Mini-review: Microtubule Sliding in Neurons

Abstract

Microtubule sliding is an underappreciated mechanism that contributes to the establishment, organization, preservation, and plasticity of neuronal microtubule arrays. Powered by molecular motor proteins and regulated in part by static crosslinker proteins, microtubule sliding is the movement of microtubules relative to other microtubules or to non-microtubule structures such as the actin cytoskeleton. In addition to other important functions, microtubule sliding significantly contributes to the establishment and maintenance of microtubule polarity patterns in different regions of the neuron. The purpose of this article is to review the state of knowledge on microtubule sliding in the neuron, with emphasis on its mechanistic underpinnings as well as its functional significance.

Introduction

There has been a school of thought over the years that the establishment and maintenance of cellular arrays of microtubules can be explained almost entirely by their assembly dynamics. During various phases of research progress, this theory dominated for the mitotic spindle, simple interphase microtubule arrays and even more complex ones such as in polarized epithelial cells or post-mitotic neurons. In the 1980s, another phenomenon called microtubule sliding was posited to occur in the axons of neurons [1]. At the time, no direct evidence was put forth, but the idea was that microtubules move, as polymers, within neurons and that this movement significantly contributes to the establishment, organization and preservation of the microtubule array of the axon. Also termed microtubule transport, the sliding of microtubules was posited to occur against other microtubules or against structures such as the actin cytoskeleton and was posited to be fueled by energy-burning machinery that would later be known as molecular motor proteins [2]. Figure 1 shows a schematic illustration of a hypothetical microtubule array that expands either by microtubule assembly alone (with the microtubules in the array remaining stationary but getting longer), by microtubule sliding alone (with the microtubules not getting any longer but moving their positions), or by both assembly and sliding together.

Figure 1.

Hypothetical microtubule (MT) array expanded by either microtubule assembly alone, microtubule sliding alone or a combination of the two together.

In the wake of the early speculations on this matter, skepticism ensued for decades as to whether microtubules really slide in the axon, with doubts mainly arising due to struggles in visualizing the movements [3–5]. Meanwhile, microtubule sliding was identified as crucial to the formation of the mitotic spindle as well as separation of the half-spindles, and in various other phenomena in different cell types [6–9]. Today, the work has come full circle, with neurons not only providing some of the best evidence for microtubule sliding, but also providing insights into the functions of the sliding as well as details on its mechanistic underpinnings.

The need to slide

Because insufficient synthesis of tubulin (the protein that compromises the microtubule) occurs in the axon and because diffusion from the cell body is incapable of delivering enough tubulin down the length of the axon, there must be an active transport mechanism to convey the tubulin. Early leaders in the axonal transport field posited that the machinery that actively transports tubulin down the axon does so by moving microtubules rather than free tubulin subunits [1, 2]. This was based not only on the logic but also on indirect evidence from radiolabel analyses of the kinetics of newly assembled proteins in the neuron. Very little was known about the transport machinery at the time, and live imaging of microtubule behaviors in living cells was only beginning to become possible. The transport of tubulin, as revealed by biochemical approaches, occurred at rates manyfold slower than later studies would reveal molecular motor proteins to move. Alternative models that posited entirely stationary microtubules [10–13] failed to provide a satisfactory explanation for how tubulin could be actively transported down the axon, if not as microtubule polymers.

Transport of tubulin as polymers also provided an explanation for another important need of the axon, which is to organize the microtubules with a plus-end-out polarity orientation [14]. This could be accomplished by transporting the microtubules with their plus end leading into the axon and anterogradely down its length. While structural factors such as nucleation sites likely also contribute to the neuron’s microtubule polarity patterns [15–17], movement of the polymers as a means of configuring them offers a powerful and flexible mechanism to generate the patterns, preserve them, and re-establish them after insult.

