic growth cones transform into large and extremely dynamic growth cones upon wave arrival (Fig. 6A). Growth cones that were dynamic prior to wave arrival also underwent a moderate, albeit significant increase in dynamics (Supplementary Fig. 7). These data suggest that waves increase growth cone size and dynamics, two features correlated with axon specification.
To visualize actin in live-cell experiments, we utilized monomeric Red Fluorescent Protein (Campbell et al., 2002) fused to actin (mRFP-actin). When expressed in hippocampal neurons mRFP-actin incorporates into the highly dynamic actin structures of growth cone filopodia and lamellipodia. In waves, mRFP-actin was dynamic and upon arrival at the growth cone it incorporated into actin structures within growth cones resulting in a net increase in their fluorescence intensity (Fig. 7A, Supplementary Movies 8 and 9). Because waves induce a rearrangement of cortical actin in the neurite, it is possible that actin in waves does not translocate along with the wave. To address this possibility, we utilized a photo-activatable GFP and a Green-to-Red photo-convertible Dendra (Gurskaya et al., 2006) fused to β-actin (hereafter referred to as paGFP-actin and pcDendra-actin). For paGFP-actin, wave structures were identified by DIC imaging before photo-activating the GFP-actin in the wave and following it by time-lapse fluorescence imaging. At least a portion (about 50%) of the activated paGFP-actin traveled in a wave to the neurite tip over a distance of 25μm (Supplemental Fig. 8). However, some fluorescent actin remains in the neurite and some diffuses into other regions of the cell. Interestingly, we observed in hippocampal neurons with short processes that some actin diffused rapidly, at rates resembling diffusion of dextrans of similar size to actin monomers measured in Xenopus neurites (Popov and Poo, 1992).
To determine whether actin in waves behaved fundamentally differently from actin within non-wave regions, we utilized pcDendra-actin. It requires less energy for photoconversion to a usable signal than paGFP-actin, the red Dendra is quite stable for imaging, and neurons expressing Dendra-actin can be identified by their green fluorescence prior to photoconversion. By photo-converting a limited amount of pcDendra-actin within specific regions of the neurite, we showed that actin within non-wave regions diffuse rapidly and bi-directionally away from the original region of photoconversion, whereas, actin within waves remained associated with the wave over a longer period of time even while it migrates (non-diffusive) (Fig. 7B,C). Diffusion of actin within non-wave regions is supported both by analysis of fluorescence decay, which demonstrates that the peak of photoconverted-actin decays exponentially (as expected of one-dimensional diffusion-based processes, τ = 0.06 s-1) (Supplementary Fig. 9), and line-scan analysis, which demonstrates a Gaussian distribution of actin centered at the region-of-photoconversion (Fig. 7C). It is of interest to note that within non-wave regions, the distribution of actin quickly broadens beyond the original width of the photoconversion region, which is noticeable even at the first frame of 561 nm imaging (Fig. 7C). In addition, nearly 20% of the pc-actin does not diffuse away during the first 30 s, suggesting that a minor but more-stable pool of actin is also present (Fig. 7D,E). Indeed, a mixed-model of two Gaussian distributions can be used to fit the actin-distribution (data not shown). In contrast, actin within waves is transported by a non-diffusive process. This is supported by the non-exponential decay-of-peak-fluorescence of actin within waves (Supplementary Fig. 9), and by line-scan analysis, which shows that peak-fluorescence is maintained over a longer time-period and that the distribution of actin does not broaden beyond the original width of the photoconverted-region when the wave remains relatively stationary (Fig. 7E).
We also sought to determine if actin remains cohesive within the wave as it travels over longer distances and over longer time frames. A portion of photo-converted actin remained within waves over several minutes (Fig. 7D,E). However, the majority of the pc-actin fluorescence is lost and diffuses away from the wave as it travels over long distances. This indicates that the actin in waves is dynamic and interchanges with the freely diffusing actin subunits and cortical actin in neurite. In support of this, pc-actin from nearby regions in the neuron amalgamated into a nearby wave. In addition to delivering actin to the terminal tip of the neurite, we also observed waves delivering actin to shaft filopodia, which supported the rapid growth of the protrusion into a potential branch-point (data not shown).
