Contributions of Microtubule Dynamics and Transport to Presynaptic and Postsynaptic Functions

Abstract

Microtubules (MT) are elongated, tubular, cytoskeletal structures formed from polymerization of tubulin dimers. They undergo continuous cycles of polymerization and depolymerization, primarily at their plus ends, termed dynamic instability. Although this is an intrinsic property of MTs, there are a myriad of MT-associated proteins that function in regulating MT dynamic instability and other dynamic processes that shape the MT array. Additionally, MTs assemble into long, semi-rigid structures which act as substrates for long-range, motor-driven transport of many different types of cargoes throughout the cell. Both MT dynamics and motor-based transport play important roles in the function of every known type of cell. Within the last fifteen years many groups have shown that MT dynamics and transport play ever-increasing roles in the neuronal function of mature neurons. Not only are neurons highly polarized cells, but they also connect with one another through synapses to form complex networks. Here we will focus on exciting studies that have illuminated how MTs function both pre-synaptically in axonal boutons and post-synaptically in dendritic spines. It is becoming clear that MT dynamics and transport both serve important functions in synaptic plasticity. Thus, it is not surprising that disruption of MTs, either through hyperstabilization or destabilization, has profound consequences for learning and memory. Together, the studies described here suggest that MT dynamics and transport play key roles in synaptic function and when disrupted result in compromised learning and memory.

Introduction

The synapse has long been known to be a primary point of rapid communication between presynaptic axons and postsynaptic dendrites. Both presynaptic and postsynaptic sites are specializations in the axon and dendrite, respectively, composed of distinct membrane receptors, adhesion proteins, scaffolding proteins, signaling molecules and cytoskeletal components (Krueger-Burg et al., 2017; Sheng and Hoogenraad, 2007). Broadly, synapses are divided into inhibitory and excitatory types. Within the central nervous system (CNS), excitatory synapses are generally associated with dendritic spines, small mushroom-shaped protuberances arrayed along dendrites, while inhibitory synapses occur on dendrite shafts and spine necks (Bucher et al., 2020). In this review we will concentrate on the cytoskeletal contributions to both presynaptic and postsynaptic compartments of central nervous system excitatory neurons, such as the pyramidal cells in the cortex and the principal neurons in the hippocampus. Specifically, this review will focus on the dynamic microtubule (MT) cytoskeleton and the emerging role it plays in synaptic function.

Both presynaptic and postsynaptic specializations require cytoskeletal components to function properly, including actin filaments (f-actin), MTs, intermediate filaments and septin filaments. Although we will focus on MTs in this review it is worthwhile introducing actin filaments as well. Indeed, the most heavily concentrated cytoskeletal polymer in postsynaptic dendritic spines and presynaptic axonal boutons are actin filaments. Actin filaments are comprised of globular actin subunits (g-actin) that polymerize into a helical structure approximately 7 nm in diameter, which can elongate several microns in length. Neurons contain a myriad of actin-associated proteins that affect actin filament geometry, stability and motility (Dent et al., 2011). Actin filaments are generally bundled together in either parallel or antiparallel arrays or are part of a branched actin network. These forms of actin networks play distinct roles in dendritic spines and axonal boutons. We will touch on some of these actin-associated proteins and how they contribute to f-actin/MT associations at synapses.

Actin forms filaments from actin monomers, while MTs polymerize from tubulin dimers, composed of an alpha and a beta subunit. These alpha-beta tubulin dimers associate end-to-end forming protofilaments, which themselves associate with one another to form a tube composed of 13–16 protofilaments that is approximately 25nm in diameter (Roll-Mecak, 2020). Within most cells, including neurons, MT can elongate to over one hundred microns in length (Brown et al., 1993). Like actin filaments, MTs can also bundle into parallel and antiparallel arrays and branch in coordination with MT-associated proteins. Both actin filaments and MTs are polar polymers, with each end having different polymerization and depolymerization kinetics. However, unlike actin filaments, which generally polymerize from one end (the plus or barbed end) and depolymerize from the other end (minus or pointed end), MTs undergo a process termed dynamic instability (Mitchison and Kirschner, 1984). Dynamic instability is a stochastic process by which the plus end of MTs undergoes cycles of growth and shrinkage. The transition of growth to shrinkage is termed catastrophe, while the transition from shrinkage to growth is termed rescue. Dynamic instability allows MTs to quickly reorganize and efficiently explore all regions of a cell, which is particularly important for neurons, because of their extremely polarized morphology (Figure 1).

Figure 1. Orientation and composition of microtubules in dendrites and axons.

A. Image of a pyramidal neuron (orange), showing apical and basal dendrites, which are studded with spines. The axon is shown below the cell body (orange blending into green) with axonal boutons, which form en passant synapses with other neurons. B. A magnified region of the dendrite showing the distribution and composition of MTs. In the dendrite, the bulk of MTs form two populations. One population is the dynamic, plus-end-out MTs (yellow with a small purple stable region at the minus end) that are concentrated close to the plasma membrane. A second population are the stable, minus-end-out MTs (purple, with a small yellow dynamic region at the plus end) that are positioned centrally in the dendrite. In the box below, the dendrite (orange) is shown as a slice with the two populations of MTs shown “end on” as purple (stable) and yellow (dynamic) dots However, it is likely that there is a smaller population of MTs that maintain the opposite polarity (dynamic plus-end-in and stable plus-end-out). C. A magnified region of the axon and expanded synaptic bouton. In the axon MTs contain both an acetylated/detyrosinated stable region and a tyrosinated dynamic region. At present it is unclear if there are distinct populations, like those in the dendrite. Note that at the axonal bouton there is a concentration of both plus and minus ends.