Skepticism about microtubule sliding in neurons, especially in the 1990s, arose in large part from studies conducted mainly on cultured neurons in which fluorescently-tagged tubulin was introduced into the neuron, allowed to assemble into its microtubules, and then bleached in discrete regions of the axon. It was reasoned that if microtubules march down the axon relatively coherently at the slow rate suggested by the radiolabel studies, then the bleached zone should be observed to move down the axon at that rate. No such movement was observed in repeated permutations of the same experiment over many years [3–5]. This was puzzling because several other experimental regimens revealed indirect evidence for microtubule movement [14, 18–22]. A breakthrough was made when the bleached zone was lengthened and the frequency of observation was shortened, such that fluorescent microtubules from the flanking regions could be seen moving through the bleached zone [20, 23, 24]. The movements were rapid, bi-directional, infrequent, and occurred at the rates of known motor proteins, but with most of the microtubules apparently stationary at any given time. These results resolved the controversy in the minds of most researchers in the field, not only revealing the movement but also explaining why the previous attempts failed to reveal movement. In addition, these results also revealed new and surprising characteristics of the movement and launched the next two decades of study on the topic.

One reason why many researchers found models based on microtubule dynamics so appealing is that microtubule dynamics are known to underlie plastic changes in cellular morphology and thereby could potentially explain such neuronal phenomena as growth cone motility and navigation, axonal branch formation, axonal retraction and pruning, and neuronal migration [2]. Microtubule sliding was originally seen as less consistent with these phenomena because it was usually conceptualized to be a very slow persistent march of polymers [1]. The newer observations [20, 23–25] demonstrating that microtubule sliding is rapid, bi-directional, and asynchronous revealed it to be an equally appealing mechanism to underlie these various phenomena, as well as dendritic differentiation. While none of this is to say that microtubule dynamics are not a powerful contributor, there is growing recognition in the field that microtubule sliding is an underappreciated phenomenon that merits greater attention.

Motor proteins are clearly able to move microtubules relative to non-microtubule substrates, as illustrated by work in which the motors are adhered to a glass coverslip. Additional in vitro experiments demonstrated the capacity of molecular motors to drive microtubules to slide against one another either in the case of parallel microtubules (i.e., microtubules of the same orientation) or anti-parallel microtubules (i.e., microtubules of the opposite orientation). A great deal of theoretical and computational work has revealed how such microtubule movements can be driven, depending on the configuration of the motor protein and the relative proportions of motor molecules of each potential configuration associating with the microtubules [26]. Figure 2 illustrates some of these possibilities. Living cells are more complex than in vitro models in that they involve a number of different motor proteins acting simultaneously as well as modifying proteins that may, for example, inhibit microtubule movements. Nevertheless, it is difficult to fathom that neurons, with their long projections, would not take advantage of such a powerful mechanism for conveying and organizing microtubules.

Figure 2.

Theoretical demonstration of how motor-based microtubule (MT) sliding can affect microtubule distribution and organization. Panel A shows how a plus-end-directed motor adhered to less moveable substrate (such as a glass coverslip) causes a microtubule to move with minus-end leading, while a minus-end-directed motor causes a microtubule to move with plus-end leading. Panel B illustrates various hypothetical arrangements of microtubules moving relative to other microtubules via (i) a plus-end directed motor or (ii) a minus-end directed motor. The microtubules are driven with different ends leading, depending on the motor and whether the microtubules are parallel (the same orientation) or anti-parallel (opposite orientations). Movements result from most or all of the motors having their motor domain on the same microtubule of the two. Non-movements result from equal numbers of motors having their motor domain on each of the two microtubules. Elements of this figure were inspired by [26].

Cytoplasmic dynein

To the best of our knowledge the first mention of cytoplasmic dynein as the potential driver of plus-end-out microtubules into and down the axon came in a review article published in 1991, on the basis of this motor having the correct directionality to do so [27]. A formal proposal came a few years later, based on new data on the rate of transport of cytoplasmic dynein itself down the axon, as assessed by the radiolabel approach [28]. The idea was that cytoplasmic dynein could affiliate with the actin cytoskeleton, probably indirectly via the cargo domain of the motor, leaving the motor domain available for microtubules to slide against. Because the motor walks toward the minus end of the microtubule, the microtubule would move with the plus end leading. Subsequent cell biological studies on cultured neurons provided support for the idea, using a variety of methods to deplete or inhibit cytoplasmic dynein [18, 24, 29, 30]. Direct studies to test whether actin filaments are the substrate against which microtubules slide revealed that microtubules in neurons can slide against either actin filaments or other microtubules [20]. The proposed mechanism had great appeal because of its remarkable similarity to events known to occur during mitosis, especially the sliding of astral microtubules against the cortical actin cytoskeleton during separation of the half spindles [31, 32].