Other proteins found in waves are involved in axon growth (and neurite growth in general) and are likely also transported in waves. For example, as previously mentioned, ADF/cofilin proteins are known to influence neurite growth, growth cone dynamics, pathfinding and axon formation (Meberg and Bamburg, 2000; Kuhn et al., 2000; Aizawa et al., 2001; Endo et al., 2003; Wen et al., 2007; Garvalov et al., 2007). Fluorescently labeled cofilin behaved similarly to mRFP-actin and was transported to growth cones in waves in cultured neurons (Supplementary Fig. 10). In other cases, we observed a gradual increase in cofilin-GFP in growth cones (data not shown).
Although waves are prevalent in cultured neurons (Ruthel and Banker, 1998, 1999; Heidemann et al., 2003; Rosso et al., 2004; Toriyama et al., 2006; Tursun et al., 2005), it is possible they are an artifact of culturing neurons on a two-dimensional substrate. To determine if waves are physiologically relevant, we examined fixed and live hippocampal slices in which neurons were fluorescently labeled either with the lipophilic dye, DiI, or via the expression of Thy1-YFP (Feng et al., 2000). Although we focused these studies on neurons in the dentate gyrus and the mossy fiber tract because previous work showed neurogenesis, neurite formation and outgrowth occur robustly in these regions early in postnatal development (Dailey et al., 1994; Knoll et al., 2006), other regions of the hippocampus including CA3 and CA1 as well as entorhinal cortex and cerebral cortex were also observed. Wave-like structures were seen in neurons from all of these regions, but waves were especially robust in the dentate gyrus from the Thy1-YFP transgenic mice as observed in vivo in fixed sections (Supplementary Fig. 11).
To confirm that these structures were waves, we performed 3-dimensional time-lapse confocal microscopy of individual neurons in organotypic hippocampal slices. We observed growth cone advance of DiI and Thy1-YFP labeled mossy fiber axons occurs at rates between 0.9μm/min and 2.5 μm/min calculated from 2D projected image stacks (Supplementary Fig. 1). In these slice preparations we also observed waves in neurons in the dentate gyrus and CA3 regions of the hippocampus. In DiI labeled neurons, we observed waves traveling in 3-dimensions in the dentate gyrus (Supplemental Fig. 12a, arrow, Supplementary Movie 7). We also observed waves in Thy1-YFP neurons in CA regions of the hippocampus (Fig. 8). In brain slices we also observed that wave arrival at the distal neurite increased growth cone size and advance (Fig. 8), similar to observations in culture. Wave velocity in slices is similar to that measured in dissociated neuronal culture, ranging between 0.8 μm and 5.6μm/min with an average of 2.3 μm/min. This velocity is similar to that of rapid growth cone advance and within the range of slow axonal transport. In addition, neurons in slices expressing GFP-actin contained dynamic actin in waves that appeared similar to growth cones (Supplementary Movie 7, Supplementary Fig 12b). In hippocampal slices, waves also induced neurite branching via the engorgement and extension of shaft filopodia (Supplementary Movie 7).
Significantly, we rarely observed waves in long axons of the mossy fiber tract, demonstrating that waves occur preferentially early in neuronal development. This finding agrees with observations of Ruthel and Banker (1999) and our observations in dissociated neuronal culture where wave frequency was high in the first few days (0.25–3 div) as neurites rapidly extend, but decreased later in culture (4–5 days), when neurite growth has declined (data not shown).
In the decade that has passed since they were first described (Ruthel and Banker, 1998, 1999), waves have received surprisingly little attention from the research community. Our work has expanded on the understanding of growth cone-like waves in axonogenesis, neurite branching and actin transport as well as confirming previous observations made on growth cone-like waves (Ruthel and Banker, 1998, 1999; Heidemann et al., 2003; Rosso et al., 2004; Toriyama et al., 2006; Tursun et al., 2005). Importantly, we also provide the first evidence that waves occur in ex vivo brain slices, suggesting a role during in vivo neuronal development.