The polarized structure of neurons also suggests they may require more decentralized nucleation and growth of MTs. MTs nucleate from tubulin dimers, forming a ringlike structure that continues to polymerize to form a tube, from the centrosome through the actions of the gamma tubulin ring complex (γTURC) and remain attached to this organelle (Akhmanova and Kapitein, 2022). The centrosome is positioned in the cell body, but the length of dendritic and axonal processes can be hundreds or thousands of times the diameter of a cell body. Electron microscopy studies of neurons have shown that most MTs are not attached to the centrosome (Baas and Black, 1990) and individual MTs do not extend the length of dendrites or axons (Baas et al., 1989). In neurons, although MTs polymerize from the centrosome, they are severed and can be transported throughout the neuron (Rao and Baas, 2018). Because the minus ends of MTs in neurons are generally not associated with the centrosome they can stabilized by other proteins. These minus-end binding proteins (-TIPs) include microtubule associated proteins (MAPs) such as the CAMSAP/Patronin family, end binding proteins (EBs), severing proteins, such as katanin, the minus-end motor cytoplasmic dynein and γ-TURCs not associated with the centrosome (Akhmanova and Steinmetz, 2019; Kuo and Howard, 2021). These proteins regulate the dynamics of the minus ends of MTs and thus play important roles in the overall MT dynamics in neurons. These -TIPs have not been studied as extensively as plus end binding proteins (+TIPs), especially in neurons, and deserve further study.

Several studies have demonstrated that MTs can nucleate and polymerize from other cellular structures in axons and dendrites. In dendrites, MTs can nucleate at Golgi outposts (Horton and Ehlers, 2003; Horton et al., 2005; Ori-McKenney et al., 2012) and a recent study in C. elegans demonstrated local MT nucleation in dendritic growth cones in association with Rab11-containing endosomes (Liang et al., 2020). Nucleation and polymerization of MTs from these endosomes were shown to be a source of minus-end-out MTs in dendrites. The recent discovery of MT nucleation from existing MTs in neurons, through the activity the HAUS/augmin complex and γ-tubulin, provides yet another structure onto which new MT nucleation and growth can occur in both axons and dendrites (Cunha-Ferreira et al., 2018; Sanchez-Huertas et al., 2016). Such growth of new MTs from existing MTs results in MT branching that is synonymous with the well-known actin nucleation and growth by the Arp2/3 complex, which results in a branched actin network (Pollard, 2007). Thus, MTs can nucleate throughout the neuronal cytoplasm and undergo cycles of growth and shrinkage, making for a very dynamic MT network.

Perhaps, more than any other advance, the discovery and imaging of MT plus-end binding proteins allowed for the appreciation of MT dynamics in neurons (Stepanova et al., 2003). Although many different types of proteins can associate with the growing ends of MTs (Akhmanova and Steinmetz, 2008) a canonical member of this +TIP group of proteins is the family of end-binding proteins (EB1–3) (Akhmanova and Steinmetz, 2015). When EB3-EGFP was first visualized in neurons it was surprising to see how dynamic MTs were in the thin processes of axons and dendrites (Stepanova et al., 2003). EB3 cycles on and off the growing ends of MTs, creating a comet-like structure that moves through the cytoplasm at approximately 5–15um/min (Dixit et al., 2009; Dragestein et al., 2008). Previous studies of MTs in neurons required injection of labeled tubulin (Dent et al., 1999; Dent and Kalil, 2001) or expression of GFP-tubulin (Kimura et al., 2005), which is incorporated throughout the MT polymer. Since MTs are packed together in dendrites and axons, MTs labeled with tubulin resulted in a continuously fluorescent axon and dendrite. Only where processes flattened, such as at the neuronal growth cone or at axon branch points, could individual MT dynamics be discerned (Dent et al., 1999; Dent and Kalil, 2001). However, upon imaging EB3-EGFP it was obvious that MTs were dynamic throughout axons and dendrites. Further work showed this was indeed the case in cultured neurons, in hippocampal slices and in vivo (Yau et al., 2016). Surprisingly, even in mature neurons (63DIV) MTs remained dynamic, polymerizing and depolymerizing throughout dendrites (Hu et al., 2008). Thus, imaging EB3 comets has shown that MTs continue to polymerize in well established, stably polarized neurons throughout their lifetimes. Interestingly, EML2-S was recently discovered to concentrate near the ends of depolymerizing MTs (Hotta et al., 2022). Imaging this protein, in conjunction with EB3, may allow the construction of a more complete picture of MT dynamics in neurons.

In addition to MT dynamics, MTs serve the important function of acting as a substrate for long-distance transport of material throughout neurons. The polarity of MTs is detected by motor proteins that either move toward the plus end (kinesins) or the minus end (cytoplasmic dynein) of MTs (Cason and Holzbaur, 2022). There are many different kinesin proteins in neurons, but one major form of cytoplasmic dynein. Each type of kinesin binds a specific set of adapter proteins, which results in the transport of specific cargo, including vesicles, proteins and mRNA, throughout the neuron (Cason and Holzbaur, 2022). Kinesin and dynein motors are also capable of transporting MTs themselves within neurons (Del Castillo et al., 2019; Rao and Baas, 2018). If motor proteins are bound, through their cargo-binding domain, to stable structures, such as the plasma membrane, submembranous actin cytoskeleton or to other less compliant MTs, they can continue to motor along MTs. This results in kinesins translocating MTs minus-end-leading, while cytoplasmic dynein translocates MTs with their plus ends leading. Therefore, the motor-based transport of MT polymers in neurons provides a further mechanism by which MTs remain dynamic in neurons. Together, MT polymerization/depolymerization and motor-based transport function in coordination to translocate material throughout axonal and dendritic processes.