In light of this, the question arose as to what motor might transport microtubules retrogradely in the axon and also the purpose of the retrogradely-moving microtubules [33]. After years of potential candidates falling by the wayside, a model emerged based on a phenomenon called “polarity sorting” [30, 34, 35]. The model derives from in vitro analyses in which microtubules are applied onto a glass coverslip coated with a molecular motor protein. As a result, the microtubules move apart, either with plus ends leading or minus ends leading, depending on the motor (figure 3A). Either way, the sorting process results in separate groups of microtubules of uniform polarity orientation. In living cells, to do this, a motor must interface via its cargo domain with a less moveable structure in order for its motor domain to be available in this fashion (figure 2A). In the axon, shorter microtubules are more readily transported than longer microtubules, with the imaging experiments discussed earlier thus far only revealing the rapid and concerted movement of microtubules shorter than 10 microns in length. According to the polarity-sorting model, the anterogradely-moving microtubules are plus-end-out, while the retrogradely-moving microtubules are minus-end-out (figure 3B). The former serve to populate the axon while the latter are cleared back to the cell body so that they do not corrupt the microtubule polarity pattern of the axon. As of yet, there is no direct evidence of the polarity orientation of microtubules as they move within the axon, but this model is consistent with a battery of experimental tests indicating that cytoplasmic dynein is responsible for both the anterograde and retrograde transport of microtubules in the axon and therefore meets the expectations of a polarity-sorting motor [30]. Moreover, when cytoplasmic dynein or its partner proteins are depleted, mutated or inhibited, microtubule polarity flaws arise in the axon, as predicted by a dynein-based polarity sorting model [30, 36–38].

Figure 3.

Schematic illustrations of how a dynein-driven microtubule-polarity sorting mechanism observed experimentally on glass coverslips (panel A) can function in the axon to generate and preserve a nearly uniform plus-end-out pattern of microtubule polarity orientation by selectively transporting minus-end-out microtubules retrogradely into the cell body. Elements of this figure were inspired by [30, 35].

If microtubules longer than 10 microns are entirely stationary in the axon, the polarity-sorting mechanism would be dependent upon microtubule-severing enzymes to break long stationary microtubules into shorter mobile ones. In support of this, the number of short microtubules visualized in transit increases or decreases coordinately with experimental elevation or diminution, respectively, of the microtubule-severing enzyme spastin [39]. However, a mechanism for targeting long minus-end-out microtubules for selective removal from the axon by completely severing them into short microtubules seems hard to fathom. Another possibility is that minus-end-out microtubules are transported effectively back to the cell body, regardless of their length, and that the movement of long microtubules of this orientation has thus far escaped detection with available imaging techniques. Some evidence for the transport of long microtubules (albeit in the anterograde direction) under specialized circumstances comes from observations of individual microtubules in the growth cone [40, 41] as well as ensembles of microtubules in the distal region of the axon during growth [42]. We suspect that whether, when, and where long microtubules can be transported depends on the distribution and properties of specific static crosslinker proteins, as discussed in the next section.