Previous studies showed that actively propagating waves behave similarly to growth cones in their rate of advance (1–6 μm/min) and in undergoing arrest and collapse in the presence of cytochalasin and nocodazole, but not in the presence of Brefeldin A, suggesting both microtubules and actin, but not Golgi-derived vesicles, are necessary for their propagation (Ruthel and Banker, 1998; Bradke and Dotti, 2000; Ruthel and Hollenbeck, 2000). We identified in waves additional regulators of growth cone motility and actin dynamics, including cofilin, Lim kinase, Slingshot, Arp3, Rap1, Rac1 and cdc42 and showed that waves contain newly assembled microtubule ends containing tyrosine-tubulin but no acetylated tubulin. All of these components suggest that waves are truly growth-cone structures that migrate along the neurite shaft where acetylated (stable) microtubules predominate.
The formation of wave-like structures is not restricted to neurons and insight into their formation has come from studies with the slime mold Dictyostelium (Bretschneider et al., 2009). Dictyostelium forms wave structures on the substrate attached membrane of the cell that propogate at about 6 μm/min along the membrane using an actin treadmilling mechanism that contributes to membrane protrusion when the waves reach the cell perimeter, much like the burst in neurite extension we see when waves reach the growth cone. Actin waves in Dictyostelium are induced during recovery after washout of the actin sequestering compound, Latrunculin A, suggesting that they form spontaneously from the self-organization of the actin cytoskeleton. Dictyostelium waves are devoid of myosin II but contain a single headed myosin (MyoIB) at the wave front, Arp2/3 complex in a three dimensional network linked to MyoIB via the adapter protein CARMIL, and coronin at the rear of the wave in close apposition to the substratum (Bretschneider et al., 2009). Occasionally these waves collapse or reverse direction, similar to what we observe for growth-cone like waves. A disruption in the localization or inactivation of some key self-organizing component presumably underlies decreases in wave motility; in neurons this eventually leads to wave collapse into the neurite shaft, which we observed in some instances. This may be similar to growth cone collapse, which can be induced by changes in actin regulating proteins (Gallo and Letourneau, 2004).
Since Dictyostelium waves can be induced during recovery from actin disassembly by Latrunculin A, it is likely that localized disruption in actin organization at the neurite soma junction could be involved in wave formation in neurons. Because cofilin is a major component of growth cone-like waves, its activity at the neurite soma junction might be the wave initiator. Furthermore, during neuronal development much of the cofilin is transported from the soma to distal regions of the neurites and a decline in its concentration or activity in the soma could lead to the observed decrease in wave frequency as neurites elongate. Studies to examine these questions are in progress.
Although waves occur indiscriminately in all of the minor processes of a stage 2 neuron, there is a >2 fold increase in wave frequency and a greater impact of waves on outgrowth in the presumptive axon during the stage 2–3 transition. This suggests that individual waves may have fundamental differences (all waves are not equal) or that the growth cones of different neurites have pre-existing differences producing varied responses upon wave arrival (all growth cones are not equal). Although we observed that waves are diverse in their size and advance rates, our results cannot rule out that inherent diversity of growth cones biases differential responses to wave arrival, especially considering that wave-induced changes in growth cone dynamics depended on the preexisting dynamic state of the growth cone. We also observed that abrupt, short-term increases in wave frequency provided greater impetus for neurite extension than waves arriving in isolation. In one case we observed the arrival of 5 waves in rapid succession immediately preceding a burst of growth and the subsequent development of the axon.
The observation that waves occur more frequently and have greater impact on growth in the developing axon suggests that waves provide some impetus for axon differentiation. To further study the involvement of waves in axon development we looked at the influence of two factors that influence axonogenesis on wave frequency and impact. The myosin II inhibitor, blebbistatin, greatly enhances axon growth, and even increases the percent of neurons with supernumerary axons. Myosin II inhibition also increased wave frequency and augmented the impact of waves on neurite growth. Likewise, the extracellular matrix protein, laminin, a signal that promotes axon growth in hippocampal neurons, also enhanced wave frequency, supporting our hypothesis that waves enhance axon specification.
Branched neurite networks are another important aspect of brain development in vivo. In developing cortical neurons, axonal branching can be inhibited by treatment with drugs that disrupt the dynamics of either microtubule or actin but at concentrations below those that affect axon outgrowth (Dent and Kalil, 2001). Similar treatments also impedes wave propagation (Ruthel and Banker, 1999) suggesting a possible link between waves and branching. We have observed both the bifurcation of waves at a branch point supporting outgrowth of both branches and waves choosing one branch over the other enhancing elongation of only one branch. During in vivo development, axon outgrowth and branch extension occur independently such that branch growth occurs while axons are stalled (O’Leary and Terashima, 1988; Luo and O’Leary, 2005).