Microtubule dynamics and transport in presynaptic axons and boutons

Axons are long, thin processes extending from the cell body of a neuron that branch extensively in the neuropil. CNS axons contact many other neurons through axonal swellings termed axonal boutons (Figure 1). Axons are composed of bundles of MTs extending along the length of the axon. However, each MT is shorter than the axon but overlaps proximally and distally with other MTs, creating a continuous array of MTs throughout the axon and all branches (Baas et al., 1988). Due to their polarity, MTs are not uniform structures, rather they contain both dynamic and stable regions. To mark these stable and dynamic regions researchers have exploited the fact that the α-tubulin subunit in the tubulin dimer is tyrosinated at the C-terminal end of the protein. Newly polymerized MTs are therefore described at tyrosinated and can be identified as such by antibodies that specifically recognize this posttranslational modification of tubulin. Over time, MTs become both acetylated, at separate residues (K40 being the most studied), and detyrosinated at the C-terminus. Thus, older, presumably more stable MTs can be labeled with antibodies that recognize acetylated or detyrosinated tubulin (Figure 1C). Interestingly, when labeled immunocytochemically with these antibodies there is a fairly sharp demarcation between acetylated and tyrosinated regions of MTs in neurons (Brown et al., 1993). Almost all MTs in axons are oriented in a plus-end-out fashion (Yau et al., 2016). This parallel array of MTs has implications for motor-driven transport of cargo. In axons, anterograde transport of cargo is exclusively the domain of plus-end-directed kinesins, while retrograde transport is carried out by cytoplasmic dynein. Two important cargos that are transported throughout the axon and are related to synapses are synaptic vesicle precursors (SVPs) and dense core vesicles (DCVs).

Early studies employing electron microscopy suggested that synaptic vesicles were associated with MTs at presynaptic sites (Bird, 1976; Gordon-Weeks et al., 1982; Gray, 1975). Moreover, several studies have identified tubulin in presynaptic fractions (see references in (Waites et al., 2021)), while others have shown MTs play an important role at the Drosophila neuromuscular junction (see references in (Dent, 2020)). However, relatively little research has focused on the connection of MTs and vesicles at the synapse in mammalian neurons, until recently. Several contemporary studies have shown that MTs are likely critical for the delivery of both SVPs and DCVs to presynaptic compartments at axonal boutons (Figure 2). One study showed that DCVs containing synaptotagmin-IV (Syt-4) are trafficked by the Kinesin-3 motor KIF1A in axons (Bharat et al., 2017). As these vesicles transit down the axon, increases in the activity of the protein kinase JNK near synapses phosphorylates Syt-4, releasing the DCVs from MTs. After releasing from MTs, the DCVs are captured by presynaptic actin filaments. Although MT dynamics were not analyzed directly in this study, it is likely that DCV release occurs at MT plus ends, as has been shown in the postsynaptic spine (see below) (McVicker et al., 2016).

Figure 2. Microtubules assist in concentrating synaptic vesicle precursors and dense core vesicles at synaptic sites.

An axon and synaptic bouton are shown above, with the outline of a dendritic spine shown below. There are three different mechanisms for concentrating vesicles at the synapse. A. Dense core vesicles (DCVs) are transported anterogradely along MTs by the motor protein KIF1A. When a DCV nears a synapse JNK phosphorylates synaptotagmin-IV (Syt-4) cargo, which releases the DCV from the KIF1A motor. B. Synaptic vesicle precursors (SVPs) are transported anterogradely along MTs by KIF1A, which has a decreased affinity for the GTP-tubulin cap at the plus ends of MTs. The decreased affinity releases it from the MT and it is captured at the actin-rich synapse. These GTP-tubulin-rich MT plus ends are concentrated in the bouton. C. Dynein transports SVPs retrogradely toward the minus end of MTs, which are also concentrated at the synapse. When dynein reaches the minus end, it pauses due to the presence of the immobile kinesin VAB-8/KIF26 and releases the SVP at the synapse. Thus, both anterograde and retrograde transport of SVPs can contribute to accumulation of SVPs at synapses. P – proximal axon, D – distal axon.

Two studies discovered complementary mechanisms by which MTs contribute to the transport and delivery of SVPs to synaptic boutons. One study, by the Holzbaur group, showed that in mature axons MT polymerization oftentimes originates and terminates in presynaptic terminals at en passant synapses of cultured hippocampal neurons (Guedes-Dias et al., 2019). Disruption of MT dynamics by pharmacological means resulted in decreased delivery of SVPs to synapses. Interestingly, like DCVs mentioned above, the SVPs used KIF1A for transport in the axon. The authors discovered that KIF1A had a decreased affinity for the “GTP cap” that exists at the plus ends of dynamic MTs. The decreased affinity of KIF1A as it reached the plus ends of MTs and data showing MT plus ends are often associated with synapses suggest that this is a mechanism for concentrating SVPs specifically at synapses. A second study showed that MT nucleation preferentially occurs at presynaptic sites in mature hippocampal neurons in culture and in slice preparations (Qu et al., 2019). This presynaptic nucleation was activity-dependent and required γ-tubulin and the HAUS/augmin complex, indicating that new MT nucleation and potentially MT branching (Cunha-Ferreira et al., 2018; Sanchez-Huertas et al., 2016) is important for delivery of SVPs to en passant synapses. These newly nucleated MTs served as conduits for transport of SVPs between boutons. Most significantly, loss of these nucleated MTs resulted in impaired high-frequency neurotransmitter release (Qu et al., 2019). Together, these results suggest that MT nucleation, dynamic instability and motor driven transport play fundamental roles in concentrating DCVs and SVPs to axonal en passant synapses.