Static Microtubule Crosslinkers

The term microtubule crosslinker is used in different ways in the literature, but generally refers to a protein that either on its own or in the form of a dimer or tetramer or in combination with some other protein binds simultaneously to two different microtubules, thus linking them together structurally. A static crosslinker is a protein that does so via ATP-independent microtubule-binding sites. Static does not mean the crosslinks are non-dynamic because they would have to be relatively transient in order for axonal transport of various cargoes such as mitochondria to move fluidly through the axon [43]. In addition, microtubules in the axon spend most of their time at most places along their length too far apart to be crosslinked, fortifying the view that crosslinks are dynamic - forming and breaking throughout the life of a microtubule. By these criteria, KIF11 (see next section) is a crosslinker but not a static crosslinker because it interacts with two different microtubules via ATP-dependent motor domains, but KIFC1 (see next section) is a static crosslinker because it interacts with two different microtubules via ATP-independent microtubule-binding domains. Simply crosslinking two microtubules does not necessarily immobilize them because the two could move in unison, but crosslinking multiple microtubules, such as those aligned in the axon, would create a superstructure that resists rapid concerted movements (figure 4). In vitro studies suggest that most crosslinkers, at least in purified preparations, have a preference for microtubules of the opposite orientation (i.e., “anti-parallel” microtubules) but can also crosslink microtubules of the same orientation [44, 45]. In theory, immobilizing microtubules of the same orientation would assist microtubule polarity sorting in the axon, while immobilizing microtubules of the opposite orientation would negate polarity sorting.

Figure 4.

Schematic depiction of how static microtubule crosslinkers function in the axon to tamp down potential movements of long microtubules. Shown is a hypothetical dimeric crosslinker with a microtubule-binding domain and a projection domain. Due to its ability to pivot, the crosslinker can crosslink either parallel or anti-parallel microtubules. Short microtubules remain mobile because they are unlikely to have a crosslink immobilize them. Shown also is a severing event (by katanin or spastin) that creates short potentially mobile microtubules by cutting them from the longer immobile ones.

Computational modeling has predicted that a static crosslinker is necessary to explain certain aspects of the data discussed above, especially the fact that longer microtubules resist sliding [30, 34, 35]. The idea is that long microtubules, simply by virtue of being long, have a greater propensity to be immobilized by static crosslinker proteins than short microtubules. For example, if a static crosslinker binds to the microtubule, on average, every 10 microns, a significant portion of the microtubules shorter than 10 microns would be mobile due to the lack of a static crosslinker holding it back. In support of this idea, TRIM46 has been identified as a static crosslinker that when depleted from certain kinds of neurons results in microtubule polarity flaws as well as greater mobility of long microtubules [30, 44]. If TRIM46 specifically crosslinks microtubules of common orientation (and some data suggest that it does [44]), long minus-end-out microtubules would not be immobilized and hence would be cleared from the axon in the same fashion as short minus-end-out microtubules. As noted above, the fact that this has not yet been visualized may be due to limitations in imaging that still persist to this day. In the absence of static crosslinkers, the computational modeling suggests that microtubule movements would become less orderly, leading to polarity flaws in the axon. This is because another important function of static crosslinkers is to tamp down the ability of motors other than cytoplasmic dynein to slide the microtubules. Providing precedent for this idea, when an isoform of MAP4 called oMAP4 is depleted from muscle cells, aberrant sliding of microtubules by Kinesin-1 occurs, which leads to microtubule disorganization (i.e., microtubules out of proper alignment) [46].

oMAP4 is a newly discovered splice variant of MAP4. MAP4 is best known as the tau/MAP2 family member present in nonneuronal cells, but MAP4 (including the oMAP4 isoform) is also present (albeit poorly studied) in neurons, where it was originally known as MAP3 [47–49]. The ubiquitous form of MAP4 is termed uMAP4. oMAP4 contains a unique 48 kD projection domain spliced out of uMAP4. It is through that unique projection domain that oMAP4 behaves as a static microtubule crosslinker. Some scientists refer to uMAP4 as well as tau and MAP2 as crosslinkers but caution is due because none of these proteins has the unique domain of oMAP4 that permits it to act in this fashion. Depletion of tau can cause axonal microtubules to drift apart [50] but this may be due to electrostatic attractions as opposed to a formal crosslinking of neighboring microtubules. Depletion of tau from axons does not cause microtubule polarity flaws [39] but does cause an increase in the numbers of short microtubules in transit [39]. The latter appears to be due to tau’s role in suppressing access of the microtubule to microtubule-severing enzymes, as opposed to diminished crosslinking [51, 52].