Collateral branches can form along the axon shaft in response to positive extracellular signals (Tang and Kalil, 2005). We have observed waves inducing collateral branching from the neurite shaft by causing the engorgement of a stable shaft filopodium, resulting in its conversion to a new branch. Thus, waves can generate a new growth cone from the engorged filopodium by delivering actin and other molecules required for growth cone function. Waves may also direct the fusion of TI-VAMP vesicles at branch-points, providing membrane necessary for branch elongation and serve as sites for calcium transients that also are correlated with branch growth (Hutchins and Kalil, 2008). Other indirect evidence suggesting waves contribute to branch formation is the correlation between enhanced branching and wave frequency we observed following treatment of neurons with blebbistatin or laminin. Inhibiting myosin II activity may relax the rigid cortical F-actin allowing increased actin dynamics, filopodia formation, engorgement and branch formation. Laminin may also stabilize shaft filopodia increasing the opportunities for waves to induce a branch.
During axon formation, the growth cone of one of the multiple neurites enlarges and displays increased dynamics (Bradke and Dotti, 1997, 1999; Ruthel and Hollenbeck, 2001; Kunda et al., 2001), which is associated with more rapid outgrowth (Gallo and Letourneau, 2004). This neurite undergoes rapid elongation forming the axon. We observed over a 2.5 fold increase in growth cone size and over a 1.6 fold increase in growth cone dynamics following wave arrival at the growth cone. Although wave arrival increased the dynamics of growth cones that were previously active, waves had a much greater effect on growth cone activity when merging with less active growth cones. Thus, waves influence axon development by increasing growth cone size and dynamics, a prerequisite for axon specification. Theoretically, if size is the only thing that matters, wave arrival can deliver a bolus of material for the extension of about 20 μm of neurite shaft if growth cone area expands from 50 μm2 to 150 μm2. This is based on our measurement where 10 μm of shaft < 50 μm2 surface area and we commonly observed > 100 μm2 increase in the size (area) of a growth cone following wave arrival.
Thirty years ago it was observed that neurofilament proteins, tubulin and actin synthesized in the soma traveled in the slow component of axonal transport in mature neurons (Black and Lasek, 1979) yet three decades later, exactly how these proteins are transported remains controversial. Most microtubules are relatively stationary in axons (Ma et al., 2004) but in cultured neurons some short assembled microtubule pieces can move rapidly along the stationary microtubules in both directions. This movement has a bias toward the anterograde direction accounting for “slow” net transport rates (Wang and Brown, 2002), a similar mechanism to what was observed for movement of some neurofilaments in the “stop and go” hypothesis (Wang and Brown, 2001; Brown et al., 2005). With this as a model, it has been suggested that actin is also transported in a filamentous form along microtubules (Baas and Buster, 2004) but direct evidence for this is lacking. Indeed, in chicken sciatic nerve actin is transported with proteins that are associated with unassembled or dynamic actin (Mills et al., 1996). Furthermore, the strong bias for newly assembled tyrosinated microtubule subunits at the distal end of a growing axon supports microtubule assembly at the distal tip. Taken together these results suggest that multiple mechanisms exist for transporting cytoskeletal proteins down axons.