Both studies mentioned above suggest that SVP anterograde movement along plus-end-out MTs and their unloading at en passant synapses are one mechanism to concentrate SVPs at synapses. A very recent study has shown that retrograde transport of SVPs via dynein can occur at the minus end of MTs in C. elegans (Balseiro-Gomez et al., 2022). In this study, the authors showed that VAB-8/KIF26, an immotile kinesin, localizes to MT minus ends and promotes dynein pausing to facilitate SVP delivery to synapses (Figure 2). Interestingly, the levels of VAB-8/KIF26 at MT minus ends are determined by neurexin and frizzled signaling. In combination with the studies mentioned above, it appears that SVP accumulation at presynaptic sites can occur through both anterograde and retrograde transport, with offloading occurring at either the plus or minus ends of MTs, respectively. Together, these studies suggest that both plus and minus end positioning of MTs within the axon is critical for SVP targeting to synapses. However, these studies do not preclude the likelihood that presynaptic actin and myosin motor proteins also play important roles in trafficking recycling synaptic vesicles within the axon (Chenouard et al., 2020; Gramlich and Klyachko, 2017). Nevertheless, MTs appear as important players in both the long-distance transport and specific targeting of SVPs and DCVs to presynaptic specializations.

Microtubule dynamics in postsynaptic dendrites and spines

Dendrites are generally shorter and thicker processes, compared to axons, and in excitatory neurons their dendritic array is studded by small mushroom-shaped growths termed spines. Importantly, the MT array of dendrites is constructed in a different way than axons (Figure 1). Unlike axons, where the MTs are entirely plus-end-out, in mature dendrites MTs are arranged in an antiparallel fashion, with approximately equal numbers of plus-end-out and minus-end-out MTs (Baas et al., 1988; Stepanova et al., 2003; Yau et al., 2016). However, in Drosophila and C. elegans dendrites contain mostly minus-end-out MTs (Goodwin et al., 2012; Stone et al., 2008). It is unclear why this difference exists, but it clearly has implications for both MT dynamics and transport. Because of the different arrangement of MTs in axons and dendrites plus-end-directed kinesins only move cargo anterogradely in axons, while they can move both anterogradely and retrogradely in dendrites. Similarly, dynein moves only retrogradely in axons but can move in both directions in dendrites.

In addition to their orientation, MT composition has been a focus of study. Many classic studies, primarily from the Baas and Black groups, have shown that MTs in both axons and dendrites have a composite structure, with a stable region toward the minus end of the MT and a dynamic region toward the plus end (Baas and Black, 1990; Baas et al., 1991; Brown et al., 1993; Li and Black, 1996; Rochlin et al., 1996; Wang et al., 1996). However, it should be noted that these studies used cultured sympathetic neurons from the superior cervical ganglia and analyzed MTs via immunocytochemistry (ICC) and electron microscopy (EM). Over the last 30 years this model of composite MTs has generally dominated the literature. Moreover, it was generally assumed that this same composite model of MTs occurred in other types of neurons, such as hippocampal and cortical neurons.

However, recent work by the Kapitein group, studying MTs in dendrites of cultured hippocampal neurons, has suggested that this model may need some revision (Katrukha et al., 2021; Tas et al., 2017). In two recent studies this group used super resolution microscopy, coupled with a motor-PAINT technique, as well as ICC and expansion microscopy to interrogate the structure of MTs in dendrites. Implementing these techniques, they showed that minus-end-out MTs are concentrated in the central region of the dendrite, are acetylated, and are more stable than the dynamic, tyrosinated MTs that are localized around this central MT core, near the plasma membrane (Katrukha et al., 2021; Tas et al., 2017). The model based on the above studies from the 1990’s posits a random distribution of composite MTs, each containing acetylated and tyrosinated regions. These new data suggest a new model whereby MTs in hippocampal dendrites are primarily either acetylated or tyrosinated and organized in a center-acetylated/peripheral-tyrosinated fashion (Figure 1). However, the Kapitein group was not able to follow individual MTs in their entirety, even with expansion microscopy and super-resolution imaging. Thus, they could not exclude the possibility some or most of the MTs in mature hippocampal neurons are composite in nature. Moreover, such a model where minus-end-out MTs are entirely stable would predict few, if any, retrograde EB3 comets in dendrites. However, it has been demonstrated clearly in several studies, using EB3 imaging of growing MTs in dendrites, that retrograde EB3 comets originate throughout the dendritic tree of hippocampal neurons (Hu et al., 2008; Stepanova et al., 2003; Yau et al., 2016). Moreover, if all the stable MTs are minus-end-out the kinesins that favor stable MTs would only be able to transport material in the retrograde direction. Thus, a combination of these models is likely to be a more accurate representation of MT organization in mature neurons. In such a model acetylated, stable MTs are concentrated in the central region of the dendrite but are also capable of polymerizing in the retrograde direction and thus contain a composite structure, tipped by a region of tyrosinated tubulin (Figure 1). Additionally, the plus-end-out MTs are likely to be mostly tyrosinated and more dynamic than the centrally located MTs, but may contain a shorter, stable, acetylated region toward their minus end. Moreover, there are likely small populations of dynamic minus-end-out MTs and stable plus-end-out MTs in dendrites as well. This model of MT composition and organization in dendrites has clear implications for both MT dynamics and movement of motors along MTs in dendrites and spines.