TRIM46 is famously enriched in the axon initial segment (AIS) of mature axons that have a formal AIS [53], but TRIM46 is also present along the length of the axon, at least in immature axons without a formal AIS [30]. When TRIM46 is experimentally depleted from relatively immature cultured neurons, the appearance of microtubule polarity flaws is accompanied by an increase in the mobility of long microtubules [30]. However, the increase in long microtubule mobility is not as dramatic as the computational modeling predicts, suggesting that other static crosslinkers participate in addition to this one. Relevant to this point, the mechanisms that determine the axonal microtubule pattern cannot be housed exclusively in the AIS, given that the pattern can be re-established in segments of the axon that are physically cut well beyond the proximal region of the axon [54].

Mitotic kinesins in post-mitotic neurons

For several years, while neuroscientists were debating the existence of microtubule “transport machinery” in the axon, mitosis researchers were already identifying specialized kinesins that had the demonstrated ability to slide microtubules and/or regulate the sliding of microtubules. It was presumed that these kinesins, which had names at the time such as CHO1/MKLP1 and Eg5, were mitosis-specific and had no business being expressed in terminally post-mitotic cells such as neurons. Musings in the early 1990s were that these mitotic motors provided proof-of-principle that microtubule-sliding motors existed in other cells and hence that similar motors might well exist in neurons [31]. Then, in a significant breakthrough, it was discovered that neurons express the very same motors that were previously believed to be mitosis-specific [25, 32, 55–66]. Moreover, powerful phenotypes in neurons resulted from depleting these motors or inhibiting them with drugs. For example, depletion or inhibition of Eg5 (now known as KIF11 in vertebrates, the only member of the kinesin-5 sub-family of kinesins in vertebrates) or KIF15 (the only member of the kinesin-12 sub-family of kinesins in vertebrates) caused axons to grow faster and lose their ability to navigate [60–62]. Depletion of CHO1/MKLP1 (now called KIF23, one member of the kinesin-6 sub-family of kinesins in vertebrates) or KIF15 caused dendrites to lose the minus-end-out microtubules needed for their mixed pattern of microtubule polarity orientation and concomitantly take on axonal characteristics [55, 63–65]. In Drosophila, depletion of the kinesin-6 subfamily member known as Pavarotti yields phenotypes similar to those observed in vertebrate neurons [67], as does depletion from vertebrate neurons of KIF20b, another member of the kinesin-6 subfamily [57].

KIF11 and KIF15 were shown to act as brakes on microtubule sliding in the axon [25, 60], while KIF23 (and Pavarotti in Drosophila) was shown to act in the cell body as a brake against microtubules sliding into the axon [65]. KIF23 and KIF15 were shown to promote the sliding of minus-end-out microtubules into dendrites but not axons [65]. Most of these effects were upon microtubule-microtubule sliding, but KIF15 was also shown to impact microtubule-actin sliding in the growth cone [25]. No evidence has been reported that inhibition of these mitotic motors results in microtubule polarity flaws in the axon, which accentuates the fact that a brake on sliding is not the same as a crosslinker. True crosslinkers are static linkages, whereas these “power” brakes depend on the energy-burning properties of the motors. The brakes presumably function by creating microtubule movements that are so slow that they are tantamount to non-movements and this prevents other motors from moving the microtubules any faster. The situation may be more complex, however, in that KIF15 has been shown in vivo to be able to drive microtubules to slide along one another in a manner that leads to the formation of parallel microtubule arrays [68], so there is wisdom in keeping in mind that each of these molecules has its own unique properties.

Only recently has KIFC1 (also called kinesin-14a or HSET in humans) been studied in neurons, despite the fact that it is the most highly expressed of any “mitotic” kinesin known to slide microtubules. The kinesin-14 subfamily of kinesins is different from the others in that it has the dynein-like property of walking toward the minus end of the microtubule (rather than the plus end, like the other kinesins). In non-neuronal contexts, KIFC1 can slide apart oppositely-oriented microtubules and slide together commonly-oriented microtubules by drawing microtubules together as it moves toward their minus ends [69]. Especially interesting about KIFC1 is the fact that it can crosslink microtubules via ATP-independent microtubule-interacting sites (one in the tail domain and one in the motor domain), and thereby prevent the sliding of the microtubules by other motor proteins [69, 70]. Indeed, the effects of experimental KIFC1 inhibition or depletion show many of the hallmarks of crosslinker inhibition, with greater mobility of long microtubules and the appearance of microtubule polarity flaws [58]. A related protein, KIFC3, cannot crosslink microtubules directly but does so indirectly through interaction with minus-end-capping proteins, specifically in dendrites [59].