Here we directly observed the anterograde movement of fluorescently labeled actin in waves. Because we could not rule out the possibility that as waves move forward they cause a reorganization of the cortical actin in the neurite shaft and result in no net translocation of actin, we used photo-activatable (pa)GFP-actin and photo-convertible (pc)Dendra-actin, which were only photo-activated/converted within the wave, to ascertain if actin was traveling with the wave. We observed a portion of this actin reaches the growth cone after traveling down the neurite shaft at a rate consistent with slow transport. However, not all of the actin arrived at the neurite tip. Indeed, we often observed that pcDendra-actin in a wave could incorporate into filopodia along the neurite, while the wave continued, suggesting that some subunits from the dynamic actin in the wave exchange with more stationary actin structures while others are recycled back into actin filaments in the advancing wave. This is reminiscent of the situation within growth cones. When paGFP-actin or pcDendra-actin in growth cones is photo-activated, the actin initially remains associated with dynamic filaments and most subunits released from treadmilling filaments get reincorporated but some subunits are lost by diffusion or exchange into other actin structures that remain behind and the fluorescence eventually declines. The tendency for actin to remain associated with the growth cone is positively correlated to its growth rate, which is related to the continuous reutilization of subunits in treadmilling filaments (Marsh and Letourneau, 1984; Letourneau et al., 1987). If paGFP-actin is activated in growth cones where advance does not occur, fluorescence is found to dissipate rapidly and appear back in the soma within 5–10 min (unpublished observations). Conversely, waves that contribute to greater neurite outgrowth may also deliver more actin and other important growth promoting factors. Although a portion of the actin remains associated with the wave as it moves, the actin composition in the wave does not need to remain unchanged as a wave travels forward in order for the wave to, in effect, “deliver” actin. In short neurites without waves, freely diffusing actin subunits can reach distal neurites relatively rapidly but it is unlikely to contribute to growth spurts or branch initiation. Conversly, waves can deliver to distal neurites or branch points a large bolus of actin and actin assembly regulatory molecules such as cofilin, previously identified to be transported with actin in slow axonal transport (Mills et al., 1996) and important for neurite outgrowth, axon development, growth cone motility and pathfinding (Meberg and Bamburg, 2000; Abe et al., 2003; Garvalov et al., 2007; Wen et al., 2007). Not only is cofilin transported with waves, but it is the dephosphorylated (active) form that is most prevalent, suggesting its role in wave motility is similar to its role in growth cones. Another axon-promoting molecule, shootin, was also recently shown to be transported in waves (Toriyama et al., 2006). Thus, waves likely represent one mechanism for transporting large amounts of growth promoting molecules to growth cones, thereby stimulating axonogenesis.
Although an attractive hypothesis for mediating outgrowth spurts, waves are certainly not the only means for transport of these molecules. We have observed a number of neurons (~20%) that do not exhibit waves over a time-course of hours in culture, yet still have active neurite outgrowth and periodic fluctuations in growth cone size and motility. In addition, we have observed gradual accumulations of actin and cofilin in growth cones without any obvious wave transport. Thus actin transport, of which waves are but one component, appears to occur through multiple redundant mechanisms. Waves are unique in that they can supply large amounts of materials directed to particular locations such as growth cones and branch points and can be employed during early neuronal development when rapid phases of neurite outgrowth and axonogenesis occur.
Because previous studies have identified waves only in dissociated cultured neurons, their relevance to neuronal development in vivo is rightfully questioned. Thus we sought to determine if waves occur on neurons developing within ex vivo organotypic hippocampal slices, a more in vivo-like setting. Three different labeling methods were used to visualize waves in neurons within slices, membrane labeling with DiI or Thy1-YFP and adenoviral-mediated cytoplasmic expression of fluorescent proteins. We observed waves in developing neurons with each of these methods. We also observed wave-induced growth cone enlargement, neurite growth enhancement and the induction of new branch-points in slices, all similar to what we observed in dissociated neuronal cultures. Interestingly, we never observed long axons with active growth cones that also exhibited waves. This may suggest that waves are important only during the initial, rapid phase of neurite growth and axonogenesis; however it might also reflect the limited number of long axons we examined.
In summary, waves do appear to have important functions in neuronal development. Future work will further elucidate the molecular underpinnings of wave initiation and their role in other processes such as axon guidance and regeneration.
This work was supported by the National Institute of Health (NS40371, DK69408, (J.R.B), NS48660 (K.C.F), the Max Planck Society, and the Deutsche Forschungsgemeinschaft Grant SFB 391 (F.B). We thank Barbara Bernstein and Laurie Minamide for technical assistance and helpful suggestions with experiments. We also extend gratitude to O’Neil Wiggan, Boyan Garvalov, Christian Gonzalez-Billault, Michael Tamkum, Stuart Tobet, Kathryn Partin, Mike Maloney, Janel Funk, Richard Davis and Shay Perea-Boettcher for helpful discussions. The Thy1-YFP mice used in this study were kindly supplied by Kristy McClellan and Stuart Tobet.