MT dynamics in dendrites and spines will be covered first. We will then touch upon motor-driven transport and try to integrate these two processes. Although dendrites and axons must maintain their integrity, oftentimes for the life of the organism, MT turnover is constantly occurring through dynamic instability and severing. Since MTs can explore the entire cell through cycles of polymerization and depolymerization, it suggests they would be able to explore all regions of a neuron including dendrites and dendritic spines. Curiously, standard preparatory and EM imaging techniques result in micrographs where MTs are confined to the dendritic shaft, with no MTs entering dendritic spines (Harris et al., 2022). However, by using unusual preparatory techniques (see Dent, 2017 for more detail) George Gray showed that MTs enter dendritic spines in the cerebral cortex and can extend into the post synaptic density (Gray et al., 1982; Westrum et al., 1983; Westrum et al., 1980). Were these studies just artifacts of the preparation? To answer this question several labs used fluorescently labeled tubulin and EB proteins to determine the dynamics of MTs in dendrites of mature hippocampal and cortical neurons in culture. Consistently, four independent groups showed that MTs continued to polymerize in dendrites over the lifetime of the cultures and explored the entire neuronal cytoplasm, including dendritic spines (Gu et al., 2008; Hu et al., 2008; Jaworski et al., 2009; Mitsuyama et al., 2008). Further studies by these and other labs have confirmed that MTs polymerize into and out of dendritic spines.

MT entry into dendritic spines does not appear to be a random process. Although dynamic instability is a stochastic process (Mitchison and Kirschner, 1984), suggesting MTs polymerize and depolymerize in a random manner, MT entry into dendritic spines appears to have some specificity. One interesting association of MT entry of spines is that it is activity dependent. MTs appear to target dendritic spines more often in synaptically active neurons, compared to less active neurons (Hu et al., 2008). Moreover, both BDNF application and chemical long-term potentiation (cLTP) enhance MT invasions of dendritic spines (Hu et al., 2011; Merriam et al., 2011; Merriam et al., 2013), while chemically induced LTD results in fewer invasions (Kapitein et al., 2011). These studies also showed that MT invasions were dependent on NMDA receptor activity.

How does synaptic activity regulate MT invasions of spines? It is well known that synaptic activity, in the form of long-term potentiation (LTP), increases calcium influx in dendritic spines through AMPA and NMDA receptors (Fortin et al., 2010; Lu et al., 2001). LTP-induced calcium influx activates several downstream pathways resulting in increased actin polymerization in individual spines that are potentiated (Bosch et al., 2014; Fukazawa et al., 2003; Honkura et al., 2008). A parsimonious explanation of how actin filaments (f-actin) could influence MT invasion of spines is that f-actin could serve as a “gate” or “ramp” in the axon shaft and could physically deflect MTs into specific spines that are potentiated (Figure 3). A more complex model would include both MT plus-end-associated proteins and actin-associated proteins interacting with one another to position MTs to polymerize into dendritic spines. The current data suggest that both processes could be functioning at the intersection of the dendrite shaft and the neck of dendritic spines (Merriam et al., 2013; Schatzle et al., 2018). The earlier study by Merriam and colleagues showed that actin polymerization after chemical LTP played a central role in MT polymerization into spines (Merriam et al., 2013). Additionally, this group showed that the actin-binding protein drebrin, which has been shown to interact with EB3 (Geraldo et al., 2008), decreased MT invasions when knocked-out and increased invasions when overexpressed. Consistently, the subsequent work by Schatzle and colleagues confirmed that f-actin in the dendrite shaft was instrumental for MT invasion (Schatzle et al., 2018). However, this group tested several actin-associated proteins, including drebrin, and found only cortactin knockdown decreased MT entries into spines, suggesting that cortactin, rather than drebrin, was necessary for MT entries. Moreover, this group suggested, based on using a mutant EB3 construct without a C-terminal acidic domain, that EB3 binding to other proteins is not required for MT entry. Together these results suggest that f-actin is required for MT entry of dendritic spines and activity-dependent actin remodeling in the dendrite shaft, below potentiated spines, plays an important role in deflecting actively polymerizing MTs into dendritic spines. Furthermore, actin-associated proteins are likely to be indirectly involved in MT entry of spines by sculpting the f-actin network at the dendrite shaft/spine neck junction.

Figure 3. Two potential mechanisms for MT-based entry and exit of material in dendritic spines.

A. The dynamic MT direct deposit model posits that the dynamic, peripherally localized MTs in dendrites are the primary source of polymerizing MTs into dendritic spines. Thus, motor proteins that prefer movement along dynamic MTs, such as KIF1A, use these dynamic MTs to transport cargo into dendritic spines. An example of such cargo are vesicles containing Syt-4 protein (likely DCVs). Dynein could then use the transiently present MTs to transport material out of spines along these MTs. B. The stable MT/kinesin to actin/myosin hand-off model suggests that material that is transported along the stable MTs in the central region of the dendrite by motors such as KIF5 cannot directly enter dendritic spines, since stable MTs do not enter dendritic spines. Rather, such cargo would need to engage myosin motor proteins, such as Myosin 5, which translocates cargo, such as the endoplasmic reticulum (ER) shown here into the spine (mitochondria also use this mechanism). In the example shown, KIF5 is transporting the ER on a stable plus-end-out MT, even though most of the stable, centrally located MTs are minus-end-out. With the bulk of the stable MTs being minus-end-out, dynein could hand-off material to myosin from distal regions of the dendrite.