As for phenotype, inhibition or depletion of KIFC1 causes the collapse/retraction of the axon [58], which is the opposite of the phenotype of inhibition or depletion of the “brake” kinesins [25, 60]. This may speak directly to the difference between a brake and a static crosslinker but might also have to do with the motor properties of KIFC1 because inhibition of TRIM46 does not cause the axon to collapse [30, 44, 58]. In addition, the mitotic kinesins can lend additional signaling-responsive control over the sliding behavior of microtubules underlying growth cone guidance [71]. This is apparent in the case of KIF11, which regulates the asymmetric dynein-based invasion of microtubules into one side of the growth cone or the other [61, 62] and in the case of KIF15, which integrates microtubules with actin filaments during growth cone guidance [25]. By contrast, inhibition or depletion of KIFC1 results in a looped bundle of microtubules in the growth cone [58], somewhat similar to a paused/stalled growth cones occasionally observed in cultured neurons [72]. All of this is a reminder that microtubule-related proteins are complicated and any one of them may simultaneously have properties of sliders, brakes, microtubule-microtubule crosslinkers, and microtubule-actin crosslinkers, as well as influence still other properties of microtubules such as their assembly dynamics [58].

Kinesin-1

Sage advice in the motor field is to never underestimate the capacity of any motor to do just about anything, but in general for a motor to be able to slide microtubules relative to one another it would most likely rely on ATP-independent binding sites that can bind to one microtubule while its cargo domain interacts with the other one. This can also be accomplished via accessory proteins, but ATP-independent binding sites are usually a strong indicator that a motor can slide microtubules. Kinesin-1 (KIF5, conventional kinesin), the predominant plus-end-directed motor for transporting membranous organelles, has recently been shown to have an ATP-independent microtubule-binding site that enables it to act in this fashion [8]. A variety of studies in different cell types indicate that Kinesin-1 slides microtubules or is at least capable of doing so under certain conditions [8, 46, 67, 73–78]. The pioneering work on this was done by Gelfand, who has focused on anti-parallel sliding of microtubules by Kinesin-1. However, given that axonal microtubules are predominantly parallel, it seems reasonable that the most profound role for Kinesin-1-based sliding in the axon may be to influence sliding of parallel microtubules (see figure 2B). An exciting possibility is that MAP7, a microtubule-associated protein that activates Kinesin-1 and draws it to the microtubule [79, 80] may promote Kinesin-1’s capacity to slide parallel microtubules.

Studies on Drosophila neurons suggest that the sliding of microtubules by Kinesin-1 in the cell body may contribute to the formation of axons [77, 81] and at the cut tip of the axon may contribute to its regeneration after severing [75], while studies in C. elegans suggests that Kinesin-1-based sliding of microtubules may contribute to formation of the dendritic microtubule array [78, 82]. At present, there is no information on a potential role for Kinesin-1 in regulating microtubule sliding in vertebrate neurons. Live-imaging of microtubule transport in the axon suggests that an opposing motor tugs on the microtubules in the opposite direction of cytoplasmic dynein, and one possibility is that this opposing motor is Kinesin-1 [30, 34, 35]. If it is, Kinesin-1-based sliding of microtubule may be responsible for the polarity flaws observed in the axon when cytoplasmic dynein, its partner proteins or crosslinker proteins are depleted or inhibited. Such a scenario is reminiscent of findings in vertebrate muscle cells wherein microtubule sliding was shown to occur via cytoplasmic dynein but could occur via Kinesin-1 when cytoplasmic dynein (or the crosslinker oMAP4, see earlier) is depleted [46]. In a study on axons and dendrites of C. elegans, a great deal of Kinesin-1-driven microtubule sliding was reported when microtubule linkages to the membrane were disrupted, and the resulting sliding caused significant microtubule relocations [78]. Two obvious questions arise, the first of which is how this sliding occurs in the case of microtubules (such as those that dominate in the axon) that are mainly of the same orientation rather than of opposite orientation. The second question is what purpose the Kinesin-1-based sliding of microtubules serves if it must be suppressed to avoid problems. It may be, as suggested by the invertebrate work, that certain events in the neuron, such as growth-related challenges at the axon’s tip or dendritic differentiation, benefit from the forces of microtubule sliding by Kinesin-1 but only when allowed to happen in a tightly controlled fashion. As mentioned earlier, parallel microtubule sliding by Kinesin-1 might be promoted by MAP7.

Schematic summary of regulation of microtubule sliding in the axon

The upper portion of Figure 5 is a schematic summary of our current thinking on how microtubule sliding is regulated by molecular motors and crosslinkers in the axon in a modified dynein-based polarity-sorting model. The table below the schematic lists relevant players as we currently understand them, their properties, their roles, and the experimental evidence. At present, because there is no experimental evidence on Kinesin-1’s role in sliding microtubules in vertebrate axons, our speculations are derived from the work on invertebrates discussed above.

Figure 5.

Mechanistic model for various factors contributing to the motor-driven polarity sorting of microtubules in the axon. The upper part of the figure schematically illustrates how the various factors shown in the lower part of the figure contribute to generate or regulate microtubule movements in the axon. The lower part of the figure shows in the form of a table the relevant players, their properties, their role, and the experimental evidence supporting the model. Elements of this figure were inspired by [30, 35]. Kinesin-6 motors are not included in the schematic because they act at the level of the cell body, not within the axon itself.

Questions and challenges for the future

Most contemporary work on microtubule sliding in neurons is being conducted either in rodent neurons in culture or Drosophila, with common themes across disparate species. There are differences, however, in what individual motor proteins (as well as crosslinker proteins) seem to do, and this may also apply to different kinds of neurons in the same animal. In addition to providing a transport mechanism for tubulin and organizing microtubules relative to their polarity, microtubule sliding plays critical roles in axonal and dendritic branch formation, growth cone motility and guidance, axonal and dendritic pruning, dendritic spines, and synapses [21, 71, 83, 84]. Emerging work suggests that microtubule sliding is also key to regulating the trajectory of neuronal migration [83, 85] . Interestingly, there appears to be an enormous capacity for microtubule sliding in the neuron that is mostly suppressed but can be activated when and where needed. That being the case, a significant goal for future studies will be to determine how the sliding of microtubules is controlled by signaling pathways responsive to moment-by-moment challenges in the life of the neuron [86, 87]. Also important is the need to develop new and better ways to image the sliding of microtubules, which remains a technical struggle given the three-dimensional shape of the neuron and high density of microtubules in most of its compartments. Finally, more needs to be done on the sliding and crosslinking of microtubules with actin filaments [88, 89], which is of significant importance to many aspects of neuronal cell biology.

Highlights.

• Microtubule sliding (also called microtubule transport) is an unappreciated mechanism that contributes to the establishment, organization and preservation of microtubule arrays in many types of cells.

• Microtubule sliding in the axon, once controversial, is now supported by a great deal of evidence, including new findings on its underlying mechanisms.

• Microtubule sliding is necessary to supply the microtubule mass needed for growth and upkeep of the axon, as well as for the establishment and preservation of the nearly uniform polarity orientation of the axonal microtubule array.

• Microtubule sliding underlies a process known as polarity sorting, in which plus-end-out microtubules are transported anterogradely down the axon to build and maintain the axonal microtubule array, while minus-end-out microtubules are transported back to the cell body to ensure that they do not accumulate in the axon.

• Microtubule-sliding forces generated by cytoplasmic dynein are theoretically sufficient for the polarity sorting of axonal microtubules, but complex challenges in the life of the axon require the participation of other players such as static crosslinking proteins and molecular brakes that regulate the capacity of microtubules to slide.

• Microtubule sliding is also important for differentiation of dendrites, neuronal migration, branching of axons and dendrites, and growth cone motility and navigation.