The studies mentioned above would suggest that regulating MT polymerization kinetics would affect MT entries into dendritic spines. Indeed, a previous study incubated hippocampal neurons in low concentrations of nocodazole, which does not depolymerize MTs but decreases their dynamics (growth and shrinkage) and showed a decrease in MT entries of spines (Jaworski et al., 2009). However, one must consider that bath application of this compound will affect all MTs in the neuron, both presynaptically and postsynaptically, as well as in any glial cells in the culture. Thus, it is quite a non-specific treatment. However, a recent study used a conditional knockout (cKO) mouse to specifically affect postsynaptic MT dynamics (Zheng et al., 2022). They discovered that conditionally depleting the Kinesin-13 protein KIF2C/MCAK, a known MT depolymerizing protein (Ogawa et al., 2004), resulted in increased MT entries into dendritic spines under basal conditions (Zheng et al., 2022). Moreover, they showed that the presence of tyrosinated MTs in dendritic spines increased after BDNF treatment, consistent with a previous study (Hu et al., 2011), and this increase was dependent on KIF2C (Zheng et al., 2022). However, they also demonstrated that MT dynamics in the dendrite shaft were affected in KIF2C cKO neurons. Thus, it is not clear where KIF2C is acting, in the dendrite shaft, spine or both. Future work may be able to sort out this enigma.

Transport of cargo along MTs in dendrites and spines

As mentioned above, a fundamental function of MTs is to provide a substrate to transport cargoes long distances within cells. Thus, an additional question that arises is what might MTs be transporting into and out of dendritic spines and is there any specificity to this transport? MT polymerization rates are on the order of 5–15μm/min and MTs generally remain in spines that they target for tens of seconds to several minutes (Hu et al., 2008; Jaworski et al., 2009). Although this is a relatively short period of time, the rate of transport of motor proteins are generally on the order of 0.3–1μm/sec (18–60μm/min). Thus, the time MTs are present in spines would allow substantial transport of cargo into or out of spines. However, the motor proteins that enter spines along MTs would have to favor dynamic MTs. Markers for new/dynamic MTs include tyrosinated tubulin, whereas acetylated or detyrosinated tubulin is a marker for old/stable MTs. Thus, motors entering spines along MTs would likely prefer tyrosinated MTs. Generally, Kinesin-1 motor proteins prefer to track along acetylated MTs (Cai et al., 2009; Konishi and Setou, 2009), while Kinesin-3 motors are less selective, but prefer tyrosinated MTs (Guardia et al., 2016; Lipka et al., 2016). These data are consistent with the finding that KIF1A, a Kinesin-3 family member, was discovered to transport synaptotagmin-IV-containing vesicles (likely DCVs) directly into dendritic spines of cultured hippocampal neurons (McVicker et al., 2016). This group referred to MTs delivering cargo directly into spines as the “direct-deposit” model (Figure 3).

However, there are other mechanisms for trafficking cargo in and out of dendritic spines. Cytoplasmic material can move in and out of spines via regulated diffusion. Translocation of cytoplasmic proteins is a complex process, both spatially and temporally, depending on the type of protein (Bosch et al., 2014) and beyond the scope of this review. Two other mechanisms for trafficking membrane-associated cargo and organelles include membrane diffusion and MT motor to actomyosin hand-off (Dent, 2017). If MTs directly enter spines, why would it be necessary for a handoff of cargo to myosin motor proteins that would move the cargo into the spine along f-actin? One possibility is that MT entry into dendritic spines is transient. MTs are present in spines for only a few minutes. Thus, the MTs that enter spines are inherently dynamic, composed of tyrosinated tubulin. Any motor that does not use tyrosinated MTs to transport material could therefore not enter spines directly on these dynamic MTs (Figure 3). For example, cargo that is transported by Kinesin-1 (i.e., KIF5 motors) is likely to use stable, acetylated/detyrosinated MTs in dendrites, as mentioned above (Cai et al., 2009; Konishi and Setou, 2009). Moreover, recent data indicate many stable MTs in dendrites are located in the center of the dendrite shaft, in regions of dendrites near the cell body (Katrukha et al., 2021; Tas et al., 2017). For both reasons, cargo transported by kinesins that prefer stable MTs would require interaction with a myosin motor to be transported into dendritic spines along actin filaments. Interestingly, several organelles, including endoplasmic reticulum (ER) and mitochondria, which are known to use KIF5 (Cai et al., 2007; Tanaka et al., 1998), do not enter spines directly on MTs (Figure 3). Rather, these organelles require Myosin-Va to transport into dendritic spines (McVicker et al., 2016; Wagner et al., 2011). Similarly, recycling endosomes, which contain AMPA receptors, require Myosin-Va/b to enter spines (Correia et al., 2008; Esteves da Silva et al., 2015; Wang et al., 2008). Nevertheless, recent evidence indicates that KIF13A, a Kinesin-3 motor, is required for delivery of endosomally-localized AMPA receptors to synapses during LTP (Gutierrez et al., 2021). Although this study did not specifically test if these endosomes entered spines along MTs, they favor a Myosin-V-based handoff mechanism, based on the previously mentioned studies.

These aforementioned studies have focused on cargo delivery into dendritic spines. What about cargo removal from spines? Less is known about removal of material along MTs that enter dendritic spines. Based on their polarity, cytoplasmic dynein would be required for transport of material out of spines along MTs. Although such movement has not been shown directly, one study demonstrated that puncta of dynein intermediate chain (DIC) and neuroligin moved out of dendritic spines in tandem (Schapitz et al., 2010). However, MTs were not imaged in this study, so it is unclear if a MT was present in the spine during this event. Thus, these puncta may represent vesicles containing dynein, but they could move along actin filaments via myosin-based transport. Indeed, AMPA receptor endocytosis during LTD requires Myosin VI in cerebellar Purkinje neurons (Wagner et al., 2019). A recent study also demonstrated the MT motor protein KIF5A/C and adapter protein protrudin work together to remove endosomes containing Rab11 and AMPA receptors during LTD (Brachet et al., 2021). It is likely the interaction of endosomes, KIF5A/C and MTs occurs in the dendritic shaft due to the affinity of KIF5 motors for stable MTs (Cai et al., 2009; Konishi and Setou, 2009) and the location of stable MTs in the center of dendrites (Katrukha et al., 2021; Tas et al., 2017), as mentioned above.

However, motors can also influence MT dynamics. A recent study by the Shen group showed that dynein heavy chain (DHC-1), has an additional function to its well-documented involvement in cargo transport in C.elegans (Yogev et al., 2017). In their study, the authors showed that in DHC-1 mutants MTs have increased dynamicity and MT looping, resulting in cargo accumulations at the tip of the dendrite. This outcome observed in DHC-1 mutants is suggestive of an important role in stabilizing the MT structure and preventing aberrations in cargo transport. The study also indicates that dynein motors are not just cargo transporters, but help maintain the stability of the MTs in C.elegans dendrites. It is intriguing that MTs, which serve as railroads for transport of cargo via motor proteins, are also regulated by the same motor proteins.

Together, these studies indicate a complex and dynamic network of events governing entry and exit of MTs, material and organelles into and out of dendritic spines (Figure 3). Furthermore, there is clearly much more work to be done to determine both how the cytoskeleton functions in delivery of material to and removal from dendritic spines. Although these mechanisms are inherently interesting and important to study, how might they affect the role of spines in complex processes such as learning and memory?

Microtubule dynamics in learning and memory

It is becoming increasingly evident that MT dynamicity plays a critical role in typical brain function, particularly learning and memory formation. Long term potentiation (LTP), an electrophysiological means of assessing the formation of memories ex vivo (Nabavi et al., 2014), has been shown to be affected by alterations in MT dynamics. Hyper-stabilization of MTs with the chemotherapy drug paclitaxel results in a deficit of LTP in murine brain slices (Shumyatsky et al., 2005). Additionally, destabilization of MTs with the MT-depolymerizing agent nocodazole has been reported to reduce LTP within hippocampal mossy fibers (Barnes et al., 2010) and Shaffer collaterals (Jaworski et al., 2009) in brain slices. Therefore, it appears that disrupting the dynamic nature of MTs either by stabilization or destabilization disturbs brain function. However, it should be noted that studies that rely on pharmacological disruption of MTs result in both presynaptic (axonal) and postsynaptic (dendritic) disruption of MT dynamics. Thus, it is unclear whether presynaptic MT dynamics, postsynaptic MT dynamics and/or the subset of postsynaptic MTs that enter dendritic spines are the cause of these deficits.

Assessment of learning and memory with behavioral testing is another means by which the effects of MT dynamicity have been examined. MT destabilization with nocodazole is associated with both a general reduction of dendritic spines and decreased memory, assessed by contextual fear conditioning (Fanara et al., 2010). This study suggests that the structural plasticity of dendritic spines that is required for learning may be controlled through MT dynamics, specifically MT turnover. Another group detected a reduction in contextual fear conditioning if the MT stabilizing drug paclitaxel was injected into the dentate gyrus 30 minutes after training, whereas paclitaxel administration 8 hours after training enhanced contextual fear conditioning (Uchida et al., 2014). Moreover, injection of nocodazole 8 hours after training disrupted fear conditioning. These results led these researchers to posit that there is an early phase of hyper-dynamic MTs and a later phase of hyper-stable MTs after contextual fear conditioning. The idea that MT stabilization may contribute to learning was also reported in a study where the MT-stabilizing drug Epothilone D rescued spatial learning in an Alzheimer’s mouse model, as assessed by Y maze (Zhang et al., 2012). This result is consistent with the belief that Alzheimer’s results from the destabilization of MTs. Additionally, it is important to note that these general pharmacological treatments may be affecting MT-dependent transport in addition to MT dynamicity. Therefore, the disruption of LTP, memory formation and/or learning may be from altered MT dynamicity, disrupted MT-dependent transport, or the compounded nature of altering both processes.

When discussing the role of MT dynamics in learning and memory formation, it is important to consider the effects of MT-associated proteins (MAPs) on MTs. The MAP stathmin alters MT dynamicity through several mechanisms. Dephosphorylated or hypo-phosphorylated stathmin can weakly bind to tubulin heterodimers, sequestering them and preventing MT assembly (Di Paolo et al., 1997; Gavet et al., 2002). However, triple phosphorylation of stathmin at three key serine residues will inhibit its binding affinity for tubulin, allowing for MT assembly to commence (Di Paolo et al., 1997; Gavet et al., 2002). Additionally, dephosphorylated/hypo-phosphorylated stathmin appears to promote MT catastrophe by direct binding to protofilaments at the minus ends of dynamic MTs (Gupta et al., 2013). In addition to altering MT dynamics, stathmin has been identified as a contributor to neurogenesis, spinogenesis, and memory formation within the dentate gyrus of adult mice (Martel et al., 2016). Thus, stathmin may be a critical component for regulating MT dynamics in typical brain function.

Considering the regulatory function of stathmin on MT dynamicity, it is not surprising that several studies suggest synaptic plasticity and learning may be associated with stathmin. Knockout animals devoid of stathmin display deficits in LTP when assessed by cued and contextual fear conditioning, but not spatial learning tasks, such as the Morris water maze (Shumyatsky et al., 2005). This is likely a reflection of the expression profile of stathmin within the adult murine brain, where it is highly expressed within the lateral nucleus (LA) of the amygdala, as opposed to the hippocampus. Another report aligned with this proposed expression, in which stathmin knockout animals presented with enhanced extinction of cued fear conditioned memories, which was associated with decreased amygdala activity (Martel et al., 2012). Additionally, transgenic animals expressing an unphosphorylatable form of stathmin display a reduction in contextual fear memory, suggesting that phosphorylation of stathmin, which regulates MT dynamicity, may be the underlying mechanism driving its role in memory formation (Uchida et al., 2014).

Stathmin appears to be bi-phasically regulated by learning. This regulation appears to be driven by changes in phosphorylation and transient binding affinity for tubulin (Uchida et al., 2014; Uchida and Shumyatsky, 2015). In the early stage of fear conditioning, 15–60 minutes after behavioral training, stathmin is de-phosphorylated at its three key serine residues: Ser16, Ser25, and Ser38. This de-phosphorylation results in dephosphorylated or hypo-phosphorylated stathmin with increased MT binding affinity and MT-destabilizing activity. Throughout the second, late phase of fear conditioning, approximately 8 hours after training, there is a resurgence of phosphorylation at the Ser16 and Ser38 residues, which decreases stathmin’s MT-destabilizing activity and results in hyper-stabilized MTs. Local injection of paclitaxel during the early phase disrupts fear memory, however injection during the late phase enhances fear memory. These results confirm that it is the stathmin-mediated alterations to MT stability, rather than other effects of stathmin, that contributes to its function in this learning paradigm.

As discussed above, MTs play a large role in both presynaptic and postsynaptic trafficking of cargo within mature neurons. Therefore, understandably, stathmin-dependent regulation of MT dynamicity may have downstream effects on other processes like intracellular transport. The late phase of stathmin activity, associated with hyper-stabilized MTs, results in an increase of GluA2 AMPA-type glutamate receptor subunits at the synapse (Uchida et al., 2014). The hyper-stabilized MTs may be enhancing dendritic transport as a means of promoting memory formation, as it is well understood that AMPA-type glutamate receptors of critical for both contextual memory (Mitsushima et al., 2011) and synaptic plasticity (Correia et al., 2008). Interestingly, it has recently been suggested that the molecular motor KIF5C contributes to both spatial memory and contextual fear memory through transportation of RNA protein complexes near dendritic spines to allow for local translation (Swarnkar et al., 2021). The motor protein KIF2C also contributes to synaptic plasticity with conditional knockout animals displaying impaired working memory and memory maintenance, as well as conditioned fear learning, but through a different mechanism (Zheng et al., 2022). The MT depolymerizing properties of KIF2C are critical for its role in synaptic plasticity, suggesting that it may employ a mechanism similar to stathmin following learning. However, follow up studies are still necessarily to elucidate if KIF2C is affecting MT-dependent transport or affecting synaptic plasticity in another manner.

Conclusions and Future Directions

The studies described above clearly indicate that MTs play a heretofore underappreciated role in both pre- and postsynaptic structure and function. MTs have long been known to provide ‘railways’ for transport of cargos throughout dendritic and axonal arbors. However, until relatively recently MTs have not been implicated in synaptic function, either in the presynaptic bouton or the postsynaptic dendritic spine. The studies described here tell a different story. MTs appear to play seminal roles in neuronal synaptic function and plasticity. MT dynamics and MT-based transport of material must be coordinated for proper communication between neurons and disruption of these processes results in compromised learning and memory. However, there have been relatively few studies that have specifically focused on how MT dynamics and motor-based transport are coordinated. Several studies mentioned in this review describe how motor proteins prefer either more stable or more dynamic MTs, but relatively little is known about how changing MT dynamics, either at the plus or minus end, might specifically affect transport of cargoes in the neuron. Future work will undoubtedly shed new light on how these two processes are intertwined.

Although we did not discuss developmental and neurodegenerative diseases, such as Alzheimer’s Disease, research is beginning to show disruption of MT dynamics and transport in the underlying etiology of such conditions. Further understanding of the cell biology underlying the role MTs and their associated proteins play at the synapse will undoubtedly provide important foundational information that can be applied to the treatment of such devastating diseases.

Highlights.

• Microtubule dynamics and transport play important roles at the synapse

• Microtubule dynamics play a role in learning and memory

• Microtubules are associated with synaptic and dense core vesicles presynaptically

• Microtubules polymerize into and transport material into dendritic spines